Big infrastructure, big uncertainty

One of the peculiar features of debates about big monolithic infrastructure projects, such as universal broadband networks and high-speed rail lines, is the way their supporters talk about them in public. To advocates, the wisdom of these projects is obvious. You can never have too much broadband! High-speed rail is the future! Why can’t we be like the visionaries who built the Snowy Mountains Scheme?!? $50 billion, $100 billion? “Chicken feed” is what we’ll call it in 20 years!

And indeed some opponents of these projects take the same attitude from the other side of the fence. Everything’s fine as it is! This new thing will be an enormous white elephant, obsolete before it is finished, you can bet on it!

What we don’t usually talk about is Knightian uncertainty – that is, risk you don’t know enough to quantify, or sometimes even recognise.

This was what former US defence secretary Donald Rumsfeld was talking about when he famously spoke of “unknown unknowns“. That Rumsfeld’s comment is frequently ridiculed just shows how alien considerations of uncertainty are in the public discourse*. Almost no-one wants to say, “well, we just don’t know”. There are things you might do if you just don’t know. They rarely get talked about.

But if you’re having a serious conversation about infrastructure, you have to talk about uncertainty. Uncertainty matters a lot for big monolithic infrastructure projects, such as universal broadband networks and high-speed rail lines: the time they take to complete and the technology environments in which they operate mean there’s great uncertainty about their ultimate benefits and ultimate demand. By the time you deliver them, the world may be in the process of passing them by. Demand risk is always a factor in these projects, but here we seem to have substantial demand uncertainty, which is worse.

This may be particularly true over the next few years, because predicting infrastructure demand growth seems a tough job right now. In the broadband business, you have to make a guess about the capability of mobile broadband and about willingness to pay for ultra-high-definition video streaming. In more traditional infrastructure, car and power use are currently and unexpectedly declining, and we don’t know whether these developments are structural or cyclical (and if they’re structural, we don’t know whether they are new trends or one-off adjustments).

My gut feel is that the shape of our transportation and communications networks is less predictable now than at any time in the past 50 years. Maybe I’m wrong. But that’s Knightian uncertainty for you: you can’t really size it up.

A few people are looking at the effects that technology forecasting has on infrastructure projects. From rail transport analyst Christian Wolmar:

One of the reasons for my scepticism about [UK high-speed rail project] HS2 is on the basis that it does not take into account future development of technology. Just look at how technology has changed since 1993 when mobile phones had barely taken root, Google, Facebook and Twitter were but twinkles in their founders’ eye and digital TV was just starting. Will there really be enough people wanting to pile into what are likely to be expensive trains in 20 years time to justify the huge expenditure on this project?

And here’s where I stick my neck out. The next big technology, one with such huge implications that it is impossible to being to predict them, is driverless cars … Perhaps they will start by being driven only on motorways but even that would have enormous consequences. It would combine many of the advantages of train travel with the flexibility of car use. Think trucks, too. The economics of transport would change as radically as they did when the railways were first developed. The time frame may be a decade or two, but the consequences will be much more far reaching than, say, the much talked-about electric cars. The driverless car – or rather motor vehicle – is the innovation that we ought all to be taking into account in our future thinking.

From University of Minnesota mathematician and network capacity expert Andrew Odlyzko:

How many times have you seen predictions and promises that better communications, such as faster Internet access, will stimulate telecommuting and decrease road congestion?
Such predictions are almost certainly wrong. At least they have been consistently wrong for about two centuries. Many, often very knowledgeable, observers, thought that transportation and communication were substitutes for each other. But the uniform experience to this age has been that they are complements, and grow in parallel. Yes, you may work from home, but chances are that you will make more trips to meet clients, or for family and other reasons, and in the end will travel more than before (barring major upsets, such as astronomical rise in price of fuels).

So what can we conclude? Most of all, that the future is hard to predict, so we have to prepare for the unexpected. Second, though, we should keep in mind that among the most common unexpected phenomena is the resilience of old technologies, services, and business methods, and their propensity to adopt some of the innovations that we work on.

If we conclude that big monolithic infrastructure projects need to factor in uncertainty, what do we do?

The simplest course is to raise our estimation of the risks. But how much? Bent Flyvbjerg recommends a technique called “reference class forecasting“, which predicts the outcome of a planned action based on actual outcomes in a “reference class”. But what’s the right reference class, and how useful are past projects, when the environment is changing fast?

Another option is to try to break down the project into smaller units, which we can tackle one by one. That only goes so far with high-speed rail, where we essentially have just three big legs (Melbourne to Canberra, Canberra to Sydney and Sydney to Brisbane). It does make a lot of sense for a fibre-to-the-premises broadband network, since such a network is by its nature mostly local in nature. It’s why we should be improving our existing broadband networks suburb by suburb, as suggested by Joshus Gans in a paper for CEDA back in 2007.

We should also, of course, adjust our thinking about such monolithic infrastructure – “megaprojects”, as Flyvbjerg has dubbed them. If the most recent study of the Melbourne-Sydney-Brisbane high-speed rail system says the environmental benefits mostly emerge from around 2057, for instance, we should probably avoid making very much of those benefits.

But first, we need to talk about big infrastructure as if uncertainty mattered. The unknown unknowns are real, and we need to acknowledge them.

* Note: You don’t need to agree with Rumsfeld about the conduct of US defence policy in order to think he was, on this occasion, talking sensibly about risk. As John Quiggin has noted in his own defence of Rumsfeld’s comments, the bloke could have usefully thought more and earlier about the repercussions of Knightian uncertainty for US attempts to reshape the Middle East through a war in Iraq.

About David Walker

David Walker runs editorial consultancy Shorewalker DMS (, editing and advising business and government on reports and other editorial content. David has previously edited Acuity magazine and the award-winning INTHEBLACK business magazine, been chief operating officer of online publisher WorkDay Media, held policy and communications roles at the Committee for Economic Development of Australia and the Business Council of Australia and run the website for online finance start-up eChoice. He has qualifications in law and corporate finance. He has written on economics, business and public policy from Melbourne, Adelaide and the Canberra Press Gallery.
This entry was posted in Economics and public policy and tagged . Bookmark the permalink.

36 Responses to Big infrastructure, big uncertainty

  1. conrad says:

    You could take even smaller steps for some types of infrastructure like rail, which means you might reduce the uncertainty in estimating the big cases. Surely small legs that connect the big cities and their close satellite cities would be worthwhile just to disperse the population and reduce housing pressures a bit (e.g., Geelong, Melbourne, Ballarat or Sydney, Newcastle, Wollongong). A similar argument might be made for telecommunications, where sticking fiber in the big cities would be much cheaper than chasing silly political goals like trying to connect every single last person, no matter how expensive. Once a few million people were on it, presumably services to take advantage of it would pop up and give you some idea of the demand for this sort of stuff.

  2. David Walker says:

    “Sticking fiber in the big cities would be much cheaper than chasing silly political goals like trying to connect every single last person.”

    Funnily enough, we did this for the CBDs and universities some years ago. The connectivity in any major Australian city centre is huge. In a sense, we have long had the NBN; it’s just been restricted to places like the CBDs. “Services to take advantage of it” have indeed popped up, and they have been very useful but not radically transformative. That should tell you something about the NBN.

    This build-out also happened without any need for a “national network”. That’s because broadband is a local-scale technology – the costs mostly come in the “last mile”, so you can build suburb by suburb as need dictates. The NBN publicity pretends none of this is true, which is why its rollout started in rural Tasmania.

    “Surely small [rail] legs that connect the big cities and their close satellite cities would be worthwhile”

    Rail is the opposite case to universal high-speed broadband; it’s all or nothing. The Sydney-Canberra leg would indeed be built first. But for high-speed rail, short legs are actually less compelling than we might think, because the train spends all its time speeding up and slowing down. The best economics come from minimising the number of stations so that your railway can compete with aircraft. Also, most of the cost is in the stations and the underground tunnels leading to them (especially in Sydney). Oversimplifying, the terminuses are the project.

  3. conrad says:

    “But for high-speed rail, short legs are actually less compelling than we might think, because the train spends all its time speeding up and slowing down”

    There are lines like this around Paris, where you go from nice-little-city to the city centre of Paris (and other centres), and they work really well (I sometimes get Aix-Marseille, and it takes almost no time at all, so I don’t see the problem with speeding up and slowing down. Perhaps you don’t hit 330ks, but it’s certainly fast enough). You just need the political guts to say we only have two stops. Our cities are also worse than Paris in terms of sprawl and other transport options — Especially for Sydney and Melbourne, it would be faster for many people to take a 15 minute train to the city centre than come from even zone 2 places in suburbia (or, say, 20 minutes if the last few ks are really slow because you don’t want the most expensive option).

    As for tunnels, I agree they are expensive, but the benefit would be reduced congestion and hence less need to spend oodles on infrastructure (I seem to remember the Vic government estimated 70 billion just to stick flyovers on all the train crossings), so even if you went over budget, there are great gains to be made, especially because all other attempts to create a second city centre for Melbourne have failed (stretching my memory back miles, I seem to remember the Cain government spending money trying to promote Dandenong for this purpose). I also don’t see why trains are such big unknowns — they’ve been around 100 years, and even fast train technology is old hat now.

    • David Walker says:

      Sorry, Conrad, I should probably clarify further that my rail comments go to the sort of high-speed rail line recently scoped for the federal government, which offers three hours Melbourne-Sydney and takes many years to build. It’s not so much the fast train technology that may be unpredictable in this scenario – it’s a number of things external to the train system.

      • conrad says:

        Yes I realize — that’s why I suggested we should start with comparatively small ones in potentially high value areas. At least for transport, these would be of immediate value vs. the grandiose dreams spruiked by politicians.

        • Patrick says:

          The Victorian government could do few better things with its infrastructure expenditure than put just the south-eastern and Bell St railway lines off-road with either tunnels or bridges. There are probably less than 10 really critical junctions (both the Bell St ones, the Burke Rd one and Elgar Rd would have to be the starting point) and fixing these would have a massive impact on congestion.

  4. James says:

    There’s some argument that money should be spent straightening the curves and generally improving existing rail infrastructure rather than building a whole new HSR, but generally the conclusion is the improvements you could get on Australia’s rail are fairly marginal and it quickly becomes more cost-effective to just build a whole new line.

    The NBN is needed for one reason alone, HD video on-demand. Remember that Netflix is 1/3 of bandwidth use in North America. Any other benefits are pure gravy.

    I am looking forward to self-driving cars, although I think they’ll affect commutes more than inter-city travel, even if you put a bed in a car and it drives you through the night most people would rather catch a plane or train.

    Flyvbjerg of course co-wrote the book Megaprojects and Risk.

  5. john r walker says:

    It is difficult to make predictions, especially about the future
    Yogi Berra

  6. Patrick says:

    It’s very hard to imagine that there is nearly enough uncertainty in the whole world for one to imagine that high-speed rail could make sense in Australia.

    If one takes the market cap of Qantas as a proxy, to the nearest couple of billion, for the expected future profits of (semi-)rapid transport between Melbourne, Sydney and Canberra, that’s about $4bn. So the suggestion is that high-speed rail would be worth about 30 times that??????

    Rumsfeld was a paragon of cautious, risk-weighted and sensible decision-making by comparison.

    If you wanted to help, you could complete the East-West link in Melbourne (which they actually are doing apparently), build a railway line next to or above or below it and make an outer loop extending along the existing Eastlink and up to the airport, and also extend the existing railway to the airport from Essendon. Sydney you would just build another runway.

    • Ken Parish says:

      You could also upgrade the Brisbane-Sydney Melbourne freight rail corridor to reasonably high speed status for a fraction of the cost of a VFT passenger line. Doing that would have huge economic benefits, as well as getting many of the big trucks off the highways and enhancing safety. A VFT passenger line is an absurd proposition from almost any angle.

  7. Ian Milliss says:

    Potentially driverless cars exist already and are on the market, our new car has cruise control that simply follows the car in front at set distance, monitors the lanes and warns you if you are drifting out, registers speed signs, parks itself, etc. I’m sure if I google I will find that it is designed to be semi driverlesss but some parts of the software are simply absent for legal and liability reasons. So no futurism involved there.

    And after 35 years of IT involvement I will say with complete confidence that 1Gbs will be far too little well before 2020, I simply point out the fact that you can now buy a 3Tb external hard drive for under $120 to demonstrate what every IT person knows, the appetite for memory and bandwidth is only unpredictable in the extreme degree of its insatiability and capacity is always consumed faster than it can be created.

  8. Ian Milliss says:

    And you should probably read this blog post by Marc Westbury about his experience of already being on the NBN to understand why it is not at all about local connection in fact it is the opposite. I think you really are thinking about it incorrectly.

  9. Gummo Trotsky says:

    Ian’s last comment raises another ‘unknown unknown’ to consider – the new social arrangements that will result from the interplay between the infrastructure project and other social & political factors. For example

  10. john r walker says:

    Ian if you go to the NBN roll-out map and type Tarago NSW you get
    this result:

    New Development Fibre | Service available – There are now NBN services available in your postcode.

    However if you move the little google map pointer a few meters to the Tarago school
    you get this result:

    Lot 1 Braidwood Road
    Tarago NSW 2580
    Not currently available
    The NBN is coming to your place; however construction hasn’t commenced in your area as yet

    If you then zoom out you will eventually discover that there is a small part of a greenfield housing site in Goulbourn (about 40ks away) that is fibered up – and that is it for the entire post code.

    The current NBN is best if you prefer puff to working. This has to be a case of fraud, surely?

  11. Ian Milliss says:

    a stuff up rather than a conspiracy I’d say. “There are now NBN services available in your postcode” seems to be technically correct, just not available for the whole postcode area.

    • john r walker says:

      Its much the same in Waterloo and Alexandria.

      The thing that gets me is why didn’t they first focus on connecting every school, hospital, medical facility and the like. A large proportion of Australians are over 65 and/or can’t afford it anyway, so why this obsession with the last 500 meters why so rigid on connecting every building all at once?

      • Ian Milliss says:

        I can’t believe you are asking that when the answer is so simple, without the last 500 metres it just doesn’t work, you may as well not build it at all. It’s like building a house cheaper by leaving the roof off. Or building a phone system based on phone booths at the end of each street – who are you going to call? This is not going to cost anything more than people are already paying especially when you consider that it will replace other services they are already paying for – fixed phone lines, cable tv, etc. Surely this has been rehashed multiple times and doesn’t need repeating. So what if it is still patchy, saying they should just connect hospitals, schools etc is rather like saying railways or roads should only connect hospitals and schools, once you do that everybody else may as well use it anyway.

        And shockingly, most people are under 65, around 86% are so really you can’t design a future around them. In fact most people are under 38 years of age and have lived with ubiquitous computerised communication their entire adult life. The NBN denialists, like most other current types of denialists, are mostly aging people clinging to an obsolete understanding of how the world works and what is needed to maintain and improve it. I’m 62 but maybe its because I’ve made my living in IT and communications for most of my life that I don’t have any difficulty whatsoever in imagining in broad outline the things that will become possible and in fact routine.

        • john r walker says:

          Fiber to every home type NBN has proven to be hard to deliver on. And I think claims about its costs/benefits when it has already cost a lot to deliver little in actual connections are arguable.
          In Canberra the NBN has had to resort to giving money to ISPs, so that they can offer $100 inducements to people to sign up and use the NBN in suburbs were it is available such is the demand for it.

          Pragmatically it is either FTTN or zero.

          Making the last bit of fibre user pays, as opposed to taxpayer funded, is a dumb, short-sighted decision in my view, but no big deal. It’s certainly better than dismantling the whole thing, as would have happened if Malcolm Turnbull hadn’t ensured that Tony Abbott WAS for turning.

          As I understand it the FTTN does not preclude connecting the last bit in the future, no?
          I.e It is not leaving the roof off, it is more like our parents great depression generation- you bought the final fittings for the house, bit by bit, when you had saved enough cash.

          Ps the ‘age’ figures are interesting , so the talk of a coming tsunami of age care costs is, puff?

  12. Ian Milliss says:

    If it is FTTN or nothing then go for nothing. That is the cheapest option of all and we will be no worse off.

    • john r walker says:

      Are you saying that the claim that the node system can be extended to direct connection, as needed, in the future, is not true?

      • john r walker says:

        2 danks st Waterloo is 3.8ks on foot from Martin Place. It is a median size complex of commercial premises .

        The NBN roll out map says of this address;

        2-6 Danks Street
        Waterloo NSW 2017
        Not currently available
        The NBN is coming to your place; however construction hasn’t commenced in your area as yet.

        With respect are you really surprised that many are tired of the it has to be perfect… even if it never arrives?

      • Ian Milliss says:

        John they have barely started building it. I should get Wendy to give a god lecture on how complex and time consuming major infrastructure projects are, especially national ones. You are caarrying on like a little kid going “are we there yet, are we there yet?”. We aren’t there yet.

      • Ian Milliss says:

        Re can FTTN be extended, yes it can, but at greater total cost than doing it properly in the first. The reason it isn’t worth doing FTTN at all is because it is an enormous expense to end up with a system that will for most people be almost imperceptibly different from the broadband they have already. So surely you aren’t arguing that it is a good idea to build a system that won’t work so that later we can do it all a second time in order to get something that does work? Just do it properly the first time.
        Two things about this whole debate really annoy me. Firstly I despise the underlying argument for mediocrity, the apparent belief that we should never expect anything to be first rate or even just done properly. Apparently large parts of the population are arguing that they are themselves crap and they only deserve to get crap. I just don’t share that abjectness, particularly at as time when we are one of the richest countries on the planet with the strongest economy.

        Secondly, by far the overwhelming number of IT industry members are saying what I’m saying and we are saying it because we have the technical expertise to understand the proposals being put up and to understand that the Libs proposal is worse than substandard, it is a technical absurdity. And as with climate change denialism, our expertise is derided and ignored by Liberal supporters basically in order to score cheap political points on the one hand and to protect the business interests of the mainstream media on the other.

        • john r walker says:

          “a small country with big distances”.
          George Seddon.
          There are real valid reasons why our roads rail and other similar infrastructures cannot be the same build ‘quality’ as in Europe or japan or the US. Indeed that was the point of the article these comments are attached to.

  13. conrad says:

    “The reason it isn’t worth doing FTTN at all is because it is an enormous expense to end up with a system that will for most people be almost imperceptibly different from the broadband they have already”

    I’d just love a connection that doesn’t have the bad habit of dropping out now and then and slowing down in the school holidays (as would my neighbors). Given I live 10ks from the centre of Melbourne, this is hardly a big ask. That difference wouldn’t be impercetible at all as I wouldn’t have to go in to work in some instances.

  14. Ian Milliss says:

    “dropping out now and then and slowing down in the school holidays” is exactly what will continue to happen with FTTN because of physically deteriorating copper network for the final connection and because you will be sharing with everyone else connected to that node, rather like an old style party line phone. It will be just like it is currently here in Linden, ADSL becomes unusable after 3.30 in the afternoon when the kids get home from school and is particularly unusable on bad television nights as everyone starts surfing the web instead.

  15. john r walker says:

    Firstly its is my understanding that wireless connections over short distances OK and are definitely the new norm, sales of fixed desk top type devises are falling. So is there that much diff between the ipad that is connected to the wifi in the basement and the ipad that is connected to the WIFi node at the bottom of the street?

    • James says:

      John: there is a big difference – the wifi in your house is supplied bandwidth by your ADSL or VDSL (FTTN) or FTTP. This means you can get 12 or 25 or 100Mbps guaranteed, for all of your devices. The LTE node at the end of the street (actually, it’s more like your suburb – if you wanted one per street, you’d have so much fibre to run you could do FTTP anyway) can give a few people 100Mbps in perfect conditions, or 10Mbps realistically because it’s providing service to a thousand people. Plus the cost of data over LTE is two orders of magnitude more expensive, as well as being an order of magnitude slower in practice. iiNet can give you 1TB of data for $100/mo at 100Mbps, you’ll get 15GB at 10Mbps for $100/mo on Telstra LTE.

      • john r walker says:

        It is looking like the NBN was one of those ‘bids’ made to get the cross-benches on side, rather than a carefully thought out, real world proposal.
        I know things like this take time… but after 3 years it hasn’t even walked the 4 ks from Martin Place to Danks street Waterloo, something I do in a hour.
        And I trust Turnbull’s “kind” estimates of the true costs of the governments proposal much more than the governments ability to estimate anything financial.

        Therefore it seems to me that the governments NBN is another half baked nice idea = ill judged debacle in practice.

        • James says:

          If you costed Turnbull’s plan using his costing for the ALP NBN, you’d add $10bn or $20bn, since he’s assuming worst case for everything for FTTP, but not for his FTTN. He also thinks Telstra will hand over the copper for free, and hasn’t accounted for the maintenance cost of the copper – this is so high Telstra thinks it’s $1b of NPV better off from decommissioning it.

  16. john r walker says:

    “unknown unknowns”, as I said Turnbull ‘seeems’ more creditable, but who really knows.

    The NBN is far from the only ‘snowy mountain authority’ project on the agenda.
    And the long running psychic wealth approach to paying for the total cost to government needed for proper funding of NBN, Gonski, NDIS ,Carbon schemes and on and on , as working realities, is not convincing.

  17. john r walker says:

    The Fin today reports that the Government has significantly reduced equity funds to the NBN

    In 2013-14, the government will tip $5.1 billion into the NBN, compared to forecasts of more than $6 billion in last year’s budget and mid-year outlook. In 2014-15, funding is now expected at $6.3 billion, compared to $6.7 billion in the mid-year outlook.

    Lots of cups and thimbles but only one pea.

  18. Ian Milliss says:

    So John you are changing your complaint from “they shouldn’t be building it but they aren’t building it fast enough” to “it’s too expensive but they aren’t spending enough money on it”? Because you are a friend I will take the tolerant attitude that you are trying to demonstrate F Scott Fitzgerald’s dictum that “The test of a first-rate intelligence is the ability to hold two opposing ideas in mind at the same time and still retain the ability to function.”

  19. john r walker says:

    As it stands the NBN is very hard to see, has a roll out map website that is not very informative – a lot of money has gone in but exactly where it is is hard to say.

    To the economist embezzlement is the most interesting of crimes. Alone among the various forms of larceny it has a time parameter. Weeks, months or years may elapse between the commission of the crime and its discovery. (This is a period, incidentally, when the embezzler has his gain and the man who has been embezzled, oddly enough, feels no loss. There is a net increase in psychic wealth.)
    Galbraith -The Great Crash 1929.

Leave a Reply

Your email address will not be published. Required fields are marked *

Notify me of followup comments via e-mail. You can also subscribe without commenting.