FINANCE 14 APRIL
Much of the controversy over the Private Finance
Initiative (PFI), particularly in the NHS, has arisen because
of the high costs, which then translate into major service reductions
and distortions in planning and accountability. The government's
standard response to these public concerns is that the increased
costs of the PFI are offset by two things: private sector efficiencies
which generate better value for money than the public sector alternative;
and the transfer of risk from the public to the private sector.
But does this really bear out in practice?
To date, these claims have rested on assertion
rather than evidence. With the recent PFI project failures in
new IT systems for the Passport Office and Benefits Agency and
growing concerns about affordability and value for money in PFI
projects for health and education, the Treasury case looks increasingly
In making the case for PFI, the government has
been relying heavily recently on a report commissioned from Arthur
Andersen and Enterprise LSE, Value for Money Drivers in the Private
Finance Initiative. Described as "the first broadly based
survey of operational PFI projects", it has been widely cited
by ministers and others.
Most of the attention that the report has received
relate to two key "findings": first, that PFI is 17
per cent cheaper than conventionally financed public projects;
and, second, that 60 per cent of these "savings" derive
from risk transfer. These finds have been presented as drawing
a line under the debate on PFI and value for money. Given the
political use that is being made of the report in the context
of the public-private partnership proposals for the London Underground
and the mayoral election for London, the analysis deserves closer
Both claims were made on the basis of an analysis
of 29 PFI Full Business Cases (FBCs). Unfortunately, it is impossible
to evaluate this analysis satisfactorily since no information
is given on the relevant spending departments (eg, the Department
of the Environment, Transport and the Regions), the sectors involved
(eg local authority or NHS), or even on the project types involved
(eg, building, IT). We made two formal requests to the Treasury
Taskforce for information about the sampling methods used, the
sectors from which the projects came and the identity of the FBCs.
These were refused on the grounds that "the report is intended
to stand on its own". We will now examine whether it does.
Turning first to the claim that PFI is 17 per
cent cheaper than conventionally financed public projects. It
is important to note that Arthur Andersen is not saying that PFI
reduces the cost of services by 17 per cent; it is recording the
fact that on average PFI was judged to be 17 per cent cheaper
than a notional public sector equivalent scheme, the Public Sector
The economic appraisal methodology used in these
comparisons contains at least two disputable components: discounting
and the costing of risk transfer. For appraisal purposes, the
net present cost of the PFI and the PSC are discounted at a fixed
rate (usually 6 per cent real) over a period of up to 60 years.
The NPC of publicly financed projects tends to be high because,
by accounting convention, capital expenditure is scored in the
years in which it takes place, which tend to be the early years.
PFI payments, on the other hand, spread capital
costs over a longer period, so that the application of a discount
rate lowers its NPC by a greater amount. The higher the discount
rateand even the Treasury says that the 6 per cent rate
is "the top of the range"the greater the potential
advantage to PFI schemes. So PFI projects with considerably higher
cash costs can still have lower NPCs than their Public Sector
Furthermore, the 17 per cent figure is questionable
as it is an average calculated from all the business casesthe
percentage differences between the net present cost of the PFI
and the PSC were added up and the average worked out. When the
individual projects are analysed, we find that more than half
the total savings come from just one projectProject N (the
Prime Project to transfer the estate of the Department of Social
Security t the private sector). It and two other projects (S and
Y) together account for 80 per cent of the putative cost "savings".
When these projects are taken out, the average
difference between PFI and PSC reduces to the considerably less
impressive figure of 6 per cent. This margin is small enough to
suggest that, for most projects in the report, value for money
simply results from the frontloading of the PSC and the level
of the discount rate.
The Public Accounts Committee, in reviewing
the first Design-Build-Finance-Operate road schemes in 1997, recommended
that a range of discount rates be used in future economic appraisals.
It would be helpful if the government were to publish a comparison
of NPCs using such a range. This would help establish the extent
to which the value-for-money conclusions drawn in this study are
dependent on the discount rate used.
The second main finding of the reportthat
risk transfer valuations accounted for 60 per cent of forecast
cost savingsis equally tendentious. As the report's authors
point out, only 17 of the 29 projects examined included data on
risk transfer: the 60 per cent figure refers only to these cases.
In the remaining cases, which collectively accounted for more
than two-thirds of the NPC of all the projects examined, it was
impossible to say what part, if any, of the savings were attributable
for risk. As with the 17 per cent "savings", the amount
of risk transferred was highly sensitive to a small number of
projects. One project, project Y, on its own accounted for 80
per cent of the savings ascribed to risk transfer. It is striking
that this was the NIRS2 system for handling National Insurance
accounts, supplied by Andersen Consulting.
The project is currently running three years
late, and the extra cost to the taxpayer has been put at £53
million, according to the National Audit Office, including £37
million in compensation for wrong payments. Risk transfer is estimated
in the report at £8.7 million. Andersen Consulting has paid
£4.1 million in compensationfor late deliverybut
no further payment will be sought. This project has been the subject
of criticism in two reports by the Public Accounts Committee.
The moral of this tale is that value-for-money analysis is ex
ante: savings identified may not be realised in practice.
It is impossible to say whether the Passport
Agency's IT system, procured under PFI from Siemens Business Services
(SBS), was also included in the study (all projects are anonymous
but many can be identified from government reports). The cost
of that debacle was £12.6 million. But SBS has agreed to
pay just £2.45 million of the costs. The real bearers of
the risk in this case have been the public: the cost of overhauling
the Passport Agency following last summer's chaos has resulted
in the price of an adult passport being increased by £7.
(The Immigration and Nationality Directorate IT scheme, provided
by the same company, is currently running two years behind schedule.)
The reported "savings" for all 29
projects was £1,062.5 millionbut only £214.5
million (20 per cent) of this could safely be attributed to risk
transfer, and even that amount was heavily dependent on a small
number of projects. The only individual risk identified was construction
cost overrun: it accounted for less than 1 per cent of the total
savings. In other words, the source for 80 per cent of cost savings
could not be identified. These figures are useful indications
of how little is actually known about the putative cost savings
in the projects under review. They certainly do not support the
opinion of project managers surveyed for the report that the main
savings were in the risk transferred.
If we step back from the report's spin, a number
of basic facts stand out. First, the study does not substantiate
the central claim that PFI shows 17 per cent savings on conventionally
financed projects. Secondly, FBCs seem to be a poor source of
information about value-for-money drivers. Finally, in view of
the absence of key data about the sample, the findings are not
generalisable to other PFI projects.
It is remarkable that, after eight years of
PFI procurement, the Arthur Andersen report appears to be the
most the government can offer in response to widespread concerns
about the initiative. In the real world beyond business cases,
it is far from clear that PFI has passed the value-for-money test.
It is time the evidence base for the policy was subjected to thorough