I’m not going to repeat the excellent Guest Post review of the book Human Transit – other than to note from what I’ve read of the book so far I agree with pretty much everything in the review. It is a very good book. It is a book that should be widely read by those involved in public transport planning, as well as by anyone who find themselves interested by public transport, how it works and how it could work better.

One part of the book that I found particularly interesting, perhaps because it fits with some of the critiques I’ve made of Auckland’s bus network in previous posts (for example: here and here) relates to the question of how much emphasis we should put on peak travel against all day travel. The question occupies chapter 6 of Human Transit, which also provides some useful insights into this issue.

At a basic level, most public transport has times of peak demand which correspond with the beginning and end of the working day. Human Transit notes that the peaked systems (those with the biggest difference between peak and off-peak demand) are almost always systems serving low-density suburban areas – where driving all day is easy and it’s only congestion (or parking costs downtown) which encourage people onto public transport during peak hours. Auckland’s North Shore is a pretty good example of a highly ‘peaked’ public transport market, with very strong demand for public transport to the CBD during peak hours (to avoid congestion and having to pay for parking) but pretty weak demand outside the peak because the public transport system isn’t particularly useful for trips other than those heading to the CBD.

There’s nothing particularly new about all this. Where life becomes interesting is when we start to look at the cost-effectiveness of adding services at different times of the day, or alternatively where we decide to cut service if we need to save money. Human Transit tells us an interesting story in this regards:

In the early 1980s, when I was an undergraduate intern in the planning department of Portland’s TriMet, I remember a day when the manager of scheduling was tearing his hair out in frustration. The agency faced budget cuts and was having to cut service, but the general manager (the chief executive officer of the agency) had instructed them to cut only outside the peak. “Don’t cut the peak,” he had said. “The peak is our bread and butter!”

But as it turned out, there was no way to protect the peak service from cuts and still save any money. They could cut midday service, but this would turn all-day shifts into peak-only shifts, which would make those shifts more expensive to run. This effect was so pronounced that it cancelled out most of the cost savings from the service cut.

There are many reasons why peak services are more expensive to run than off-peak services. Labour is one matter, getting the buses and trains back to the depot when empty (those dreaded “not in service buses”) is another, but perhaps the most significant is the cost of the bus or train fleet that must be owned to provide that ‘peak of the peak’ service frequency. Increasing off-peak frequency generally only requires a bit more fuel and staff cost (and may not even require that if staff are heading back to the depot anyway, or have contracts requiring minimum shift times). Increasing peak frequency typically requires buying more vehicles (be they buses or trains), hiring more drivers, running more ‘out of service’ kilometres and entering into increasingly complex and inefficient rostering procedures. The book picks up on this issue:

Fleet size depends on how many vehicles you need at once, and that, of course, depends on your peak service, not your all-day pattern. Many transit agencies must purchase, license, store and maintain a vehicle that makes only one round trip per day. That’s a huge inefficiency compared to an all-day operation whose fleet may work 10 to 20 hours each day.

I think it was Paul Mees’s book “Transport for Suburbia” which compares Vancouver’s West Coast Express peak-only commuter train with the Skytrain system on this issue. On average, each West Coast Express train has its seats filled around twice a day – one trip in during the morning and one trip out during the evening. In contrast, each Skytrain has every seat filled something like 50 times a day because they operate at high frequencies all day long, providing service for far more trips than just commuting ones. That makes the Skytrain system incredibly efficient to run, even if overall it may not seem as “crowded” during very peak times as the West Coast Express train.

What the book goes on to highlight is that if we are looking to improve the cost-effectiveness of our system, we shouldn’t focus on cutting services outside the peak (even if the buses or trains are fairly empty), but rather look at services during the peak that aren’t at maximum capacity:

…if you see a transit vehicle running nearly empty during the peak period and in the peak commute direction, you may be seeing some actual waste. Suppose that a commuter express bus, doing a long run from a distant suburb into the city, carries only fifteen people, less than half a seated load. If the transit agency is looking at its cost-effectiveness fairly, this situation should look much much worse than a bus with fifteen people at noon, running on an all-day, two-way line. For this one commuter express run, the agency has spend hundreds of thousands of dollars to buy the bus, thousands more for the land on which to store it, and thousands per year to maintain and operate it, including penalty rates for the driver who works a brutal schedule and is kept from spending time with family most days – all to serve only fifteen people!

In this previous post, I compared bus frequency across a number of central isthmus routes: 

Perhaps the most interesting observation of the above (along with Dominion Road’s excellent inter-peak frequencies) is how service levels drop away dramatically after 6pm, which encourages people to travel home during the evening peak. A thorough assessment of whether each and every one of the peak hour services is needed may end up freeing up resources to run a huge number of inter-peak and evening services – at the cost of just a few peak time buses.

The 15 minute frequency all the time of the Outer Link bus is a pretty extreme example of an ‘un-peaked’ service, but is a key element of the route’s success. We are able to afford to run a good frequency service around much of the inner isthmus, even on a Sunday night, because we don’t have the extremely high costs of additional peak services (aside from a couple, to ensure the route can keep to its timetable in peak hour traffic). While those Sunday night buses might generally be empty, over time the guaranteed frequency of the service means that people will learn to trust it and rely on it – influencing land-use patterns, decisions people make about where to live and decisions people make about how many cars to own. The Outer Link’s constant timetable and its regular frequency are all made possible by squeezing the peak a little bit harder (lower frequencies at peak time on some services in the area than there used to be), so we have achieved a really high-quality service at relatively low cost. I suspect though that some additional peak time services, for capacity reasons, will be required inevitably.

This is not to suggest that we should go cutting peak time services ‘willy nilly’, because – after all – the TriMet General Manager was correct in saying that serving the peak is the ‘bread and butter’ of public transport, to an extent. However, it’s worth keeping in mind the cost differences of running peak and off-peak services, so you don’t get quite so grumpy next time you see a predominantly empty off-peak bus, or have to squeeze onto a peak time bus. Even though such an outcome looks inefficient, looks can be deceiving.

Share this


  1. This does seem to have been the logic followed by ECan with the revitalisation of Christchurch bus services over the past twenty years. In addition to adding more off peak services the routes were reconfigured to provide a fairly standard length of about 4km of route with bus stops with outer suburbs having main road express service through the inner route zone allowing outer suburbs to have almost the same service frequency as inner suburbs using only a few more buses than with the older all-stops system popular with the cost-minimising mentality of the 1970s and ’80s.

  2. TriMet is a LRT system subject to the same constraints as a bus system in terms of operation while Skytrain is a fully automated system. The beauty of such systems is their operational flexibility. You can run the as many trains as you want at any time of the day with a small supervision staff in the control room. The additional cost to run extra service is essentially power and wear and tear of the system. Offering attractive off-peak headway will certainly facilitate modal shift as it improves the overall quality of service from the user point of view. However automation implies a fully segregated right of way which makes them more expensive to procure and thus better suited to fulfill medium to large transportation demand.

Leave a Reply

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