STRENGTHENING EVIDENCE-BASED POLICY IN THE AUSTRALIAN FEDERATION  -  17-18 AUGUST 2009

Henry Ergas and Alex Robson

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. These problems are aggravated by the fact that the spenders

themselves are not a monolithic entity. Even if spenders as a whole face the collective

consequences of their decisions, each individual spending unit (such as a Minister, a Department

or a territorial level of government) may view the stock of available public funds as a ‘common

pool’ (like an open seas fishery), which it can draw on at a fraction of the resulting opportunity cost

while still garnering for itself all or the bulk of the political benefit. The scope to transfer the costs of

wasteful projects to future generations, which have little or no voice in the political process, as well

as to future governments (which will bear the political consequences of ‘pulling the plug’ on failed

ventures), then makes the risks of inefficient outcomes all the greater.

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

ex ante rules that shape taxing and spending powers; 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 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.

The specific focus is on the processes used in the economic evaluation of major infrastructure

decisions. Particularly since the election of the Rudd Labor Government in 2007, very significant

increases have occurred in public infrastructure outlays. According to the Minister for

Infrastructure, Transport, Regional Development and Local Government, Anthony Albanese:

And to these amounts must be added the Government’s commitment to the construction of a new

national broadband network, which could involve a cost of up to $43 billion.

Many of these decisions involve individual projects whose costs exceed a billion dollars; if those

projects’ costs exceed their benefits, the result is to make future generations poorer. The public

stake in proper project evaluation is therefore great and indeed, has been stressed by the

Government itself. Thus, in its 2008-09 Budget, the Government committed 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

of project evaluation in Australia. Although there is a long history of use of Cost-Benefit Analysis

(CBA) (see Dobes 2008 for a survey), recent years have seen increased emphasis placed on

other, less rigorous, approaches to project evaluation (see Dobes and Bennet 2009 and Ergas

2009). And some important infrastructure decisions appear to have been taken by Australian

governments without any systematic assessment of their costs and benefits.

So how robust are our project evaluation processes? In examining this question, we proceed as

follows. We start by setting out the nature and role of CBA, and especially its bearing on efficient

resource allocation and on the control of principal-agent problems in government. That discussion

highlights just how important CBA 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 CBA has been, or will be, undertaken of this

decision, we carry out such an assessment, using an engineering cost model to estimate the

project’s incremental costs. We also estimate a range for consumer Willingness to Pay and its

evolution over time. We then define some counterfactual scenarios, including one that involves

postponing deployment of the NBN. 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. We

outline some major trends in transport CBA 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 CBA, had benefits that were not far above costs. Our

examination of that CBA raises a number of concerns, including double counting of benefits and

substantial difficulties with the approach the CBA adopts to the calculation of the project’s ‘wider

economic impacts’ (essentially, pecuniary externalities associated with the project).

Overall, our review suggests the following conclusions:

investment, or more broadly, that would focus on securing greater efficiency from the

existing capital stock. Simply put, infrastructure investment appears to be viewed as a

benefit, rather than a cost;

are then compounded by evaluations that are too vulnerable to ‘fudge factors’. In a

Gresham’s Law of evaluation, bad evaluations (often by consultants) can drive out good,

given that they trade at equal values.

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 pre-determined 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 CBA 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 CBA processes alone can counteract these powerful

trends. Nonetheless, we think three changes would have merit:

not been subjected to economic project evaluation;

and post-project completion. In contrast, there is little or no such audit currently, and in

many instances, CBAs are not even updated, maintained or properly archived after the

initial ‘go/no go’ decision is taken;

government, preferably in an independent entity, such as the Productivity Commission.

The Little/Mirrlees rule (discussed below) suggests that the value of proper project appraisal is

at least 10 percent 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.

THE NATURE AND ROLE OF COST-BENEFIT ANALYSIS

We start our survey of Australian project evaluation by setting out some background to the role of

Cost-Benefit Analysis (CBA) in the public finance process.

In essence, CBA 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.

CBA can be viewed in four, complementary, perspectives.

First, CBA is related to (though not identical with) the basic equi-marginal condition for overall

efficiency in resource allocation. Thus, given a cardinally measurable objective function and

perfect knowledge of the effect on welfare of any decision, it is a condition of an optimal set of

Henry Ergas and Alex Robson

decisions that the marginal dollar of public expenditure has a benefit equal to that of the marginal

dollar of private expenditure (thus assuring that the overall level of public expenditure is optimal)

and that the benefit of a marginal dollar of public expenditure is equalised across programs,

projects and project elements. Because CBA aggregates willingness to pay across agents with

different marginal valuations of income, it is not a perfect measure of underlying utility (and hence

cannot be treated as an ideal social welfare function); nonetheless, taking that important caveat as

given, one would at least question whether a set of public decisions was optimal if it did not

maximise the aggregate benefits obtainable for given aggregate costs or minimise the aggregate

costs required to obtain a given aggregate benefit, in each case, measured using CBA.

Second, set against the backdrop of a given portfolio of projects, CBA can be used to evaluate

whether one or more public projects should be added to or removed from that portfolio. In other

words, CBA is a tool that can be used to asses whether wealth (the difference between the

aggregate valuation of outcomes and the cost of obtaining those outcomes) would be increased

by the decision to (say) proceed with a particular project, compared to the relevant alternatives

(which may involve doing nothing, deferring or otherwise varying the project, or proceeding with

an alternative project). Whether enhancing wealth in this sense is either necessary or sufficient for

a project to be worthwhile is a complex issue. Without going into the details of that discussion, it

seems reasonable to suggest that projects that fail properly specified cost-benefit tests should be

looked at very carefully, and found to have other, significant, redeeming features, before they are

allowed to proceed. By the same token, if a project has benefits that (evaluated in willingness to

pay terms) clearly exceed its costs (i.e. the net benefits of forgone alternatives), it seems

reasonable to presume that absent compelling reasons to the contrary, society would gain were it

to proceed.

Third, CBA is an instrument that the principals in public sector governance can use to improve the

CBA to be properly conducted, the evaluator has to go through the key elements of proper policy

analysis, namely, specification of the objectives being sought, identification of the alternative

options for achieving those objectives, an analysis of the likely consequences of each such option,

and consideration of the risks (including of error in evaluation) that attach both to each option and

to the assessment as a whole. Moreover, whether a properly specified CBA has been undertaken

should be readily verifiable. As a result, the requirement to carefully assess, and report, the costs

and benefits of decisions can improve the quality of decision-making and reduce the information

asymmetry between principals and agents. In doing so, it can:

interests of the principal, come to be determined either by the agenda of self-interested third

parties or by the agent’s own interests and aspirations. Capture risks are especially great in

areas such as infrastructure where the benefits of greater spending are highly concentrated

while the costs of that spending are very widely spread;

policy commitments that differ from (and may undermine) those of the public, as can occur

when traditions dictate a particular response to particular types of problems, even though

the circumstances which may have made that response appropriate in previous times no

longer hold;

4 See for example Adler and Posner (2006), Posner (2001), Spence and Cross (2000).

Henry Ergas and Alex Robson

decisions as would be warranted;

tendency to confuse the unfamiliar with the improbable, and more generally to underestimate

the risks of a course of action when its benefits seem large, and under-estimate

which decisions are taken and by highlighting anomalies, such as differences between

project appraisals in the valuation of common elements; and

identification of variances from those expected values and encouraging analysis of the

causes of those variances, which (importantly) can facilitate learning and continued

improvement in the decision-making process.

Ultimately, all of these effects mean that cost-benefit analysis is never merely an analytical tool:

rather, as Aaron Wildavsky emphasised many years ago (Wildavsky 1966), it is inevitably an

instrument in shaping bureaucratic structure and process, both within each public sector body and

between that body, the other elements of the public sector with which it interacts, and the wider

political system.

Fourth and last, CBA can be anchoring device that reduces undesirable policy instability. Thus,

consider a situation in which government alternates between two majorities, 1 and 2, whose

pivotal voter preferences are for project portfolios A and B respectively, as indicated in Figure 1.

Majority 1

Majority 2

5 As Thomas Schelling observes in his introduction to Roberta Wohlstetter’s classic study of Pearl Harbour, “There is a

tendency in our planning to confuse the unfamiliar with the improbable. The contingency we have not considered

seriously looks strange; what looks strange is therefore improbable; what is improbable need not be taken seriously.”

Schelling in Wohlstetter (1962) at p. vii.

Henry Ergas and Alex Robson

Although A and B are on the welfare frontier (which is convex because asymmetric information

makes all social transfers costly, i.e. imposes a deadweight loss or excess burden of taxation), the

expected value of the sequence of policy choices is the interior point C. Systematic application of

CBA could then lead to a point such as D, which falls short of the frontier (because CBA is not a

perfect measure of underlying utility and in any event, will involve errors in application) but which

is still superior to C (Laffont 2000). In essence, in this setting, CBA acts to give greater legitimacy

and wider credibility to policy choices, and hence makes them less vulnerable to displacement.

The resulting gains are obviously all the greater if, in the alternative world, policy instability leads

to programs being initiated, causing significant costs to be incurred, only to be terminated before

their benefits are obtained.

For all of these reasons, CBA has been widely recognised as a central component of proper

project appraisal. We therefore now turn to consider its role in recent infrastructure decisionmaking,

beginning with telecommunications and then proceeding to transport.

TELECOMMUNICATIONS

We start by explaining the relevant context and then examine recent decisions in the light of costbenefit

analysis.

The background to recent telecommunications decisions is set out in Ergas (2008). Two trends

dominated the period leading up to the 2007 change in government.

First, an impasse developed in relations between Telstra and the Commonwealth government

over the issue of upgrading the Australian telecommunications network to higher broadband

speeds. Simply put, Telstra was unwilling to undertake that upgrading without substantial

regulatory reforms that would have protected the required investments, while the government was

unwilling or unable to convince the regulator (whose discretionary powers would have been

significantly curtailed, as they had been in the energy industries) to accept the proposed reforms.

Second, the Commonwealth engaged 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 was 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 $1,387 million over the project life (Hardin and Ergas 1999). Despite this, the

Howard government proceeded with the quality of service requirements and made them even

more stringent in subsequent years.

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

continued under the Rudd government. Upon being elected, the government launched a request

for proposals for a network that could deliver a capability of at least 12 Mbit/s to 98 per cent of the

Australian population. That process closed at the end of January 2009 without any of the

Henry Ergas and Alex Robson

proposals received being considered satisfactory. Following the failure of that process, the

government announced on April 7 2009 that it had decided to build a National Broadband Network

(NBN) which – at a cost that could reach $43 billion – would provide fibre optic connections (at

speeds of 100 Mbit/s) to all premises in towns of 1000 people or more, which equates to coverage

of some 90 percent of the population. ‘Next generation’ wireless and satellite would be used to

deliver 12 Mbit/s to the remaining 10 percent, with simultaneous deployment of the new network

(which is to operate on a ‘wholesale only’ basis) in urban, regional and rural areas. Additionally,

the government announced that it would initially own at least 51 percent of the new network,

although it could “sell down” its equity interest within 5 years of the network being fully

constructed.

As regards evaluation of the costs and benefits of the proposals, Communications Minister,

Senator Stephen Conroy, when asked by the Opposition whether a cost-benefit study had been

carried out of the proposed expenditure, said (according to a report in Communications Day of

May 13th, 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.” Senator Conroy also said,

still in responding to Opposition queries, that “We don’t need any more studies, any more cost

benefit analyses, to know that this is an infrastructure investment that this country is crying out for.

How many reports do they [the Opposition] need before they just accept that they have been

responsible for leading [to] Australia falling behind the rest of the world?” The Finance Minister,

Lindsay Tanner, then confirmed that no cost-benefit study had been carried out and that none was

envisaged, because “We just formed the view that in effect we had to make the clear decision that

said this is the outcome we are going to achieve come hell or high water because it is of

fundamental importance to the future of the Australian economy’’ (Bartholomeusz 2009).

Since then, one study, by Professor Joshua Gans, has been submitted as evidence to a Senate

inquiry into the NBN (Gans 2009). 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. They understate costs, as they exclude

obvious cost items such as operating and maintenance expense, depreciation and backhaul. They

overstate likely revenues, as they confuse wholesale and retail prices and outlays. They also

confuse benefits and transfers, treating reductions in profits as a social gain (when they are

merely a transfer). These deficiencies are summarised in Appendix A. Even more seriously,

however, Professor Gans’ submission uses the wrong test for assessing whether a project is

worthwhile: 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. Even if the total benefits (as measured by aggregate willingness to pay)

exceeded total costs (which is far from clear), this would say nothing about whether society gained

by undertaking the project, as the project’s incremental benefits might be less than its incremental

Before turning to examine the project’s costs and benefits, however, 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 which

6 Under reasonable conditions, the correctly measured willingness to pay for a quality improvement that raises the

Bradford and Hildebrand 1977.

Henry Ergas and Alex Robson

might be available for addressing it. As a result, it is reasonable to ask what the precise problem

the NBN is intended to resolve is, and what other means might have been used to do so.

The government’s primary concerns appear to be with the availability of broadband access and its

price. However, the data the government has cited as to availability 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 percent of PSTN lines being connected to ADSL2+ enabled exchanges

and close to 50 percent of copper lines being short enough deliver very high speeds. Moreover,

competing hybrid fibre coax networks (which currently deliver up to 30 Mbit/s but which can, at

relatively low cost be upgraded to much higher speeds) either pass or run very close to some 60

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

access.

A similar picture emerges as regards 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

Nor is there evidence that suppliers of social services lack access to high speed services –

indeed, the opposite is the case. Thus, according to the Commonwealth Government’s own

National Baseline of School Broadband Connectivity 2008, “The majority of schools in

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

also reported using fibre (46.5 per cent).” However, despite having high speed access, 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.” In other words, availability does not

appear to be the constraint the NBN deployment assumes.

7 Low incremental costs for HFC upgrade are discussed in Soria and Hernández-Gil 2009, as well as in Telstra 2008A,

HFC network in New Zealand to 100 Mbit/s 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.

8 Thus, according to Telstra’s most recent annual results, released on August 13 2009, Telstra's wireless broadband

subscriptions doubled over the year to reach over 1 million (this does not include customers with 3G handsets; rather,

it solely covers the number of data card subscribers). In contrast, Telstra's high speed services (20Mbit/s 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.

9 Moreover, as we note below, residential mobility and new household formation rates in Australia are relatively high. As

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

and incur low incremental costs from doing so. Moreover, 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.

Henry Ergas and Alex Robson

As to prices, Australian broadband prices are in 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 third party access to

low in urban areas, while take-up of ULLS has increased very rapidly. Thus, in the 18 months to

March 2009, the number of ULLS lines in all areas more than doubled to well over 600,000. This

increase has been accompanied by a growing presence of ULLS access seekers in Telstra

exchanges, with the number of Telstra exchanges with three or more competing ULLS access

seekers rising from 155 to 300, while the number of exchanges with less than three access

seekers shrank. It seems unlikely that retail prices that substantially exceeded costs (i.e. high

mark-ups over input charges, which are low and regulated) could persist in the face of such largescale

competitive entry; rather, the more logical inference is that retail prices for high speed

broadband broadly reflect costs, which are relatively high in Australia due to population

support.

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

Thus, these areas have already benefited from large scale subsidies both to supply and to

demand, but the gap in penetration levels persists (though it has tended to narrow). Tasmania,

which the government has selected as the first location for deployment of the NBN, is a case in

point. State and Commonwealth initiatives have seen well over $100 million spent in Tasmania on

high speed access facilities in non-metropolitan areas (such as fibre deployments in Ballarat,

Cooma and the Green Triangle, as well as fibre optic deployment in new ‘sea changer’ estates)

appear to have low utilisation and in some cases, are poorly maintained and hence are being run

down. Overall, these outcomes, like those above, suggest that the 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 the 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 timeinconsistency,

10 The average spacing between residential premises in metropolitan areas in Australia is nearly twice that in the United

States and over three times that in Europe.

11 This includes the funding secured by Senator Harradine in 1996 in exchange for his support for the initial tranche of

the Telstra privatisation. Since then, there has been little transparency in the Commonwealth funding allocations, so

the actual outlays may be significantly greater. There has also been considerable State government spending on

broadband, including the core funding for the TasCOLT FTTP trials. As for take-up levels, while the gap between

Tasmania and the rest of Australia has narrowed in recent years, it remains substantial (with Tasmanian broadband

penetration about half that in the ACT, which has the highest levels in Australia), reflecting lower income and

education levels and a higher average age in the population.

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

maker) ex ante turns out not to be the optimal policy ex post. If the policymaker cannot commit to a policy, it may then

find itself wanting to change its policy ex post (say, after a regulated firm has made an irreversible investment

decision), regardless of what it promised ex ante. Such an approach to policy is said to be time-inconsistent – see

Henry Ergas and Alex Robson

reforming the regulatory arrangements (along lines already adopted in the energy industries) so

as to provide greater investor confidence, and then seeing whether socially desirable investment

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 what material has been 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

model describes the rollout of a fibre to the home (FTTH) network with a footprint covering 90 per

cent of the Australian population by modelling the construction cost of new infrastructure.

The model’s geographic coverage is differentiated at a high level between metropolitan and nonmetropolitan

areas. A bottom-up modelling approach is used, building the required infrastructure

up from assumptions on the:

From these assumptions, the model derives the average size of the metropolitan and nonmetropolitan

exchange areas and the length of main and distribution cables.

Current exchange areas are further aggregated to account for the substantially longer reach of

fibre technology compared to copper. Based on the maximum fibre length defined in ITU-T

standard (G.984.1), the aggregated exchange area is determined and the main cable is extended

from estimated current exchange locations to the aggregated exchange.

The core transmission network is not explicitly modelled, but investment costs are inferred from

annual lease cost data and from assumptions on required contention ratios, or Committed

Information Rates.

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

Kyland and Prescott 1977. Specific applications of the concept to regulated industries can be found in Evans,. Levine

and Trillas 2008, Guthrie 2006 and Levine, Stern and Trillas 2005. A test of whether ACCC decisions in

telecommunications are time-inconsistent (with the conclusion that they are) is in Ergas 2009.

13 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. A systematic discussion of the conceptual issues involved can be

found in Jullien, Pouyet and Sand-Zatman 2009.

14 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.

Henry Ergas and Alex Robson

infrastructure. Use of any other cost of capital than that for the private sector alternative will distort

resource allocation as 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 and last, investing in a new broadband network has a high level of systematic

risk. As a result, the Arrow-Lind conditions for use of the risk free 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 instance, and the cost of the project to taxpayers must

asset beta of 0.825, which is consistent with the upper bound of Telstra’s beta estimate of the

CAN for a range of services. We believe this value is conservative, as it presupposes that returns

from the new network are no more sensitive to economic conditions than Telstra estimates for its

legacy copper network.

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

Some of these variables are:

Setting these variables to their base case levels (which involves a GPON 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 as between metropolitan and non-metropolitan

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

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

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

15 In a classic article, Arrow and Lind 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. See Arrow and Lind 1970.

Henry Ergas and Alex Robson

initial service quality to 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 non-content payments for telephony

and broadband service (i.e. the sum of the monthly rental and of the non-content component of

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

involves continuation and some upgrading of the current copper-based network alongside

progressive upgrading of the HFC, with copper delivering speeds of some 20 to 40 Mbit/s and the

HFC delivering speeds of 50 to 100 Mbit/s. Incremental and selective upgrading to fibre optic

would occur over time, with fibre optic likely reaching 30 per cent of premises by the end of the

next decade, either directly, i.e. on an FTTP basis, or extending to the curb. The costs of this

scenario could be in the order of one-third those of the NBN in the metropolitan and regional

areas, up to around 80 per cent of the population. As for remaining areas, these would primarily

be served by wireless, at costs that would be around one-half those of the NBN, with speeds of 10

to 30 Mbit/s. 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 a 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 a month in metropolitan areas and

of $120 per month in regional areas) allow speeds to rise to 100 Mbit/s 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 prospect of that deployment may prevent the somewhat more limited,

but sooner in time, 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. In considering this, it is important to remember that most

currently envisaged applications function reasonably efficiently at speeds well below those

contemplated either in the NBN world or in the counterfactual. Thus, over time, advances in

compression and coding tend to reduce bit rate requirements, to some extent offsetting the

broadcasting and high definition video-on-demand have peak transmission requirements of less

than 20 Mbit/s. 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 Mbit/s and 60 Mbit/s 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 (i.e. the

16 They are even higher when compared to the access payments made by the average residential premise,

remembering that about 30 percent 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.

17 With MPEG4, for example, High Definition video can now be transmitted at 8Mbit/s, which brings it within the range of

12 Mbit/s DSL.

Henry Ergas and Alex Robson

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 (i.e. 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 time-allocation model (Becker 1965). Thus,

adapting that model, consider a consumer who spends time either working or downloading. The

consumer derives incremental benefits from an increase in download speeds, as this frees up

additional time for working, boosts the individual’s labour supply and widens the individual’s

wages (i.e. who have a higher opportunity cost of time), but all else being equal, the incremental

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 (i.e. that

exceeds the service’s start-up costs). Appendix C 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.

A cost benefit analysis should assess the incremental benefits and costs of the project, and then

compute the incremental benefits net of incremental costs. To this end, we consider a median

consumer over the next twenty years. This representative consumer has a willingness to pay

(WTP) curve for higher speeds, which (for the reasons set out above) is increasing but concave as

speeds increase. In addition, we assume that this willingness to pay curve is growing over time

with increases in income, the development of new applications, and possibly ‘bandwagon effects’

in demand.

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:

increases to 60 Mbit/s by year 6, and remains on 60Mbit/s.

(but at a slightly slower rate than the baseline) to 60 Mbit/s by year 9, and then has speeds

of 100 Mbit/s from year 10 onwards.

18 Nothing is lost in the formulation by replacing the alternative ‘work’ with a composite good, valued at the opportunity

cost of time.

19 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.

Henry Ergas and Alex Robson

the same path as the baseline. The median consumer then goes on to 100 Mbit/s at a later

time than with the project Scenario A – from year 15 onwards.

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 below.

Henry Ergas and Alex Robson

For the median consumer’s willingness to pay, we assume $50 WTP for 10 Mbit/s, increasing to

$71 for 100 Mbit/s. As these assumptions pertain to the median consumer, they are reasonably

close to current market outcomes. 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, with

annual growth of 3.6 per cent for WTP for 100 Mbit/s. The initial WTP curve for scenarios A and B

and their growth rates over time are shown in the figure below.

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. To this end, suppose that WTPs are

log-normally distributed, with the median of the natural logarithm of the distribution of WTPs set so

that the resulting distribution has the same median as under the baseline scenario (i.e. the median

Henry Ergas and Alex Robson

WTP is $50), and with the standard deviation of the natural logarithm of WTPs equal to one

quarter of the natural logarithm of the mean. Then, by construction, the top 25 per cent of

consumers will 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.

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.

Henry Ergas and Alex Robson

These are plotted in the charts below.

Henry Ergas and Alex Robson

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 ($1,273) is

the present value of the future stream of benefits that the consumer expects to receive.

4% 1,237 1,228 -9 91 90 -1

12% 612 599 -13 82 80 -2

4% 1,237 1,249 11 91 92 1

12% 612 615 3 82 82 0

4% 1,237 1,540 303 91 113 22

12% 612 695 83 82 93 11

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

Henry Ergas and Alex Robson

negative, and so accounting for the incremental monthly costs that were computed earlier (of

around $75 a 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, i.e. 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.

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. Under these new enhanced scenarios,

the consumer’s willingness to pay curve still begins at the same point for low speeds, but

to pay for increases in speeds is higher in the enhanced settings. The year one WTP curves in

the original (standard) analysis and the enhanced analysis are shown in the chart below. In the

enhanced Scenario C, the targeted, high WTP consumers are assumed to have a WTP of $120

for 100 Mbit/s speeds in the first year (compared to a WTP of $100 in the first year of the standard

analysis). In all enhanced scenarios, the growth rates of WTP are assumed to be the same as the

standard analysis.

0

20

40

60

80

100

120

10 Mbps 20 Mbps 30 Mbps 40 Mbps 50 Mbps 60 Mbps 100 Mbps

WTP in first year,

standard

WTP in first year,

enhanced

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 mid-speed tier (and hence increases the relative value of the options that involve

incremental development of the access network).

Henry Ergas and Alex Robson

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% 1,608 1,609 1 118 118 0

12% 776 753 -23 104 101 -3

4% 1,608 1,648 41 118 121 3

12% 776 787 11 104 105 1

4% 1,608 1,918 310 118 141 23

12% 776 861 85 104 115 11

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 discontinuity (the upgrade

fundamental feature of the situation, as we consider a median user, while there are substantial

numbers of users – especially in non-metropolitan areas – who have low willingness but very high

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 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 HFC networks can currently provide. Specifically, we

examine:

in year 4 and remain there.

100 Mbps by year 6 of the NBN project, where they remain.

20 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.

21 This likely reflects the fact that WTP is correlated with human capital endowment, and human capital, and especially

that associated with ‘information’ activities, tends to be concentrated in metropolitan areas. See Glaeser and Ponzetto

2008 and O’Flaherty 2005.

Henry Ergas and Alex Robson

These speed adoption paths are plotted below.

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

the figure below.

Henry Ergas and Alex Robson

0

500

1000

1500

2000

2500

3000

10

Mbps

20

Mbps

30

Mbps

40

Mbps

50

Mbps

60

Mbps

70

Mbps

80

Mbps

90

Mbps

100

Mbps

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 the chart below.

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

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 under the assumption of an 85%-15% split between urban and non-urban areas.

Henry Ergas and Alex Robson

For the NBN, under Scenario D and the assumption of a CIR 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 an 85%-

15% 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 takeup

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 takeup 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 the figure below.

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

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.

The estimates are set out in the table below. 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.

Henry Ergas and Alex Robson

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 the existing network with a

service quality (which we do through the willingness to pay calculation). It turns out that the NBN

would only satisfy this condition 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 mid-point 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 (i.e. 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

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 of course 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

22 Using 40 Mbps is especially realistic if sorting is allowed to occur – i.e. 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.

Henry Ergas and Alex Robson

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

precise specification of those benefits, it is not possible to assess whether they have any

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 the project imposes on

distortions associated with financing them through taxation are not, and need to be added to the

social costs of the project. In contrast, under the counterfactual world, 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 (i.e. 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 Commonwealth government will be both the primary investor in a major

competitor and the industry policy-maker 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 a ten to fifteen year period of time,

geographical mobility is relatively high, and consumers will sort themselves geographically in 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

23 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 (i.e. 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.

24 Claimed wider benefits such as the promotion of tele-medicine seem very difficult to credit. With respect to telemedicine,

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, Ergas,

Hannan and Arts 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.

25 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 CBA.

Henry Ergas and Alex Robson

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 to the willingness to pay for the incremental speed the project provides, 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 than incremental cost, but less than project average cost, will not consume

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 are counted towards the CBA’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 effect of the entity’s pricing

particularly severe for the NBN, as it is intended to be a ‘wholesale only’ network. This could limit

its ability to price discriminate (as it will not know or be able to directly access the willingness to

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 is a popular one,

that while “The majority of schools in metropolitan locations reported 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.”

27 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.

28 In other words, the entity’s ability to engage in Ramsey-Boiteaux pricing may be quite limited. While menus of selfselecting

charges (i.e. second degree price discrimination) can be used to approximate Ramsey-Boiteaux pricing, the

approximation is far from perfect and there are in any event additional costs involved in using self-selecting prices for

intermediate inputs – see Ordover and Panzar 1982.

29 “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, Tuesday, 12 May 2009. The Minister goes on to say that “It is worth noting that Access Economics views

this as a conservative estimate.” However, as discussed below, the comparison Access Economics make is to a world

in which only dial-up service is available (noting that as of the time of writing, 70 percent of Australian households

subscribe to some form of broadband).

Henry Ergas and Alex Robson

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. Those assumptions cannot be carried over to the

NBN for two reasons: first, the relevant comparison is to a counterfactual world in which high

speed broadband is relatively widely, though not universally, available (rather than to one in which

there is no high speed broadband); and second, the NBN will receive public funding on a larger

scale than envisaged for the Government’s first-round process.

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

productivity gains should set to one-third to one-half their initial levels, given that 70 percent 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

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 percent (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.

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 CBA 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

30 Access Economics 2009, page 20, emphasis added.

Henry Ergas and Alex Robson

NBN – suggest such an assessment would conclude that the policies at issue impose costs than

exceed the relevant benefits.

TRANSPORT

We now turn to a consideration of project evaluation in transport, where there have also been very

significant increases in outlays. We first discuss some important features of the institutional

context; and then examine one of the largest projects which the Government has decided to fund,

as a example of some of the limitations of our project evaluation procedures.

A summary of the major steps in the development of transport funding in Australia is set out in

Appendix D. Over time, there has been a trend to an expansion in the role of the Commonwealth,

with the division of responsibilities between levels of government becoming increasingly blurred.

Focussing on recent developments, the most relevant measure is the establishment of the

Building Australia Fund (BAF), which was set up on 1 January 2009 by the Nation-Building Funds

Act 2008 to finance capital investment in transport infrastructure (such as roads, rail, urban

transport and ports), communications infrastructure (such as broadband), energy infrastructure

and water infrastructure. The BAF is essentially a hypothecated fund for financing investments in

The Government also set up Infrastructure Australia (IA), a policy advisory body, which was

established by the Infrastructure Australia Act 2008 that came into effect on 9 April 2008. IA’s

composition includes representatives of the States and Territories, along with representatives of

the Commonwealth. Among other functions, IA is charged with reviewing and recommending

proposals for infrastructure projects. These assessments are to be guided by the “Building

Australia Fund Evaluation Criteria”, which, under the terms of the Nation-building Funds Act 2008,

are determined by the Infrastructure Minister. These criteria, which were tabled on 18 December

money .. (and have) an expectation of long term public benefits, taking into account

IA has issued an outline of its project evaluation methodology, which state that the first step in its

evaluation of proposals is to “assess compatibility .. to Infrastructure Australia’s strategic

priorities”. To then evaluate those proposals that are so compatible, IA says it will:

31 However, the hypothecated funding is not attached to a hypothecated tax, so this is not a Wicksell-like earmarking

scheme (where taxpayers would face the tax price of infrastructure). Rather, it is simply a ‘piggy bank’ allocation of

funds to a particular set of purposes.

32 BAF Evaluation Criteria, Tabled 18 December 2008, at Schedule 1.

Henry Ergas and Alex Robson

IA goes on to say that:

It is clear that this evaluation framework places considerable emphasis on CBA. As IA itself notes,

can be seen from Appendix E, (which lists the relevant instruments and their coverage), these

guidelines broadly require the use of CBA to evaluate all major transport projects. However, two

further trends also emerge from the IA material and from a consideration of developments in the

various Australian jurisdictions.

The first is the tendency to constrain the range of options considered to those which demonstrate

System Management in Australia, which are now the standard reference for all transport agencies

in Australia, introduced a ‘strategic merit test’. In practice, that test – which asks whether an option

is consistent with government strategy – can act to filter out projects and options that do not pass

a prior test of political suitability. The result is that in considering how best to address a given

transport need, the range of options compared in the CBA may be restricted in ways that (by

eliminating relevant alternatives) can disguise the true opportunity costs of the option ultimately

selected.

The second tendency, which is also marked in all Australian jurisdictions, is the ever greater

reliance being placed on multi-criteria evaluation (Dobes and Bennett 2009). There is, in other

words, a proliferation of evaluative criteria and approaches that, although claimed to complement

CBA, may in practice displace it. A striking instance is the so-called ‘triple bottom line’

methodology, which purports to give due weight to environmental and social considerations as

merely private) costs and benefits, the effect of reliance on this methodology is to allow the

evaluator to double or treble count benefits, and to increase the cumulative scores given to

favoured projects: for example, by including as separate criteria, which are used to rank and score

alternative projects, impacts on the environment, on bio-diversity and on greenhouse gas

abatement, counting each of these as separate and additive sources of benefits. The weights

given in these approaches to the various evaluation elements are almost invariably arbitrary,

bearing no relation to the community’s willingness to pay for the relevant benefits (or to accept for

any relevant costs) – see Ergas 2009.

The overall result is that while the form of CBA is retained, both the value of the CBA is

compromised (by the narrowing of the range of considered alternatives) and the significance of

33 Australian Government, Infrastructure Australia, 24 September 2008, pages 4 and 5, page 12.

34 Ibid, page 12.

Henry Ergas and Alex Robson

the results of the CBA reduced (as the CBA is only one element among many in the evaluation

criteria, and decision-makers pick and choose among the alternative results). The consequent risk

is that of the CBA becoming a box to be ticked but not to be taken too seriously, either in its

execution or in the weight placed on its results. This in turn reduces the barriers to rent-seeking

and to poor governance, as ‘fudge factors’ distort the policy assessment and muddy the

documentary trail leading to the ultimate decision.

So as to examine these issues in a practical context, we have analysed the CBA undertaken for

rail links between Melbourne, Geelong and the regions to Melbourne’s west, was recommended

by IA for immediate funding; indeed, taking the project as a whole, it is the largest such project.

The Commonwealth Government has since announced that it is making $3.2 billion available for

closer examination; additionally, it is one of the few recommended projects for which a CBA is

that our discussion of the CBA is not intended to suggest that this CBA is particularly poor; rather,

it highlights issues that occur, albeit to differing extents, in Australian CBA more widely.

As with many current CBAs, the first issue the EWR assessment raises is that of whether the

appropriate range of alternatives has been considered. In particular, the project is designed to

alleviate capacity constraints affecting regional rail in Victoria; however, it is by no means

established that significant new capacity is the only, much less the most effective, option in that

respect. Thus, Mees 2008a, 2008b argues that current capacity constraints on the links at issue

are primarily due to poor management and inadequate governance; and that both past experience

in Melbourne and international benchmarks suggest those constraints could be effectively

addressed by reforming work practices, improving scheduling and making minor investments in

signalling and related equipment. Mees also shows that previous capacity increments, themselves

aimed at alleviating projected capacity constraints, were based on unduly optimistic projections of

demand, leaving a costly legacy of underutilised capacity. Mees concludes that policy-makers

should have examined less expensive, albeit politically less attractive, alternatives that involve

altering the way the Melbourne rail system is run. Similar concerns about the range of options

considered are also raised, somewhat less directly, in an independent review of the project

undertaken for the Victorian government (Dotson 2008).

However, these alternative options are not considered in the CBA of the project undertaken for

(and apparently accepted by) the Victorian government (Meyrick and Associates 2008), which

broadly assumes that existing constraints would persist absent the capacity augmentation.

35 The project involves several related components, including an initial upgrade to a regional rail link and then the

construction of a tunnel. These elements, as they relate to the rail link, were combined in the published CBA and

hence are treated here on a combined basis. We understand that the relationship between these components, and

their sequencing, has undergone continued development since the CBA was completed. The project's elements are

referred to as the Regional Rail Express and East-West Rail Tunnel in the May 2009 statement "Nation Building for

the Future". It is the combined project, as assessed in the CBA but without its non-rail component, that is discussed

here.

36 Combining the East–West Rail Tunnel and the Regional Rail Express, total estimated cost is $7.3 billion; the Pacific

Highway Corridor project, which is the next largest project recommended by IA, has an estimated cost of $6.7 billion.

The Commonwealth Budget has allocated $3.2 billion to the former and $618 million to the latter.

37 Taking the IA list of “priority projects ready to proceed”, CBA’s are publicly available for the F3-Branxton Freeway

project (NSW), the Seaford Rail Extension project (SA) and the Majura Parkway (Stage 2) project (ACT). There is also

a public version of an early stage feasibility study for the Gold Coast Rapid Transit (QLD).

Henry Ergas and Alex Robson

Further concerns arise on closer assessment of the CBA. These relate, first, to the treatment of

the standard building blocks of transport sector CBA (which we refer to as the conventional

savings); and second, to the role of what are now referred to as ‘Wider Economic Benefits’ (we

explain the meaning of the term below). As can be seen from Table 5, the finding that the project

has net benefits relies significantly on each of these.

$14.4 billion $9.2 billion

1.0 1.2

Source: Meyrick and Associates 2008

Transport CBA follows a relatively standardised form, typically focussed either on examining net

social savings from a project or on assessing changes in net social surplus from a project, with the

outcomes, if properly implemented, not being affected by the choice of approach. These standard

building blocks are used in the CBA for the EWR.

However, in the EWR CBA, incremental fare revenues are simply added to the other sources of

benefit – see Meyrick and Associates 2008, Table 16. This is equivalent to double counting the

gain from the vehicle operating and travel time savings (as these are ‘paid for’ through the fares)

gross benefits, they are a large share of project net benefits (which are themselves very small).

Consistent with IA’s project evaluation methodology, the EWR CBA examines a range of ‘Wider

Economic Benefits’ that are claimed to arise from the project proceeding. Reliance on these

‘Wider Economic Benefits’ to increase gross estimates of benefits is increasingly common in

Australian infrastructure evaluation, following work initially done in the UK – see UK Department

for Transport 2006 and Joint Transport Research Centre (JTRC) 2008.

38 This can be readily seen by considering a highly simplified example. Thus, assume there are 100 identical users who

in the counterfactual drive from Geelong to Melbourne but under the proposed project, will take the train. The capital

cost of the train is $900 and it costs $5 in operating expense per user. Driving costs each user $3 in operating costs

and $10 in time costs; taking the train, they pay a $5 fare and incur a time cost of $3. The approach adopted in the

expenses = $1400, for benefit/cost ratio of 1.07. However, this greatly overstates the benefit. Thus, the net gain to

travellers, which is the difference between the $1300 cost they incur for road travel minus the $800 cost they incur

from train travel, is $500. The rail operator incurs an operating cost of $500 and has operating revenues of $500,

leaving an expected rental of zero. The current social surplus is therefore only $500, which is insufficient to cover the

capital cost of the train. (The use of the benefit/cost ratio (BCR) as a criterion is also incorrect, as it will not, in general,

lead to an efficient allocation of a given level of outlays, which requires use of a criterion that maximises the sum of net

benefits – see for example, Quinet and Sauvant 2007.)

Henry Ergas and Alex Robson

CBA has, of course, long taken account of technological externalities (such as project-related

changes in noise); ‘Wider Economic Benefits’ are essentially pecuniary externalities, that is,

effects on other parties caused by changes in the prices at which they can transact. In competitive

markets, such pecuniary effects are merely transfers of benefits from one party to another – see

for example, Meyer and Straszheim 1971, pp. 199-202. This is not the case, however, when

markets are imperfectly competitive and in particular, when markets are neither perfectly

competitive nor completely monopolised (in which case, benefits are internalised by the

monopolist – see Mohring and Williamson 1969).

A familiar example is a transport project that by opening trade between two regions, reduces the

extent of monopoly power. In the simple case in which the demand curve in each region is linear,

producers are identical and marginal costs are constant, the total benefit from the project is 1.5

times that measured in the conventional CBA. While that has long been known to cost-benefit

analysts, the general view has been that in mature economies with well developed transport

However, attention to these pecuniary externalities has been revitalised as a result of the renewal

of geographical economics, with the emergence of the Krugman-Fujita-Venables general

equilibrium model of spatial monopolistic competition, the parallel renewal of urban economics,

both theoretical and empirical, and (though less centrally) developments in the analysis of

oligopsony in labour markets. Each of these areas of work has suggested that transport changes

could have effects in reshaping locational decisions, product markets and labour markets that are

not fully captured in conventional approaches to CBA and are sufficiently material to warrant

attention.

Reflecting this, and drawing on work in the UK, it has become increasingly common for Australian

transport analysts to ascribe benefits to claimed effects of the project in allowing fuller exploitation

of agglomeration economies, reducing market power in product and/or factor markets and

increasing labour force participation. Thus, the EWR CBA ascribes material benefits to increased

agglomeration economies and to greater labour supply. Indeed, absent these benefits, the project

would not have benefits in excess of its costs.

benefit to reductions in the ‘effective economic distance’ between areas (with the benefits being

calculated using estimates of these effects for the UK). At an analytical level, it is difficult to

reconcile these estimates either with the results of urban economics or with those of general

equilibrium geographical economics. Thus, in the work-horse model of geographical economics, a

reduction in transport costs creates net gains from external economies but these are the sum of

greater gains in the ‘larger centre’ (and to immobile factors in that centre) and of smaller gains and

losses in the ‘smaller’ centre (typically, with losses to immobile factors in that smaller centre): see,

for example, Brakman, Garretsen and Marrewijk 2001, pp. 308-313. As a result, the calculation of

39 This refers to situations where improvements that increase economies of scale, or convert variable costs into lower but

fixed and sunk costs, yield potential increases in social surplus but at the expense of greater monopoly power (and

hence possibly lower consumer surplus – depending on the reference point adopted).

40 A review of the analysis of Wider Economic Benefits in the East-West projects was undertaken by

PriceWaterhouseCoopers for the Victorian government -- PriceWaterhouseCoopers 2008a. However, this assessment

omits the points covered here.

Henry Ergas and Alex Robson

of the net welfare change is generally highly sensitive to the precise structure of the transport

change and of the production and demand environment: see, for example, Baldwin et al. 2003.

Application of what amount to no more than standard multipliers for the UK is of little help in this

context, particularly for the purpose of assessing projects involving many billions of dollars of

public outlays. This is all the more the case as there appear to be identification problems in the

estimates used and as marginal and average, gross and net, effects are not separately identified.

Consideration of modern urban economics only underscores these concerns. Thus, both theory

and empirical analysis suggests that a substantial share of agglomeration economies arise from

spill-over effects in human capital – see Glaeser 2000, Acemoglu and Angrist 2001, O’Flaherty

2005, and Glaeser and Gottlieb 2008. Moving skilled people from one area to another is only

advantageous if the impacts of those skills differ across areas – resources should, in other words,

be pushed to areas that are more productive and where the elasticity of productivity with respect

to agglomeration is higher. When this is done, the resulting gain in the new equilibrium (that is, the

equilibrium once the transport project is complete and the allocation of people over space is such

that the spill-over at the margin is equalised across places) is not the gross gain in the destination

area (assuming there is such a gain), but rather the net gain taking account of the loss of

agglomeration economies in the origin area. Again, rather than demonstrating that such

rearrangements of skills are likely to occur, the EWR CBA appears to assume as much, as well as

assuming that the effects can be assessed by applying summary impact multipliers derived from

the UK.

Additionally, in the standard model of urban residential location, the welfare benefits of changes in

transport costs depend on the extent to which transport infrastructure and other goods subject to

congestion (such as schools and hospitals) are appropriately priced – see Arnott 1979. Thus,

reductions in transport costs (such as those being modelled in the EWR CBA) will alter settlement

patterns, typically inducing greater decentralisation (as people exercise their preference for larger

lots at points further along the initial bid/rent curve): see, for example, Baum-Snow 2007 and

Duranton and Turner 2008b. While this leads to a welfare gain (as those moving clearly value the

new location at more than the previous location), whether welfare rises overall depends on

incorrect to assume there will be benefits from agglomeration without taking account of any

possible offsetting costs as location patterns change.

Finally, even if there were possible gains from increased agglomeration, it would only be correct to

impute to the project that element of those gains that the project delivers at least cost. If the gains

at issue could be delivered by cheaper means – for example, by reducing restrictions on urban

development and redevelopment – and the project is an alternative to those means, then the

offset to project costs should be no greater than the cost of the foregone alternative.

None of this is to say that agglomeration economies and other changes associated with transportinduced

shifts in the pattern of economic activity are negligible. As a growing body of literature

shows – see among others Baum-Snow 2007, Chandra and Thompson 2000, Donaldson 2009,

41 Thus, in a study of ports, Cohen and Monaco 2008 find that an increase in port capital stock in one US State leads to

lower manufacturing variable costs in that State but increases the variable costs of manufacturers in neighbouring

States. The net effect depends on the balance between these.

42 Of course, in the long run spatial equilibrium, utilities must be equalised across locations. However, because

equalisation occurs in utilities (assuming one cannot live in two places at once), rather than marginal utilities, the

equilibrium is not first best.

Henry Ergas and Alex Robson

Duranton and Turner 2008a, Elhorst and Oosterhaven 2008, Fernald 1999, Heyma and

Oosterhaven 2005, Michaels 2008 – that is not the case. However, the literature also results in a

broad range of estimates of the extent of effects, the degree to which they prevail across different

undertaken, there is great merit in the conclusion reached by Glaeser and Gottlieb 2008 in their

consideration of urban policy:

We would therefore put little or no weight on the agglomeration benefits claimed in the EWR CBA.

costs as an increase in the take-home wage and applies to that wage increase a labour supply

adjust for the likely difference in labour productivity between the additional workers and the

existing workforce. Given that the added labour hours are valued at the take-home wage, it would

obviously be inappropriate to include them as a social benefit; however, it is correct to treat the tax

However, the approach adopted in the CBA is incorrect. This is because a reduction in transport

which show the shift in the budget line and in the consumption/leisure equilibrium consequent on a

reduction in travel time. In essence, while an increase in net wages changes the slope of the

consumption/leisure budget line, a reduction in travel time shifts the budget line out. The effects on

labour force participation depend on the impact on hours for those who already work, which

depends on the income expansion path, and the extent of the shift into the paid labour force for

those who would not work in the counterfactual. There is no reason to expect a conventional

elasticity of labour supply to capture these effects (all the more so one estimated for the response

of hours to wages in the UK).

43 For example, the agglomeration effects of public transport seem to be less than those for road travel – a factor not

taken into account in the EWR CBA.

44 For example, Fernald 1999 finds that US interstate highway investment had high returns in terms of increasing

productivity during the period before 1973, in the order of 10 per cent per year, and an almost zero return after. He

attributes this result to the fact that the interstate highway network was essentially complete after 1973, and that

marginal improvements to the network were no longer as productive, and also to the fact that the interstate highway

network became progressively more crowded, and to that extent more expensive to use, after 1972.

45 The elasticity is reported in the EWR CBA as -0.1 (see Meyrick and Associates 2008, page 27 and page 44); we

assume it is intended to be 0.1, or the claimed effects would not follow.

46 Assuming the increased hours are not matched by a required increase in public outlays.

Henry Ergas and Alex Robson

= −

+ −

+ −

= −

Henry Ergas and Alex Robson

+ −

+ −

+ −

Now consumer decides

to work

Consumer initially decides

not to work at all

+ −

Note: Leisure increases,

but not by as much as reduction in

(size of effect depends on income expansion path)

Henry Ergas and Alex Robson

Matters are further complicated by the fact that reductions in transport costs affect settlement

discussion above), the interactions with long-term labour supply decisions are not. In particular, it

may be that the main effect of reduced commuting times is to alter the distribution of working

hours as between days worked per year and hours worked per day (with lower commuting times

being associated with greater incidence of part-time work or more generally, short working days

but possibly more days of work per year) – see, for example, Huberman and Minns 2007. Again,

absent detailed modelling, of the kind done by Elhorst and Oosterhaven 2003, the most reliable

estimate of the long run effects on labour supply is likely to be zero.

The EWR is a very major project, involving billions of dollars of taxpayers’ money. It has been

endorsed by the Victorian and Commonwealth governments. However, there are a number of

difficulties with the CBA, at least in its published form. When the errors are corrected, and account

is taken of the marginal social cost of public funds (which the CBA ignores), project costs greatly

exceed project benefits.

Many of the difficulties associated with the EWR CBA, and especially those with respect to the

treatment of ‘wider economic benefits’, also arise with increasing frequency in other Australian

transport CBA’s. They suggest a need for significantly greater rigour in the way these benefits are

determined, if they are not to become a ‘fudge factor’ that undermines the quality and relevance of

the cost-benefit studies.

IMPROVING THE EVALUATION PROCESS

The case studies set out above 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’s Report 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 then, to a belief that the global financial crisis meant that high levels of

government spending were not only feasible, but desirable. As the threat of recession loomed,

confused reasoning lead 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

47 The observed long-term constancy of time spent commuting to work (see Vanderbilt 2008 pp.131 and follows) is

consistent with the inference that reductions in travel time are largely offset by shifts in settlement patterns. So too is

Down’s ‘fundamental law of traffic congestion’ – see Downs 1992 and Duranton and Turner 2008b.

Henry Ergas and Alex Robson

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

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 rentseeking

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 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 decision-making) and readily

become (at best) ‘tick the 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 Walle 2006, 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 BAF and of Infrastructure Australia. Although there can be merit to coordinated approaches

to infrastructure selection, there can be little doubt that the new mechanisms create significant

incentive problems. To the extent to which the projects they fund are worthwhile, that funding may

simply displace funding of those projects by the States themselves, but with higher transactions

may, in other words, be incentives for adverse selection, and then for moral hazard in project

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 Private-Public Partnerships (PPPs). While these

48 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 CBA, 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-

123. Additionally, to the extent demand expectations are reduced, this should lead to lower infrastructure investment.

49 Obviously, if the Commonwealth funding were simply matching grants associated with the pure spillover effects of the

State infrastructure decisions – i.e. 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).

50 The question of how to design multi-level funding institutions, and associated CBA 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.

Henry Ergas and Alex Robson

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 (i.e. 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 decision-maker 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 given governments

intent on poor decisions, PPPs can not only make those decisions more (privately) profitable but

A third factor, that is yet to fully play itself out, is the recourse to hypothecated funding sources for

long-term infrastructure finance, most notably the BAF. While economic theory yields ambiguous

important ways in 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.

51 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).

52 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 – see Landes and Posner 1975.

53 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.

Henry Ergas and Alex Robson

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.

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.

as a matter of course. Instead, most CBAs are never released, and those that are, are often

difficult to locate. Governments should also regularly publish, in readily accessed form, the CBA

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 CBAs routine, the fact that a CBA

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

quality of the CBAs that had been carried out.

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

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

having independent review of all CBA’s 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 ANAO, 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 postcompletion

review of projects. Although a few useful post-completion reviews of CBAs have been

undertaken (Bureau of Transport Economics 2001, Bureau of Transport and Regional Economics

2007a and 2007b, NSW Audit Office 2006, NSW Auditor General’s Report 2005, NSW Treasury

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 far-reaching

consequences. To begin with, it would force Commonwealth and State entities to more properly

54 The pervasiveness of these forms of bias in transport assessments is amply documented in Flyvbjerg, Bruzelius and

Rothengatter 2003.

Henry Ergas and Alex Robson

document and archive material related to the CBAs and the CBAs themselves. In contrast, as

matters currently stand, CBAs are typically undertaken before the final form of projects is

determined, and then never updated. Additionally, little investment is made in documenting CBAs

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 Mirlees reached in reviewing, after two decades, the impact of their great

CBA manual:

to promote a greater sense of professionalism in the 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 under-explored – such as

the implementation of CBA in the context of hypothecated funds (where congruence requirements

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 on-going training for both

practitioners and users of CBA. 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

PC 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

the list of issues identified above highlights the task that remains.

55 Little and Mirlees 1994 at page 206.

56 When decisions are delegated to agencies, and agencies are instructed to make optimal use of their budgets, the

expected growth path of agency budgets on the one hand and of investment opportunities on the other becomes an

important factor in determining the optimal pattern of outlays. When an agency regards both its current budget and its

current set of investment opportunities as representative of future opportunities – either because these regenerate

defining the choice set for evaluation, choose a set of projects and time horizon that can reasonably be regarded as

congruent. Where agencies are budget funded, it is not unreasonable to assume the current budget defines such a

set; however, this assumption cannot simply be carried over to an agency whose budget is hypothecated.

Henry Ergas and Alex Robson

CONCLUSIONS

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.

That matters should be so is by no means a new phenomenon. Thus, in Hancock’s magnificent

Australia (1930), now sadly out of print, the great historian famously said that it was a failing of

democracies, and especially of Australian democracy, to constantly confuse ends and means, and

to show too much reluctance “to refuse favours, to count the costs, to discipline the policies they

have launched”. “[The] policies therefore yield diminishing returns, until at last, they may become a

positive danger to the national purpose that called them into existence”. Nowhere was this more

marked, Hancock noted, than with public involvement in infrastructure ventures such as rail,

where Australian government was “particularly slow to confess it has got into a bad business, for

its mere entry .. has created vested interests which immediately express themselves in politics..

So.. it throws good money after bad, and hopes that something will turn up. In this way, losses

accumulate in a lump, and the crisis, when it comes, is likely to be prolonged and severe.”

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 Butlin,

Barnard and Pincus identified 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” (Butlin, Barnard and Pincus 1982, page 294). That so little should have

changed is not encouraging.

Set against that background, how great a contribution can improved project appraisal make to

securing better outcomes? Little and Mirlees, in considering ‘The costs and benefits of analysis’,

develop a simple model of the value of information in which good project appraisal yields benefits

1994, pp. 225-227). For an economy investing over $10 billion a year on its transport and

communications infrastructure, 10 per cent of project value would seem like a saving well worth

seeking. That said, the Little/Mirlees model assumes unbiased estimates and a decision-maker

who, as a benevolent social planner, maximises social welfare; it is hardly contentious those

assumptions do not hold – if they did, central planning would be a far better system than it has

ever proved to be.

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.

57 The Little-Mirlees formulation yields a value of appraisal that is at least 10 percent of standard deviation of the errors

removed by the appraisal, multiplied by the ratio of that standard deviation to the standard deviation of the errors not

removed. This ratio should be about 1, though with competent appraisal it could be much more than that. As a result,

a conservative estimate of the value of appraisal is 10 percent of project value.

Henry Ergas and Alex Robson

Each of the steps in Gans analysis is set out below and problems are identified.

1. Estimate Annual

Costs

$43 billion capital expenditure to be recovered

over the life of the network

$4.3 billion to be recovered

each year from end-users

For costs to be recovered, charges to users must include not only annual capital charges but

also taxes, deprecation, O&M, marketing and retailing as well as costs for those network

components (such as backhaul, switching etc) not provided by the NBN. Gans ignores all

these other cost items.

2. Estimate Gross

Commercial Return

Assume 3.6 million users (half of all fixed

users) taking voice and broadband services

over the NBN. Voice ARPU assumed to be

current voice ARPU less usage costs ($765

minus $238 in usage costs = $527 pa) and

broadband ARPU assumed to be the same as

current ($500 pa).

$3.7 billion recovered from

end-users annually (3.6 million

users at $1027 ARPU)

This calculation assumes that all voice and broadband revenue accrues to the NBN operator

and can be used to recover the costs of the NBN. However, prices will also need to cover

retailing and network costs incurred by the NBN’s wholesale customers (e.g. for backhaul).

Currently, only a small share of broadband revenues goes towards covering costs of the

copper access network (at most $20 per month out of around $60 per month average revenue)

which the NBN will replace. If all the current revenues go to the NBN Co (which only provides

the access loop) retail charges must increase for full cost recovery to occur.

3. Estimate Net

Commercial Return

Subtract annual costs from gross commercial

return

Gans states there is a $320

million shortfall

This calculation seems wrong. Subtracting costs from revenues above, we get a $600 million

shortfall.

4. Estimate Consumer

Welfare Gains

Assume usage costs for all end-users fall by

$400 pa and calculate the increase in

consumer surplus that results from this.

$2.88 billion in consumer

surplus gains ($400 multiplied

by 7.2 million end-users)

It is not clear why usage prices would fall by $400 under the NBN. Since Gans acknowledges

that access costs ‘may rise’, this implies calling prices (both fixed-to-fixed and fixed-to-mobile)

falling to 0 (Gans notes that access charges account for $350 out of the $765 paid by endusers

for fixed voice). The calculation ignores the fact that providing calls and other forms of

usage involves costs (for switching, transmission, network control, termination on other

networks (eg mobile networks) etc) above and beyond those incurred in the fixed line

reticulation network (which is all that is included in the estimated NBN cost). Gans seems to

assume these costs do not need to be recovered. How that can be is not explained.

The calculation is also inconsistent with the ARPU assumed in step 2 above, which assumes

that only fixed-to-fixed usage costs will be eliminated and that current prices will only fall by

$238 per user.

It assumes all fixed-line users (not just NBN users) will benefit from the same price reductions,

but it is not apparent why this would be the case.

5. Estimate Net Social

Return

Add consumer welfare gains (step 4) to net

commercial return (step 3).

$2.3 billion net benefit The ‘social return’ is in fact not a social return at all, but rather a transfer from producers to

consumers. Gans seems to assume that fixed operators currently profit from high usage

charges and that under the NBN this profit will be transferred to consumers. Even if this were

the case (and it is not clear why it would be), this would not amount to an increase in net

economic welfare. Rather, it would simply be a transfer of wealth from producers to

consumers.

Source: Gans J, ‘The right policy for telecommunications and broadband’, Submission to the Senate Select Committee on the National Broadband Network, June 2009

Henry Ergas and Alex Robson

ULLS PRICING

The regulated price of the Unconditioned Local Loop Service (ULLS) in Australia is relatively low

by international standards. In the latest published ULLS pricing arbitration, the Australian

Competition and Consumer Commission (ACCC) determined a Band 2 monthly charge of $14.30

Hungary and Estonia had lower ULLS charges (Chart 1).

For comparison, we take the monthly charge and add a share of connection costs assuming the

connection lasts for three years (ie the connection charge divided by 36). Charges for the

European countries are converted to Australian dollars using the average exchange rate for 2008

0 5 10 15 20 25 30 35

Estonia

Hungary

Netherlands

Italy

Australia

Sweden

Slovenia

Latvia

Greece

Portugal

Bulgaria

Austria

Cyprus

Romania

France

Denmark

Malta

Belgium

Spain

Germany

Slovakia

Lithuania

UK

Poland

Czech

Luxembourg

Finland

Ireland

Average monthly total cost (Euro)

Data source: ACCC; European Commission; Oanda

Since the Powertel determination the ACCC has published Pricing Principles and Indicative Prices

in the Powertel determination, at $16 per month. Although this price will not necessarily apply to

of the Trade Practices Act 1974, March 2008

59 European Commission, ‘Progress Report on the Single European Electronic Communications Market 2008 (14th

Report)’, Brussels, March 2009, Annex 2, pp121-122

60 Exchange rates sourced from Oanda

Henry Ergas and Alex Robson

any access seekers unless an access dispute is notified, we also compare this price to the

European benchmarks in Chart 2 below. Again, this price is at the lower end of the international sample.

0 5 10 15 20 25 30 35

Estonia

Hungary

Netherlands

Italy

Sweden

Slovenia

Latvia

Greece

Portugal

Bulgaria

Australia

Austria

Cyprus

Romania

France

Denmark

Malta

Belgium

Spain

Germany

Slovakia

Lithuania

UK

Poland

Czech

Luxembourg

Finland

Ireland

Average monthly total cost (Euro)

Data source: ACCC; European Commission; Oanda

ULLS TAKE-UP

Partly as a result of low and declining ULLS prices, adoption has been growing rapidly in recent

years. ULLS lines now account for 11% of all PSTN lines in CBD areas (Band 1) and 9% in

areas more than doubled to well over 600,000. This growth was dominated by Band 2, with very

little growth seen in Bands 3 and 4 (Chart 3).

This has been accompanied by a growing presence of ULLS access seekers in Telstra

exchanges. Over the same 18-month period, the number of Telstra exchanges with three or more

ULLS access seekers grew from 155 to 300, while the number of exchanges with less than three

access seekers shrank (Chart 4)

62 ACCC, 'Snapshot of Telstra’s customer access network – March 2009'

Henry Ergas and Alex Robson

0

100

200

300

400

500

600

700

Q3 2007 Q4 2007 Q1 2008 Q2 2008 Q3 2008 Q4 2008 Q1 2009

ULLS lines (thousand)

Band 4

Band 3

Band 2

Band 1

Data source: ACCC, 'Snapshot of Telstra’s customer access network'

0

20

40

60

80

100

120

140

160

180

200

1 2 3 4 5 6 7 8 9 10 >10

ULLS access seekers

Number of ESAs

Sep-07

Mar-09

Data source: ACCC, 'Snapshot of Telstra’s customer access network

Henry Ergas and Alex Robson

Greater speeds allow individuals to download the same amount of information in a shorter period

of time, or more information in the same amount of time, or some combination of these two

possibilities: i.e. more information in less time. The marginal willingness to pay for higher speeds

is the marginal change in the consumer’s indirect utility as speed changes marginally. Willingness

to pay for a discrete increase in speeds will be the accumulated marginal changes in utility over

the discrete increase in speeds.

To model this willingness to pay, consider a simple model of the allocation of time, due to Becker

(1965). The model can be extended in various directions but the basic version illustrates the main

features. Suppose that the consumer values sending and receiving information over the internet;

The consumer also values other consumption goods. Let be the quantity of other goods

consumed, which have a price of

Suppose that it takes time to download and upload information on the internet. Let be the

internet speed, measured in volume of information per unit of time (eg Mbps) Then the amount of

time devoted to downloading information from the internet is / .

This simple framework of course has its limitations. Information that is downloaded may itself take

time to be consumed. For example, in an electronic book is downloaded, it must be read, and

faster download speeds will not alter the time taken to read the book. Thus for many information

goods the amount of time actually saved with faster download speeds may not be that great if the

consumer has to take time to consume the information anyway. On the other hand, some goods

are consumed instantaneously (for example, internet games), and in this case an increase in

speeds may be more relevant to the consumer.

Similarly, other consumption goods take time to consume – for example, reading the newspaper,

drinking a cup of tea, and so on. The internet may be a substitute for some of these goods (for

example, reading a newspaper), but may be a complement to others (for example, drinking a cup

of tea). If internet speeds increase, then the consumer will switch out of the former and towards

the latter.

In any case, suppose that the consumer has lump-sum income and earns labour income at a

nominal wage rate of w from working. Let be the number of hours worked. The time and

budget constrains for the consumer are therefore:

if 0

if 0

⎧ = ⎪⎪

=⎨⎪

+ >

⎪⎩

and

63 We consider other pricing schemes below. In particular, one would be interested pricing schemes that set flat monthly

access fees, and zero per unit download charges up to a certain download limit.

Henry Ergas and Alex Robson

⎛ ⎞

+ ⎜ − ⎟ = +

⎝ ⎠

The consumer maximises utility from consumption goods and the quantity of downloads, subject

to the constraints above. At an interior solution we have:

⎛ ⎛ ⎞ ⎞

⎜ ⎜ − ⎟ + − ⎟

= = = ⎜ ⎝ ⎠ ⎟ ⎜ ⎟

⎜⎜ ⎟⎟

⎝ ⎠

At an interior solution the first order conditions equate the marginal rate of substitution between

consumption goods and downloads with the relative price:

+

=

money. The time cost comes from the loss of the wage the consumer could otherwise have

earned, multiplied by the time foregone in employment which depends on the time taken to

download one more unit of information. The direct money cost is simply the money price of

downloads. At the same time, the consumer gives up some units of consumption, which reduce

Now consider a consumer at the optimum who chooses an optimal bundle. Let the consumer’s

maximised utility be:

Then a change in s at the margin increases indirect utility by:

other variables fixed. But this is just the slope of the budget constraint, i.e. it is equal to:

⎛ ⎞

⎜ + ⎟

= − ⎝ ⎠

This means that the first term in parentheses in the preceding equation is zero (this is just the

envelope theorem), and thus:

= =

Henry Ergas and Alex Robson

The marginal utility of income at the consumer’s optimum is:

Dividing both sides by this marginal utility of income, we get:

=

The welfare effect of a marginal increase in speed is the current amount of information

downloaded, multiplied by the wage, divided by the square of the speed.

unit of information, so that downloads become less expensive in terms of their time opportunity

cost. Thus the marginal benefits of (and marginal willingness to pay for) an increase in download

speeds are likely to be higher for those consumers who have a high wage or who already

consume high download volumes. On the other hand, other things being equal, marginal benefits

fall away with higher speeds, and fall away with the square of the speed in this simple model.

The diagram below illustrates the effect on the consumer’s choices and utility of an increase in

internet speeds in this framework. Note that downloads increase, but time spent on the internet

(and therefore labour supply) may either rise or fall, depending on the elasticity of demand with

respect to changes in speed.

⎛ ⎞

+ ⎜ − ⎟ = +

⎝ ⎠

if 0

Time Constraint:

if 0

⎧ = ⎪⎪

=⎨⎪

+ >

⎪⎩

⎛ ⎞

= + ⎜ − ⎟ −

⎝ ⎠

⎛ ⎞

= − ⎜ + ⎟

⎝ ⎠

+

⎧ ⎫

⎨ − ⎬

⎩ ⎭

'

Increase in Internet Speed

Henry Ergas and Alex Robson

INTRODUCTION

The use of the Solow-Swan model to derive rough estimates of the dynamic effects of changes in

fiscal policy has a long history in the literature. Sato (1967) examines the effect of income taxes,

consumption taxes, profits taxes, wage taxes and wealth (capital) taxes in the Solow-Swan model.

Feldstein (1970), Grieson (1975) and Boadway (1979) examine tax incidence in the Solow-Swan

model assuming separate rates of taxation on labour and capital income. Atkinson and Stiglitz

(1980), Boadway and Wildasin (1984) and Jha (1998) present textbook treatments. Recently,

Gale and Orszag (2004) and Shapiro (2004) have used the Solow-Swan growth model to estimate

the growth effects of higher budget deficits (reduced rates of public savings) in the United States.

Consider a discrete-time version of the Solow-Swan model, allowing for a constant rate of

exogenous technological progress over time. The production function is:

productivity factor that grows at a constant rate, is the economy’s net capital stock at the

beginning of period t and is aggregate hours worked during period t. We assume a constant

aggregate savings rate, constant population growth, and that population growth translates into the

same rate of growth in aggregate hours worked.

hours worked, and let

capital stock evolves according to:

units in terms of effective hours of work:

where all variables are now expressed in terms of AL or “effective work hours”. The law of motion

for the capital stock per effective work hour is now:

%

( )( )

1 1

δ

In the steady state of this model, the growth rates of capital and output per effective work hour are

zero. In other words, the steady growth rate of output and capital is simply equal to the growth

rate of the population plus the growth rate of exogenous technological change. The steady state

capital per effective work hour solves:

%

( )( )

γ

Henry Ergas and Alex Robson

1

1

α

γ γ δ

= ⎜ ⎟ ⎝ + + + ⎠

This capital stock is an increasing function of the saving rate, a decreasing function of the growth

of the capital stock. The steady state capital stock per effective work hour is illustrated in the

diagram below.

In the steady state, the level of output per effective worker is constant. Thus, GDP grows at the

rate

γ

Solow-Swan model this last term would not appear.

THE SOLOW MODEL WITH GOVERNMENT SPENDING: COMPARING STEADY STATES

The NBN represents a temporary increase in government spending above that which had

previously been envisaged. To examine the economic effects of such a temporary increase, we

first examine the qualitative effects of taxation and government spending in the Solow model.

services. Then, assuming that the government has no outstanding debt, the overall saving rate as

a fraction of national income is:

Henry Ergas and Alex Robson

The first term is public saving; the second term is private saving. Suppose that the economy

begins in a steady state in which the government spends G dollars each period, which equates to

rate and that the size of government as a percentage of GDP is constant. To isolate the economic

costs of government spending, we first assume that this spending does not enter into individuals’

utility functions or production functions and that the spending is financed by a uniform tax on wage

and capital income and returned to consumers as a lump sum transfer.

assumptions, saving per effective worker is:

And so the steady state capital stock per effective worker is:

%

1 1

γ γ δ γ γ δ

= ⎜ ⎟ = ⎜ ⎟ ⎝ + + + ⎠ ⎝ + + + ⎠

Taxation reduces disposable income, which with a constant saving rate out of disposable income

reduces overall saving per effective worker even when the government balances its budget.

Lower saving reduces the steady state level of the capital stock per effective worker. This gives

us the standard result in the literature: the cost of government spending financed by income

taxation in the neoclassical growth model Solow model is that it reduces the steady state level of

capital per worker, and therefore reduce the steady state level of GDP per worker.

Henry Ergas and Alex Robson

What is the effect on steady state economic growth? Since growth is exogenous in this model

there would be no steady state or long run effects of government spending on economic growth

rates. Nevertheless we could go ahead and compute a measure of the cost of funds here by

comparing the reduction in the steady state level of output and consumption with the revenue that

the tax raises.

This cost of government spending in terms of foregone output is illustrated in the diagram below.

The imposition of the tax to fund government spending reduces output per effective worker by

cost of funds in terms of output per effective worker. The revenue raised per effective worker is

too that the total excess burden (and the total cost of funds) will be a convex function of the tax

rate here. As more revenue is required and taxes need to be raised, taxes will become more

costly at the margin.

There will also be a Laffer curve for income taxation. In what follows we shall assume that the

economy is operating on the increasing part of this Laffer curve, so that an increase in taxation

increases tax revenue.

Henry Ergas and Alex Robson

GDP Falls

Capital Stock Falls

Economic Growth Rate No Change

Wage Rate Falls

Interest Rate Rises

PERMANENT INCREASES IN GOVERNMENT SPENDING

levels. To this end, suppose that there is a permanent increase in government spending that is

financed by an increase in taxation. Computing the steady state effects of higher government

spending in each period is straightforward: higher g means higher taxes, and a lower capital stock

per effective worker, and a lower level of steady state output per effective worker.

But what happens during the transition to the new steady state? In other words, what is the

immediate effect of an increase in government spending financed by an increase in taxation?

spending. Then the capital stock per effective worker in the next period is:

%

1 1 1

1 1

δ

( )( )

1 1

α α

δ

effective work hour. In the first period the change in the capital stock per effective work hour is

simply:

%

( )( )

( )( )

1 1

1 1

1 1

α α

α

δ

γ

γ

Δ = −⎡−+ Δ + −⎤+ + ⎢⎣⎥⎦−Δ

=

+ +

In the Solow-Swan model a permanent change in tax rates and government spending has no

impact on steady state growth rates. But the above expression tells us such a change will impact

output level is lower when taxes increase, and since the capital stock falls gradually to its new

level during the transition to the new steady state, it must be the case that the economy’s growth

rate is less than the steady state growth rate between the new steady states. In other words,

64 This point has also been noted by Engen and Skinner (1996)

Henry Ergas and Alex Robson

higher government spending financed by higher taxation not only lowers the level of output; it also

lowers the rate of growth in all periods between the two steady states.

This result is shown in the diagram above for a situation in which the economy initially starts out

with no government spending. The diagram plots the average product of capital per effective

worker against the level of the capital stock per worker. The increase in spending means that the

new steady state capital stock is lower than the old one. The economy therefore has “too much”

capital stock, and must begin decumulating capital and reducing output. During this transition the

growth rate of the capital stock must therefore be less than what it was in the old steady state,

which means that along the transition path the growth rate of output must also be less than what it

was in the old steady state. Thus, an increase in government spending reduces the rate of

economic growth during the transition [see, for example, Engen and Skinner (1996)].

The implications for the path of the levels of GDP and the capital stock are illustrated in the

diagram below. The government permanently increases government spending as a fraction of

path toward the new (lower) steady state levels.

Henry Ergas and Alex Robson

IMPLICATIONS FOR TEMPORARY CHANGES IN GOVERNMENT SPENDING

All of this has implications for temporary tax-financed changes in government spending. To

analyse such a temporary change, we simply repeat the above analysis but reduce spending at

some later point in time before the economy reaches its new steady state. The effects of reducing

spending back to previous levels are simply the reverse of the effects outlined above. After the

initial increase the economy is on a transition path to a new (lower) steady state. But once

spending is reduced back to previous levels it begins a new transition path back to its old steady state.

Since the economy has been decumulating capital following the temporary increase in

government spending, when spending suddenly returns to its old level it must now have less

capital than in the old steady state. To get back to this old level it has to start accumulating capital

again, which means that once government spending falls, economic growth must now exceed the

steady state growth rate. The cost of this temporary increase in government spending is simply

the sum of the reduction in output that occurs, relative to the level that would have occurred but for

the temporary change in government spending.

reduces spending back to its previous level, putting the economy back onto a different transition

path back to the old steady state. (Note that the diagram is not to scale – it is drawn as if the

decline in output and capital per effective worker is the same absolute size, but obviously this is not correct).

Henry Ergas and Alex Robson

The analysis above assumed that the fraction of government spending borne by savings was

reasonable assumption, since permanent income does not change by much as a result of the

temporary change. To model the possibility that a greater fraction of the government spending

in government spending that is borne by savings. If the spending change is temporary, we would

%

( )( )

( )( )

1 1

1 1

1 1

α α

α

λ δ

γ

λ

γ

Δ = −⎡−+ Δ + −⎤+ + ⎢⎣⎥⎦−Δ

=

+ +

ASSUMPTIONS, CALIBRATION, SCENARIOS AND QUANTITATIVE ESTIMATES

To analyse dynamic effects in the Solow-Swan growth model, one has to make assumptions

about whether the economy is initially in a steady state (a situation in which all variables are

growing at the same growth rate) or not. The model economy can then be shocked and we can

analyse the effect on different variables of interest and see how long it takes to get back to the

steady state. Alternatively one can just take the economy as one finds it, impose the shock, and

investigate deviations from the baseline.

In exercises of this kind some compromises must be made. To analyse dynamic effects of this

policy change we assume that the economy is initially in its steady state.

Henry Ergas and Alex Robson

The following parameter assumptions are used but can easily be changed. The capital share is

cent per year. The growth rate of productivity is set equal to the long run average of the last 40

years, at 1.1 per cent per year. These latter assumptions give a steady state growth rate of 2.31

per cent per year, which is consistent (for example) with Treasury’s 2007 Intergenerational Report

assumptions regarding growth rates over the next 20 years. We assume a depreciation rate of 5

per cent, which is the roughly equal to the average rate of 5.3 per cent in the National Accounts

over the last 5 years. The saving rate is initially set at 20 per cent and the level of federal

government spending as a fraction of output is 24.5 per cent (this is the 2008/09 figure used in

MYEFO). These parameters loosely replicate Australian aggregate data, giving a steady state

capital/output ratio of 2 and a gross marginal product of capital of 18 per cent (the actual

capital/output ratio in the data is 2.94 per cent).

The full amount of NBN spending is assumed to take place in the first year of the experiment. An

alternative approach would be to spread the spending out over a number of years; this makes only

The results of the experiment with the baseline Solow-Swan model are reported in the main text

hypothetical model economy, computing the percentage output loss relative to the baseline of “no

policy change”, and then translating the results into dollar values.

The costs are estimated as the net present value of the cumulative effect on GDP out to 2020,

assuming the policy is implemented at the beginning of 2009. Costs are discounted back to

greater costs. Intuitively, as the temporary change in government spending is increasingly borne

by savings and investment, the greater are the future costs since the NBN spending and taxation

needed to fund it “eats up” more investment and therefore more capital stock.

nation’s productive capital stock. Let h be the fraction of the change in government spending that

brings about a net increase in the nation’s productive capital stock. Then the change in the

productive capital stock brought about by the temporary change in government spending is

simply:

α

λ

γ

− Δ

+ +

borne by consumption, and all spending on the project adds to the capital stock. This is a highly

unlikely scenario for two reasons. First, there is a very real possibility that this new public capital

expenditure could partially or completely crowd out new private sector expenditure. There is

much evidence in the macroeconomics literature for this crowding out effect. For example,

Aschauer (1989) finds that an increase in non-military public capital accumulation induces as

much as a dollar-for-dollar reduction in additions to the private capital stock. In the other direction,

there is evidence of “crowding in” of private investment as public investment has fallen. Makin

65 If the spending is spread out over eight years, the cost estimates are slightly higher, since output per effective worker is

falling less sharply but for a longer period, and the same spending increase represents a slightly higher fraction of GDP.

Henry Ergas and Alex Robson

(2003) notes for Australia that since the 1990s relatively lower public capital spending has been

more than offset by relatively higher private capital expenditure in the economy. The Reserve

Bank of Australia has also recently noted that in 2008 private business investment in Australia

reached growth rates not seen since the 1970s.

Second, there is the very real possibility that this capital stock will not be productive. Aschauer

(1989) finds that public investment is productive, as a general proposition. However, Otto and

Voss (1998) find no evidence of excessive returns for public investment in Australia, estimating

that the average real investment return for both private and public capital is around 9 per cent.

The Solow-Swan model also measures other parameters, including real interest rates and wages.

However, these are not additional costs – just different ways of reporting the same costs.

Results from one run of the model are presented below in diagrammatic form. The graphs below

Henry Ergas and Alex Robson

Henry Ergas and Alex Robson

INCORPORATING POSSIBLE PRODUCTIVITY EFFECTS

The above discussion assumed that there were no productivity effects of the NBN. In modern

growth models there are several approaches to modelling the supply-side effects of productive

government spending. One strategy is to assume that the flow of government spending enters

directly into the economy’s production function. This approach is used, for example, by Barro

(1990). Another approach is to assume that the government-owned capital stock enters into the

economy’s production function and that additions to this stock (government investment, net of

depreciation) augment the economy’s physical capital stock. Both approaches require estimates

of an aggregate production function, with government investment estimate included as a separate variable.

In this note we use a different approach by assuming that the project increases the growth rate of

year of the project for the first eight years. This means that there is a permanent increase in the

level of productivity after the project is completed. In any case, since the NBN technology could

become obsolete or the infrastructure may be built but take-up rates may be low, it cannot be

assumed that every dollar of additional spending will add to the nation’s productive capital stock.

Therefore a range of productivity increases and crowding out scenarios were examined, with the

net present value of the cumulative effect on GDP out to 2020 calculated.

The results of show that even if the NBN permanently increases the economy’s total productivity

level over the first eight years of the project, these gains may be outweighed by the project’s

economic costs. Because private savings are diverted from other productive uses in the earlier

years of the project, GDP may actually be lower than it would otherwise be in those years. Over

the period to 2020 the increase in the net present value of GDP depends on the sum of these

effects, and could be negative even if a significant increase in productivity is assumed. If the

increase in the economy’s productivity level is assumed to be a more modest 0.5 per cent, then

Henry Ergas and Alex Robson

the results show that the net effects on GDP are negative under a range of crowding out

scenarios.

Henry Ergas and Alex Robson

State Governments. By the late 1920s, most States had established State road authorities to

administer ‘road funds’ for the construction and maintenance of main roads. In addition, State

Governments collected a per ton mile tax and licence fees from interstate trucking operators from

the 1930s until 1954 when these charges were declared invalid. The per ton mile tax was then

replaced with a ‘maintenance tax’ of one third of one penny per ton mile and in response to

increasing evasion, the maintenance tax was replaced with diesel franchise fees in the early 1980s.

The Commonwealth government began providing funds to the States for road infrastructure

projects in the 1920s in the form of annual tied grants which were financed by the partial

hypothecation of customs and excise duties on fuels, and taxes on vehicle chassis.

Commonwealth assistance for specific roads was introduced in the 1940s and continued until they

were incorporated into annual road grant funding in the early 1970s.

In the late 1940s and throughout the 1950s, the Commonwealth greatly increased its financing of

road construction. In 1959, the decision was made to break the nexus between road grants and

fuel tax revenues. Commonwealth grants paid in the five years following this decision increased

considerably (in acknowledgement of, amongst other things, the trend toward heavier and faster vehicles).

By the mid-1970s, the Commonwealth had assumed full responsibility for funding the construction

and maintenance of ‘National Roads’ (the major links between the State and Territory capital

cities) and became considerably more involved in road expenditure decisions.

A substantial upgrading of the road network, particularly of National Roads, was undertaken in the

1980s. This was funded by the full hypothecation of an additional surcharge on the existing fuel

Development Trust Fund Act 1982 (ABRD Act) to substantially upgrade the road network,

particularly National Roads. The ABRD program and surcharge ended in 1988.

Transport Development Act 1988 from 1989 to 2000, the Commonwealth Government has set

road funding in the budget process since 1991-92.

Throughout the 1990s, the Commonwealth Government continued funding National Roads. It

increased the coverage of the network and also began funding urban links. State and Territory

Governments formally accepted responsibility for funding arterial roads and Local Governments

for funding local roads.

Accompanying this formalisation of responsibility, Commonwealth Government local road funding

was untied in 1991-92 and thereafter provided as general purpose assistance. Arterial road

assistance to the States and Territories was similarly untied in 1994, then absorbed into GST

payments to the States and Territories in 2000.

Henry Ergas and Alex Robson

In 2004-05, the national land transport policy ‘AusLink’ was implemented to achieve more

consistent national land transport funding and investment decision-making across the modes.

Over three quarters of Australian Government directed land transport funding and investment is

now undertaken through AusLink.

AusLink has the following core components:

intermodal connections. This includes major road and rail links connecting capital cities and

major industrial centres (including connections through urban areas), links to ports and

airports and other rail, road and intermodal connections.

and integrating the National Network, and the investments it will make.

Roads to Recovery, Strategic Regional and Black Spot programmes.

arrangements with the States and Territories and the private sector to share the costs of

some projects in the AusLink Investment Programme.

Funding provided independently of AusLink includes ‘Identified/Untied Local Road Grants’; some

grants tied to South Australian local roads; the Federation Fund; Australian Rail Track Corporation

(ARTC) grants; an upgrade of the mainline interstate railway track in Victoria; and the Eyre

Peninsula rail upgrade.

The most recent road funding-relevant measure is the Building Australia Fund (BAF) which was

established on 1 January 2009 by the Nation-building Funds Act 2008 to finance capital

investment in transport infrastructure (such as roads, rail, urban transport and ports),

communications infrastructure (such as broadband), energy infrastructure and water

infrastructure.

The Government has also established Infrastructure Australia (IA), which was set up by the

Infrastructure Australia Act 2008 that came into effect on 9 April 2008. IA is a policy advisory body

with the role of providing advice to Australian governments about infrastructure gaps and

bottlenecks that hinder economic growth and prosperity. It will also identify investment priorities

and policy and regulatory reforms that will be necessary to enable timely and coordinated delivery

of national infrastructure investment.

In particular section 1 of the Act states:

Henry Ergas and Alex Robson

IA is also charged with guiding the national audit and infrastructure priority list of the Building

Australia Fund announced in the 2008-09 Federal Budget. Thus section 2 of the Act states:

IA is also required to produce an Annual Report (section 26).

Regarding Ministerial directions to IA, the extent of these directions is limited. Section 6 of the Act

states:

Henry Ergas and Alex Robs

IA is comprised of 11 members and a Chair (Sir Rod Eddington). The 11 members are comprised of:

Section 28 of the Act states that IA is to be assisted by an Infrastructure Coordinator, who will

lead a small professional Office of Infrastructure Coordination within the Infrastructure,

Transport, Regional Development and Local Government portfolio:

on

8

Henry Ergas and Alex Robson

Methodology

Bilateral Auslink agreements between

Commonwealth and States

National Guidelines for Transport

System Management in Australia

Volume 3

Infrastructure Australia’s

Prioritisation Methodology

National Guidelines for Transport

System Management in Australia

Volume 3

Handbook of Cost-Benefit Analysis

Infrastructure Australia to apply

prioritisation methodology which

includes CBA to determine priority list

for funding of national projects

Only for Auslink funded

projects

Henry Ergas and Alex Robson

for the Submission of Business Cases

and Gateway Reports (TC08/07)

WWG Guidelines for Privately

Financed Projects

Section 95 of Auslink bilateral

agreement with NSW

NSW Government Guidelines for

Economic Appraisal

WWG Guidelines for Privately

Financed Projects

Determination of Appropriate

Discount Rates for the

Evaluation of Private Financing

Proposals

Economic appraisal required for all

individual projects with a total cost in

excess of $1 million.

Full appraisals are required of projects

over $10 million and summaries only

required for projects worth between $1

million and $10 million. Appraisals to

be submitted to Treasury.

Treasury then makes

recommendations to Budget

Committee based on review of

appraisal.

More generally this economic

appraisal is part of a requirement on

agencies to submit a business case to

Treasury.

SOCs (State Owned Corporations)

generally required to submit economic

appraisals for PFPs (privately financed

projects) only.

National PPP Guidelines applicable to

other types of PPPs than PFPs.

Projects likely to have potential to

provide value for money using a PPP

delivery method are those with a total

capital value exceeding $50 million

and therefore projects of these value

trigger evaluation of PPP as

procurement method.

“Selection of major projects

undertaken by an agency

should be subject to expost

evaluations as should

major ongoing programs

which may involve a series

of smaller project. These

involve:

re-evaluation of benefits

and costs of selected option

to assess whether the

anticipated benefits were

realised and the forecast

costs kept to;

reconsideration of

alternative options;

examination of the project

design and implementation

to assess scope for

improvement to option

adopted. “

Henry Ergas and Alex Robson

Section 82 of Auslink bilateral agreement

with Vic

1) Investment Lifecycle Guidelines

(ILG): Options Analysis and

Business Case

2) Partnerships Victoria requirements

supplementing National PPP

Guidelines

3) Dept of Transport Guidelines for

cost-benefit analysis (consistent

with National Guidelines for

Transport System Management)

Requirements under ILG mandatory for

major investments, defined as

investments requiring more than $5 million

in funding.

Dept. of Transport Guidelines apply to all

‘significant new projects’ subject to review

by Dept of Transport’s Project Review

Committee (expenditure of at least $10

m).

Yes in ILG

Yes for Auslink funded

projects

Henry Ergas and Alex Robson

Value for Money Framework (for PPPs)

Section 104 of Auslink bilateral agreement

with Qld

Project Assurance Framework: Cost

benefit analysis guidelines

National PPP Guidelines

Project Assurance Framework requires

Department of Infrastructure and Planning

to produce documents accompanying six

generic project stages:

- preliminary evaluation

- business case development

- supply strategy development

- source suppliers

- establish service capability

- deliver service.

CBA is required at stage of analysis of

project options.

Yes for Auslink funded

projects

No for other projects

Henry Ergas and Alex Robson

Treasurer’s Instructions No. 17

Section 86 of Auslink bilateral agreement

with SA

Guidelines for the Evaluation of Public

Sector Initiatives

Cabinet approval is required of any

initiative with an estimated cost equal to or

greater than $11 million. This will be

based on evaluation prepared in

accordance with Guidelines. The

Evaluation to include

Identification of the Service Need;

Identification of Options to Deliver the

Service;

Substantiating the Project;

The Determination of Funding and Post

Implementation Review.

Yes as set out in Guidelines.

Yes for Auslink funded

projects

Henry Ergas and Alex Robson

Acemoglu, D. and J. Angrist 2001, ‘How large are human-capital externalities?: Evidence from

Cambridge, Mass.; London, England.

Economic Review, 92(4), pp. 1126–1137.

Analysys Mason for BSG 2008, ‘The costs of deploying fibre-based next-generation broadband

infrastructure’, Final Report, 8 September.

Arrow, K. J. and R. C. Lind 1970, 'Uncertainty and the evaluation of public investment decisions',

American Economic Review, 60(3), pp. 364-378.

177–200.

171-188.

Working Paper, 2144, The World Bank.

and Oxford.

of development, MIT Press, Cambridge, Mass.; London, England.

Economy, 98(5), pp. S103-S125.

pp. 775–805.

Studies, 46(3), pp. 505-511.

Henry Ergas and Alex Robson

Economics, 8(2), pp. 111-131.

trade, location and growth, Cambridge University Press, Cambridge; New York.

pp. 129-144.

Public Economics, (75)3, pp. 315–40.

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

Annales d’Economie et de Statistique, INSEE : Paris (English version in Champsaur, P. (ed.) 1990,

Essays in Honor of Edmond Malinvaud: Vol. 2, Macroeconomics, The MIT Press, Cambridge, Mass.).

in Twentieth Century Australia, George Allen & Unwin, Sydney.

Chandra, A. and E. Thompson 2000, ‘Does public infrastructure affect economic activity?: evidence from

Cohen, J. P. and Monaco, K. 2008 ‘Ports and Highways Infrastructure: An Analysis of Intra- and

Dobes, L. 2008, ‘A century of Australian cost-benefit analysis’, Working papers in cost-benefit analysis,

Commonwealth Department of Finance and Deregulation.

Dobes, L. and Bennett, J. 2009, ‘Multicriteria analysis: "good enough" for government work?’,

Donaldson, D. 2009, ‘Railroads of the Raj: Estimating the Impact of Transportation Infrastructure’,

London School of Economics.

assumptions, data and analysis, Final Report.

Washington, D.C.; The Lincoln Institute of Land Policy, Cambridge, Mass.

and so little good, Oxford University Press, Oxford.

Elhorst, J. P. and J. Oosterhaven 2008, ‘Integral cost-benefit analysis of Maglev projects under market

______2003, ‘Forecasting the impact of transport improvements on commuting and residential choice’,

Journal of Geographical Systems, 8(1), pp. 39-59.

Henry Ergas and Alex Robson

Engen, E. M. and J. S. Skinner 1996, ‘Taxation and economic growth’, NBER Working Paper No. W5826

pp. 153-163.

and Reform, 16(3).

Sydney.

A Journal of Policy Analysis and Reform, 15(4), pp. 35-58.

Ergas, H. and A. Hardin 1999, ‘Annuity and levelisation issues in forward looking cost models: the case

of telecommunications’, paper presented for the ACCC Depreciation Forum, 30 September.

Australia's Broadband Future: Four doors to Greater Competition, CEDA.

Estache, A. and D. Martimort 1999, 'Politics, transaction costs, and the design of regulatory institutions,

World Bank Policy Research Working Paper No. 2073, available at SSRN:

http://ssrn.com/abstract=620512

Evans, J., P. Levine and F. Trillas 2008, ‘Lobbies, delegation and the under-investment problem in

Journal of Economics, 88(4), pp. 551-573.

Review of Economic Studies, 41(4), pp. 505-513.

Fernald, J. G. 1999, ‘Roads to prosperity? Assessing the link between public capital and productivity’,

American Economic Review, 89(3), pp. 619–638.

Florio, M. 2007, 'Multi-government cost-benefit analysis, shadow prices and incentives', in Florio,

Cambridge University Press.

Economic Activity, vol 2004, No. 2, pp. 101-187.

Gans, J. 2009, ‘The right policy for telecommunications and broadband’, Submission to the Senate

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

Glaeser, E. L. and J. D. Gottlieb 2008, ‘The economics of place-making policies’, NBER Working Paper

No. 14373.

Papers on Urban Affairs, pp. 101-149.

Henry Ergas and Alex Robson

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

Economics, 1975, 4(1), pp. 75-85.

Literature, 44, pp. 925-972.

Heyma, A. and J. Oosterhaven 2005, ‘Social cost-benefit analysis and spatial-economic models in the

planning models in the Netherlands: theory, application and evaluation (pp. 155-175), Ruimtelijk

Planbureau/NAi-uitgevers, Rotterdam.

Huberman M. and Minns C. 2007, ‘The times they are not changin’: Days and hours of work in Old and

275.

and micro transport planning and investment tools: summary and conclusions, Round Table, 25-26

October 2007, Boston, Discussion Paper No. 2008-6, January OECD-International Transport Forum.

Jullien, B., J. Pouyet and W. Sand-Zantman 2009, 'Public and private investments in regulated network

industries: coordination and competition Issues', mimeo.

Economic Journal, 68(2), pp. 310-329.

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

Journal of Political Economy, 85, pp. 473-92.

Laffont, J. J. 2000, ‘Etapes vers un état moderne: une analyse économique’ in Etat et Gestion Publique,

Actes du Colloque du 16 Décembre 1999, La Documentation Française, Paris, pp. 117-150.

Journal of Law and Economics 18, pp. 875–901.

Levine, P., J. Stern and F. Trillas 2005, ‘Utility price regulation and time inconsistency: comparisons with

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

University Press, Cambridge [England]; New York, NY.

Makin, T. 2003, 'The changing public-private infrastructure mix: economy-wide implications’,

Australian Journal of Public Administration, 62(3), pp. 32-39.

Henry Ergas and Alex Robson

Mees, P. 2008a, ‘Australia’s future infrastructure requirements: avoiding wasteful expenditure in urban

transport’, submission to Infrastructure Australia’s discussion paper: Australia’s future infrastructure

requirements.

School of Global Studies, Social Science & Planning, RMIT University, Melbourne.

Institution, Transport Research Program, Washington D.C.

Transport Plan, Prepared for Victorian Department of Transport.

Michaels, G. 2008, ‘The effect of trade on the demand for skill - evidence from the Interstate Highway

Mohring Jr., H. and Williamson, H. F. 1969, ‘Scale and "Industrial Reorganisation" Economies of

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

of Economics, 104(2) pp. 399-415.

23(3), pp. 659-675.

42, pp. 47-66.

Paolucci, F., H. Ergas, T. Hannan and J. Aarts 2009, ‘The effectiveness of health informatics’, in press.

Posner, E. 2001, 'Controlling agencies with cost-benefit analysis: a positive political theory

pp. 525-51.

economic benefits, report prepared for the Department of Premier and Cabinet, Victoria.

Quinet, E. and A. Sauvant 2007, ‘Une méthode d’optimisation des programmes d’investissements de

des investissements de transport, Economica, Paris.

Shapiro, M. D. 2004, ‘Discussion of Engen and Hubbard ‘Federal government debt and interest rates’’,

NBER Macroeconomics Annual Conference, April.

Soria, B. and Hernández-Gil, F. 2009, ‘Exploring potential natural monopoly properties of broadband

access networks’.

Henry Ergas and Alex Robson

Journal, 89, p. 97.

Finance (pp. 241-361), The Brookings Institution, Washington D.C.,

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, 17 December 2007, Schedule A.

_____ 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.and D. A. Wittman (eds.),

Oxford handbook of political economy, (chap 26), Oxford University Press, N.Y.

Wildavsky, A.. 1966, ‘The political economy of efficiency: cost-benefit analysis systems analysis, and

Funds Act 2008,

http://www.comlaw.gov.au/ComLaw/Legislation/LegislativeInstrument1.nsf/0/903949216C558D72CA257

52600833469/$file/BAF+Criteria+Instrument+(FINAL).doc

projects: Case study 1: Wallaville Bridge.

Bypass.

three years.

September.

Henry Ergas and Alex Robson

NSW Treasury 2008, ‘Circular: Revised project size/risk thresholds for the submission of business cases

and gateway reports’, NSW TC 08/07.

Paper, June.