Melbourne Airport Rail: a patronage discussion


Overview

Melbourne Airport Rail is becoming a reality. Current plans have it opening in 2029, four years after Melbourne’s Metro Tunnel.

I won’t go into the project details here, but there are some concerns about them. The rail flyover at Albion and use of existing tracks from Metro Tunnel to Sunshine leave many questions unanswered. Other plans are on hold while Airport Rail is prioritised. We haven’t seen a viable business case, and I wonder if we ever will.

Here, I’m focusing on the future patronage of Airport Rail, and its effect on other ways of getting to and from the airport.

Through the busiest hour of the day, I predict airport trains will average about 220 passengers each to and from the Airport; less than half their seating capacity of 500. 

This will be in 2029, provided air travel gets back onto a growth trajectory post-COVID. Airport Rail patronage might match, but not exceed, that of SkyBus. Together they’ll do little to stem the growth of airport-related road traffic, which will still have to cater for two-thirds of the demand.

However, even at this seemingly low level of usage, Melbourne Airport’s overall transit mode share will be very high by Australian – and even some international – standards.

Pre-pandemic:

  • 70% of air travellers were interstate and international visitors, and two-thirds were on leisure trips
  • 85% of air travellers were to/from Greater Melbourne, and over half were to/from inner Melbourne
  • Airport area workers mostly lived close to the airport, and nearly all drove to work
  • Transit took 20% of trips to and from the airport, mainly on SkyBus.

Amongst Australia’s main airports, transit already takes a high share of Melbourne Airport’s landside travel demand. That’s despite its being much further away from the CBD than most, and not having rail access.

My calculations suggest that Airport Rail could increase transit’s share of airport landside travel to over 40% after 10 years of operation, as long as air travel’s pre-COVID growth trajectory resumes (albeit with a few years’ lag).

That’s a significant shift, but it won’t be enough to halt absolute growth in car/taxi use. Traffic queues on the Tullamarine Freeway won’t be going away because of it.

Post-COVID, domestic air travel will probably bounce back strongly, but international air travel will take much longer to recover. However, there are some big unknowns:

  • Will cheap, high-volume international air travel be possible, in a net-zero emissions world? I doubt that it’ll ever be as cheap as it was pre-COVID.
  • Will high speed rail ever eventuate on the east coast? If so, it could take ten years’ worth of air travel growth away from Melbourne Airport.

Let’s delve into a bit more detail.

Pre-COVID airport travel

In 2019, when the Airport handled 37 million passengers a year, there were about 134,000 landside trips to and from the airport on an average weekday. 33,000 of these were airport area workers, leaving 101,000 trips to and from the airport for travel reasons daily.

The mode share of trips to and from the airport was as below. 20% of all trips were by bus, of which the vast majority were on SkyBus.

42% of all Airport weekday trips (57,000 trips) were to and from inner Melbourne, with mode shares as below. 31% of trips were by bus, but taxis carried the largest share (44%).

Trip distribution and growth

The maps below show the distribution of trips to and from the Airport, using data from a study I was involved in a few years ago, for airport passengers, and ABS journey to work data for airport area workers. The concentrations of air passengers to/from inner Melbourne, and airport area workers to/from places close to the airport, are clearly evident.

The maps also show my estimates of future changes, due to population growth and associated trip re-distribution. I’ve shown the amount of air travel activity as well (mmpa = million movements of passengers per annum): this might be a better indicator than the year, going forwards.

Despite strong growth in population in the west, outer east and north, these areas contribute small overall percentages to air travel. This is not surprising, given that 70% of air travel is from international and interstate visitors, and most of them travel to and from Inner Melbourne.

Time of day of travel

There’s a ‘lag and lead’ between the time of day that passengers arrive and leave by air and the time that they make their landside journeys, because of check-in, customs clearance and baggage collection times. Allowing for this (and adding movement of airport area workers) enables profiles of typical daily landside travel demand through the day. My estimates are shown in the graphs below.

Trips arriving at the airport in the mornings show a pronounced peak due to airport area workers arriving at the same time as peaks in both domestic and international air travellers. Trips leaving the airport are more spread out across the day, giving rise to a lower and broader peak in the afternoon.

The busiest hour (6-7am) sees about 6,300 trips to the airport, just under 5% of the daily two-way total.

The future for air travel is by no means certain

At the moment, post-COVID recovery is expected to add 3-4 years’ delay to pre-COVID population growth projections in Australia. On top of that, air travel will take longer to recover from the massive effect of COVID, and re-couple itself to that background growth. I think the overall effect could be to add 4-5 years to pre-COVID air travel growth projections.

However, a big unknown is how a global shift to zero emissions might change air travel in the longer term. If the world gets its act together on climate, will international air travel go down again? Will climate pricing put air travel out of our reach? Will high speed rail ever eventuate in Australia, providing an emissions-free alternative for intercity travel? These and other game-changers (including another pandemic, or severe climate deterioration if we don’t manage to prevent it) could have big implications for the future of air travel in the years following Airport Rail’s completion.

Effect of Airport Rail

I’ve built a simplified mode choice model to predict the effects of Airport Rail, accounting for its attractiveness as well as travel times, fares and service levels compared with those of other modes of transport.

Using this model, I predict that, soon after opening in 2029, if Melbourne Airport reaches 48 million air passenger movements a year, Airport Rail would attract similar patronage to bus. Overall transit mode share could increase to 33% (from 19% pre-COVID).

Ten years after opening, with continued air travel growth (to 69 million passenger movements), Airport Rail could exceed 50,000 trips a day, and bus a similar amount. Overall transit mode share could increase to 41%.

Airport Rail will attract patronage from car/taxi and SkyBus, as shown in the graph below.

Even with Airport Rail, growth in air travel activity means that car/taxi use could be 12% more than pre-COVID in 2029, and 37% more in 2039. Airport Rail won’t halt growth in car/taxi use, so queues on the Tullamarine Freeway will persist.

Of the 28,000 trips a day on Airport Rail in 2029, about 1,300 trips will be towards the airport in the airport peak-hour (6-7am). With the proposed 10-minute service, each train will carry about 220 passengers, at their busiest.

This isn’t much for a 1,380-passenger, 500-seat High-Capacity Metro Train. Airport trains will work harder than this on the Dandenong line; they’ll bring a lot more people into the city than they will to the Airport. Maybe there’s a case for a 15 or 20 minute service instead of 10 minutes, freeing up more trains to serve the rest of the Sunbury line instead.

We’ll have to wait until 2029 to see if my projection is right (and longer still, if air travel doesn’t pick up again by then), but it doesn’t seem like a big result for $10 billion of taxpayer funds. Incidentally, that’s about the same cost that the much more complex Metro Rail Tunnel was (in 2015 dollars), before the cost increases seen during its construction.

How does it work elsewhere?

Is it possible to gauge how Melbourne’s Airport Rail link might fare compared to other cities? Fortunately, there’s a reasonable amount of data available from other airports; I’ve assembled some of it below, from a few different sources.

Comparisons are odious; rail arrangements vary enormously from one city airport to the next, as do general public transport networks in their host cities. 

The airports I’ve found data for aren’t a statistically reliable sample, and I can’t be sure that the travel metrics are defined consistently between them. Nonetheless I think they give some useful insights.

Firstly, I’ve compared airport transit mode shares (with and without rail) with overall city transit mode shares in the graph below. I’ve shown Melbourne as it was in 2009 (for comparison with the other cities) and 2019 (just before COVID), and as I predict it will be, with Airport Rail, in 2029 and 2039.

The graph shows a big spread around the linear trend.

Predictably, US and Australian cities are clustered in the low transit mode share range, European cities are in the middle and Asian cities are high (although my sample of them is small).

Airports without rail access are generally in cities with low overall transit shares, but they still mostly achieve higher transit mode shares than their host cities.

Adelaide, Perth and Brisbane airports all have transit mode shares lower than their city averages.

Brisbane is a special case worthy of further comment. Airtrain was privately built in the early 2000s and struggled to attract patronage initially. The operating agreement prevented the Government from sending any more transit to the airport; they built a new toll road instead, thus further disadvantaging Airtrain (Airport Link, the toll road, has also failed to achieve originally predicted revenues). Airtrain attracts only 8% of airport travel demand, in a city whose overall transit mode share is over 10%. The City of Brisbane now envisages that their new Metro transit system might serve the airport in future, after the Airtrain concession ends in 2036. I wonder if Airtrain will survive!

Melbourne is the only Australian city whose airport has an appreciably higher transit mode share than the city-wide average, although Sydney’s comes close (and its airport rail mode share has grown since 2009).

Amongst cities without airport rail, Melbourne’s airport achieved a high transit mode share pre-COVID. At 20% in 2019, it was nearly double the average for the city as a whole. This was mainly due to SkyBus, which had grown its share of airport travel considerably in the preceding years.

My predictions suggest Airport Rail would put Melbourne Airport outside the US/Australian ‘pack’ when it opens in 2029. Ten years after that, it would be much closer to European levels of airport (but not city-wide) transit shares.

If anything, this suggests that my predictions are on the high side. It would put Melbourne in a class of its own, where airport transit mode share is well over double the city-wide average.

Another way of looking at this is to compare transit mode share with airport passenger throughput, as below.

Although this isn’t a statistical sample of airports, it’s interesting that Melbourne is on the average trend line of the group. By 2019, thanks to SkyBus, it already had a higher transit mode share than Sydney and JFK New York apparently had in 2009, both similar in passenger throughput but also both served by rail.

When Airport Rail opens, my prediction of Melbourne’s transit share will approach that of similar-sized European airports, and by 2039 it’ll up there with London Heathrow and Paris Charles de Gaulle’s 2009 figures. This looks pretty optimistic, suggesting again that my forecasts, despite not filling up the trains much, could well be on the high side.

Conclusions

My prediction is that Airport Rail would attract about 28,000 trips a day when airport activity is at 48 million passengers a year. This could coincide with Airport Rail’s opening year of 2029, if air travel recovers from COVID’s effects.

This would be 16% of all trips to and from the airport. Added to SkyBus, overall transit mode share would be 33%, compared with about 20% pre-COVID.

At this level, there’d be about 1,300 passenger trips on trains to the airport in the busiest hour. This is 220 per train with the proposed 10-minute service, so airport trains would be running with less than half their seats occupied (there’d be plenty of room for luggage!).

Even with the proposed myki fare, Airport Rail would only match, rather than exceed, SkyBus patronage levels in future years. This is assuming that SkyBus will be upgraded to keep pace with demand growth as well, which would be the best way to increase overall transit mode share for airport trips.

Incidentally, shutting down SkyBus to get more patronage on Airport Rail would be a very bad move. Airport Rail won’t attract all Skybus users; most of them would probably choose taxi or rideshare services instead. I estimate that overall transit mode share would go down by as much as ten percentage points, compared to my predictions herein.

When (and if) the Suburban Rail Loop reaches the airport from the east, airport trains would then carry people on orbital journeys as well. Easy interchange would be required between SRL and airport trains at the Airport, to help attract people moving between the northern and western suburbs. I hope the station design will allow for this.

I’ve written this post to put my forecast on record. If a business case is published (without redactions!) we’ll see how my figures compare with Government predictions. It’ll be 8 years before we can see the actual result, and of course things could change dramatically again, at any time…

Infrastructure Victoria’s 30-year strategy: confusing and incomplete


Overview

INFRASTRUCTURE VICTORIA released its new draft 30-year infrastructure strategy for community comment in December 2020. Comments are invited before 26 February. A final strategy will then be presented to the State Government, who’ll have 12 months to respond and develop a 5-year infrastructure plan (based on the strategy, one hopes!).

There’s little to explain how its 95 recommendations were arrived at. There are very few stated targets or directions, nor is there a list of KPIs to monitor as time passes, and drive adjustments to the strategy.

IV’s ten stated objectives are very broad, motherhood statements, like ‘Foster healthy, safe and inclusive communities’, or ‘Protect and enhance natural environments’.

The problem with such vague objectives is that nearly all Government strategies (and many corporate ones) have them in one form or another: they’re not specific to an infrastructure strategy. Recent major project business cases and EESs have them too; in that case they’re used to overstate positive effects whilst downplaying negative ones. ‘Optimism bias’ is the polite term for this; I prefer to call it ‘deliberate distortion’.

Sadly, IV’s strategy has fallen victim to this problem as well.

Even against the nebulous objectives, there’s no indication that the strategy is well-balanced, that it’ll achieve its purpose, or that it’ll avoid undue and unacceptable trade-offs. These are vital attributes of a well-founded, evidence-based and responsible strategy these days.

If anything, the far-from-perfect scorecard indicates that the strategy puts economic objectives ahead of social and environmental ones. It’s time we stopped doing that, once and for all.

Overall, there’s little to demonstrate:

  • where it’ll get us to in 30 years, on measurable trajectories (like emissions reduction/elimination, air quality improvement, energy efficiency/renewability, community safety/health, economic performance); or
  • what’ll be required if sudden shocks (like a pandemic!) or new imperatives (like fast-tracking emissions elimination, or adaptation to rapid climate change) arise.

As I understand it, the State Government is free to accept or reject IV’s strategy. Maybe they’ll cherry-pick from it, and/or modify it to suit their purposes. They might just ignore it completely.

IV doesn’t state that the strategy is a cohesive whole and should be adopted in its entirety. I suppose that’s not surprising, because the strategy isn’t a cohesive whole anyway.

In this day and age, we need to plan thoroughly and carefully. Our strategies and plans must include initiatives that have been properly tested against all alternatives, and against strategic objectives that contain concrete targets or directions (avoiding imprecise words like ‘improve’ or ‘maximise’) and eliminate negative impacts and trade-offs.

All our planning now needs to fall in line with the overriding imperative of avoiding catastrophic climate change. If it doesn’t, then it will, by definition, only contribute more to the problem.

The draft IV strategy – bewildering structure

The draft strategy contains 95 recommendations, variously grouped under:

  • Ten objectives (plus one principle)
  • Four sections (and sixteen sub-sections)
  • Ten sectors

The names of the objectives, sections and sub-sections all start with an action verb. Look at the table above – someone’s tried to ensure that every verb is different, although ‘align’, ‘build’ and ‘foster’ each crept in twice.

Although the verbs are action-oriented, they’re also largely non-specific. ‘Advance climate change mitigation and adaptation’, ‘Align social infrastructure with better service delivery’ – what do these actually mean?  How would one gauge when they’ve been fulfilled?

The ten stated objectives are generalised and motherhood in nature and there’s little to show that they’ve been developed or derived from a demonstrable evidence base. They cover all the usual, laudable principles, such as ‘Foster healthy, safe and inclusive communities’, ‘Reduce disadvantage’, or ‘Lift productivity’.

The four main sections: ‘Confront long-term challenges’, ‘Manage urban change’, ‘Harness infrastructure for productivity and growth’ and ‘Develop regional Victoria’, seem to have little meaning and a lot of overlap. It’s unclear why they were chosen, and how they relate to the objectives.

The structure of the document is endlessly confusing. Why not establish some definitive objectives (real outcomes, not motherhood principles) and why they exist, then discuss what actions are needed to achieve each? If any actions satisfy more than one objective, that can be explained. If any actions detract from any of the objectives, then say so, and/or omit or change them, so that they don’t; this is a major omission from the document as it stands (more on this later).

The recommendations – a grab bag

The 95 recommendations cover a wide range of things. Some are projects, some are policies, some are simply actions (like ‘Publish Victoria’s transport plan’).

Surprisingly few recommendations encapsulate tangible targets or directions. By tangible I mean a defined target in absolute or percentage terms, with a firm date. I’ve found seven of these:

  • Zero net greenhouse gas emissions by 2050 – this is already legislated and is cited in many of the recommendations.
  • Phase out internal combustion engine vehicles during the next 30 years (recommendation 1).
  • All new homes to be 7-star energy rated by 2022, 8-star by 2025 (recommendation 4).
  • Achieve legislated 2032 public transport accessibility targets (recommendation 44).
  • Renew at least 50% of all older public housing (low rise apartments and 3-bed detached houses) by 2031 (recommendation 57).
  • 30% tree canopy coverage in new growth areas (recommendation 71).
  • Reach at least 4.5 social housing dwellings per 100 households by 2031 (recommendations 73 and 94).

Many recommendations stipulate timeframes (some very short) for implementing new policies and process reforms, but don’t attach physical targets or objectives to those changes.

The strategy frequently mentions the legislated target for Victoria to achieve net zero greenhouse gas emissions by 2050. However, there are no targets stated for the contributions of each infrastructure sector to that total, nor to any interim steps along the way. There’s no evidence that the strategy is designed to help achieve net zero emissions, nor do I have any confidence that it will.

How were the recommendations arrived at?

There’s little demonstrated rationale behind the choice of each recommendation.

Most of the evidence base used to support the recommendations relates to existing shortcomings (for example, public housing provisions, or public transport accessibility). Some projections are made, but they’re mostly looking at ‘business-as-usual’ situations.

A possible exception to this is contained in a supplementary report on transport and land use modelling, which I’ve just started reading. I’ll cover it shortly in a follow-up post. I’ll be looking for what the transport modelling tells us about the future, and evidence that the strategy’s recommendations have been informed by this in some way.

A few choice issues to get to grips with…

I have issues with just about every recommendation. However, I’ll concentrate on a few particular ones here.

Circular economy (recommendations 28-31)

IV states that: ‘Also known as a ‘closed loop’ system, a circular economy aims to reduce the environmental impacts of production and consumption by avoiding waste and reusing or recycling materials.’

This is incorrect. A circular economy is one that eliminates waste and the constant use of resources. The Ellen MacArthur Foundation has a lot of detail about the circular economy. Here are their definitions (bolding is mine):

THE CIRCULAR ECONOMY
An economy that is restorative and regenerative by design. In a circular economy economic activity builds and rebuilds overall system health. The concept recognises the importance of the economy needing to work effectively at all scales – for big and small businesses, for organisations and individuals, globally and locally. It is based on three principles:
1. DESIGN OUT WASTE AND POLLUTION: What if waste and pollution were never created in the first place? A circular economy reveals and designs out the negative impacts of economic activity that cause damage to human health and natural systems. This includes the release of greenhouse gases and hazardous substances, the pollution of air, land, and water, as well as structural waste such as traffic congestion.
2. KEEP PRODUCTS AND MATERIALS IN USE: What if we could build an economy that uses things rather than uses them up? A circular economy favours activities that preserve value in the form of energy, labour, and materials. This means designing for durability, reuse, remanufacturing, and recycling to keep products, components, and materials circulating in the economy. Circular systems make effective use of bio-based materials by encouraging many different uses for them as they cycle between the economy and natural systems.
3. REGENERATE NATURAL SYSTEMS: What if we could not only protect, but actively improve the environment? A circular economy avoids the use of non-renewable resources and preserves or enhances renewable ones, for instance by returning valuable nutrients to the soil to support regeneration, or using renewable energy as opposed to relying on fossil fuels.

Source: Ellen MacArthur Foundation

Of course, the Victorian Government can’t go it alone, nor can the Federal Government; it’ll take a global effort. Nonetheless, there are some vital principles that IV must take seriously if they want to use the ‘circular economy’ term.

IV’s definition refers only to reducing waste, and using it more effectively. Whilst this is laudable (and IV’s recommendations here are good, but incomplete), it all falls well short of a circular economy and shouldn’t be termed as such.

IV says ‘In early 2020, the Victorian Government publicly committed to transitioning to a circular economy in its new policy framework Recycling Victoria: A new economy.’ I haven’t read this, but if it only covers recycling then it can’t be said to cover all aspects of a circular economy.

Sorting out our broken recycling system is vital, but it’s nowhere near enough.

Throughout the strategy document, IV suggests questions it would like our responses to. For the circular economy, on page 95, they are:

  • What other cost-effective actions can the Victorian Government take towards a circular economy?
  • What more can the Victorian Government do to enable more private investment in recycling facilities?

These are loaded questions, driven by the severely limited definition of a circular economy that IV has used. Why should circular economy actions be limited to things that are ‘cost-effective’ (apparently it used to be cost-effective to export our waste to China, for them to recycle it!)? Why is more private investment needed in recycling facilities? Why not public investment, subsidy or legislation to achieve circular economy outcomes?

Governments must consider how to reorientate all their spending decisions (and revenue sources) to circular economy activities. A move towards a circular economy can only be made through this lens. Positive actions rather than negative ones, eliminating trade-offs, would be a good place to start.

I think this whole area of discussion is vital and deserving of much more than IV has put into it. IV (and the Victorian Government) mustn’t misuse or cheapen the circular economy term.

Transport pricing confusion (recommendations 45-55)

There are a lot of recommendations that address transport pricing:

  • Adopt peak and off-peak public transport fares
  • Price each public transport mode differently
  • Abolish the free tram zone
  • Remove annual charges while introducing distance-based pricing for electric vehicles
  • Appoint an independent transport pricing adviser
  • Increase and extend the Melbourne Congestion Levy on parking
  • Incorporate congestion pricing for all new metropolitan freeways
  • Trial full-scale congestion pricing in inner Melbourne
  • Trial demand-responsive pricing on parking in inner Melbourne
  • Price parking at major public transport hubs, all train stations and park-and-rides
  • Phase out fixed road user charges and introduce user pays charging

This is trying to address several issues at once, and not very well, in my view. They affect:

Public transport revenue (positively) and use (negatively):

  • Peak and off-peak fares are worthwhile, not to reduce peak crowding but to encourage more off-peak use. But IV proposes increasing peak fares, rather than reducing off-peak ones. 
  • Pricing each transit mode differently will undo the benefits of integrated fares and separates the modes instead of seeing them as components of the same system. Perhaps instead of the oversimplified, politically-manipulated two-zone fare system for Melbourne, we should consider distance- or time-based fares for all public transport; that was one of the stated benefits for moving to integrated automated ticketing in the first place.
  • Abolishing the free tram zone seems popular amongst the pundits, but many cities have free transit in their CBDs. In Melbourne, many users have already paid for their tram use through using transit to commute (and this percentage will increase in future, as car use declines for CBD commuting). Will abolition really affect CBD tram crowding all that much? The free City Loop service is a separate tourist attraction and can’t safely cater for increased demand without being modernised (losing its tourist appeal in the process).
  • Pricing park-and-ride will push more users away from transit. A better solution might be to ‘gate’ access to secure station car parks using a valid myki, then charge for removing the vehicle if the myki hasn’t been used that day from that place (to deter parking freeloaders).

Compensating for revenue loss from non-fossil fuel vehicles:

  • Replacing up-front charges with distance-based ones for EVs will increase perceived costs of their use, and could slow EV take-up. Has the effect of this been taken into account?
  • The Government has already announced a distance-based charge from July 2021, of 2.5 cents per km for EVs (and 2 cents for plug-in hybrids); the rationale for this is to replace lost revenue from fuel excise and help make EVs ‘pay their way’; a debatable notion even for fossil fuelled vehicles.
  • Why are EVs and hybrids singled out for special pricing treatment? EV charge revenues go into State coffers while fuel excise is Federal, and there’s no hypothecation of either of them.
  • We need incentives, not barriers, to wider, faster EV adoption. Transition settings should allow for this as part of the move towards wider pricing reform for all transport (which is long overdue).

Managing traffic congestion:

  • A key feature of using direct pricing to manage congestion is that the more successful it is, the less revenue it raises.
  • Why use congestion pricing on new freeways (thus making them into tollways)? We need to price users onto them, not away from them (and onto other roads). Of course, we can’t do this properly while we still have toll roads as well. Rather than build more (toll) freeways, what would happen if existing toll roads were bought back, and their tolls removed? Has the effect of doing this been tested for its effects on congestion? Or on the subsequent ability to introduce a proper road pricing regime, one that prices traffic off the roads we don’t want it on?
  • I have no problem with a Government levy on off-street parking in busy areas, but I haven’t seen any proof that it has directly reduced congestion. Too many other things have happened since its implementation, and I suspect ‘congestion levy’ is a serious misnomer. Sydney has similar levies and they apply in several places – the CBD, North Sydney, St Leonards and Chatswood, for example. Why not use them in some of the other activity centres around metro Melbourne, in conjunction with transit improvements like the Suburban Rail Loop?

We must consider pricing of public and private transport together, not separately. The overall objectives of urban transport pricing must be clear and unambiguous. There are inherent conflicts between pricing to increase revenue, to recover costs or to manage demand and mode choice.

IV’s recommendations don’t represent a clear and cohesive package; collectively their main effects will probably be to push people away from both public transport and future freeways, which doesn’t make any sense to me.

The single best recommendation of the lot is to appoint an independent transport pricing adviser, provided they have teeth to make their advice stick. Their first action should be to throw out all the other recommendations and start again!

East-West Link (recommendation 67) 

Oh no, not again…

What’s the rationale behind choosing to protect an inner-city east-west road link (i.e. East West Link, modified to join up with the dodgy West Gate Tunnel)? This route has never been protected before, and much of the land it would need is already used for other, more valuable purposes (like residential areas or Royal Park). Land acquired for it by the Napthine Government has been returned by the Andrews Government to its previous purpose and sold back to others.

Although it became a costly political dogfight, I always thought the real reason for cancelling the project was its large impacts (including negative economic impact as well as property destruction and community opposition).

What’s changed?

Reintroducing it without demonstrating that it’s needed, or justified, is just inviting more political chicanery and wasted public funds.

How balanced is the strategy?

The 95 recommendations are tabulated against 11 objectives (actually the 10 described objectives plus ‘Align with land use and infrastructure integration’).

Here’s a radar chart of the number of ticks that each objective gets across the 95 recommended actions. I’ve grouped the objectives under economic, social and environmental headings:

This is a very poor way of assessing the strategy’s efficacy, but it’s the only way possible from the information given, with its inherent optimism bias.

The strategy doesn’t look very well-balanced (if it was, the radar chart would be much more ‘circular’). Economic objectives (like ‘Lift productivity’) get more ticks than Environmental ones (all of which score 40 ticks or less). The two Social ones, which overlap somewhat anyway, score quite differently (‘Reduce disadvantage’ scores low).

Obviously, some recommendations will do more than others against each objective, but no information is given on this; they simply get a tick against the objectives they contribute to, rather than a rating. Worse than this, there’s no indication of which recommendations score negatively against any of the objectives, only if they score positively (thus getting a tick). Some recommendations are bound to have negative effects, but these are apparently ignored.

It would be much more honest and informative (and a rigorous test of the strategy as a whole) if there was some sort of rating of each recommendation against each objective, on a scale of (say) -5 to +5, instead of just a tick. These ratings would be based on a more detailed appraisal of each recommendation and its consequent effects, both ‘good’ and ‘bad’. The overall performance of the strategy against its own stated objectives would then be much easier to understand.

Before any of that can be done, the objectives must be revisited to make them much more specific, and to address future imperatives properly.

Summary

IV’s draft strategy makes little, if any, coherent sense. Its objectives are vague and meaningless. It proposes many worthwhile reforms, but there are too many loopholes, contradictions and inconsistencies between them. Together, they won’t help us get to net zero emissions by 2050.

We must urgently eliminate emissions, extinctions, inequities, health risks, waste and outdated technologies. We must rein in our collective footprint to match the planet’s capacity to sustain us, and we must do it quickly. The science is in, on all of this.

Every strategy we develop and implement must make meaningful and measurable contributions to these imperatives. A proper strategic approach is needed, not just a motley collection of initiatives born from years of inaction.

We deserve much better than Infrastructure Victoria has given us. I’d like to see a complete re-write.

We also need a firm commitment from the State Government to adopt and implement a properly done strategy (i.e. not this draft) in its cohesive entirety, as a centrepiece to the new direction. This means no more secrecy, left-field projects or pork-barrelling at election time, from anyone. Without this kind of change, there’s no point in going further.

Comments are welcome, as always. If you feel inclined, then write to IV as well, and feel free to use my thoughts! Submissions to IV close on 26 February 2021.

Induced travel: how quickly do people adapt?


INDUCED travel due to changes in the transport network is a well-known phenomenon. For many years, it was ignored in transport modelling, by accounting only for re-routeing of trips rather than mode shifts or origin and/or destination changes.

Nowadays, most modelling includes it, thus producing variable trip matrices between scenarios rather than fixed ones.

A key point is that induced travel makes a project (especially a road project, where travel times are very sensitive to the amount of traffic) a ‘victim of its own success’. A road improvement attracts more traffic than is explained by re-routeing alone. This extra traffic reduces the travel time savings that a project might provide, which in turn reduces its economic benefit.

I’d argue that the only reason induced travel is regarded as a distinct phenomenon is because of the history of transport modelling. Modelling techniques started with simple assignment of a fixed number of trips to a network representing a single mode (mainly road). Thus, it only captured route changes. Multimodal capability was added later, bringing mode shift effects to the mix. Also, by looping back through trip generation and distribution, the effect of a network change on the number and distribution of trips was also estimated.

Most four-step transport models have stopped there; only a few of them attempt to calculate changes to the time of day of trips (to predict peak spreading), and none are capable of modelling land use changes due to a project; this is usually done outside the transport model.

There are a number of land use-transport interaction (LUTI) models around, but they’re very difficult and costly to set up and calibrate. Historically, their usefulness was limited because of this, but now it seems they’re gaining renewed interest.

The pace of change

There seems to be a view, expressed in recent business cases, that it takes a long time for induced traffic effects to materialise because people are thought to be slow to change their origin, destination, travel mode and route choices after a new project opens.

Conveniently, the longer induced traffic takes to grow, the larger the scheme benefits are in the early years of the benefit stream. This is where it matters most, before the discount rate kicks in too much.

Evidence produced to support this view is based on flawed, old studies. One that still gets quoted a lot in Melbourne is ‘Long run economic and land use impacts of major infrastructure projects’ (SGS Economics & Planning, 2012). Commissioned by the Victorian Department of Transport, this study looked at post-opening effects of CityLink, Western Ring Road and the City (rail) Loop. It was mainly concerned with induced land use changes, and its basis and results were highly questionable. For example, its estimates of the effects of these projects on population was completely at odds with Census data. CityLink was supposed to have induced population growth of 21,500 in Prahran and 13,000 in Hawthorn from 2001 to 2011, but ABS says they only grew by 6,500 and 4,700 in total, respectively. Western Ring Road supposedly caused a 2,800 population increase in Broadmeadows from 1996 to 2011, whereas ABS data shows it went down by 2,000! There were many more inconsistencies.

A major flaw in many of the arguments put forward is that they confuse induced traffic arising from the immediate network effect of a new project with that from longer-term changes in land use. Therefore, quoting studies like the (flawed) SGS one in relation to non-land use associated induced traffic is completely incorrect anyway.

In the West Gate Tunnel business case for example, induced traffic effects were phased in gradually over a full 10 years. I reckon this increased the user benefit stream by about 7% (compared to a more reasonable, but still conservative, 3 years’ phasing in). They were behavioural shifts only, represented by comparing a fixed-trip model run with a variable one, using the same land use projections in each. Land use changes can conceivably take that long, but they were excluded. Moreover, the transition between the fixed and variable model results was treated as a straight line instead of some sort of inverse exponential relationship with a much larger shift in earlier years then gradually reducing (this is the usual profile for toll road ramp-ups, for example).

Until then, I’d never seen that particular ‘wrinkle’ used to boost benefits. All the other cost-benefit analyses I’ve done, or reviewed, used variable-matrix runs exclusively, thus allowing for induced traffic effects from day one (or, more correctly, year one). They often included a ramp-up allowance in the first year or so, especially (but not only) for toll road projects.

I’ve already written about the Zenith model and its illogical ‘single loop’ method. This produces an arbitrary trip distribution (by using base year demand on a future year network) and then omits to refine it by purposely leaving it out of the model looping and convergence process. The Zenith model also assumes that public transport has unlimited capacity, and that only toll prices (but no other prices in the modelling) will go down in future years due to increased prosperity.

Given these distortions, the resulting estimates of induced travel, like most of the other model outputs, are highly questionable.

So here we have something that materially affects the benefit stream, with very little basis in fact and calculated using very dodgy modelling. This is of course symptomatic of the ‘optimism bias’ (aka deliberate distortion) which infects every major project appraisal that I’ve examined of late.

How often do people relocate?

Prompted by all these thoughts, I’ve been thinking about other ways of getting a handle on how quickly people might change their travel habits. It occurred to me to try and see whether we’re talking about the same people, which seems to be an underlying assumption in the business cases (except when considering land use change).

This led me to look at the pace at which population changes in an area, as opposed to the net growth.

Population growth comes from three things:

  1. Natural increase (births minus deaths)
  2. Net internal migration (arrivals minus departures)
  3. Net overseas migration (arrivals minus departures)

ABS gives all the component numbers for this (publicly-available at SA2 level throughout the country). The data doesn’t show where they move to and from, only where they appear and disappear. At the time of writing it’s available for three consecutive years (ending June 2017, 2018 and 2019).

Using this data, I’ve calculated two measures:

  • Annual population growth comprises natural increase plus net internal migration plus net overseas migration (arrivals minus departures), divided by year-start population to give a percentage.
  • Annual population change is the number of new people at the end of the year. It comprises births plus internal arrivals plus overseas arrivals, divided by the year-end population to give a percentage. It can be calculated for a single year, but it can’t be easily extrapolated from one year to the next because there’s no way of knowing how many of the new people at the end of a given year are still there at the end of the following year (although one could make some assumptions about that).

Here are some numbers for Greater Melbourne as a whole, averaged over the three years of data:

  • Average annual growth was 2.5%. Of this, natural increase contributed 29%, net internal migration 5% and international migration 66%.
  • Average annual change was 6.1%. Of this, births contributed 20%, internal arrivals 30% and international arrivals 50%.

Looking at the average figures for all the SA2s in Greater Melbourne (of which there are 302), things get more interesting. Average annual growth is the same, but annual change increases to 14%, due to the large numbers of people relocating within Greater Melbourne every year. Of this, births contributed 9%, internal arrivals 70% and international arrivals 22%.

Internal movement turns out to be by far the biggest component of population change in smaller areas (except in places where lots of new housing is being built). It keeps the real estate market afloat and, in Victoria, provides the bulk of State Government stamp duty revenue on property transactions (especially as new dwellings are exempt).

Looking at how these averages are distributed around the city gets even more interesting!

The two maps below show the annual average percentage population growth and change respectively, for the two years 2017-2019, by ABS SA2 area in Greater Melbourne. I’ve rendered them both to the same blue-to-red heat map scale.

Map Description automatically generated
Map Description automatically generated

The first map shows a lot of blue, reflecting that net growth is quite slow in many areas of Melbourne which are already well-established. The urban fringe growth areas to the south-east, north and west are the only red areas. There’s also a ring of middle-level growth (around 5% pa) in inner suburbs around and north of the CBD.

The second map is almost entirely red. In most areas, change is 2-3 times higher than growth. There’s a large population ‘churn’ in inner suburbs around the CBD (probably reflecting larger numbers of rental apartments, and students), and a few distinct pockets in the eastern suburbs (at activity centres), as well as the urban fringe growth areas.

It would be nice to be able to do similar maps for employment and education (the destination areas for many journeys, especially in peak hours), but that data isn’t available to this level of detail. I’m sure it would illustrate similar, high degrees of change.

Evidently, people are frequently moving home. In this process, many are also likely to change where they work, school and shop. In 3 to 4 years, many areas could see more than half their populations change. This is more than enough to keep travel choices changing regularly and reduce widespread, entrenched habits.

If one adds to this the effects of increasing use of GPS-aided travel (with congestion-aware apps) and autonomy of vehicles, it’s likely that travellers will become more and more responsive to changes in transport supply.

The idea that travel behaviour changes (changes to origin, destination, mode and route) take anything more than 1-2 years to materialise after a new project opens seems very unlikely. Land use changes will take longer, and should always be assessed separately, unless you’re using a LUTI model.

COVID effects

All this analysis uses pre-COVID data. It’ll be very interesting to see what happens to the pace of population change, as well as growth, in the wake of the pandemic.

The real estate market has slowed, but not stopped; many have been buying properties sight-unseen, or with online walkthroughs. Also, it looks as though quite a few people are moving out of Melbourne altogether, to move into regional areas or interstate.

As well as a slowdown in population growth and a change in internal relocation patterns, the pandemic has also temporarily halted a lot of travel. As people return to a new ‘COVID-normal’, some shifts in habit will undoubtedly occur. There are doubts that public transport will return to pre-COVID crowding levels (at least, not before a vaccine is available). Some authorities are taking the initiative to encourage more sustainable transport modes and travel choices. Predictions suggest that working from home will become much more widespread.

All of these factors will change previously established relationships between demographics and travel. We’ll need to re-base our transport models to account for this.

The ABS population change data will be very interesting over the next couple of years; I’ll continue to monitor it.

North East Link – the perfect storm


Summary

A PERFECT storm is brewing in Melbourne’s north-east.

The $16 billion North East Link is touted as ‘the biggest road transport project in Victoria’s history’.

It’s also shaping up as the biggest mistake of the Andrews Government’s autocratic approach to transport infrastructure.

Here are just a few of the problems with NEL:

  • It was announced before being scoped or costed in any way
  • The assessment was done by the very organisation set up to deliver it (the NEL Authority)
  • Its business case was based on excessive traffic growth and incorrect cost-benefit analysis (it probably only returns 70 cents per dollar invested, not $1.30).
  • Despite including a busway, it’ll reduce public transport mode share overall
  • Its massive impacts on people and the environment are unacceptable to the local community
  • It’ll involve extensive tunnelling, with unanswered questions about soil disposal
  • Toll road investors have walked away from the project, so the taxpayer will take the financial risk (the scale of which will probably be kept secret).

In the aftermath of COVID-19 the whole thing makes even less sense than it did before.

The Andrews Government must stop, take a deep breath and reconsider their transport priorities. They must re-evaluate the scale of the problem and review the nature of the solutions. They must also get the transport modelling and economics done properly.

It’s time for a massive reality check. While infrastructure investment is important for post-COVID jobs and economic recovery, we must pick better projects than the ones being chosen by our politicians.

Overview

NEL is another West Gate Tunnel in the making, except it’s a far bigger project with even more risk to the taxpayer.

It’s justified by the same erroneous transport modelling and another overstated benefit-cost ratio. All this for a $16 billion (and counting) mega-project that destroys valuable habitat, taking trees, houses and large amounts of land.

The road will be built and run under an ‘availability’ contract; the operator will be paid a pre-agreed fee for keeping it open to traffic, over a predetermined number of years. This is like the Peninsula Link Freeway set-up, but with one big difference; users will have to pay tolls for using NEL. Toll prices will be set – and revenue collected – by the Government.

Toll revenue won’t cover the fees paid by the Government to the operator. Thus, the operator will be shielded from the risk of future traffic variations; the taxpayer will take the risk.

No doubt, like Peninsula Link, the deal will be ‘commercial-in-confidence’ and the Government won’t tell us how much profit the operator will make out of it, nor how much it will cost the taxpayer over the concession period.

Regardless of whether it’ll produce sustainable transport outcomes (it won’t), the sheer scale of NEL and the haste with which it’s been assessed and designed, plus the unknown aftermath of COVID-19 on all the assumptions behind it, mean that it MUST be re-evaluated properly, and re-scoped accordingly, before a contract is signed.

It’s irresponsible in the extreme to do otherwise.

My involvement

As an expert witness during the NEL EES hearings in 2019, I reviewed the transport modelling and strategic planning-related aspects of the project.

My expert witness report raised an initial set of issues. We then had a ‘conclave’ with the transport modelling team. Some issues were clarified, but further digging simply revealed more concerns.

I gave a presentation at the EES hearings to throw more light on this. It then took over two weeks for NELA to lodge a response from the transport modellers, just a day or so before our legal team’s time slot was up.

I wrote a further response to this reply, but I was refused permission to submit it to the EES review panel. The panel concluded, in their final report, that the traffic modelling was ‘probably good enough’ for the purposes of the EES.

I disagree with their assessment. The distortions in the traffic forecasts are substantial and will have a material bearing on its justification, as well as its impacts and effects.

If you’d like to read the paper trail, here are the links one by one (feel free to ask me if you have any queries):

  1. My expert witness report
  2. The traffic conclave report
  3. My presentation slides
  4. A report on the convergence issue in 4-step models
  5. Response from NELA’s transport modellers
  6. My final rebuttal to (5)

The EES terms of reference didn’t allow the Panel to question the fundamental basis for the project. They were mainly concerned with its physical impacts and how to mitigate them. To them, it didn’t matter that there were flaws in the preceding business case. They even felt that the excessive traffic might end up conservatively over-estimating some of the effects.

The business case is supposed to be the document that determines the worth of a project. However, in this twisted world, business cases are done after the political decision is made to proceed.

This approach is enabled by the Major Projects Facilitation Act. Designed to speed up the process, the legislation allows politicians to pick and announce a project, set up an implementation authority and instruct them to prove its worth. The implementing authority has an inherent bias because it’s their job to get their project through the hoops, to politically determined timeframes.

Election platforms have become the place for determining priorities, and the public service is actively prevented from intervening or influencing them. Politicians take free rein in dreaming up what they think is needed, based purely on political gain. This is done without a strategic plan to determine long term targets and objectives.

Anyway, enough ranting; let’s dig into the NEL business case.

NEL traffic modelling

You’d be forgiven for thinking that I’ve got it in for the Zenith model and its creators, Veitch Lister Consulting. Far from it; I think it has the potential to be the best model of its type out there, if only they’d fix it up properly. 

I’d welcome the opportunity to help them.

Like West Gate Tunnel and East West Link previously, the Zenith transport model was used for traffic forecasting. It still had the illogical ‘single loop’, this time with more documentation defending it.

After the EES traffic conclave, a ‘full loop’ model run was provided, so we could see the differences. However, I don’t believe it was done properly; it almost certainly didn’t converge to within acceptable mathematical tolerances. Its results can’t be relied on any more than the ‘single loop’ method.

The table below compares some key model statistics.

Differences between the total number of person and vehicle trips modelled in Greater Melbourne were small, but the differences in vehicle-km and vehicle-hours were larger. The ‘full loop’ method gave 4% less vehicle-km and 9% less vehicle-hours in 2036.

Notice also that NEL results in fewer public transport trips, even though it includes a busway. This is also despite Zenith assuming no limit to future public transport carrying capacity, which shouldn’t be done for project appraisal.

The changes due to NEL were affected more significantly, and also rather strangely. From the numbers, it looks as though the ‘full loop’ produced a 4% larger increase in car trips, a 64% larger reduction in public transport and a 13% smaller reduction in active transport (walk and cycle) trips due to NEL.

It also resulted in an 18% smaller increase in vehicle-km, but a 110% larger decrease in vehicle-hours due to NEL. Behind these overall numbers, there were large differences between the vehicle-km and vehicle-hour changes for cars and commercial vehicles.

These results are greatly concerning. They clearly show that ‘single loop’ produces very different results to ‘full loop’. However, one key problem is that the degree of convergence of BOTH methods is unknown. Without iterating to achieve convergence, there’s no way of knowing how much model ‘noise’ is affecting the results. Given the very strange numbers in the table above, I suspect that it’s having a huge effect.

Proper model convergence is a fundamental requirement of modelling done for projects like NEL, to ensure that the model isn’t producing unwanted side-effects.

If this ‘full loop’ result was used for cost-benefit calculation, I think it’d produce completely unbelievable results.

Here are some traffic flow results from the two model runs.

‘Full loop’ did indeed produce less traffic than ‘single-loop’. It also attracted 7% less traffic to NEL in 2036, which meant less traffic relief on surrounding roads (as shown above).

‘Single loop’ probably overestimates 2036 traffic by as much as 10%, but it’s impossible to know for sure because the ‘full loop’ run was also wrong.

Toll discounting

As on West Gate Tunnel and East West Link before it, the modelling assumed that future toll prices would go down in real terms, because of increased future prosperity. However, tolls were the only price that was lowered for this reason, in the modelling. If we’re all going to be more prosperous, all prices ought to be adjusted, not just the tolls. All this assumption does is make the toll road more attractive compared to everything else.

I expect that the incorrect reduction in future toll prices would probably attract 10% more traffic to NEL.

Including the general over-estimating of ‘single loop’, this probably means that NEL traffic in 2036 is overestimated by about 20%.

These distortions are solely introduced by illogical and erroneous modelling methods and assumptions.

Cost benefit analysis

The cost-benefit analysis in the business case gave NEL a benefit-cost ratio of 1.3 (1.4 including wider economic benefits). 

Percentage-wise, the wider economic benefits were smaller than the Western Distributor’s, which seems reasonable; the Western Distributor serves central Melbourne much more directly than NEL. Wider economic benefits are most noticeable in big activity clusters like CBDs.

The economic results in the NEL business case are summarised below.

Some of the economic benefits flow from changing things in the economic calculations without also changing them in the transport modelling. This is procedurally wrong. The economic benefits are intended to reflect perceived or behavioural prices. If something is changed which would affect behaviour, it should’ve been changed in the traffic modelling too.

A key example is the perceived extra cost of congestion. If this was a real phenomenon, and it was somehow added into the transport model, it would probably divert traffic away from congested links, increase public and active transport mode shares, and/or shift journeys from peak to off-peak time periods. However, in this case, as I’ve already said in Tunnel Tales 3, theoretically the transport model has already been calibrated to take full account of perceived values of time and costs associated with travel, congested or not.

The travel time reliability is another item in which additional benefit is claimed from users’ perception of increased network reliability (reproducible travel times from one day to the next) due to the project. Again, the concept of reliability ought to be already reflected in the transport model calibration; it would be influencing travel choices every day.

The benefit stream in the cost-benefit analysis was inflated to account for increased values of travel time in future years due to increased prosperity). However, this wasn’t included in the transport model (in which future values of time were constant in real terms, as per Government guidance of the day). If it had been modelled, it would’ve changed future traffic patterns, and given a different result for NEL. It’s incorrect to escalate a behavioural value in the cost-benefit analysis without also doing it in the modelling.

Finally, since the business case was published, the capital cost of NEL has increased. The business case used a cost of 12.2 billion (2017 dollars). The latest State budget puts it at $15.6 billion, which is 8% more in real terms (after allowing for escalating construction prices in the interim).

The table below summarises my estimate of the impact of these issues on the benefit-cost ratio.

The outcome is to reduce the benefit-cost ratio from 1.3 to 0.7, or from 1.4 to 0.8 including wider economic benefits.

Conclusions

While this is the best I can do at the moment (without getting the traffic modelling re-done!), I think it shows that:

  • It’s imperative that the economic appraisal and the traffic modelling both contain the same pricing assumptions throughout
  • The traffic modelling needs to be done properly (i.e. without ‘single loop’, and with clear evidence that the demand model converges sufficiently, in all modelled years and time periods).

Changes since the business case, particularly COVID-19, mean that NEL shouldn’t proceed until it’s been re-assessed properly, including taking full account of the pandemic’s possible effect on future growth and travel patterns.

If this changes the scope, timing and overall need, so be it. The present project, like so many others, has been concocted from a political standpoint rather than a demonstrated need.

North East Link is the biggest road project in Victoria. Despite that, its justification was unduly rushed and highly questionable. Its construction will probably be rushed as well. It threatens to cost us many times more even than the West Gate Tunnel will.

Tunnel Tales (3 of 3) – Rubbery figures: why the West Gate Tunnel doesn’t stack up


THE sorry tale of my West Gate Tunnel experiences is a long one. I’ve broken it up into three parts:

  1. The overall story
  2. The transport modelling
  3. The cost-benefit analysis (this post)

In this final part, I’ll cover the heavily distorted cost-benefit analysis and the exaggerated economic value that resulted.

Overview

When my concerns about the Western Distributor (WD) business case were reported in 2017, then-Roads Minister Luke Donellan said in response: “it stacks up, and we’re getting on with it”.

If by ‘stacks up’ he was referring to a positive economic return on investment, he was sorely mistaken. The assertion was based on a flawed and biased business case, produced by the same people who did one for East West Link under the previous government.

The Andrews Government signed up to a project that doesn’t stack up at all, either economically or financially – even without the COVID effect.

The WD business case lumped Transurban’s West Gate Tunnel proposal with other initiatives, which made it look better than it is. It was based on flawed traffic forecasting, and the cost-benefit analysis had some additional distortions. It also misled the Federal Government by misrepresenting Infrastructure Australia’s requirements for economic appraisal.

Including cost increases so far, the economic return of the West Gate Tunnel to Victoria is more like 60-70 cents per dollar invested, not the $1.30-$1.60 stated in the WD business case.

Toll revenue from the West Gate Tunnel would probably only cover about 40% of the capital and running costs. Transurban’s been given additional toll revenue from CityLink, and $2.6 billion of public funding, to make it financially viable. In the appraisal, this was even regarded as a ‘strategic point of difference’ in the project’s favour: that Transurban was the best entity to do the project, because they could leverage CityLink tolls to help pay for it!

Of course, all this would matter a lot more if investment decisions were actually made taking the economic and financial returns into account. The Andrews Government had already decided – even before being elected – that they were going to back Transurban’s proposal. The business case was deliberately distorted to support that decision.

Cost-benefit analysis

The accepted measure of a project’s economic worth is the benefit-cost ratio. 

The benefit-cost ratio is essentially the present value of a stream of economic benefits, compared with that of the construction and running cost, over a given period. If it’s more than one (i.e. the benefit stream is worth more than the cost stream) then it’s said to be economically viable. This means that its benefits to the economy are worth more than its costs.

The cost-benefit analysis only includes effects that are ‘monetisable’. This means that they can be translated into something that reflects their impact on the economy, by expressing them in dollar terms.

When cost-benefit analysis was first used in transport, it really only included construction and maintenance costs weighed up against users’ travel time, accident and operating cost benefits. Since then, further research and increased complexity of analysis have led to many other elements coming in. Cost-benefit analyses for transport projects now routinely include allowance for economic effects of transport emissions, induced traffic, network resilience and ‘wider economic benefits’ (which cover effects on jobs and productivity).

The science is still out on wider economic benefits. Because of this, cost-benefit ratios are usually expressed both with and without them. However, recently several other elements have crept which are even more questionable.

Western Distributor business case

The cost-benefit analysis in the WD business case, and the many pages describing the assumptions and methods behind the numbers, shows how complex such calculations can be. Unfortunately, this also makes them pretty opaque to non-specialists (including the end clients for the work, especially the politicians who use the results to ‘sell’ their decisions).

I had the job of reviewing the cost-benefit analysis while it was being prepared, when I was working at the Victorian Department of Economic Development, Jobs, Transport and Resources (DEDJTR) under a contract arrangement.

I was immediately concerned with the quality of the work being done, and I raised several queries about it. My concerns were ignored, and I was taken off the project. 

The cost-benefit analysis, and the business case document that surrounded it, were full of so-called ‘optimism bias’. This is a polite term for what I consider to be deliberate distortion to produce a favourable case. It can’t be regarded as an impartial assessment.

The cost-benefit calculations

The centrepiece of the WD business case, the cost-benefit analysis, showed that “The Project” had a benefit-cost ratio of 1.3 to 1.6 (the latter including wider economic benefits) when supposedly done to Victorian guidance. This range became 1.9 to 2.2 when calculated using a misleading interpretation of Infrastructure Australia guidance, as it stood in December 2015.

The table below lists the items making up the costs and benefits to reach this result. You can find the original in the 2015 business case document, reachable via the West Gate Tunnel project page (open the document library under the ‘Library’ menu and keep going down the list until you find it). If you can’t find it, let me know.

Problems with the calculations

There are quite a few problems with the calculations. The biggest ones (not in order of significance) were:

  1. Misleading representation of Infrastructure Australia guidance: IA guidance was purported to exclude induced traffic effects, which was incorrect. Furthermore, the Victorian and IA numbers were given without any commentary on which might have been closer to the truth; this was unhelpful and misleading.
  2. Extended project scope: The package of initiatives assessed did not limit themselves to Transurban’s proposal; they included other projects, including the Monash Freeway upgrade east of Warrigal Road.
  3. Incorrect treatment of induced traffic: The user benefit stream phased in the effects of induced traffic over a full 10 years from opening, thus keeping early-year benefits higher than they should have been.
  4. Distortions caused by Zenith modelling: Zenith modelling probably overestimates traffic on the West Gate Tunnel by 20-25% (see my previous post).
  5. Overstated user benefits from congestion relief: Spurious benefits have been added for car users (‘Perceived cost of congested travel time’ in the table above).
  6. Skewed analysis of network resilience: Calculations that exaggerated the reduced reliance on West Gate Bridge.
  7. Cost increases since the business case: According to the latest State Budget papers, the cost of the project is now 31% higher than it was in the business case.

I’ll step through these one by one, with my reckoning of the cumulative effect on the benefit-cost ratio.

Misleading representation of Infrastructure Australia guidance

The business case gave two sets of economic figures. One was supposed to be concordant with Victorian guidance, and the other with IA guidance at the time.

The differences were down to three factors. It was claimed that IA guidance:

  • called for a 30-year appraisal period, while Victoria’s said 50 years;
  • had a different method of calculating residual project value at the end of the appraisal period; and
  • did not need allowance for induced traffic (a well-known and widely accepted phenomenon).

Despite using a 30-year appraisal period for IA, the residual value at the end of 30 years was equated to the benefit stream from the next 20 years after that. This was stated to be the recommended method of calculating residual value in IA guidance. Therefore, the IA appraisal appears to have used 50 years’ worth of benefits in any case.

This left the treatment of induced traffic as the only real difference between the two sets of numbers.

I reviewed IA’s guidance at the time over the induced traffic issue. Whilst there was no specific mention of it in the main text, an appendix clearly said that it should be included (or reasons given why not). Furthermore, there was a draft revision of the guidance available at the time (adopted soon afterwards) that clearly said, up-front, that induced traffic must be accounted for.

The reason given for using IA figures was to enable comparison with other supposedly IA-compatible project benefit-cost ratios. No attempt was made to ascertain whether all other projects presented to IA had excluded induced traffic.

Furthermore, the report didn’t comment on whether one set of guidance was better than the other. No professional opinion was given either way. Given the wide range of the results, how does this help the decision makers?

Induced traffic is a well-known phenomenon. No benefit-cost ratio should be without it these days, unless the project is so small or unusual that it doesn’t generate additional travel and mode shift. This is certainly not the case for a large road project. 

As the numbers themselves illustrate, ignoring induced traffic increases the benefits by about 63%.

We can safely ignore the ‘IA guidance-compatible’ figures as pure fiction. I only hope that IA did the same.

Extended project scope

The WD business case assessed a package of initiatives bundled together:

  1. West Gate Freeway widening
  2. The new Western Distributor
  3. Cycling and pedestrian facilities
  4. Webb Dock access road improvements
  5. Monash Freeway upgrade

Transurban’s ‘market-led proposal’ comprised items 1 and 2 only. The others were included as necessary accompanying works, which is reasonable for items 3 and 4. However, item 5 (the 40 km-long Monash Freeway upgrade) is 20-60km away from the West Gate Bridge, in the eastern suburbs. The influence and interrelationship between the two initiatives is very small, and the justification for including it in the business case is tenuous.

Further down in the business case, a series of sensitivity tests indicate that the BCR of the Monash upgrade is 4.2, while that of the remaining initiatives together is 1.1. I estimate, for this to be true, the cost of the Monash upgrade would’ve been $220 ± 55 million in present value, equating to $350 ± 90 million in nominal dollars (the spread in these numbers allows for a range to the rounded-off benefit-cost ratios in the business case).

Effect on the benefit-cost ratio: reduced from 1.3 to 1.1.

Dubious treatment of induced traffic

The phenomenon of induced traffic is well known. When a new road is opened, it not only attracts re-routed traffic, but also gives rise to new trips.

In 2011, the Victorian Auditor-General provided a useful definition of six different ways that a road project can attract traffic, reproduced below. I’ve added which of these are included in the traffic modelling for West Gate Tunnel.

Like most four-step models, Zenith as used on the West Gate Tunnel doesn’t calculate either changes to time of travel (at least, not those who move from one modelled time period to another) or changes in land use patterns resulting from a project. It does cover the other types of induced traffic, but with distortions because of its illogical methodology.

Induced traffic is generally quite a small component of total traffic, but it has a large effect on the benefit stream because of the added congestion it can cause.

The business case explains how induced traffic was accounted for in the cost-benefit analysis. It was done by using a ‘fixed matrix’ (i.e. no induced traffic) approach in year 1 and a ‘variable matrix’ (i.e. including induced traffic) approach in year 10. In years 2-9, the benefit stream was calculated by interpolating between the two.

A graph was provided in an early draft of the business case to illustrate this approach (see below). Note that the blending between ‘fixed’ and ‘variable’ outputs means that the benefits in early years are much higher than if ‘variable outputs were used from the outset.

I think that 10 years is highly excessive for this blending assumption. The Zenith modelling doesn’t take account of the land use effect; the other, included forms of induced traffic generally take much less than 10 years to materialise. 

I consider 3 years is more realistic (I’ll devote another post to this soon), and I’ve been able to simulate the effect that this would have on the benefit stream, as shown below. It reduces the present value of the benefits by 7%.

Effect on the benefit-cost ratio – reduced from 1.1 to 1.02.

Distortions caused by Zenith modelling

As explained in my previous post, the traffic modelling for the project contains significant distortions. I reckon that these result in a 20-25% overestimate of traffic using the West Gate Tunnel. The effect of this can be looked at in two ways, either by assuming slower traffic growth (and hence lower benefits) over the life of the project, or by factoring the benefits down by a suitable percentage.

In the absence of properly-done traffic modelling, I’ve chosen to reduce the benefits by 15%. This probably equates to about 5-6 years of traffic growth.

Effect on the benefit-cost ratio – reduced from 1.02 to 0.87.

Overstated user benefits

There was a strange item on the benefit list – the so-called ‘perceived cost of congested travel time’.

This significant item was 12% of the benefit stream. It has since been included in Victorian economic appraisal guidance, but I seriously question its validity, and the method of calculating it.

The WD business case contains a lot of discussion about it, essentially arguing that drivers will place a higher value on their travel time when they’re in congested traffic, and that this extra value is proportionally added to the modelled travel time savings when a road reaches a volume-capacity ratio of 70%, reaching its maximum when the volume-capacity ratio is at 100%.

I have several problems with this method.

  • The choice of the additional perceived value ($5.30, or about 30% of the underlying value of travel time) is based on surveys done in Sydney in the early 2000s. Arguably, congestion was less widespread then and may have been less ‘accepted’ than it is today, yet no attempt was made to verify the value that might apply in Melbourne conditions in 2015.
  • The assumption that it kicks in when traffic reaches 70% of the road capacity and increases linearly from there to 100% is extremely arbitrary. Like most of its type, Zenith doesn’t explicitly model road intersections (the source of virtually all traffic congestion in urban areas); the theoretical capacity of the road network is simulated by using speed-flow curves. There’s no proof that congestion (of the type that supposedly increases perceived costs) starts at a volume-capacity ratio of 70%.
  • Most important of all, if this was indeed a real perception, then it ought to have been included in the traffic modelling as well. The reality is that the traffic model is supposed to be calibrated to reflect observed behaviour, so the user costs that come out of it already take full account of the perceived value of times and costs associated with travel, congested or not.

It’s completely illogical to add this extra item of benefit into an economic appraisal derived from a behaviour-based transport model like Zenith.

I therefore think this item should be removed from the benefit stream, at least until it can be categorically proven to be outside the benefits already implicit in the traffic model, and to exactly what degrees of congestion it applies, in Melbourne.

Effect on the benefit-cost ratio – reduced from 0.87 to 0.76.

Skewed analysis of network resilience

The assessment of network resilience is based on estimating how often the West Gate Bridge suffers a typical, significant traffic-delaying incident, and the likely disbenefit associated with such an event.

The busines case has some extreme hyperbole about this. Apparently, a major incident on the West Gate Bridge “can … generate congestion across the entire road transport network” and “… would be calamitous for the economy”. I think we can all agree that these are serious exaggerations.

This aside, the analysis implicitly assumes that the West Gate Tunnel:

  • will substantially reduce the effects of such incidents; and
  • will be free of any such incidents itself.

These are both clearly untrue. I’ve been a little generous, though, and only reduced this benefit by 50% instead of removing it altogether.

Effect on the benefit-cost ratio – reduced from 0.76 to 0.73.

Cost increases since the business case

Since the business case in 2015, the cost of the initiatives has increased significantly. 

In the WD business case, the total nominal-dollar cost of the project was $5,226 million.

In the 2019/20 State Budget, the costs total $6,646m, broken down as follows:

  • West Gate Tunnel: $6,302m
  • Webb Dock access: $61m
  • Monash Freeway upgrade: $283m (I’ve excluded Stage 2 of this, which costs another $684 million in the budget but has since blown out even more).

Excluding the Monash Freeway upgrade, this is a 31% increase in project costs.

Effect on the benefit-cost ratio – reduced from 0.73 to 0.56.

Assuming that wider economic benefits remain the same proportion of total benefits for the project, the benefit-cost ratio including WEBs would be 0.72.

Conclusion

The benefit-cost ratio in the WD business case suggested the project would return $1.30-$1.60 to the Victorian economy per dollar of cost.

Removing the unwarranted inclusion of the Monash Freeway upgrade, the biggest distortions and optimism bias would reduce this to $0.60-$0.70 for the Transurban project elements (West Gate Tunnel and West Gate Freeway widening).

There’s no doubt in my mind that the West Gate Tunnel will not return a positive effect on the Victorian economy.

What’s more, it doesn’t even stack up financially; Transurban needed State funding as well as ramped up and extended tolls on CityLink to pay for it.

We’ve been sold a pup.

Tunnel Tales (2 of 3) – transport modelling and the ‘single loop’


THE tale of my West Gate Tunnel experience is a long one! I’ve broken it up into three parts:

  1. The overall story
  2. The transport modelling (this post)
  3. The cost-benefit analysis

This second post is wordy and somewhat technical. Hopefully, my overview gives you the salient points before diving into more detail. The opinions herein are my own; I encourage you to investigate further for yourselves if you’re interested.

Let me know what you think. I’d be particularly pleased to hear from experienced transport modellers (including Veitch Lister Consulting!) about these issues.

I have some loaded questions for four-step transport modellers reading this post:

  • Have you ever checked the convergence of your demand models (in all modelled years and time periods), to see if they meet ATAP requirements (the 0.1% difference in the ‘demand/supply gap’), and to ensure that changes between model loops were less than 10% of the changes due to the project you were assessing?
  • If so, what were your results, and what did you do, if anything, to improve them?
  • Did you do this through your own initiative, or at the request of a reviewer?
  • If you did this and your model did not meet these requirements, did you make that clear to your client, or to those using your results for further analysis?
  • Have you ever felt pressured by others for your modelling to produce ‘favourable’ results (perhaps especially when working for toll road bidding consortia)?
  • Do you feel resentful or defensive when your hard work is questioned by others?

Overview

Transport modelling for major projects like West Gate Tunnel is a complex process. Good practice has developed over many years, and growth in computing power has made larger models possible.

Here in Victoria, the two most-used strategic, multimodal transport models are:

  1. The Victorian Integrated Transport Model (VITM), developed by the State Government; and
  2. The ‘Zenith’ model of Victoria, developed by Veitch Lister Consulting (VLC).

VITM is used more for public transport and smaller road projects, while Zenith is used mainly for toll roads, because it’s thought to have better toll diversion modelling. Zenith was used for East West Link, West Gate Tunnel and North East Link.

Zenith uses an unusual and illogical method for future year forecasting, called ‘single loop’ distribution. Essentially it calculates trip distribution (more on that below) by combining today’s travel ‘demand’ with future transport ‘supply’. It also does this only once in the iterative process that the overall modelling method requires, instead of looping through for successively better results.

This means that:

  • Zenith significantly overestimates future car-kilometres of travel;
  • the model procedure doesn’t comply with established guidance; 
  • its mathematical stability and reliability can’t be proven, nor even calculated properly; and
  • its outputs therefore shouldn’t be relied upon for cost-benefit analysis.

The single loop method assumes that people make their trip origin-destination decisions as if the transport network was more developed than it is. Conversely, you could say that trip origin-destination decisions are made as if the travel demand on the network had stopped growing sometime in the past.

For most projects, the main forecasting year is 20-25 years into the future, so this effectively means that trip distribution is calculated using the travel demand from 20-25 years earlier. In fast-growing Melbourne, the effect of adding about 2 million more people is therefore ignored during that step of the modelling.

A lot of documentation has been produced in business case and EES paperwork to defend this method, but I still disagree with it. The copious explanations don’t stand up to close scrutiny.

A second problem with Zenith (as used on the East West Link, West Gate Tunnel and North East Link) is that future-year toll prices are lowered, based on an assumption that real wages growth over the last 15 years or so will continue unabated into the future, making prices more affordable. Not only is this a very optimistic projection, it also means that, in the model, toll road users are the only ones who’ll benefit from this, when in reality everyone will. Why not reduce all the other costs (fuel, fares, values of time, etc, etc) in the modelling by the same amount?

In my estimation, these two issues alone mean that forecast traffic on the West Gate Tunnel in 2031 could’ve been overestimated by something like 20-25%, and we can’t assess the reliability of these numbers anyway (due to the convergence issue). The only way of knowing for sure is to re-do the modelling properly.

This error flows through into the economic appraisal, which I’ll cover in part 3.

As it stands, Zenith modelling shouldn’t be relied on for project economics, nor EES impact appraisal. VITM is probably better (it uses the four steps properly, at least), but both models need improvement before being used further.

Most clients, especially Governments, don’t really care about these niceties (in fact most of them won’t even understand them!), especially when the modelling gives them a ‘good result’ for their projects. Unless they do independent, concurrent peer reviews of the work, they have to trust the modellers.

The onus is on transport modellers to do their work properly, and to comply fully with established guidance. We need to lift the game.

What is a ‘transport model’?

In the context of projects like West Gate Tunnel, a transport model is a computerised method of forecasting movements of people and vehicles.

Essentially, a transport model predicts the travel patterns associated with given land uses (demand) and transport provisions (supply) in an area. It enables testing of future changes in both.

The results from such models are used in strategic studies (area-wide), project planning (including business cases) and impact studies (like EESs). They are the main basis for predicting the economic, social and environmental effects of initiatives. Thus, they are central to the selection, justification and environmental appraisal of transport projects.

The Australian Transport and Planning Assessment (ATAP) Guidelines (especially under ‘Tools and Techniques’) contains a detailed description of the minimum requirements for developing and using such models. They result from decades of use and development of transport models worldwide. They have a lot in common with the UK’s Transport Analysis Guidance (TAG).

The four computational steps in a typical transport model are:

  1. Trip generation – estimating how many trip-ends will be produced and attracted by all the different land uses in an area;
  2. Trip distribution – joining up the trip-ends to create a pattern of trips to and from all the different areas;
  3. Mode split – predicting which modes of transport trip-makers will choose for each trip; and
  4. Trip assignment – allocating the trips to the area’s transport network so that the demand can be totalled for each ‘link’ in the network.

Each step is calculated using mathematical algorithms which are designed to echo observed travel behaviour. These algorithms use a range of input variables, including travel times/costs, population and job numbers.

Calculation using a four-step model requires iteration, or looping, through the four steps. Each loop uses the results from the previous loop to refine the travel times and costs. A correctly set up model will ‘converge’ mathematically, until the differences between the results from one loop to the next are within specified small tolerances. This is a fundamental requirement of the process. A model which doesn’t converge sufficiently is unreliable and shouldn’t be used for project appraisal.

Click image for full-size version

Before the model can be used to predict future travel, it must be very carefully and thoroughly calibrated to observed conditions, using data collected through physical surveys. These include travel times, traffic counts, public transport patronage figures and household travel surveys. The model’s algorithms have coefficients and equation forms that are carefully adjusted until each of the four steps – and the model as a whole – is as accurate as possible.

It takes a lot of time and money to set up, calibrate and validate a four-step model. In Australia, most jurisdictions have their own, in-house models developed progressively over many years. Some use models developed by private companies, notably the ‘Zenith’ models of Veitch Lister Consulting (VLC).

The Victorian Government has its own in-house model, called the Victorian Integrated Transport Model (VITM), but it also uses VLC’s Zenith model of Victoria. In recent years, the tendency has been to use Zenith for major road projects – especially toll roads – and VITM for public transport projects and smaller road projects. The main reason for this is that VITM doesn’t have a well-developed toll road forecasting capability.

The ‘single loop’

Using Zenith over VITM for toll road modelling means that several unusual quirks of Zenith get incorporated into the process. The biggest of these is the use of an illogical short cut when modelling future years, which I’ll call the ‘single loop’ distribution.

The diagram below shows the difference between the normal four-step process and Zenith’s single loop.

Click image for full-size version

The single loop method combines base year travel times and costs with future year transport network conditions when calculating the trip distribution. It also does this only once. Thus, the future year model only loops through the mode split and assignment steps in subsequent iterations. This conflicts with established practice. It also means that the convergence of the four-step modelling process can’t be calculated, so there is no way of knowing how reliable the future year model runs are.

The single loop method is illogical because it is an artificial situation created by calculating trip distribution on a future-year network by using base year travel times and costs. In a typical modelling exercise, the future years can be 20-30 years ahead of the base year. Why spend all that time and effort to calibrate and validate the model against base year survey data, then use it in a completely different way in future years? 

I was told more recently that this has been a feature of Zenith for a long time. The Zenith model has a very large number of zones, which makes its computation times very long; this would’ve been a much bigger factor with the computers of 25 years ago when Zenith was first set up. Was it originally a time-saving measure?

The convergence issue is important for many reasons:

  • ATAP guidance states that the demand model’s convergence should be measured using a specific mathematical test (the ‘demand/supply gap’) which cannot be calculated unless the entire model is looped properly. The results of this test should be less than 0.1% between the final two iterations.
  • ATAP also says that the change in network user costs due to a project should be tested using the last two iterations (or by running an additional iteration) to gauge the possible margin of uncertainty in the project benefits, and also to see how much of this difference is due to model ‘noise’. This makes a lot of sense; if the model is too large geographically compared with the project being tested, small but multitudinous differences can crop up in parts of the model which are nothing to do with the project itself. This is rather like background noise in the model which can drown out the effect of the project itself. 
  • UK’s TAG goes a little further; it says that the difference in user costs between the last two model iterations must be less than 10% of the change in user costs due to the project being evaluated.

To meet these requirements, these convergence measures should be calculated and reported for each time period the model uses, and for each year the model is run for. They are the acid test of a model’s reliability and validity for project evaluation, yet I know of no-one in Australia who provides such information properly.

In the UK, where concurrent peer reviews are standard practice (and where much of the Australian guidance is copied from), such reporting is mandatory and commonplace. Without it there is no way of knowing whether a model’s results are reliable, nor whether background noise is drowning out the project’s effects. This is fundamental; without it the entire modelling process is open to serious question.

Aside from all this, there’s another major problem with the single loop method. It produces a lot more – and longer – car trips in future years than the correct way of doing it. During the West Gate Tunnel work we were able to compare Zenith with VITM in this regard (VITM loops through the distribution step properly). The models’ results are not too different in the base year (2011), but Zenith produced much more growth than VITM twenty years into the future (2031). When Zenith was run properly (i.e. not single loop) its results were much more consistent with VITM’s.

Click image for full-size version

Note that Zenith covers a larger geographical area than VITM (it includes regional areas around the city), so the last two measures above (km per trip and per capita) are the most relevant for comparing the models.
However, I also question the fact that Zenith has 42% higher car trip lengths than VITM in 2011; this would mean that the average trip length in the regional areas would be nearly 50km, which is definitely excessive, given the vast majority of regional area trips take place within towns and small cities.

The single loop method produced 10% more traffic (car trip-km) than the normal method in 2031. Average car trip lengths increased by 8%, and car trip-km per capita by 10%. This would make the 2031 road network more congested than it ought to have been. Therefore, more traffic would use the West Gate Tunnel.

Notice that, in VITM, car trip-km per capita went down by 9% between 2011 and 2031. In single-loop Zenith, it went up by 3%. 

At the time (i.e. in 2015), VLC argued that car trip-km per capita would increase in the future. This was mainly attributed to Melbourne’s continued urban sprawl. 

A few other reasons were given as well, in this report. During the EES, I wrote some notes on the modelling and related issues for an objector (they were a work-in-progress, and raise many questions which remain unanswered).

This trip-km growth is counter to recent trends. In 2015, BITRE Yearbook data showed that Melbourne’s annual car-km per capita had peaked in 2004 and had decreased steadily since then. The latest BITRE Yearbook, released in December 2019 (just after North East Link’s EES, and before COVID-19), shows that it has continued to decrease. Furthermore, other cities show similar trends.

Click image for full size version

The graph above shows the BITRE data, and I’ve shown where the trends in the VITM and Zenith modelling would put Melbourne in 2031. Which one do you think is more on-trend?

I wonder how much longer VLC will argue that their single loop modelling is right, while data continues to show that it’s wrong?

As I said already, I was told that the single loop method has been a feature of Zenith for a very long time. Was it originally adopted as a way to shorten computing times, especially in future year runs (of which there can be many, for a given project)?

Deeply buried in the East West Link business case documents (page 1964 of 2022 in the ‘September update’ of 2013), VLC shows that single loop distribution was used on that project, in 2012. No reasons were given for it, nor was it queried by the peer reviewers at the time.

The fact remains that, whether you believe trips will start lengthening again or not (I don’t believe they will), the single loop method doesn’t comply with Australian guidance. It’s a distortion of proper four-step modelling. By definition, a single loop model can’t converge, and its results can’t be relied on for project appraisal. I’m sure that experienced transport modellers would agree with me on that (and I’ve asked a few).

Discounted toll charges

There are other questionable settings in the Zenith model. One particularly problematic one for me is the fact that tollway charges were reduced in future years. This was done in the East West Link, West Gate Tunnel and North East Link models. 

The stated reason for this was that future effective earnings (wages growth ÷ prices growth) were expected to escalate by 1.8%pa over the 20 years from 2011 to 2031, using the same annual growth rate observed from 1995 to 2011. However, the only setting in the model that was changed to reflect this was the toll charges, which were reduced by about 25% for cars and 30% for commercial vehicles in 2031.

This incorrect distortion artificially favours toll road users over everyone else. If earnings are going to increase, everyone will benefit, not just toll road users. All costs in the model, not just the toll charges, should therefore be adjusted to allow for it.

Apart from all that, it’s highly questionable whether the next 20 years will be as prosperous as the last 15, even if we hadn’t had a pandemic. When VLC did the same calculation a few years later for North East Link, the result was to reduce tolls by only 1.55%pa instead of 1.8%pa, because the economy had slowed between 2011 and 2015.

Documents show that this toll charge reduction increased East West Link traffic by 15%. I haven’t found the same information for West Gate Tunnel (the corresponding document hasn’t been released, as far as I can see), but I’d expect a similar result.

Conclusion

The traffic forecasts for West Gate Tunnel were distorted by two main factors:

  1. An illogical and non-standard modelling method that produced about 10% more traffic in the model in 2031, compared with the proper way of doing it.
  2. An incorrect 25-30% discounting of toll charges in 2031 modelling, making toll roads more attractive than they ought to have been.

I estimate that these issues could have resulted in an 20-25% overestimate of West Gate Tunnel traffic in 2031.

Furthermore, there is no confidence that the modelling is mathematically stable. Because it doesn’t follow normal practice, the key indicator of sufficiently tight model convergence can’t even be calculated.

The over-optimistic traffic forecasts for West Gate Tunnel in turn gave much higher economic benefits for users of the project, thus inflating the benefit-cost ratio.

Some might say that this isn’t important because the future is so unpredictable. I disagree, because these problems are produced through modelling errors. Appraisal tools must be as accurate and error-free as possible so that we can rely on their results, and their stability. Future variations in behaviour, the economy, technology and other things can then be tested by changing the input assumptions.

This brings me to the economic appraisal, in which there are many other distortions apart from those produced by the traffic modelling. I’ll cover these in the next post, in which I’ll also attempt to ‘reverse-engineer’ the West Gate Tunnel’s benefit-cost ratio.

Stay tuned!

Tunnel Tales (1 of 3)


THE tale of my West Gate Tunnel experiences is a long one. I’ve broken it up into three parts:

  1. The overall story (this post)
  2. The transport modelling
  3. The cost-benefit analysis

Overview

I was closely involved in the Government’s process of assessing Transurban’s West Gate Tunnel proposal. I was retained to review the transport modelling and economic appraisal, and quickly found major problems with the work being done. Frustrated with resistance from the appraisal team, I eventually alerted the Treasurer to the problems. This was a bad move on my part; I was promptly removed from the work.

I did get a chance to tell my story a little while later, but it was too late. The project was approved, and it progressed through to construction, where many new problems have arisen.

West Gate Tunnel should never have been approved by the Victorian Government. Like East West Link before it (and North East Link after), the business case was distorted to make the project look much better than it is, through erroneous traffic forecasting and cost-benefit analysis.

To cap it all, Transurban’s CityLink concession will be extended as part of the deal, on top of extra escalations of CityLink tolls AND another $2.6 billion of public funds. Obviously, the project doesn’t even stack up on its own financially.

I predict that, if and when it eventually opens, the forecast conditions that it depends on won’t materialise. It won’t deliver the claimed benefits and it’ll stand as an ugly monument to our broken planning system.

Similar problems exist with North East Link. I’ll cover that in another post soon. In the meantime, I implore the Andrews Government to avoid what could be an even bigger mistake, and re-evaluate North East Link properly. COVID-19 provides the circuit-breaker for an urgent and honest re-think of the priorities.

Independent, concurrent peer reviews should be mandatory. They must be ‘open book’, and show how the reviewers’ requirements have been satisfied. A final report must be released documenting the process, listing outstanding problems and describing the fitness for purpose of the work.

Such reviews cannot be done after the work is complete (which is what happened later, on North East Link).

Beginnings

Back before the 2014 election, which hinged in part on Labor’s promise to scrap East West Link, I was invited by Tim Pallas, along with a few others (far more distinguished than I!), to join a planning think tank. We discussed the need for evidence-based, sustainable land use and transport planning, trying to help Labor’s election platform.

During this process we talked about the East West Link cancellation and I remember being told: “Don’t worry – we have another idea that’ll take shape after the election”.

In November 2014, Labor won the election and East West Link was quickly cancelled.

Peer review

Early in 2015 I got an assignment under contract to the Department of Transport (when it was DEDJTR – Economic Development, Jobs, Transport and Resources), reviewing business cases for transport projects.

A flurry started in April 2015, when Transurban’s proposal to build West Gate Tunnel (then called Western Distributor) was revealed, and the Government put the wheels in motion to assess this ‘market-led proposal’.

An assessment team was set up within DEDJTR from the remains of the Linking Melbourne Authority (LMA had been implementing East West Link after taking it through a business case and environment effects appraisal under Denis Napthine’s government). Consultants were brought in to do transport modelling and the economic appraisal (the same ones who’d done that work for East West Link). I don’t know whether this consulting work was competitively tendered for.

I was allocated to review the economic appraisal, and an external peer reviewer was appointed for the transport modelling. Because of the many overlaps between the two, we worked together very closely.

Meetings started; the assessment team seemed very keen to get the project through. I heard someone say words to the effect of: “we didn’t get East West Link; we’re determined to get this one”. Hardly an impartial appraisal of a private company’s proposal.

The business case work was fast-tracked, with insufficient time and meetings for us to review progress as it happened. Nonetheless we immediately found some real problems with the work. Traffic modelling was producing much higher growth than we expected, and economic benefits were being overstated.

When we raised our concerns, we were told that they’d be looked into, but no promises were made because of the tight timeframe. We were even told that the issues might not get addressed after the business case was completed. This seemed ridiculous to me.

Finding it impossible to do our job under these circumstances, we documented the problems as much as we could, as more detail was progressively – but too slowly – provided to us.

After much frustration over several weeks, I took it on myself, rightly or wrongly, to alert the Treasurer, Tim Pallas, to the problems we were encountering. I warned him that the work being done was as bad, if not worse, than in the business case for East West Link. He thanked me and told me that he’d get someone to follow up with me.

I heard nothing more, and about a week later both the peer reviewer and I were taken off the project. No explanation was given to me for this; I was transferred onto another piece of work within the Department to use up the rest of my allocated budget.

At that stage I was afraid to speak up because nobody had said a word to me about my talk with Tim Pallas. I decided to lay low and see what happened next.

I’ll go into more detail in the next couple of posts, but in a nutshell, these were our main concerns at the time:

  1. The traffic forecasting model used a non-standard, illogical shortcut to create future year projections. Called the ‘single loop’ method, it arbitrarily combined present-day travel demand with the future-year transport network when calculating future-year travel patterns. This demonstrably produced much more traffic growth than the proper way of doing such modelling. It also meant that the model’s mathematical stability (a fundamental requirement) could not be calculated in accordance with established guidance and best practice. We had quite a few other problems and queries, but this was the biggest one.
  2. The economic appraisal had many distortions and add-ons, which got the benefit-cost ratio over the magic threshold of 1.0. Many of them were non-standard and highly questionable. Added to these, the business case assessed a double-act, namely West Gate Tunnel AND an upgrade to Monash Freeway on the other side of the city. The latter was there mainly to boost the benefits of the former; there was no interdependence between them. Finally, the appraisal had two benefit-cost ratios, one for Infrastructure Australia and one for the Victorian Government. The first was larger, because it was calculated without any allowance for induced traffic; the team (wrongly) claimed that IA’s guidance at the time didn’t require it.

I watched as the business case was released in late 2015 and the Government announced their support for the project, which was really a foregone conclusion. This included a new development; the intention to extend Transurban’s CityLink concession to give them 10 years’ more revenue. I wondered, if West Gate Tunnel was such a great project, why didn’t it stand up financially without this extra revenue?

The problems we were concerned about were not addressed in the published business case. They were also carried through, unchanged, into the environmental effects statement.

Senate inquiry and EES

In 2017, a Senate inquiry was announced into toll roads in Australia. I made a submission focussing on my West Gate Tunnel experiences. The inquiry committee told me they’d had a request that my submission be kept secret, but after some discussion they agreed to release it anyway (I’d used publicly-available data to demonstrate my points). I was also interviewed at the hearings in Melbourne. The transcript of this was also made public, and the story was out.

This coincided with the start of the EES hearings.

During the EES hearings, a lot of effort went into persuading the Government to release the transport modelling peer review report (which I knew had never been properly completed), but nothing was forthcoming. I gave some behind-the-scenes advice to one of the objectors, writing extensive notes for them in response to the EES documents and evidence presented during the hearings.

It was all too late to make any difference, but at least I’d got it off my chest.

There was a lot of obfuscation during the EES hearings and many major objections were ignored or discounted when the project was approved.

Construction

Construction of West Gate Tunnel started in 2018. It soon ran into difficulties, the largest of which was the extraction, treatment and disposal of PFAS-contaminated soil.

Contaminated soil awaiting safe disposal (The Age)

As of August 2020, this issue is ongoing; Transurban is battling it out with the builder, and the Government is watching from the sidelines denying any responsibility, at least so far. We shall see how that plays out.

Conclusions

From the outset, the appraisal and approval process for West Gate Tunnel was biased in favour of the project proceeding.

As I’ll show in my next couple of posts, the transport modelling and economic appraisal for the business case contain distortions which combine to provide a highly optimistic view of the project’s effectiveness. Removing these shows that the project is not economically justified, and should not have been approved.

The project’s lack of viability is further proven by the fact that it doesn’t stack up financially. Transurban can’t make it work without extra toll revenue on City Link AND a substantial injection of public funds.

Construction problems are forcing significant delays and extra costs. Transurban has reportedly considered cancelling their agreement with the Government to build it. I wonder what will happen next.

Independent, concurrent peer review processes should be mandatory on all major project appraisals. The reviews must take place in an ‘open book’ fashion as the work proceeds, with clear audit trails to show how the reviewers’ requirements have been satisfied. A final peer review report must be released documenting the review process, listing any outstanding problems and giving a clear opinion on the fitness for purpose of the work.

It’s entirely unacceptable for peer reviews to be done after the work is complete, because if any serious shortcomings are identified it’s too late to do anything about them.

Victoria’s broken transport planning ‘system’


WE’RE all watching how our governments respond to COVID-19. We’re heartened by politicians’ new-found reliance on expert research and opinions as they struggle with enormous, unexpected challenges. We don’t know what the lasting effects will be, but we hope that things will change for the better, not only in emergencies but in all other areas of concern.

The virus has shone a harsh, tragic light on many of our shortcomings. It’s incumbent on us to reflect on this and campaign for change in our areas of expertise. This is my attempt to contribute in mine.

My primary interest for the last couple of decades has been in transport planning in Victoria, Australia.

Let me be very clear about this – the system of rational and objective planning for transport is completely broken in Victoria.

It’s a systemic problem. Politics has trumped planning when it comes to investment decisions. Nowadays, projects are picked – or dreamt up – solely by politicians and their advisors, especially during election campaigns. They’re then post-rationalised through flawed appraisal and approval processes.

There’s no guarantee that the investments are the right ones, nor that they’ll provide worthwhile economic, social and environmental improvements. Political gain is the primary benefit.

It’s nothing short of pork-barreling in the extreme, with billions of our dollars.

There’s no strategic planning, no binding targets or objectives (apart from motherhood ones) and the principles of the Transport Integration Act 2010 are ignored, twisted or openly flouted.

Infrastructure Victoria prides itself on its independence but so far, its outputs have been either warmed-over reordering of the same old lists of projects, or research pieces on new technologies and trends. They have little or no demonstrable influence on decision-making in transport.

Victoria’s Department of Transport, like much of the public service in this state, has been steadily neutralised and dumbed-down to become a mere facilitator of political decisions. The vital concept of ‘frank and fearless advice’ has been destroyed.

The Victorian Auditor-General’s Office has thrown some light on all this in its reviews, but they are, by definition, after the event. Despite some strong criticisms in their reports, very little has changed. I’d argue that things are actually getting worse, despite VAGO’s presence.

The last independent, comprehensive and detailed transport study in Victoria was the East West Link Needs Assessment study, led by Sir Rod Eddington and published thirteen years ago. It was a rare example of a study into a problem (east-west travel in Melbourne) which came up with some actions clearly aimed at solving it. It wasn’t perfect, but it took a lot of effort to get it there, I can tell you! Some of its actions were even implemented.

As I’ve said, nowadays we’ve got a politically-generated list of priorities. The appraisal process for these is a box-ticking, post-rationalisation exercise. We have flawed, biased business cases and environmental appraisals; they make a mockery of established guidance and principles.

East West Link protest in 2013

Community consultation through these processes is highly variable and prevents any meaningful debate on the basic needs – it’s nowadays limited to comments and submissions during EES hearings, many of which are ignored or discounted. The adversarial nature of the hearings relies on verbal gymnastics by lawyers, trying to prove their clients’ points to well-meaning but often under-qualified panel members.

Many genuine issues, concerns and objections don’t survive this process. Those that make it into the panels’ recommendations then risk being ignored, belittled or criticised by the Planning Minister when rubber-stamping the projects, often with little or no real changes.

Recently we’ve seen big objectors (local Councils, in the cases of East-West and North-East links) resort to court proceedings to try and get their concerns addressed. I’m not aware of any real successes there. It’s a questionable use of ratepayers’ money, and it shouldn’t even be necessary.

Once these poorly-conceived and under-planned projects get started (Victoria’s ‘Big Build’), we’re seeing unexpected problems, cost blowouts and time overruns. These further erode the economic and financial viability of the initiatives.

Yes, we have a big backlog of work due to underfunding and neglect in previous years, but that doesn’t justify the unseemly haste and shoddy planning we’re seeing these days.

In future posts, I’ll record my own experiences in these processes to substantiate these observations. I’ll cover many major projects, including:

  • East West Link
  • Melbourne Metro Rail Tunnel
  • West Gate Tunnel
  • North East Link
  • Airport Rail
  • Suburban Rail Link

As I write about these and other experiences in the coming months, I’ll update this post as well. Links will appear here to new posts, where I’ll expand on my comments in this one.

Hopefully, it’ll all hang together!

In the meantime, I’d welcome your thoughts and comments. Am I right in my observations, or is it all just sour grapes? How can we bring our politicians to account? What are your personal experiences of this? Is the Victorian situation echoed in other states, and other countries? Where are people ‘doing it right’, and what does that look like?

First post – what’s this all about?


WELL, this is the first post to my blog! Welcome to all readers.

I’ve decided to start this site as a place I can post articles, analyses and commentary on transport planning and my experiences.

Some posts will be topical and current, while others might go back in history and highlight some particular projects, studies and other activities I’ve been involved with. The sky’s the limit, really.

I’m hoping it’ll become a lasting record of my life and career experiences, as well as some personal memories. We’ll see how it evolves.

Here’s a short summary of my life and career milestones, for starters…

I was born in September 1954 in Malden, Surrey. I grew up in Orpington, Sundridge and then Sevenoaks, all in Kent. I went to school at Crofton Infants, Dulwich College and Sevenoaks School.

My father was the sales manager for a marine paints company in London, and my mother was a magazine journalist; she was fiction editor, then editor-in-chief, of the UK’s Woman’s Weekly magazine.

My career started when I finished school and started my bachelors degree course in civil engineering, at City University in London, in 1972. It was a ‘thin sandwich’ course, with 6 months working and 6 months studying every year. I worked with R Travers Morgan and Partners in London and my first 6-month office stint, as a school leaver, was in transport planning! I was allocated to the team of economists and planners who were doing a big study into the future of London Docklands. Subsequent work experiences included traffic surveys, site investigations and feasibility studies, mostly to do with road planning and construction.

After finishing university in 1977, I worked in the London office for a while, then had a brief but fascinating stint in Amman, Jordan, seconded to a local firm. After this I moved to North Wales to join the resident engineers’ team for the A55 North Wales Coast Road, a large and very complex project to upgrade the main road along the coast.

It was during this time that I decided to move to Australia. Travers Morgan already had offices here, specialising in transport planning and economics, so it was a natural step for me; several of my friends and colleagues from my uni years had already moved here.

I left the UK in 1985 and arrived in Perth, Western Australia. My first major study was into new route options for the Great Eastern Highway through the hills area of Perth, followed by a study into mass transit options for the Mitchell Freeway corridor in the northern suburbs, which ultimately led to construction of the Perth-Joondalup rail line.

In 1988 I moved to Melbourne to run the Travers Morgan office there. The recession hit soon after, unfortunately, and I had the tough job of laying off most of the team as we struggled to find work.

After those difficult few years, I moved to Sydney in 1991 to do a major study for the NSW Roads and Traffic Authority. The fascinating ‘Road Transport – Future Directions’ study took 18 months and gave me a great introduction to Sydney’s transport planning scene.

After a couple more busy years doing transport studies in NSW, I left Travers Morgan to spend time freelancing and also setting up and running a non-profit organisation involved in dolphin and whale research and conservation. I had many amazing adventures over the next 4 years, but I had to get back into full-time transport consulting again, to make ends meet.

Travers Morgan had closed down after a merger, and many of my ex-colleagues had joined Sinclair Knight Merz, a well-known Australian engineering consultancy. I started there in 1997, in the Sydney office.

In 2001 I moved back to Melbourne with SKM, to run another big study – the North Central City Corridor Study. This gave me a new start in Melbourne and led to many other interesting assignments, including the Metropolitan Tram Plan, East West Link Needs Assessment (Eddington Study), National Heavy Vehicle Charges assessments and a Victorian public transport DDA Strategy. I ran the transport planning team in the Melbourne office and was fortunate to attract many fantastic people who now play significant roles in transport planning in Victoria.

I was invited to move to SKM’s London office in 2008 to help build the UK transport planning business. Sadly things got off to a bad start – the GFC hit London a couple of months after I did, and winning work became very difficult. We were lucky to have a large project (the Vauxhall-Nine Elms-Battersea Transport Study) to keep the London team busy.

I returned to Melbourne in 2010 and continued with SKM, picking up more transport studies including the Rowville and Doncaster Rail studies, Melbourne Airport Landside Access Strategy and many others.

Merger moves were afoot at SKM, and in 2013 it became part of US-based Jacobs. As a shareholder I did OK financially from the deal, but along with many others I left the business soon afterwards.

Since 2014, I’ve been freelancing as a sole trader. I got a contract in the Victorian transport department, reviewing business cases and their transport modelling and economics, including Melbourne Metro Rail Tunnel and the West Gate Tunnel, amongst others. After this, I spent a few months with the Grattan Institute, then continued as a sole trader. Several expert witness and review roles followed at EES hearings into major projects, including Melbourne Metro, West Gate Tunnel and North East Link.

In 2020 I joined Knowles Tivendale’s Movement & Place Consulting as a part-time Senior Associate. It’s been great to get back into a team again, although of course the COVID-19 situation has prevented us from getting together face-to-face much, so far.

%d bloggers like this: