Concept Economics

Australian Government spending on infrastructure projects has increased rapidly in recent years, especially over the course of 2009. In this paper, we examine the processes for project evaluation in the light of the Government’s commitment in the 2008–09 Budget to ‘[infrastructure] decision making based on rigorous cost–benefit analysis to ensure the highest economic and social benefits to the nation over the long term’ and to ‘transparency at all stages of the decision making process’. We find that, contrary to this commitment, significant projects have been approved either with no cost–benefit analysis or with cost–benefit analysis that is clearly of poor quality. Moreover, despite the commitment to transparency, very little information has been disclosed as to how most projects were evaluated.

In the longer version of this paper we also examine the cost–benefit assessment undertaken for the second largest infrastructure project the Government has committed to, which involves the construction of a rail link in Victoria. We find that lower-cost alternatives to the project were not taken into account in the evaluation, in particular the option of increasing capacity through improved efficiency and better governance of the rail network. Even taking that exclusion on board, we find that the appraisal that was approved by Infrastructure Australia (or, at least, the only version of that appraisal that has been made available) is seriously flawed, including errors of double counting and manifestly incorrect estimates of project benefits. Absent these errors, the project would generate benefits that fall well short of its costs.

We conclude by noting that high-quality project evaluations will not be made if governments do not see value in them. This appears to be the case in Australia, especially with respect to major projects. Nonetheless, we advance a number of proposals for improving the process, including transparency (which is now largely lacking), serious audits and reappraisal of projects at predetermined milestones, and steps to introduce greater rigour into key aspects of the analysis.

‘The core of public finance’, as Jurgen von Hagen (2006, p. 464) has succinctly put it, ‘is that some people spend other people’s money’. This separation of spenders and payers gives rise to a wide range of problems of accountability and control (which economists typically analyse under the rubric of ‘principal–agent’ problems), reflecting divergences of interest between these parties and the inability of voters and taxpayers to costlessly and perfectly discipline the behaviour of those who spend money on their behalf.

There are broadly three sets of control mechanisms that are commonly used to limit these risks:

i)        ex ante (based on forecasts rather than actual results) rules that shape taxing and spending powers;

ii)       budget processes that signal the opportunity cost of public funds and manage resource allocation so as to control, if not prevent, externalities between spending agents (including those associated with common pool problems); and

iii)      political competition and accountability that, however effectively or ineffectively, discipline ‘poor’ uses of resources and reward ‘good’ uses.

The role of formal project appraisal within these control mechanisms, and the effectiveness with which it is implemented within the Australian federation, is the central concern of this paper.

So how robust are our project evaluation processes? In examining this question, we start by setting out the nature and role of cost–benefit analysis, and especially its bearing on efficient resource allocation and on the control of principal–agent problems in government. That discussion highlights just how important cost–benefit analysis is to serious project appraisal, and to helping to control the risks inherent in a situation where very large projects, offering highly concentrated benefits but with very diffuse costs, are being vigorously advocated by powerful private interests.

On that basis, we examine the situation in telecommunications. In essence, neither the Howard Government (1996–2007) nor its successor placed any weight on systematic analysis of the costs and benefits of major telecommunications decisions.

The most spectacular recent instance is, of course, the decision to build a ‘National Broadband Network’ (NBN) with significant taxpayer funding. As the Government has stated that no cost–benefit analysis of this decision has been, or will be, undertaken, we carried out such an assessment. Our results suggest that the incremental benefits of the NBN, when compared to the counterfactual scenarios, do not justify the incremental costs.

Given that evaluation of project decision making in telecommunications, we turn to transport in the longer version of this paper. We outline some major trends in transport cost–benefit analysis in Australia, including those resulting from the creation of the Building Australia Fund and the establishment of Infrastructure Australia as a policy advisory body. To assess the quality of the evaluation processes, we undertake a detailed analysis of the east–west rail project in Victoria.

Although that project involves several components, some of which are not now proceeding (or have been deferred), it remains extremely large and has now received very substantial funding from the Commonwealth. However, this is a project which, even in its sponsor’s cost–benefit analysis, had benefits that were not far above costs. Our examination of that cost–benefit analysis in the longer version of the paper raises a number of concerns, including double-counting of benefits and substantial difficulties with the approach the cost–benefit analysis adopts to the calculation of the project’s ‘wider economic impacts’ (essentially, pecuniary externalities associated with the project).

In essence, cost–benefit analysis is a technique for evaluating collective decisions that hinges on the comparisons of the costs of a proposal to its benefits, where costs and benefits are valued in monetary terms. Cost–benefit analysis asks whether the sum of the amounts the individuals who comprise the community at issue would be willing to pay for the project to proceed exceeds the costs of that project. Generally, a project enhances wealth — in the sense of the aggregate monetary valuation of the community’s resources — if it meets a properly specified cost–benefit test.

Cost–benefit analysis can be viewed from four complementary perspectives.

We start by explaining the relevant context and then examine recent decisions in the light of cost–benefit analysis.

Ergas (2008a) sets out the background to recent telecommunications decisions. Two trends dominated the period leading up to the 2007 change in government.

First, an impasse developed in relations between Telstra and the Australian Government over the issue of upgrading the Australian telecommunications network to higher broadband speeds.

Second, the Australian Government engaged in a wide range of spending programs (with appropriations totalling close to $4 billion, in 2008 prices) aimed at promoting service upgrading, usually in regional areas, and implemented an ever broader and more draconian range of quality of service regulations.

None of these spending initiatives or quality of service regulations were ever

subjected to proper cost–benefit analysis (or if such analysis was undertaken, it was

never disclosed). However, an analysis by one of the authors found that in 1999 the

total benefits associated with addressing claimed service quality problems

(including in terms of consumer gains and network-related cost savings) were

between $644 million and $713 million in present value terms over the length of the

project life. These benefits were outweighed by the costs which (again in present

value terms) were estimated at $1387 million over the project life (Ergas and Hardin 1999).

The lack of attention to systematic evaluation of the costs and benefits of policy

initiatives has continued under the Rudd Government. The Minister for Broadband,

Communications and the Digital Economy, Senator Stephen Conroy, when asked

by the opposition whether a cost–benefit study of the proposed expenditure had

been carried out, said (according to a report in the Communications Day of 13 May

2009), that there was ‘no need’ for such a study, as ‘Labor’s commitment to build a

high speed broadband network has been clear … A range of studies have been

carried out all over the world that have investigated the economic impact of broadband.’ (Bartholomeusz 2009)

Since then, one study, by Professor Joshua Gans (2009), has been submitted as

evidence to a Senate inquiry into the NBN. Although its author notes that the

calculations are essentially back of the envelope, the submission suggests that the

social benefits of the NBN will exceed the costs. However, these calculations are

seriously flawed. These deficiencies are summarised in Appendix A of the longer

version of this paper. Even more seriously, however, Professor Gans’s submission

uses the wrong test for assessing whether a project is worth while: it compares total

costs and benefits, when the correct test is whether the incremental gains from the

project (relative to network capabilities in the base case) exceed the associated incremental costs.

Before turning to examine the project’s costs and benefits, it is useful to undertake a

wider consideration of the relevant decision. In particular, it is uncontroversial that

sensible policy evaluation requires a specification of the problem to be addressed

and of the policy options for addressing it. As a result, it is reasonable to ask what

precise problem the NBN is intended to resolve, and what other means might have been used to resolve it.

The Government’s primary concerns appear to be with the availability of broadband

access and its price. However, the data on availability that the Government has cited

actually refers to take-up of broadband services, and hence might be more indicative

of the demand for broadband than of its supply. This is all the more probable given

that broadband availability appears to greatly exceed demand, with some 80 per

cent of PSTN lines being connected to ADSL2+ enabled exchanges and close to

50 per cent of copper lines being short enough deliver very high speeds. Moreover,

competing hybrid fibre-coaxial networks (which currently deliver up to 30 Mbps

but which can, at relatively low cost, be upgraded to much higher speeds) either

fixed services account for a relatively small share of total broadband

3 Low incremental costs for hybrid fibre-coaxial upgrade are discussed in the Soria and

Hernández-Gil paper (2009), as well as in the Telstra documents (2008a, 2008b and 2008c). It is

its hybrid fibre-coaxial network in New Zealand to 100 Mbpss for NZ$10 million. The cost of

deploying the proposed FTTP network in those coverage areas is likely to be at least 10 to 20 times greater.

4 According to Telstra’s most recent annual results (released on 13 August 2009), Telstra’s

wireless broadband subscriptions doubled over the year to reach over 1 million (this figure

includes data card subscribers only; it does not include customers with 3G handsets). In

contrast, Telstra’s high-speed services (20 Mbps plus) had 241,000 high-speed subscribers in

June 2009, up from 160,000 the previous year. This represents about 10 per cent of Telstra’s

broadband customers. See http://www.telstra.com.au/abouttelstra/investor/docs/tls685-fyr2009esultsannouncement.pdf .

consumers’ low valuations of the incremental benefits of higher speed fixed network access.

A similar picture emerges regarding business access to high-speed broadband.

Competing, ubiquitous fibre networks cover all of the capital city central business

districts (CBDs). Larger business premises outside the CBDs are almost always on

direct fibre optic connections, even in non-metropolitan areas, as are premises such

as hospitals and government offices. Smaller businesses have access to business

parks, which are almost invariably on fibre access networks, and those smaller

businesses that operate in activities where high-speed communications are an

important element tend to locate in those business parks (where they can also

benefit from other economies of agglomeration). Symmetric high-speed services

over copper (such as BDSL) are available in virtually all urban locations and in

many regional centres. There is, in short, no evidence of any absence of business

There is also no evidence that suppliers of social services lack access to high-speed

services — indeed, the opposite is the case. In other words, availability does not

appear to be the constraint the NBN deployment assumes.

Australian broadband prices are in the upper half of OECD comparisons. However,

prices in a number of countries are distorted by subsidies, and those subsidies would

need to be added back, along with a mark-up to reflect the marginal social cost of

funds, for a welfare comparison to be made. Additionally and importantly, there is

significant competition in Australian broadband supply and key input prices are

regulated. Service supply to CBDs and business parks is intensely competitive, as is

the wiring of new residential estates. As for established premises in metropolitan

areas, broadband is widely provided by Telstra’s competitors using Telstra’s

Unconditioned Local Loop Service (ULLS), a regulated service that provides thirdparty

access to the copper pair. As is shown in Appendix B of the extended version

of this paper. Australian regulated ULLS charges are relatively low in urban areas,

while take-up of ULLS has increased very rapidly.

As for non-metropolitan areas, the case that supply is failing to keep up with

demand is also weak. Overall, these outcomes, like those above, suggest that the

5 Residential mobility and new household formation rates in Australia are relatively high. As a

result, consumers who value high-speed access will tend to move to locations at which access is

available and will incur low incremental costs from doing so. We are unaware of any evidence

of a residential housing price premium associated with access to high-speed broadband. These

elements suggest that latent demand, and welfare losses from lack of access, are likely to be low.

primary obstacles to take-up may lie in low customer demand, which implies low

customer valuation of any new network.

This is not to say that there are no issues with respect to investment in, and

upgrading of, Australia’s telecommunications network — the opposite is true. As

argued in Ergas (2008a, 2008b), the current telecommunications-specific access

regime vests enormous and unwarranted discretion in the regulator. In this industry,

as in others, such discretion creates a risk of time inconsistency; that is, of

reforming the regulatory arrangements (along lines already adopted in the energy

industries) so as to provide greater investor confidence, and then seeing whether

where service is commercially unviable, these could have been dealt with at

relatively low cost through a voucher scheme, which would have the merit of being

technologically and competitively neutral (Ergas and Ralph 2008). There is,

however, no evidence, at least in material disclosed to date, that the costs and

benefits of those options were assessed relative to the option of simply building a new network.

What then can be said about the costs and benefits of the new network? To examine

the underlying economics, we have used a cost model developed by Concept

network with a footprint covering 90 per cent of the Australian population by

modelling the construction cost of new infrastructure.

6 Simply put, time inconsistency refers to situations where a policy that is optimal (from the point

policymaker cannot commit to a policy, it may then find itself wanting to change its policy

Prescott 1977). Specific applications of the concept to regulated industries can be found in the

literature (Evans, Levine and Trillas 2008; Guthrie 2006; Levine, Stern and Trillas 2005). Ergas

(2009b) provides a test of whether ACCC decisions in telecommunications are time-inconsistent

(with the conclusion that they are).

7 Obviously, some care is required in the design of such an option. In particular, if there remains a

material threat of the Government expropriating the returns on that investment, for example by

subsequently building a network of its own, then socially desirable investment may be deterred.

Jullien, Pouyet and Sand-Zantman (2009) systematically discuss the conceptual issues involved.

8 The model was developed by Dr Dieter Schadt, and we are grateful for his assistance in this

respect. Obviously, he bears no responsibility for our use of the model’s results.

As regards capital costs, we have assumed a weighted average cost of capital

(WACC) in which the cost of equity is determined according to the capital asset

pricing model. This reflects three considerations. First, this investment substitutes

for private sector investment in competing infrastructure. Use of any other cost of

capital than that for the private sector alternative will distort resource allocation

between the public and the private sector (see, for example, Steiner 1974). Second,

the Government has confirmed on a number of occasions that it intends the project

to earn a commercial rate of return, suggesting that it values capital devoted to this

project at that rate of return. Third, investing in a new broadband network has a high

level of systematic risk. As a result, the Arrow–Lind conditions for use of the riskfree

rate as the discount factor (which depend on the assumption that the benefits of

the investment are independent of variations in overall incomes) do not hold in this

The model is designed to allow testing of the sensitivity of the results to a range of

variables. Setting these variables to their base case levels (which involves a GPONgigabit

passive optical networking-architecture), we estimate a final retail cost per

customer (on a nationally averaged basis) of just over $170 per month. This amount

is the cost of the access network plus the cost of backhaul to the service provider’s

network, and an allocation for usage and other retail costs. It is, in other words,

broadly comparable to the charge for a broadband service, minus the cost of any content.

While both the input assumptions and the outcomes are broadly consistent with

studies undertaken in other countries (see, for example, Analysys Mason 2008), the

cost estimates are sensitive to a range of assumptions, including, with respect to

consumer take-up rates and cutover arrangements, the extent of aerial deployment,

the project cost of capital, achievable operational efficiency improvements and the

quality of service provided. Variations in those parameters lead to a possible range

for unit per customer costs of between $125 per month and $225 per month. There

is also very significant variation in costs between metropolitan and nonmetropolitan

areas. Thus, for the most likely estimate of $170 per month, unit costs

in metropolitan areas are $133 per month, while those in non-metropolitan areas are just under $380.

9 In a classic article, Arrow and Lind (1970) showed that if a government project is ‘small’ (in

relation to the total wealth of taxpayers) and ‘the returns from a given public investment are

independent of other components of national income’, then the social cost of risk for project

flows that accrue to taxpayers tends to zero as the number of taxpayers tends to infinity. The

required assumption, in other words, is that the returns from the project are not related to (in the

sense of being dependent on) income from other investments in the economy.

Given these sensitivities, we have run a variant which seeks to minimise unit costs,

including by assuming that eventually all premises will subscribe to the service.

This variant, which also sets initial service quality at relatively low, but perhaps not

inappropriate, levels (in terms of the Committed Information Rate used to

dimension backhaul) and somewhat reduces the WACC, only slightly reduces unit

retail costs in metropolitan areas but could reduce unit retail costs in nonmetropolitan

areas to around $280 per month. Nonetheless, even these costs are

high compared to current charges. They are about double the level of current noncontent

payments for telephony and broadband service (that is, the sum of the

monthly rental and of the non-content component of DSL charges) in metropolitan

These costs need to be compared to alternatives. The most straightforward

counterfactual involves continuation and some upgrading of the current copperbased

network alongside progressive upgrading of the hybrid fibre-coaxial, with

copper delivering speeds of some 20 Mbps to 40 Mbps and the hybrid fibre-coaxial

delivering speeds of 50 Mbps to 100 Mbps. The costs of this scenario could be in

the order of one-third those of the NBN in metropolitan and regional areas, up to

around 80 per cent of the population. Remaining areas would primarily be served by

wireless, at costs that would be around one-half those of the NBN, with speeds of

10 Mbps to 30 Mbps. Regulatory reform that increased investment certainty would

make the progressive upgrading that took place in this counterfactual both quicker

and more extensive.

Incremental cost-based retail network charges for broadband service per

connectable premise under the counterfactual would therefore be in the order of

$50–$70 per month in metropolitan areas, rising to around $80–$100 per month in

regional areas, with a difference relative to the NBN scenario of around $75 per

month in metropolitan areas and of $120 per month in regional areas (noting that

the regional areas have less population coverage than is envisaged for the NBN, so

that the like-for-like comparison involves assuming a regional cost-based rate in the

NBN of around $210). Broadly speaking, the additional outlays (of $75 per month

in metropolitan areas and of $120 per month in regional areas) allow speeds to rise

to 100 Mbps in one step. However, this benefit is somewhat qualified by the fact

that deployment of the new network may take 7 to 10 years (if not longer), but the

10 They are even higher when compared to the access payments made by the average residential

premise, remembering that about 30 per cent of households do not subscribe to any form of

broadband service. Relative to those current average monthly payments, they are more than

twice the current average monthly payments in metropolitan areas and about four times those in

non-metropolitan areas.

prospect of that deployment may prevent the somewhat more limited, but sooner,

upgrades that would otherwise have occurred from occurring.

The question then is whether the valuation of the incremental speed associated with

the NBN outweighs the incremental costs. While there are some symmetric services

(such as very high-quality videoconferencing) that could benefit from higher

speeds, the difference in delay and overall service quality between (say) 30 Mbps

and 60 Mbps would only rarely be discernible. This is all the more so as once the

access network operates at reasonably high speeds; the relevant constraints on

service quality are likely to come from performance in the core network (that is, the

links between the first point of traffic aggregation and the global Internet), with

further increases in access network speeds having little effect. Holding all else

constant, it is therefore reasonable to expect the valuation of further reductions in

download time to decline as average download times themselves decline (that is, as

speeds increase). The median consumer’s willingness to pay (WTP), taken as a

function of service bit rate, would, in other words, increase more slowly for

successive increases in speed.

This can be illustrated using the standard Becker (1965) time-allocation model.

Naturally, the incremental benefits are higher for those earning higher wages (that

is, who have a higher opportunity cost of time), but, all else being equal, the

applications, the valuation of speed will therefore be significantly concave, though

the location of the valuation curve will shift over time, as ‘bandwidth-hungry’

applications develop and as a greater number of consumers attain a utility level

from access to broadband that induces them to obtain the service (that is, that

exceeds the service’s start-up costs). Appendix C of the extended version of this

paper details the model.

Given these considerations, we have undertaken an assessment of the costs and

benefits for the project. As with any such assessment, a substantial number of

assumptions need to be made. In this section, we explain the approach we have

adopted.

11 Goolsbee and Klenow (2006) use Becker’s framework to compute the consumer benefits of

access to the Internet, but they do not examine the welfare effects of greater download speeds.

Computing incremental benefits of a project requires specification of a baseline

scenario with which to compare the project scenario. We consider three such

scenarios, which entail the following alternative comparisons.

gradually increases to 60 Mbps by year 6, and remains on 60 Mbps.

increase (but at a slightly slower rate than the baseline) to 60 Mbps by year 9,

and then has speeds of 100 Mbps from year 10 onwards.

consumer is on the same path as the baseline. The median consumer then goes

on to 100 Mbps at a later time than with the project Scenario A — from year 15

onwards.

with a relatively high willingness to pay — those consumers in the top quintile.

This is, in other words, a targeted version of the project, with the aim of serving

only high WTP areas.

These speed adoption paths are plotted in figures 6.1 and 6.2.

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

For scenario C, we assume the same WTP curves, except that the relevant consumer

that is targeted when the project is built has a much higher WTP. By construction,

the top 25 per cent of consumers are assumed to have initial valuations exceeding

$100, and we take this consumer as the representative consumer that is targeted by

the project under scenario C. We also assume that the growth rate of this

consumer’s WTP is 5 per cent per year (figures 6.3 and 6.4)

0

20

40

60

80

100

120

140

160

10 20 30 40 50 60 100

0

50

100

150

200

250

300

10 20 30 40 50 60 100

The next step is to combine the speed adoption path and the WTP curves to

calculate a WTP curve for the baseline and the project under each scenario, and also

compute the difference in the path of WTPs under each scenario. This gives us the

incremental WTP curve — it is the path of benefits that the representative consumer

would receive if the project went ahead, instead of the baseline.

These are plotted in figures 6.5 to 6.7.

EVIDENCE-BASED

POLICY

-20

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

WTP Baseline WTP NBN WTP Difference

0

20

40

60

80

100

120

140

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

WTP Baseline WTP NBN WTP Difference

0

50

100

150

200

250

300

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

WTP Baseline WTP NBN WTP Difference

We then compute the present value of the stream of benefits under each scenario,

using a range of discount rates. The numbers in the tables are the present value of

the consumer’s WTP, expressed in dollars per month. Thus, the number in the first

row of the first column ($1273) is the present value of the future stream of benefits

that the consumer expects to receive.

The tables also compute the ‘monthly constant equivalent’, which is the constant

amount that a consumer with the relevant discount rate would be willing to pay in

each and every month over the next 20 years to receive the given stream of benefits.

So, for example, under scenario A, a consumer with a 4 per cent discount rate

would be willing to pay $0.52 every month (rounded up to $1 in the table) for the

next 20 years to not have the NBN, and instead receive the benefits under the baseline.

To arrive at a final assessment of costs and benefits, we subtract the incremental

costs computed earlier from these incremental benefits. Note that under scenario A

the incremental benefits are negative, and so accounting for the incremental

monthly costs that were computed earlier (of around $75 per month in metropolitan

areas and of $120 per month in regional areas), the NBN has incremental net

benefits that are negative. For all the other scenarios, the incremental benefits of the

NBN are far below the incremental costs; indeed, it is difficult to conceive of

credible scenarios for the NBN that would make its incremental costs fall below the

incremental benefits (that is, result in the project yielding net benefits to Australia).

Indeed, in all of the scenarios, the incremental upgrading path is always the most

socially beneficial.

% $ $ $ $ $ $

4 1237 1228 –9 91 90 –1

8 846 834 –13 86 85 –1

Scenario A

12 612 599 –13 82 80 –2

4 1237 1249 11 91 92 1

8 846 852 6 86 87 1

Scenario B

12 612 615 3 82 82 0

4 1237 1540 303 91 113 22

8 846 1002 156 86 102 16

Scenario C

12 612 695 83 82 93 11

NPV = net present value

To what extent do these results depend on the willingness to pay curves? To

examine this question, we have conducted a sensitivity analysis on the WTP

assessment, by examining ‘enhanced’ WTP curves in each of the three scenarios.

The results of the enhanced WTP analysis are very similar to the standard analysis.

The ranking of the three scenarios remains unchanged, with the delayed project

(scenario B) and the targeted project (scenario C) becoming slightly more attractive

from an incremental benefit point of view. The incremental benefits under scenario

A actually fall and become more negative under the enhanced WTP setting. In other

words, increasing the willingness to pay for higher speed reduces the attractiveness

of the NBN option, essentially because it also increases the density of demand in the

midspeed tier (and hence increases the relative value of the options that involve

incremental development of the access network).

Put slightly differently, the enhanced WTP curves have higher marginal WTP at

lower speeds relative to the original analysis. Under the NBN the consumer misses

out on those relatively high marginal gains in the early years, even though the

consumer eventually receives high absolute benefits. This fact, combined with the

logic of discounting, means that scenarios B and C become more attractive, while

scenario A becomes less attractive.

% $ $ $ $ $ $

4 1608 1609 1 118 118 0

8 1087 1070 –17 111 109 –2

Scenario A

12 776 753 –23 104 101 –3

4 1608 1648 41 118 121 3

8 1087 1108 21 111 113 2

Scenario B

12 776 787 11 104 105 1

4 1608 1918 310 118 141 23

8 1087 1247 160 111 127 16

Scenario C

12 776 861 85 104 115 11

NPV = net present value

Overall, the results are relatively robust because WTP is concave in speed, network

coverage and in the rate at which upgrades are deployed, while costs are convex at a

understate the consequences of this fundamental feature of the situation, as we

consider a median user, while there are substantial numbers of users — especially in

In the counterfactual, the loss incurred on these users is limited by the more limited

coverage of the upgrading; in the NBN, these costs are incurred in full and relatively soon.

To examine the net benefits and costs of the NBN, we examine a scenario

(scenario D) that is intentionally conservative as far as service quality is concerned,

as it involves speeds under the base case rising to only 20 Mbps, which is less than

the hybrid fibre-coaxial networks can currently provide.

12 Costs are, in other words, concave in speed up to 30–60 Mbps and then leap at the discontinuity.

Costs are always likely to be convex in the geographical breadth of deployment and in the speed

of deployment, while the WTP gains in each of these dimensions are likely to be concave.

13 This probably reflects the fact that WTP is correlated with human capital endowment, and

human capital — especially that associated with ‘information’ activities — tends to be

concentrated in metropolitan areas (see O’Flaherty 2005).

to 20 Mbps in the fourth year and remain there.

increase to 100 Mbps by the sixth year of the NBN project, where they remain.

These speed adoption paths are plotted in figure 6.8.

0

20

40

60

80

100

120

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

Baseline NBN

For consumer willingness to pay, we assume a monthly WTP of $50 for 10 Mbps,

increasing to $104 for 100 Mbps. To estimate aggregate willingness to pay, we

assume that all consumers are alike. We also assume an annual growth rate of

3 per cent in WTP at the lowest speed, but assume that the growth rate increases as

we move up the WTP curve. Thus, we assume an annual growth of 3 per cent for

WTP for 10 Mbps, with the growth rate rising to 3.9 per cent for 100 Mbps. The

initial annual WTP curve for scenario D and its growth rate over time is shown in figure 6.9.

Our next step is to combine the speed adoption path and the WTP curves to

calculate a WTP curve over time for the baseline and the project, and also compute

the difference in the path of WTPs under each scenario. This gives us the

incremental WTP curve — it is the path of benefits that the representative consumer

would receive if the project went ahead, instead of the baseline. These are plotted in figure 6.10.

0

300

600

900

1200

1500

1800

2100

2400

2700

10 20 30 40 50 60 70 80 90 100

Under the scenario D baseline, we assume that retail prices are $30 per month in

metropolitan areas, and $50 per month in non-metro areas, which gives a national

monthly cost recovery retail price of $32.90 (assuming an 85%–15% split between

urban and non-urban areas).

For the NBN, under scenario D, and the assumption of a CIR (Committed

Information Rate) of 1 Mbps, the engineering cost model provides estimates of

break-even retail prices of $128 per month in metro areas, and $313 in non-metro

areas, for a national average cost recovery price of $155 (again assuming a 85–15

per cent split between metro and non-metro areas).

To compute aggregate costs and benefits, an assumption must be made about the

path of demand. Under scenario D, the NBN engineering cost model assumes an

S-shaped take-up pattern over time, with 50 per cent of the population taking up the

service by year 6 and a saturation rate of 80 per cent. For the baseline case, we

assume a slightly more rapid take-up rate, with the same starting percentage as

under the NBN but with a final saturation rate of 90 per cent. These two demand

profiles are shown in figure 6.11.

0

500

1000

1500

2000

2500

3000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

WTP Baseline WTP NBN WTP Difference

0

1,000,000

2,000,000

3,000,000

4,000,000

5,000,000

6,000,000

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

NBN

Baseline

Finally, we can put all of this together and compute aggregate costs and benefits

under the baseline and the NBN, and compute the present value of net incremental

benefits of the NBN (table 6.3). The numbers in the table are the estimated present

value of the net incremental benefit of the NBN, relative to the baseline. The

estimates suggest that undertaking the project will result in a social loss in present

value terms of between $13.9 billion and $20.4 billion, depending on the discount rate chosen.

% $(2009) billion

6 –20.4

7 –19.2

8 –18.1

9 –17.2

10 –16.2

11 –15.4

12 –14.6

13 –13.9

Since we have assumed that the willingness to pay for the NBN far exceeds that for

the baseline, it is clear that the key drivers of the NBN’s social losses are the large

capital and operating costs of the project.

In fact, the central result of our modelling can be expressed in terms of the familiar

condition for replacement investment. More specifically, it is economic to replace

existing network, in each case adjusted for relative service quality (which we do

through the willingness to pay calculation). It turns out that the NBN would satisfy

this condition only if the present value of the additional cost of deploying and

operating the NBN, compared to even the 20 Mbps scenario, were no more than

$14 billion (evaluated at a discount rate of 13 per cent) to $24.7 billion (evaluated at

a discount rate of 6 per cent).

Put slightly differently, assuming a midpoint discount rate of 10 per cent, it is

irrational to spend more than $17 billion on the NBN, even if the alternative is a

world in which the representative consumer cannot obtain service in excess of

20 Mbps and even if demand for high-speed service is rising relatively quickly. This

amount of $17 billion is well below current estimates of the costs the NBN will

involve, especially if it is to serve non-metropolitan areas. Alternatively and more

realistically, if the base case (that is, the alternative to the NBN) is one in which the

representative consumer is assumed to ultimately have access to 40 Mbps (rather

than 20 Mbps as above), then it is inefficient to proceed with the NBN if the present

value of its incremental costs of deployment and operation, evaluated at a 10 per

cent discount rate, exceed $10.6 billion, which is below the lowest bound of the

It may be thought that these estimates understate the gains from the project because

they do not take account of wider economic and social benefits. While it is likely

that use of higher speed access lines will allow productivity gains, we would expect

those gains to be reflected in consumers’ and businesses’ willingness to pay for that

use. As a result, treating the productive efficiency gains as an added benefit

amounts to double counting. As for wider social benefits, it is unclear what they

consist of, and whether they are indeed greater under the project than under the

counterfactual. Moreover, to the extent such social benefits exist, there must be the

question of whether the project is the most efficient means of ensuring their

Rather, it is our view that the estimates understate the likely project-related social

costs. Thus, it seems probable that, evaluated at a rate of return that reflects the risks

themselves are a transfer, the distortions associated with financing them through

taxation are not, and need to be added to the social costs of the project. In contrast,

14 Using 40 Mbps is especially realistic if sorting is allowed to occur — that is, if account is taken

of the fact that suppliers will target those customers who place a high value on speed, and that

those customers will have incentives to choose locations (for instance, at which to site offices) that offer such access.

15 If these benefits can be obtained at lower cost under some alternative option, then the cost

increase from forgoing the use of that lower cost option (that is, from using the NBN to deliver

those benefits, rather than the cheaper alternative) is a net cost to the project and should be

treated as such in the analysis.

16 Claimed wider benefits such as the promotion of tele-medicine seem very difficult to credit.

With respect to tele-medicine, it is not clear what residential medical applications require access

to residential fibre optics, short of a future being projected in which individuals will have CAT

scanners in their homes. As for GPs and medical centres, there is no evidence that network

access costs and speeds have any effect on their use of tele-medicine; Paolucci et al. (2009)

survey the literature on the effectiveness of tele-medicine and do not find such evidence.

Finally, hospitals are generally already connected to high-speed access networks and would be

so under the factual and counterfactual alike.

17 Of course, the project might be profitable were it given a monopoly or regulatory protection

from competition (say, through an exemption from the merger laws that allowed it to acquire

assets that would otherwise act as an effective competitive constraint). However, were that the

case, then the efficiency costs of such a monopoly would need to be brought to account in the

cost–benefit analysis.

under the counterfactual, taxpayer outlays would be limited to any vouchers used to

subsidise demand by consumers in high-cost areas. Moreover, the prospect of

taxpayer financing of the project’s losses can lead to moral hazard, as well as to

direct political interference in project decisions, diminishing the productive

efficiency with which the project is pursued. Our estimates, however, do not gross

up financing costs for the difference in the value of private and public income (that

is, for the marginal social cost of funds) and assume the project is deployed and

operated at least cost.

Additionally, the NBN project, whatever its merits, will create risks to the integrity

of the regulatory system. First, the Australian Government will be both the primary

investor in a major competitor and the industry policymaker and regulator, creating

sovereign risk for private investors and introducing potential distortions to policy

and regulatory decisions. Second, the NBN may involve some form of joint venture

between entities that would otherwise have the scope to compete on a head to head

basis, with the associated dangers of collusion. Third, there will be strong pressures

for geographically uniform pricing, which can add distortions not only to resource

allocation but also to competition (for example, if restrictions or taxes on bypass are

used to protect the flow of cross-subsidies). These costs are not taken into account

in our estimates.

At the same time, our estimates of the project benefits do not take account of

offsetting equilibrating processes and therefore tend to overstate them. In particular,

it is clear that, in the counterfactual, those consumers that place the greatest value

on high-speed access will generally have such access, for two reasons: first,

suppliers will have incentives to provide it, including through geographically

targeted upgrades; and second, over 10 to 15 years , geographical mobility is

relatively high, and consumers will sort themselves geographically in a way that,

inter alia, reflects the valuations they place on different forms of broadband access.

As a result, the population that gains access to very high-speed broadband in the

NBN world relative to the counterfactual is likely to be that segment that places the

lowest valuation on broadband access. To that extent, our estimates, which do not

allow for this sorting process, exaggerate the gains from NBN deployment. This is

all the more the case as our counterfactual scenario (scenario D) assumes relatively

low speeds would be available should the NBN not proceed.

This overstatement of project gains is accentuated by our approach to estimating net

benefits, which compares the willingness to pay for the incremental speed the

project provides to the incremental cost of providing that speed. However, whether

benefits are realised depends to a significant extent on the entity’s future pricing

policies. For example, if prices are set at average costs, then some potential utility

gains will not be realised (as those consumers who value the project output at more

EVIDENCE-BASED

POLICY

than incremental cost, but less than project average cost, will not consume its

cost–benefit analysis for the road link may be undertaken on the basis of potential

social gains; however, the tolls may lead to some users whose valuations exceed

marginal costs (and hence who are counted towards the cost–benefit analysis’s

estimate of benefits) not actually using the road, causing realised benefits to fall

below assumed levels. Because we do not discount our estimated benefits for this

Finally, we have not costed the most natural alternative — which is simply to delay

the project and re-examine its economics every few years. This option to delay is

likely to have high value, particularly if it is accompanied by regulatory reform that

addresses the current disincentives to invest. Such an option would allow any public

investment to be more narrowly targeted to areas of genuine and durable market

failure and would reduce both the risk of asset stranding and of significant

deadweight losses due to the tax financing of project losses.

In short, we believe our estimates overstate the likely gains and understate the likely

costs from the NBN.

All that said, the notion of wider productivity benefits from broadband deployment

Access Economics that:

using fibre (51.6 per cent) and most schools in provincial locations also reported using fibre

(46.5 per cent)’, most schools ‘use download speeds of up to 4 megabits per second, which is

the lowest download speed range used in the FCS baseline survey. This disparity may be due to

affordability of the service or the specific contractual arrangements negotiated, throttling and

issues relating to the availability of suitable online curriculum resources and tools.’

19 Obviously, were perfect lump sum taxes and transfers available, then no such social costs would

eventuate. Project charges to users would, in such a world, be set to marginal costs, and any

fixed costs would be covered through public transfers. Unfortunately, such perfect lump sum

taxes and transfers are not available, and hence it may be efficient to impose break-even

constraints (or at least some degree of fixed cost recovery) on public suppliers. The welfare

costs of any such constraints then need to be taken into account.

20 ‘Access Economics predicts that a national high-speed broadband network would mean

economy-wide productivity growth 1.1 per cent higher after ten years compared to if the

network was not built.’ Senator the Hon Stephen Conroy, Minister for Broadband,

Communications and the Digital Economy, speech to CeBIT Australia 2009 AusInnovate

Conference, 12 May 2009. The Minister goes on to say: ‘It is worth noting that Access

Economics views this as a conservative estimate.’ However, as discussed below, the comparison

Access Economics makes is to a world in which only dial-up service is available (noting that as

of the time of writing, 70 per cent of Australian households subscribe to some form of broadband).

… economy-wide multifactor productivity levels would be around 1.1 per cent higher

in an Australian economy with HSBB [high-speed broadband] available everywhere

relative to an Australian economy without any HSBB after ten years. That is, the

average annual growth rates in productivity would be around 0.1 percentage points a

However, as the Access Economics report plainly states, these productivity gains

are relative to an economy in which only dial-up service, or similarly very low speed

access options, would otherwise be available. Moreover, it is also plain from

the Access Economics report that the numbers cited are no more than assumptions,

albeit ones Access Economics believes to be conservative for the comparison being made.

To take account of these differences, we believe that the Access Economics

estimates of productivity gains should be set to one-third to one-half their initial

levels, given that 70 per cent of households now have some form of broadband

access. Additionally, account needs to be taken of the likely crowding-out effects of

the public expenditure. We use a simple macroeconomic model with crowding out

( > 0 ) to assess the likely impacts. The results, set out in table 6.4, are expressed as

the present value of the cumulative change in GDP over a twelve year period,

discounted to the present at a discount rate of 7 per cent (the rate used by Access

Economics) and put in 2009 dollars. Broadly, the results suggest that cumulative

GDP declines, despite an assumed increase in productivity.

Table 6.4 shows the present value, in 2009 dollars, of the cumulative 12-year

change in GDP due to construction of the NBN, for a range of values of

productivity increase and of extent of crowding out of other investment.

0.3 –12 –17.1 –22.2 –27.3 –32.4 –37.5

0.4 –7.6 –12.7 –17.8 –22.9 –28 –33

0.5 –3.2 –8.3 –13.4 –18.5 –21 –23.6

Note: A discount rate of 7 per cent is used, for comparability with the Access Economics (2009) results.

This loss is not directly comparable to that derived from a comparison of

incremental project costs and consumer valuations; however, some component of it

— that part that reflects distortions due to the burden of taxation — could properly

be added to the cost–benefit analysis loss (as that loss is calculated without regard

to the difference between the private and public value of income). Unfortunately,

this component is not separately identifiable, being simply an element in the

assumed crowding-out parameter.

In short, under both the Howard and Rudd governments, important

telecommunications decisions have been made without formal, transparent

assessment of costs and benefits. Our review — both of the quality of service

regulations implemented by the previous government, and of the proposed NBN —

suggests such an assessment would conclude that the policies at issue impose costs

than exceed the relevant benefits.

The case studies set out above and in the long version of this paper suggest that at

least some important infrastructure decisions are being taken on the basis of little

evidence and in at least some instances, inadequate analysis. This is an obvious

concern given the scope poor infrastructure decisions have to reduce capital

productivity and hence lower living standards in the longer term. Mounting

evidence of inefficiencies in the way our infrastructure is run — with the search for

‘ribbon-cutting’ opportunities displacing adequate investment in maintenance,

causing a rapidly growing maintenance deficit that is well documented in Victoria

and New South Wales (NSW Audit Office 2006, Victorian Auditor General 2008)

— only adds to the concerns. What then can be done to strengthen the evaluation process?

Ultimately, the quality of evaluation depends on the value governments place upon

it. Governments that view project evaluation as merely a nuisance that stands in the

way of the decisions they want to take, and that believe they can get away with no

evaluation or poor quality evaluation, will, over time, invariably succeed in

devaluing the evaluation process. This has, we believe, occurred in Australia in recent years.

In part, this simply reflects a loosening of government budget constraints due, first,

to sustained economic growth and, second, to a belief that the global financial crisis

meant that high levels of government spending were not only feasible, but also

desirable. As the threat of recession loomed, confused reasoning led to a belief that

infrastructure investment could legitimately be claimed to be a tool of

macroeconomic policy, even though, in an economy with monetary and aggregate

fiscal policy instruments, infrastructure investment should play no role in

stabilisation policy and cyclical conditions should not affect the timing or extent of

infrastructure outlays, other than through their effects on projected demand and on

the shadow prices of inputs (effects which, properly analysed, can suggest that

infrastructure projects should be deferred, rather than accelerated, during

There are, however, also longer term forces at work. These forces reduce the

effectiveness of accountability and increase the attractiveness of infrastructure

decisions as elements in rent-seeking bargains.

The first is the ever greater blurring of responsibility for infrastructure between the

Commonwealth and the States, and the progressive loosening, by the

Commonwealth, of budget constraints at a state level. This reduces the electoral

accountability of, and electoral pressure on, State Governments, while reducing the

opportunity cost that State Governments incur for poor investment decisions. To

some extent, the Commonwealth has sought to offset the resulting moral hazard by

imposing performance obligations on the States, such as the evaluation

requirements built into Auslink. However, much as with foreign aid, these

requirements typically bear only a very indirect link to the outcome being sought

(which, in this case, is quality decisionmaking) and readily become (at best) ‘tickthe-

box’ constraints that are often easily gamed (as the quality of compliance is

rarely monitored, and, when monitored, even more rarely acted upon). Threats of

conditionality have little credibility, especially when doing so would impose a

significant political cost on the Commonwealth itself. Again, much as with foreign

aid (see Azam, Devarajan and O’Connell 1999; Brautigam 2000; Knack 2001;

Alesina and Weder 2002; Bardhan 2005; Easterly 2006; Moss, Pettersson and van

de Walle 2006; and Janus 2009), the result is a degradation in institutional quality

and in ultimate outcomes.

These issues associated with fiscal federalism have become even more complex

with the creation of the Building Australia Fund and of Infrastructure Australia.

Although there can be merit in coordinated approaches to infrastructure selection,

there can be little doubt that the new mechanisms create significant incentive

21 Bureau develops a non-Walrasian model with an external constraint, a monetary policy

instrument and fiscal policy. While no policy instrument should be thrown away, his main result

is that macroeconomic considerations should enter into the evaluation of infrastructure

investment only to the extent that the consequences of that investment are orthogonal to those of

the macroeconomic instruments. As for the impacts of cyclical factors on the cost–benefit

analysis, where public assets will compete with private assets (as in the case of the NBN), then

the costs of those public assets will rise during recessions, even in the presence of Keynesian

unemployment; see, for example, Johansson (1991, pp. 122–3). Additionally, to the extent

demand expectations are reduced, this should lead to lower infrastructure investment.

problems. To the extent to which the projects they fund are worth while, that

funding may simply displace funding of those projects by the States themselves, but

with higher transactions costs and possibly poorer monitoring and other

The second factor that has contributed to a decline in the quality of project

evaluation is the growing involvement of the private sector in the design,

construction, financing and operation of major infrastructure projects, both through

the contracting out of almost all aspects of project implementation and perhaps

especially, through public–private partnerships (PPPs). While these may have

complex, and often undesirable, impacts on the quality of public administration (see

for example, Estache and Martimort 1999). In particular, because the incentives are

high powered (that is, the private party secures substantial gains from reducing costs

under the contract), these arrangements increase the returns to rent-seeking and to

tainted deals between governments and private sector suppliers. Particularly with

PPPs, the effects are then threefold: they concentrate the gains from the project (as

some share of these is now captured by the private participant), and, by so doing,

increase the payoffs from collusion between the public decisionmaker and the

project’s private beneficiaries; they allow crucial aspects of the project to be

cloaked in commercial commerciality, thus reducing the transactions costs of

collusion; and they relax (or, more properly, are widely but incorrectly claimed to

relax) the public sector budget constraint. Each of these effects induces a

deterioration in the efficiency of decisions and overall outcomes.

Ultimately, PPPs are only as good as the governments that make them; and if

governments are intent on poor decisions, these partnerships can not only make

22 Obviously, if the Commonwealth funding were simply matching grants associated with the pure

spillover effects of state infrastructure decisions — that is, a Pigouvian subsidy — the issue of

displacement would not arise. Conversely, if the projects are so poor that they would never have

been undertaken by the states then there will indeed be a ‘flypaper’ effect and aggregate

infrastructure outlays will rise (on which see, generally, Brennan and Pincus 1990; as per

Brennan and Pincus, this is a case where the grant pushes spending to the corner solution).

23 The question of how to design multi-level funding institutions and associated cost–benefit

analysis processes so as to deal with these effects has received some attention in the EU,

although with few readily implemented results to date; see Florio 2007.

24 The ‘power’ of an incentive structure is determined by the extent to which the agent to whom

that incentive structure applies can secure for itself the gains from cost reductions (or other

improvements in performance). Incentives are said to be ‘high powered’ when the agent secures

a large share of the gains (as in a fixed price contract); conversely, they are ‘low powered’ when

the agent’s share of any gains is small (as in a cost-reimbursement contract).

those decisions more (privately) profitable but allow them to be locked in through

A third factor, which is yet to fully play itself out, is the recourse to hypothecated

funding sources for long-term infrastructure finance, most notably the Building

Australia Fund. While economic theory yields ambiguous results as to the effects of

which earmarking it can reduce the quality of public expenditures.

First, earmarking implies inflexibility in the allocation of revenues among

competing uses. If the earmarking is substantive, in the sense of being effectively

constraining, social rates of return are unlikely to be equalised at the margin across

uses. Tax rates, expenditure levels or more likely both, will be distorted as a consequence.

Second, reserving revenues to a program gives it a monopoly over those revenues,

encouraging and potentially perpetuating technical inefficiency in its supply.

Third, earmarking can facilitate rent-seeking by allowing the interest groups that

benefit from the hypothecated revenue stream to focus their activities more

effectively. Rather than competing against other interest groups for a larger share of

general revenues, the relevant groups can limit their efforts to seeking an increase in

(or protecting from erosion) the hypothecated fund. At the same time, the political

commitment they secure is potentially made more credible by the earmarking,

increasing both the ‘price’ that the interest groups are willing to pay in exchange

and the resources they are willing to dissipate in obtaining it. Rent-seeking

coalitions therefore become easier to create and sustain, and the aggregate costs to

the community from rent-seeking rise, as Kimenyi, Lee and Tollinson (1990) found

in their study of the US Highway Trust Fund.

Fourth, these adverse consequences are made all the greater by the risk created by

earmarking of fiscal illusion; that is, of the hypothecated revenues not being as

visible as other forms of public revenue and expenditure.

25 This is similar to the ‘Landes–Posner effect’, whereby an independent judiciary increases the

extent of rent-seeking by making it easier for legislators to lock in tainted deals (Landes and

Posner 1975).

26 For example, earmarking may be a way of increasing the credibility of promises, reducing the

inherent incompleteness of the implied contracts between government and the public. As well as

any direct benefits arising from greater credibility of commitments, this may allow proponents

of programs to signal the quality of the programs, of the proponents or both. Thus, in the model

of Brett and Keen (2000), a commitment to dedicate revenues to a particular use, which is of

value to the public but would not be of value to a ‘poor-quality’ politician, can support a

separating equilibrium in which politicians signal their quality to the electorate.

All of these factors create risks that the new earmarked funds, though they may

increase spending on infrastructure, could reduce the quality of that spending.

Set against these long-term forces, project evaluation is a relatively weak reed, and

the effects of changes to evaluation processes alone may well be relatively small.

Nonetheless, we would suggest three areas for reform.

should not be publicly disclosed as a matter of course. Instead, most cost–benefit

analyses are never released, and those that are are often difficult to locate.

Governments should also regularly publish, in readily accessed form, the cost–

benefit analysis rankings of those projects they have decided to proceed with and

those they have considered and rejected (as is done in Finland, for example). Were

disclosure of cost–benefit analyses routine, the fact that a cost–benefit analysis had

not been conducted on a particular project would become more obvious, as would

the relative quality of the cost–benefit analyses that had been carried out.

improving the efficiency of principal–agent relations, both by allowing principals to

better assess the outcomes of the efforts made by agents and by deterring collusion

between agents and third parties (see Mookherjee and Png 1989). The introduction

of an independent auditor, whose interests are separate from those of the party being

audited, increases the likelihood of poor conduct being detected, including when

that conduct takes the form of bias (for instance, associated with ‘excess optimism’

The auditing we believe desirable would take two forms. To begin with, there is

substantial merit in having independent reviews of all cost–benefit analyses for

‘mega-projects’ (say, projects with projected outlays in excess of $500 million).

This could be done by an office answerable to Parliament, rather than forming part

of the Executive. Such an office could be similar to the Congressional Budget

Office in the United States. Were establishing such an institution considered too

radical, at the very least adequate specialist resources should be provided to a

parliamentary standing committee to engage the kind of forensic analysis required.

This is not to cast doubt on the Australian National Audit Office, but rather to

suggest that its competence, and standard form of operation, are not especially well suited to this task.

As well as this form of review, there is a pressing need for much more to be done in

terms of post-completion review of projects. Although a few useful post-completion

27 The pervasiveness of these forms of bias in transport assessments is amply documented by

Flyvbjerg, Bruzelius and Rothengatter (2003).

reviews of cost–benefit analyses have been undertaken (BTE 2001; BTRE 2007a,

2007b; NSW Audit Office 2006; NSW Auditor General 2005; NSW Treasury 2008;

Victorian Auditor General 2009), these are ad hoc, which limits their effectiveness

both as instruments of accountability and as a means of learning from experience.

The Auslink program mandated post-completion reviews; unfortunately, this

requirement has not been rigorously enforced. We believe it should be.

Mandating systematic and transparent post-completion review could have farreaching

consequences. To begin with, it would force Commonwealth and state

entities to more properly document and archive material related to the cost–benefit

analyses and the cost–benefit analyses themselves. In contrast, as matters currently

stand, cost–benefit analyses are typically undertaken before the final form of

projects is determined, and then never updated. Additionally, little investment is

made in documenting cost–benefit analyses and in ensuring the integrity of the

documentation chain. A genuine system of post-completion reviews would require

all of those deficiencies to be addressed. At the same time, such reviews could be

used both to benchmark jurisdictions and to more effectively learn from mistakes.

In short, we would strongly endorse — and argue more should be done to

implement — the conclusion Little and Mirrlees (1994, p. 206) reached in

reviewing, after two decades, the impact of their great cost–benefit analysis manual:

If good project appraisal warrants expenditure, as we argue, so does good appraisal of appraisal.

group of people engaged in project evaluation. There are still many complex

technical issues to tackle in Australian project evaluation — including the selection

of the criterion function (where, unfortunately, the use of Benefit–Cost Ratios is

still widespread, despite its well-known deficiencies), the treatment of the marginal

social cost of funds (which is usually ignored), the determination of the discount

rate (often set in a manner that is somewhat arbitrary), the assessment of changes in

service quality and reliability (which is particularly important in public transport, as

well as in communications), the appropriateness or otherwise of corrections for

‘optimism bias’ (which, in the authors’ opinion, are likely to be ineffective at best

and distorting at worst), the role of ‘wider economic benefits’, and so on. While

many of these issues are well traversed in the literature (if not in the practice) of

project evaluation, there are other important issues that are relatively underexplored,

such as the implementation of cost–benefit analysis in the context of

the design of incentive-compatible evaluation schemes for structures such as Infrastructure Australia.

There is consequently considerable potential for cooperative research across

jurisdictions, and for using that research, and its dissemination, as an instrument of

ongoing training for both practitioners and users of cost–benefit analysis. Moreover,

that process could help give greater standing to the ‘profession’ of project

evaluation and help define a community of those involved in project evaluation

across different areas of infrastructure policy. There is an important role here for the

Bureau of Transport and Regional Economics and also for the Productivity

Commission. Thus, the Productivity Commission could, much as it did in regulation

review, issue ‘information notes’ recommending particular approaches to the

technical issues analysts face. While we do not believe there is one ‘right’ approach

to all of these issues, and hence would not favour mandatory standardisation across

the States, that should not impede the exchange of views and the fostering of

comparability of analyses across jurisdictions (so that the effect of different

approaches can be identified). Much has been done in this respect by the Australian

highlights the task that remains.

Infrastructure investment is a cost, not a benefit; a means, not an end. This proposition, which is obvious to economists, is as utterly alien to contemporary Australian politicians as the notion of comparative advantage was to their predecessors.

The costs and risks of this approach to infrastructure have also been known for many years. There are surely many echoes in current telecommunications decisions of the tendency, identified by Butlin, Barnard and Pincus (1982, p. 294) in their analysis of the development of the Post-Master General’s Department, for Australian public enterprise to provide ‘services that were too large, too quickly supplied and too cheap’. That so little should have changed is not encouraging.

To recognise this, however, is not to imply that no value should be placed on good appraisal; on the contrary, it is one of the protections taxpayers deserve to have.  Testimonials of commitment to ‘evidence-based policy’ notwithstanding, shaping an environment in which project appraisal can effectively discharge this task remains as great a challenge as it has ever been.

Overall, our review suggests the following conclusions:

In our view, these outcomes are driven by governments that see little real value in major project evaluation. They may see merit in evaluation of essentially routine decisions (such as the decision to place a new roundabout or improve a road surface) or in cost-effectiveness analysis of the options available for meeting predetermined goals (such as improving bus transit in a congested area), but not in the full analysis of objectives and options (including the option of not spending taxpayers’ money). This, we argue, reflects the impact of a perception (initially due to strong economic growth, and then to a belief that the global financial crisis justifies greatly increased outlays) that public funds have a negligible opportunity cost. This perception has been accentuated by the growing blurring of accountability in the Australian federation, which reduces the budget disciplines on the States, and the blurring also of responsibility for financing infrastructure as between the public and private sectors (which, whatever its other merits, increases the return to rent-seeking deals between governments and private infrastructure developers). Together, these trends risk making cost–benefit analysis merely a box to be ticked, rather than an exercise that has real value, not least to government itself.

We are not optimistic that changes to cost–benefit analysis processes alone can counteract these powerful trends. Nonetheless, we think three changes would have merit:

The Little–Mirrlees rule suggests that the value of proper project appraisal is at least 10 per cent of the value of projects. With Australia spending ever more on infrastructure, these are gains well worth seeking. Whether they can be achieved is obviously an open question.

Harvard University Press, Cambridge, Mass.; London, England.

Alesina, A. and Weder, B. 2002, ‘Do corrupt governments receive less foreign

Arrow, K.J. and Lind, R.C. 1970, ‘Uncertainty and the evaluation of public

Azam, J.P., Devarajan, S. and O’Connell, S.A. 1999, ‘Aid dependence

reconsidered’, Policy Research Working Paper No. 2144, The World Bank.

vol. 75, no. 299, pp. 493–517.

Stockholm, Sweden.

Brennan, G. and Pincus, J. 1990, ‘An implicit contract theory of intergovernmental

Brett, C. and Keen, M. 2000, ‘Political uncertainty and the earmarking of

Bureau, D. 1985, ‘Cohérence entre choix des projets et politique de régulation

—— 2008b, ‘Setting access prices: a critique of the ACCC’s approach in

—— and Hardin, A. 1999, ‘Annuity and levelisation issues in forward looking cost

models: the case of telecommunications’, Paper presented at the ACCC

Depreciation Forum, 30 September.

—— and Ralph, E. 2008, ‘A policy framework for a new broadband network’, in

Estache, A. and Martimort, D. 1999, ‘Politics, transaction costs, and the design of

regulatory institutions’, World Bank Policy Research Working Paper no. 2073,

http://ssrn.com/abstract=620512

Evans, J., Levine, P. and Trillas, F. 2008, ‘Lobbies, delegation and the underinvestment

Florio, M. 2007, ‘Multi-government cost–benefit analysis, shadow prices and

Northampton, Mass, pp. 1–46.

Gans, J. 2009, ‘The right policy for telecommunications and broadband’,

Submission to the Senate Select Committee on the National Broadband Network, 18th June.

Goolsbee, A. and Klenow, P.J. 2006, ‘Valuing consumer products by the time spent

Guthrie, G. 2006, ‘Regulating infrastructure: the impact on risk and investment’,

Jullien, B., Pouyet, J. and Sand-Zantman, W. 2009, ‘Public and private investments

in regulated network industries: coordination and competition issues’, mimeo.

Kimenyi, M.S., Lee, D. and Tollinson, R.D 1990, Efficient Lobbying and

Knack, S. 2001, ‘Aid dependence and the quality of governance: cross-country

Kyland, F. and Prescott, E. 1977, ‘Rules rather than discretion: the inconsistency of

Landes, W.M. and Posner, R.A. 1975, ‘The independent judiciary in an interestgroup

Levine, P., Stern, J. and Trillas, F. 2005, ‘Utility price regulation and time

vol. 57, no. 4, pp. 447–78.

Little, I.M.D. and Mirrlees, J.A. 1994, ‘The costs and benefits of analysis: project

appraisal and planning twenty years on’, in Layard, R. and Glaister, S. (eds),

England; New York, NY.

Moss, T., Pettersson, G. and van de Walle, N. 2006, ‘An aid-institutions paradox? A

review essay on aid dependency and state building in sub-Saharan Africa’,

Working Paper no. 74, Centre for Global Development.

Mookherjee, D. and Png, I. 1989, ‘Optimal auditing, insurance, and redistribution’,

NSW Treasury 2008, ‘Circular: revised project size/risk thresholds for the

submission of business cases and gateway reports’, NSW TC 08/07.

Paolucci, F., Ergas, H. Hannan, T. and Aarts, J. 2009, The effectiveness of health informatics, in press.

Posner, E. 2001, ‘Controlling agencies with cost–benefit analysis: a positive

Soria, B. and Hernández-Gil, F. 2009, Exploring potential natural monopoly

properties of broadband access networks, Paper presented at the 19th European

Regional Conference of the ITS, 20 September.

Spence, D.B. and Cross, F. 2000, ‘A public choice case for the administrative state’,

Steiner, P.O. 1974, ‘Public expenditure budgeting’, in Blinder, A.S., Break, G.F.,

The Brookings Institution, Washington D.C, pp. 241–361.

Telstra 2007a, Application to ACCC for exemption from standard access

obligations in respect of the Singtel Optus HFC network, 17 December 2007.

—— 2007b, Application to ACCC for exemption from standard access obligations

in respect of the SingTel Optus HFC Network, Schedule A, 17 December 2007.

—— 2007c, Application to ACCC for exemption from standard access obligations

in respect of the SingTel Optus HFC Network, 17 December 2007, Schedule A,

Annexure 2, Expert report of Michael G. Harris: Use of HFC to deliver

broadband services, 12 December 2007.

Von Hagen, J. 2006, ‘Political economy of fiscal institutions’, in Weingast, B.R.

University Press, N.Y, ch. 26.

Wildavsky, A. 1966, ‘The political economy of efficiency: cost–benefit analysis

vol. 26, no. 4, pp. 292–310.