Over the last few weeks I have been putting together a series of posts about Paul Mees’s new book Transport for Suburbia: beyond the automobile age. It really is an excellent book, and I am getting to the real ‘meat’ of the book about how we can actually make our public transport systems drastically better. A key part of improving public transport systems in cities with relatively dispersed trip patterns (ie. Auckland) is what Mees calls “The Network Effect”. The network effect happens when you make it easy for people to transfer from one service to another, as it brings so many more locations within relatively easy public transport access.

To illustrate the network effect, Mees uses the hypothetical city of “Squaresville”, which is illustrated in the diagram below: The Squaresville scenario is basically the worst possible situation for traditional public transport, with trips having absolutely no pattern at all. Mees explains further:

The city has a grid road network, with ten north-south and ten east-west roads, at intervals of half a mile or 800m. travel patterns are completely random, with no dominant pattern of movement. Each of the city’s 100 square blocks produces 100 trips a day: one internal trip (made on foot), and one external trip to each of the 99 other blocks of the city – giving 9900 external trips in total.

Squaresville has ten bus routes that grew up in a free market environment, with each operated by a different firm. There is one route along each north-south road, reflecting a past era when this was the dominant pattern of movement (Figure 9.1A). This means that each resident of Squaresville has a bus within 400m walking distance, but can only reach the nine other city blocks lying along her or his bus route, giving access to 900 daily trips out of the total of 9900. Assume that public transport attracts a third of the trips it can theoretically serve, this gives a total of 300 trips (a third of 900), or a city-wide mode share of only 3 per cent.

This would be a fairly good approximation of the current bus system in Auckland I think. There are many “suburbs to CBD” routes running roughly parallel with each other, but very few crosstown services linking them up. Furthermore, the cross-town services that do exist generally run at very poor frequencies, take bizarre backstreet routes (yes I’m looking at you 008) and integrate extremely poorly with the suburb to CBD routes.

Mees continues:

Now, imagine that the government of Squaresville wants to do something about the low rate of public transport use in the city. It pays the bus operators to double service frequencies on Squaresville’s ten bus routes (Figure 9.1B). With a typical demand elasticity of 0.5, this would increase patronage by half, to 450 trips per day or 4.5 per cent of the market. Occupancy rates will fall, since patronage has grown more slowly than service levels, and fare revenue will not cover the extra costs. Subsidies will rise, cost-recovery will worsen and so will greenhouse emissions per bus passenger. Public transport is still of marginal importance, but it has become less efficient in economic and environmental terms.

One could argue that this has been what has largely happened in Auckland over the past decade. We’ve spent more and more money on the existing services, and while that has certainly led to some level of patronage increase, we haven’t really made the kind of gains we would have expected. This is clearly shown in the table below, which shows that over the past decade subsidies for public transport have increased from $45m a year to $145m a year, while patronage has only grown from 44m trips a year to 58m trips a year – barely keeping up with population growth. Is there another way? Returning to Squaresville, Mess suggests so:

Imagine instead that the additional buses are used in a different way. Ten east-west routes are introduced to complement the ten existing lines and create a grid network, as shown in Figure 9.1C. The number of trips served directly doubles, to 1800 [as there are now 20 routes], but by transferring between routes, passengers can now access the entire city, so the network also serves the remaining 8100 trips. Squaresville’s planners do everything possible to make transfers convenient, providing integrated fares, convenient facilities and coordinated timetables. But since so many transport analysts say that passengers dislike transferring, let’s assume that the mode share for trips requiring a transfer is only half that for direct trips, that is one-sixth. So the total number of public transport trips is one-third of 1800 plus one-sixth of 8100, giving a total of 1950.

Under the second model of service provision, public transport’s mode share has jumped dramatically, from 3 to 20 per cent. Service has increased 100 per cent, but patronage has grown 550 per cent, giving an elasticity of 5.5. Increased revenue would more than cover the costs of the additional service and occupancies would rise substantially, reducing subsidies and greenhouse emissions per passenger.

Now while Squaresville obviously isn’t real, I think it shows that dispersed trip patterns certainly does not have to be the death-knell for successful public transport. There’s an interesting comparison in terms of how airlines operate – in that you don’t expect to be able to fly directly from New Zealand to a great number of random places in Africa – for example – because the demand would be too low for a direct service. Nobody would wait a week for a direct service when there are much higher frequencies that involve transfers at major hubs like London or Singapore. Mees uses the example of direct airline services between Australia and many European cities to make this point clearer:

There were once infrequent direct services between Australia and other European cities, but this was at a time when fares were so high that airlines could run half-empty planes and still make money. The transfer at London or Singapore is one of the costs of the dramatic fall in real airfares in the last three decades, but it has also allowed Australians easier access to a larger range of European cities.

The point is that to achieve efficiencies and allow services to be provided at regular intervals to an incredibly wide range of places, transfers are necessary. Just as they’re necessary in the airline business, they’re necessary in public transport. Of course there will be the “cost” of the annoyance of having to get off one bus and onto another (or onto a train), but if that ‘cost’ means the benefits of being able to simply walk up the road without checking a timetable, jump on a bus you know is coming every 5-10 minutes, get off somewhere and jump on another bus almost immediately and that process will get you just about anywhere, surely it’s worth it?

Share this


  1. I have seen from your previous posts that Mees has used Auckland as an example of what not to do. I would be interested to see what he would recommend Auckland did to improve its PT usage, would any council or group ever be brave enough to try?

  2. The squaresville analogy also struck me as being very appropriate for the former tram lines that spread out over the isthmus from the CBD, Sandringham, Dominion, Mt Eden especially. They still are among the best bus services in Auckland but are not much use for anyone who doesn’t want to go to the CBD. Now that the travel patterns have become more complicated because of developments such as St Lukes, New Lynn, Ellerslie offices etc the buses need to be able to match that, and the network effect is the only way this can be done. With privatized services this never works because the cross-city services will take a fair few years to start to get a good return. However because they make the old north-south services more profitable so this evens out he losses on the cross-city services.
    Air NZ and intercity buses also use this model, some services are loss-leaders. Thats why its so hard for a new player to break in to the market, no new entrant can bring in the network effect, only a monopoly.

  3. Mees does look at the question of why private transit firms don’t adopt the network effect voluntarily. His answer is that networks require cross-subsidy because real cities aren’t like Squaresville and demand is not evenly dispersed – some routes and periods of the day are more profitable than others, but all need to be run at a high frequency and standard to create the network effect. Who’s going to volunteer to run the low patronage services?

  4. The funny thing is that ‘Squaresville’ isn’t too far from the truth with Melbourne’s grid layout. It already has buses along just about any main road, just lacking any kind of reasonable frequency or connection between them. The tram routes do work this way in some parts of town, particularly around the CBD/fringe grid where frequencies are high and routes (somewhat ‘accidentally’) interconnect.

    I agree the problem is the fractured nature of private operators. Here again Mees calls for a single state agency for the planning of routes and timetabling to create a actual network out of the system. Auckland should too, it would be good if the PTMA wasn’t repealed and ARTA’s successor had the power to coordinate connections.

  5. Nick the PTMA isn’t being repealed, just altered/weakened/changed in some form or another. It will be interesting to see exactly what changes happen, and how they’re justified. It seems that the biggest benefit of the PTMA, apart from integrated ticketing, is that we should get better value for money out of the public transport system.

    I think that Joyce’s proposed changes would take away the opportunity for ARTA (and others) to prohibit commercial services. Restrictions can still be placed on them, they just can’t be banned.

  6. I guess the key question is will it allow ARTA to undertake a form of gross contracting that actually works, i.e. keep the ability to cross-subsidise and stop the private operators from keeping all the profit from the busiest routes while asking for a handout for the rest.

  7. to admin –

    I am not sure that it’s a matter of “who’s going to volunteer to run the low-patronage services”. It is a matter of who’s going to pay for it. Even if you had a centralised authority for services, like London, what you would find with these services is that they would cost a lot of money for the numbers of additional passengers they would carry. You could certainly design a set of tenders for these services, but it’s finding the money that is still the issue. If bus companies aren’t operating them commercially, there’s generally a good reason.

    Also, it’s not clear that any of the bus companies in our commercial model are making windfall or monopoly profits. A quick check of Infratil’s numbers suggests that all of the free cash they are generating from NZ Bus is then being absorbed back in the business as they invest in equipment and so forth.

  8. Ross, I think you are focussing too closely on the profitability of individual routes when the whole point is to consider the profitability of the network as a whole. The idea is that a council controlled organisation could use the profits from the more profitable routes to pay for the less profitable ones… which through network effects would attract more patronage to the system and create greater profitability overall even though there would still be a mix of more profitable and less profitable individual routes.

    Right now any profits from commercial routes are basically returned to private shareholders. While NZBus might be currently re-investing profits into equipment they must be making a net profit in the long term or they simply wouldn’t exist as a business.

    The answer to ‘who will pay for them’ is the same as it is now: we will, through bus fares and council rates. But instead of our bus fares going to NZBus shareholders via profits and our land rates going to NZBus shareholders via subsidies, we can simply pay companies like NZBus to provide a fixed service and reinvest any profits back into an improved system.

    Cosider the ‘Beeching Axe’ of rail cuts in the UK under Thatcher. They went through and closed all the ‘unprofitable’ rural and branch lines while maintaining the ‘profitable’ trunk routes, in the name of economic efficiency and cost reduction. Yet the economists in charge ignored the network effect, and quite simply without a system of rural and branch lines to feed the trunk routes patronage plummeted on the trunks. They were left with a smaller, less useful system that had a lower cost recovery ratio then before they started!

  9. No, I think you’ve missed my point:

    * My argument is that even in a council-controlled bus system, where profit was not the issue, the /net/ costs of running the system would rise significantly if you expanded services along the lines you suggest. Network effects mean that you would attract more passengers to the system, I grant you that, but not enough to cover the costs of extra service provision.

    * If NZBus is profitable, Infratil will keep owning the business, but again, I would not call their profitability excessive or monopolistic. The bus market simply isn’t strong enough to sustain monopoly profit.

    * The Beeching cuts were in the 1960s, well before Margaret Thatcher (in office 1979-1990). While there was a network effect, as you put it, it was actually slight. Where lines lost in the Beeching cuts are being restored, as in Scotland, it is because underlying market growth is justifying their re-instatement, and they’re not being put back to feed passengers into the ‘trunk’ parts of the rail system. Beeching’s cuts did remove some /trunk/ lines that today we really wish he hadn’t.

  10. Ross, the point is to adjust the system to capture more effectively the network effect – even with the same amount of resources there could be great benefits.

    Mees’s argument (did you even read it all?) is that a 100% increase in service provision would (theoretically at least) lead to a 550% increase in patronage. That’s very different to your contention that the extra patronage attracted wouldn’t be worth it.

  11. I think the issue may be that the network effect will take some time to kick in properly, so for a fair while the new services would run at a large loss. We know we cant introduce the network effect overnight, although we could probably make some major improvements by simply altering timetables, and making connections more visible.
    The question is in the short term do we concentrate on a small number of services to give a 10 min frequency, or do we improve frequencies of many other services from say 30min to 20min, or 1hr to 30min etc.
    The implementation will probably result in lower farebox recoveries for several years until the network effect really kicks in, which many councils/govt would not like.

  12. There are two ways to introduce the network effect. One is through a grid, and the other is through a “hub and spoke” system. Given Auckland’s geography I think a hub and spoke system might make a lot more sense.

    The key is fewer different routes, but better quality and higher frequency routes. If ARTA ever actually get around to rolling out their RTNs, QTNs and LCNs we could be on our way to something of a network effect.

Leave a Reply

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