Dwelling on dwell-times — Estimating the economic benefits of speeding up Auckland’s trains


If you’ve been reading TransportBlog for a while, then you may have noticed that the term “dwell-times” crops up relatively frequently. The term describes the average time that trains are stopped at stations. In several previous posts, we’ve discussed how average dwell-times on Auckland’s new electric trains are approximately 50-60 seconds per stop. In contrast, best-practice dwell-times on rail systems overseas are in the order of 20-30 seconds per stop, so 20 – 40 seconds faster than ours.

In this recent post, I suggested something “had to be done” to shorten dwell-times, to which one commenter (quite reasonably) asked “why“. I was surprised by their question, because to me the benefits of reducing travel-times by 30 seconds per stop seemed obvious. It means that we take Auckland’s trains out of the steam age and into the electric age.

Upon reflection, however, I realised that it wasn’t immediately obvious how apparently small delays of 30 seconds per stop would incur economic costs that warranted action. This realisation motivated me to write this post, in which I attempt to estimate some of the economic benefits of shorter dwell-times on Auckland’s train.

Note that this post does not consider how we might go about reducing dwell-times, and I’d like to ask people who comment not to de-rail the post with such technical matters. The point I’m trying to convey, and which I am interested in discussing, is the economic benefits that flow from running trains faster. Similar arguments apply to efforts to close small intermediate stations, such as Westfield and Te Mahia.

When I think about reducing dwell-times, four obvious sources of economic benefits spring to mind: 1) Reduced costs; 2) Existing user benefits; 3) New User Benefits; and 4) Decongestion benefits. Let’s now dive into the data (hand-waving?) ….

Cost savings: The total cost of running Auckland’s rail services is likely to be in the order of $125 million p.a. Various factors contribute to these costs, but major componenets are likely to be:

  • Staff, including drivers and train managers;
  • Operating costs, such as fuel, mainenance, and track access charges;
  • Asset depreciation, especially on the rolling stock, electrified infrastructure, and depots; and
  • Station operations, such as ticketing and security services.

Different places account for these costs in different ways, which can end up making things quite complex. To simplify things, it’s common to analyse public transport costs in terms of three underlying drivers: 1) vehicle-kilometres; 2) vehicle-hours; and 3) peak fleet requirements. These cost drivers are, more-or-less and to varying degrees, what determine the costs of the services that we operate. Shorter dwell-times won’t, of course, affect the distance that trains travel, even if they will reduce vehicle-hours and/or peak fleet requirements.

To estimate the cost savings from shorter dwell-times, I assumed that measures are taken to reduce dwell-times by 30 seconds per station. If I assume there are 15 stations on the average rail run from Swanson/Papakura/Manukau to the City, then this saves 7.5 minutes on every 55 minute run, or ~14%. I then assume that one-third of this 14% reduction in (in-service) vehicle-hours flows through to the operating cost bottom-line, that is, shorter dwell-times reduce total costs by 4.5%.

Given a total annual spend of $125 million p.a. on operating rail services in Auckland, normal practice dwell-times would reduce costs by $5.68 million p.a. If I then assume an 8% discount rate over 30 years, then this has an NPV of $69.1 million.

That’s the first example of how an apparently small number can lead to a large number when you take a network-wide, long-run perspective …

Existing user benefits: The second effect of shorter dwell-times is to expedite journeys by rail. That us, existing rail users will also benefit from faster travel-times. Current rail patronage is about 18 million journeys p.a., which is predicted to grow to 40 million in a post-CRL world. For the sake of simplicity, let us assume that Auckland averages 30 million rail passengers per year over the next 30 years.

Moreover, I now assume these passengers travel an average of 15km per rail journey. If I assume rail stations are spaced, on average, at one station per 3km, then this implies there are 4 intermediate stations per journey (NB: Remember that users will not benefit from faster dwell-times for the last station). Saving 30 seconds over 4 stations equates to a time saving of 2 minutes per journey. If I assume a value-of-time of $10 per hour, then we can monetize the value of these time savings as follows: (2/60) hours per journey x $10 per hour x 30 million journeys per year = $10 million p.a. This has an NPV of $121.6 million. Happy train users are valuable train users.

New User Benefits: Faster trains will, of course, also attract new users. Let’s assume the elasticty of demand with respect to in-vehicle travel time is -0.50. That is, a doubling in travel-time leads to a 50% decline in patronage. In turn, this means that the 14% reduction in travel-time associated from shorter dwell-times would be associated with a 7.0% increase in patronage, or an additional 2.1 million journeys p.a. If I now apply the rule-of-half, that is 2.1 million x (2/60) x $10 per hour x 0.5 = $0.35 million p.a. Or $4.3 million over 30 years.

Decongestion benefits: Some of the additional rail journeys undertaken as a consequence of faster travel-times would have otherwise been loaded onto the road network. This in turn means congestion would be higher. If I assume that 60% of new rail journeys occur in peak periods, and that 25% of these journeys would have otherwise been placed on the road network, then this suggests faster dwell-times diverts 315,000 vehicles off congested roads every year. If we again assume the average rail journey is 15km long, and that decongestion benefits are valued at $0.40 per kilometre, then I find that decongestion benefits are valued at $1.8 million p.a., or $23.0 million over a thirty year period.


Let’s summarize the estimated economic benefits of faster dwell-times:

  • Cost savings of $69.1 million
  • Existing user benefits of $121.6 million
  • New user benefits of $4.3 million
  • Decongestion benefits of $23.0 million.

Yielding a total of $217.9 million. I must acknowledge these are extremely rough and ready estimates and I could well be wrong and/or out by a decent margin of say +-50%. So don’t anyone go quoting me to the decimal point.

Notwithstanding their approximate nature, the order of magnitude of the estimate is significant. To put it in context, an economic benefit of $217.9 million is equivalent to approximately $7.3 million for every second we cut from average dwell-times. Another way to look at it: I understand the total cost of Auckland’s new trains was in the order of $500 million. Cutting 30 seconds per stop is then worth approximately half of the cost of buying completely new trains. This seems plausible to me; I suspect that a large component of the anticipated benefits of electrification would stem from faster journey times, which have not as yet materialized (NB: In some ways you could argue it’s impressive Auckland has achieved so much patronage growth despite the slow travel-times.

Anyway, the main point is to demonstrate how small time savings can quickly add-up to large dollar values when you take a network-wide, long-run perspective. This is why we harp on about dwell-times and why it’s heartening to see Council putting pressure on AT to sort this mess out. Indeed, cost savings of $69 million above would go straight to Council’s bottom-line, as would approximately $100 million in additional fare revenue over 30 years (NB: This is not an economic benefit per se at it’s simply a transfer from passengers to Council — even if it is of course a fiscal benefit). These fiscal cost savings would result in either lower rates and/or improved services.

In our increasingly resource-constrained world, I consider frugality to be not just prudent, but indeed relatively noble. This is especially true when it comes to other people’s time and money. On this topic, Benjamin Franklin, once said “The way to wealth … depends chiefly on two words, industry and frugality: that is, waste neither time nor money, but make the best use of both.” Confucius put it even more succintly when he said “he who will not economize will have to agonize”.

Or, if you prefer the words of Bruce Lee.

On that note, I suspect my marginal utility of spending more time on this post is by now lower than my next best alternative option.

Have a good ‘un.

Guest Post: EMUs Not Much Faster Than Steam

This is a guest post by Robert Finley, a qualified Civil Engineer with a lifelong interest in trains and their operation in both New Zealand and overseas. He has photographed steam trains on most of New Zealand’s rail network in his travels.

So you think Auckland’s new trains are a bit slow and perhaps they’re not all that they’re cracked up to be?  Well in some ways you’re right since some services are slower than some of the steam trains of 50 years ago!

A direct comparison of the timings of the new Electric Multiple Units (EMU) with the timings of steam hauled trains in the early 1960s reveals some interesting facts. The comparison was made with a train service that was often hauled by a steam locomotive;  scheduled to depart from Papatoetoe at 8.05 am and arrive at the old Auckland station platforms (where spare EMUs are now parked)  at 8.42 having traversed what is now called the Eastern Line through Panmure.

Steam train leaving Papatoetoe station. Photo Robert Finley

Steam Train Service

This service – for the purists, train No 112 – actually originated from Mercer, stopping at all stations. It was therefore considered a long distance train thus requiring steam heating especially in winter and thus the necessity of steam locomotive haulage. Diesel locos were also used and, while the writer does not have many details of the diesel performance, those that exist are noticeably slower than steam.

Steam locomotives used were normally the 10 year old JA class with the occasional older J class. The train consist was usually 4-5 steel panelled carriages with a guard’s van, although sometimes timber sided cars were used. This service was very popular as it enabled people to reach city offices by 9 am using the very efficient trolley bus service that was waiting for every train. The number of passengers on this steam train service would be similar to that on recent EMU runs.

The writer was a regular traveller on this service and recorded, over about 30 different runs, the actual start and stop times at every station.

However to enable direct comparison with current trains some timing adjustments are necessary due to the changed configuration of the current network:-

  • Two stations do not now exist, Mangere and Tamaki. Thus it is necessary to deduct from the overall time not only the actual time stopped here, but also the associated acceleration and deceleration (acc/dec) required;
  • Sylvia Park and Panmure stations have been relocated but no adjustment to overall timings are required
  • Timings for the EMU’s arrival into Auckland are taken as if the EMU had stopped opposite the old platforms, now known as the Strand station, as well as at Britomart itself.


EMU Trains

As the writer is not now a regular user of these services,  7 different runs were made in the latter part of 2015, generally in the mornings.  Although a smaller sample, the results were surprisingly consistent and are therefore considered sufficient  be statistically significant.

The average overall  time from Papatoetoe station to the old Strand station platforms for the EMU trains was 29m 50s and the average running time was 23m 11s. The total stopped time for 8 stops was 6m 39s, or an average of 50s.

Summary Comparison

Comparison has been made of both the overall time and the actual running time from departing Papatoetoe to arrival at The Strand.

#   measured as average of actual times

##   some services scheduled at 31 min and some at 33 min

The average adjusted running time (i.e. the time the train was in motion) for the steam service was 27m 32s compared to the current day EMU service of 23m 11s – a saving of 4m 21s . But the average adjusted overall  time for the steam service was 30m 41s and for the EMU service was 29m 50s – just 51 seconds faster.

So the average overall EMU time is 29m 50s and the fastest steam time was 29m 06s – 44s faster! In fact, about  ¼ of the steam runs bettered that average EMU time. Furthermore, the fastest steam run was even quicker than the currently scheduled EMU time, adjusted for the Britomart-Strand timing.

Side By Side Comparison

Two typical runs are compared side by side in the following table.

Dwell Time

It should be very clear that the reason for the lacklustre performance of our EMU fleet lies almost entirely with the dwell times at stations. The average time that the EMU trains are stationary is more than double that of steam hauled trains!

But there are also permanent and temporary speed restrictions, slows and general delays caused by signals, track, points condition and other effects – I counted at least 5 such restrictions on one run.

In steam hauled trains, passengers had to climb up steps, probably gripping a filthy handrail, possibly even having to open the gate on the car platform and almost certainly opening the door into the car itself. Nevertheless, some total stopped times of just 7-12 secs at some stations were often recorded. This, of course, required a considerable degree of alacrity on the part of the entire train crew to ensure that passengers were hustled on and off and the train restarted rapidly, ensuring the utmost efficiency in order to meet or improve on the timetable. On one famous occasion, a steam run left Papatoetoe 10 minutes late and arrived in Newmarket on time!

In an EMU, you just walk straight in without touching anything except perhaps the door button. But passengers are to blame for some of the delay too – they just meander along the platform towards any old door instead of getting into the nearest one, likewise when alighting they just stroll to the door. These people would just get left behind in London!  A massive education program for passengers must be implemented. Incidentally, the same lackadaisical approach has a massive effect on the running of the whole AT bus system.

Analysis shows that on average it takes an average of 10.3s for the doors to fully open after a train has stopped. There is  5 secs after the actual stop before the green button light even comes on and assuming someone pushes the button immediately, there is a further 5 secs before doors are fully open.

It then takes an average of 22.5 secs from when the main doors close before the train moves. The Train Manager checks that everyone is on or off and closes the main doors. He then gazes around at the scenery and then closes his door, and the train still doesn’t move.

So there is at least 33 secs of dead time at every single stop.

But not only is the overall service slow, the EMU’s can not even meet their own timetable. For the 7 (off peak) runs the average schedule lateness departing Papatoetoe was 120 secs. Not one single run was able to recover this and in fact all lost further time to arrive at Britomart an average of 3min 19 sec late.  If I had been  wanting to transfer to a Penrose train – due to leave 2 mins after scheduled arrival from the Eastern line I would have missed it on every occasion.  Does the small army of clipboard wielders at Britomart monitor this situation? If not, what do they do?

Incidentally, some recent  check runs indicate that there has been no change in these basic operating parameters

In the same era as these steam services were running, there was also a train that left Auckland at 5.25 pm and ran via Panmure  non-stop to Otahuhu and thence to Papakura. It, too, was frequently hauled by a steam (tank) loco and provided a schedule of just 19 mins compared with the current EMU service of about 24 mins.

Why does AT not run some non or limited stop services at peak times? Other railway operations can manage this. With the huge amount of money spent on upgrading the whole metropolitan signalling system I cannot believe that the train control system is incapable of this.

How do Other Systems Perform?

Is this typical of railway operations elsewhere?  In Brisbane, the door light comes on when you press the button as you approach the station and the doors open immediately the train stops. The average time from the doors closing to the train moving is 4 secs.

In Wellington it averages 7 secs and if it took any longer on the London Underground, the whole city would grind to a halt! And in Mumbai they don’t even bother to close the doors at all – its not certain that they even have any doors!

Adelaide, on the other hand, is a different kettle of fish. There the time lapse from doors closing is so short one would need a very accurate system to measure it – it is almost instantaneous. The Adelaide EMU’s are of almost identical configuration and age to the Auckland EMU’s. They cover  36 km from Adelaide to Seaford in 52 minutes with 22 intermediate stops. Some peak services make just 5 stops with a time of 38 minutes.

Trains from Britomart to Papakura cover 31.5 km  in 53 minutes with just 14 intermediate stops.

A measure of performance might be ‘distance x no of stops over time’.  Thus Adelaide’s performance factor is 15.2 while Auckland’s is 8.3.  Go figure!

Further, on some Adelaide trains where no passengers board or alight, the total dwell time at a station can be as short as 8 secs. There is no conductor, guard or train manager apart from the occasional ticket inspectors. The driver is in complete control, he has cameras and platform mirrors at his disposal to monitor the progress of passengers and leaves as soon as he is ready. Furthermore, if a wheelchair passenger needs to board, they wait at the marked spot on the platform which correlates to the front door of the train, and the driver gets out of his compartment and places the ramp onto the platform! There are also no security staff on trains or at stations.


The new Auckland trains are quiet, smart, smooth, comfortable (if you don’t have long legs!), clean and brisk in starting and stopping.

But there is also clearly something drastically wrong with either  the specification, design or construction of the EMU’s or, most likely, Auckland Transport’s operation of them. Either AT does not recognise it has a problem or, if it does, is unable or unwilling to rectify it.

So, with respect to the extended door closing and opening durations it seems most likely this feature is a result of AT’s specification and operation of the trains.  What is the reason for it, given that no other metro transit system appears to have it?

Is there some other ulterior motive for this feature?

And, finally, how is AT proposing to rectify this appalling situation? This must be rectified for optimum performance of these otherwise very efficient EMU’s.

Has anyone from AT ever been to other metro operations to see how they are run? Is anyone at AT aware of just how bad the performance of these new trains really is? If AT cannot even organise any of these matters  it does not bode well for managing the added complexities and options once the City Rail Loop is built.

The new trains are little faster overall than not only the diesel trains they replaced, but also the steam trains that preceded them! So come on Auckland Transport, how about you see if you can get your fancy new trains to get into town faster than a steam train!

Increasing capacity on our trains

Ridership on Auckland’s trains has experienced fantastic growth over the last few years, increasingly a staggering 70% in just three years to 17.3 million trips as of the end of September. The positive thing is that the growth remains strong at nearly 20% per annum. Growth has been so impressive that it is ahead of what was projected to occur with electrification, despite the electric trains not coming into service around two years later than expected in the business case.


The good news is the forecast for the future looks bright. The new bus network that’s just been rolled out in South Auckland and over the next 18 months will be rolled out to the rest of the urban area. That, combined with the recently implemented Simplified Fares will likely see a lot of people transferring to trains as part of their journey and should continue to help drive ridership over the next few years. No one knows just what the maximum capacity of the rail network will be prior to the City Rail Link but previous modelling has suggested somewhere in the 20-25 million range.

But there are definitely some clouds on the horizon in the form of capacity constraints. Already we get reports of services at, or close to capacity during peak times and as patronage rises that will only get worse.

Packed Train

How will trains cope with another 5 million+ trips a year on them?


Ultimately we need more trains, both to increase the length of our existing trains so more run as six-car sets and to run more services. But there are two main issues with this.

  • Even if Auckland Transport placed an order for more trains today, it will take at least two years before we see them on the tracks and carrying passengers. Based on current trends we don’t have that long.
  • To run additional services it requires the completion of the City Rail Link (and other associated infrastructure like signalling improvements). That isn’t expected to be finished until 2022/23 and until that happens, peak frequencies can’t change from what they are now. Note: Because the CRL makes services more efficient, AT have said in the past they think they have enough trains to run services once the CRL opens. They’re absolutely kidding themselves if they think that’s true.

So here are a couple of thoughts on what we can do to address this and ensure there is still enough capacity to enable ridership to grow.

Speed up the trains we have

Our trains are so stupidly slow it feels George R.R. Martin writes books faster. The sad fact is that right now, the timetable with brand new electric trains is slower than it was with the clunky old diesels. Through a combination of archaic operating process and poor technology they are often both slow between stops, especially approaching stations, and have incredibly long dwell times at stations. My perception on some of this may be clouded by being on the Western Line which often has a level crossing right next to the station.

We know that over the last year or two, numerous signalling, track and train improvements have been made to enable faster, more reliable services but so far those improvements haven’t been reflected in the timetable and sometimes feels like services are being kept slow just to avoid getting ahead of schedule.

The good news is that some improvements are coming. AT say a new timetable is due in March which should finally capitalise on the improvements mentioned above. The benefit of that is it will free up a few more three-car trains to allow more to run as six-car ones. Here’s what AT say about journey times with the new timetables:

Run times on the Southern and Eastern Lines with electric trains will be shorter than previous pre-electronic train control system (ETCS) signalling with diesel trains, with equivalent times on the Western Line due to large number of rail level and pedestrian crossings and speed limits at these crossings.

But nothing appears to be being done about the dwell times which can often exceeding 50 seconds even for minor stops. This was a video I shot while in Tokyo of one of the metro lines showing a dwell time of just 20 seconds from the time we stopped till we were underway again. Admittedly not all trains were this fast but this is what AT should be aiming for.

One process used in Japan that we could fairly easily adopt is to have the train manager in the rear driving cab of train checking the doors/platform are clear rather than the process we have of the TM closing all but one door, checking they’re closed then closing his door before the train can leave. At stations on curves giving poor visibility, screens linked to CCTV cameras can assist TM and we already have this at a couple of stations, such as Fruitvale Rd station. Even a few seconds saved at each station can add up significantly.

Saving time not only frees up more trains to boost capacity but also helps make trains more attractive in general.

Order more trains now

This one seems kind of obvious but given the two-year lead time for trains, it’s essential we order more as soon as possible. We know there’s already been talk within AT about the need for this and it also appears they’re also looking at tying it in with decisions around how to deal with Pukekohe. One solution talked about is buying up to about 10 battery powered trains to enable electric services to be extended without the high cost of extending the wires. That in turn would free up some units to allow more services to run as six-car trains. The key issue though is funding and who pays for it as these trains don’t come cheap. Each train costs something like $10 million to build so we’d probably need at least $100 million now and more after the CRL.

Spread the peak

Currently AT consider the peak to be 7am-9am and 4pm-6pm but even within those times there is a ‘peak of the peak’ and it is at those times where capacity is most constrained. AT should be looking at how they can encourage more people to travel at different times, both for addressing capacity constraints but also for attracting more people to use services. The first step will be to improve off peak services so they run at a minimum of every 15 minutes all day but AT’s own statutorily required Regional Public Transport Plan calls for even better with services at a minimum of every 10 minutes on the three major lines.

RPTP rail frequencies

We’re hoping the previously mentioned March timetable change will see improvement on this matter.

Another important thing they could do to encourage off peak travel would be to implement off peak fares, offering a discount for travelling at times when the network isn’t as busy. Of course, not everyone can shift when travel but some can and the added benefit of this is it could attract more people to use PT.

Reconfigure the trains we have

Of all the options at AT’s disposal, perhaps the fastest, cheapest and easiest might be to reconfigure the trains. Our trains are currently configured to focus on longer distance trips by squeezing as many seats in them as possible. Each three-car train has around 230 seats and of which just over 2/3rds are set to face forwards of backwards with the remaining, including the low floor section of the middle car, set up as metro style sideways seating. The trains are also said to carry 143 people standing although at a squeeze I’m sure they’ve carried more.

There’s a good reason so many PT systems use sideways seating and that’s because it enables a lot more people to fit in each vehicle and while I was in Japan I noticed even most medium distance trains travelling further than Pukekohe were configured exclusively this way. It also doesn’t lose all that many seats because sideways seating takes up some of the space otherwise needed for leg room. As a comparison, the low floor section of the middle car has 16 seats on each side, although half of them can fold up to enable bikes, prams or wheelchairs to more easily fit in, while the centre section of the two end cars has 20 seats per side.

Initially converting just the centre sections of two end cars in each train would lose just 16 seats from a train but gain a huge amount of additional space for people to stand. I also understand the seats are designed to be easily changed. If you catch a train you may notice the current directional seating is cantilevered off the walls. This means there’s are no poles to move or marks to be left on the floor and changing the seats is simply a case of changing a bracket to turn the seats around. It would probably also have the added advantage of stopping vandals from scratching the backs of seats. If needed, we could do the same with the rest of the seats on the trains.

EMU Interior June 2

This is an old image, the trains have vertical poles between each pair of seats for those standing to hold

I realise not everyone likes the sideways seats but I’m sure most would rather be able to get on a train them be so busy they’re forced to wait for the next one. We should probably consider doing this on some buses too, the red Citylink would be a prime candidate.