Just after Melbourne’s trams were privatised in 1999, some employees of the new private operator Yarra Trams suggested that tram stops in Melbourne were too close together, and that removing tram stops would help speed up people’s tram journeys. A review of all stops on Yarra Trams’ routes was announced, aimed at removing those that were ‘surplus’ according to the operator’s criteria. Press reports suggested that up to 50 per cent of all tram stops might disappear, while Yarra Trams produced glossy newsletters pushing the idea that fewer stops were good for passengers.

After the PTUA declared this the best strategy yet invented to drive passengers away from the tram system, the government intervened and stopped the cull of tram stops going ahead. However, the idea has refused to die. Careful observers have noted that the stops on the new tram extension from Mont Albert to Box Hill are around 40 per cent further apart than those on the old route just to the west – despite there being no difference in urban form or density between the old and new sections of the route. And Yarra Trams, in a reprise of its abortive 1999 exercise, recently pushed through the removal of one-third of the tram stops in parts of Collins Street, Flinders Street, Bourke Street and Victoria Parade.

Increasing the spacing between stops can help the operator run its vehicles faster – though only marginally, if at all – but it also makes passengers walk further. So all it does at best is procure an advantage to the operator at the expense of (some of) its passengers. A much better way to speed up trams is to give them proper priority at intersections and segregate them from car traffic. Higher frequencies can also help reduce delays due to overcrowding.

In Melbourne, tram stop deletions have in fact been implemented so poorly that even the potential advantage to the operator is nonexistent. Some of the first stops to be removed were in the ‘Paris end’ of Collins Street. Prior to the changes, Collins Street trams potentially had to stop in five locations between Spring and Swanston Streets, four of which were tram stops. Today, the same trams potentially have to stop in seven locations, three of which are tram stops! While deleting tram stops is generally a bad idea, it is a particularly foolish idea to do so when trams still have to stop for long periods at traffic lights.

Nonetheless, arguments are still found why removing tram stops might be to passengers’ overall benefit. Commentators point to cities that are building new ‘light rail’ systems with stops 400 metres or more apart, and note that there are tram routes in Melbourne where the stops are just 200 metres apart. Others suggest that when you trade off the ‘small’ inconvenience to passengers who have to walk further against the benefit to other passengers who get a faster ride, the overall result is good for the ‘average’ passenger.

One argument at least can be dismissed outright: that certain tram stops should be eliminated because few people appear to use them. If it’s really true that few people use a particular tram stop, then the tram should rarely have to stop there anyway. But in any case, public transport is currently only used in Melbourne for about 8 per cent of trips. This means that every tram stop in Melbourne is underutilised. If run according to world’s best practice, one can expect public transport to be used for around 25 to 30 per cent of trips in a city like Melbourne. The number of people using a tram stop currently is no guide to the number that would use the stop in such a future scenario – and removing a stop deprives us of much of the future patronage growth that might come from that location.

A belief that removing tram stops is a good idea can also arise from inadequate analysis of actual passengers’ travel habits. If tram stops are 400 metres apart, this does not mean the maximum walking distance is 200 metres. Removing a tram stop in Collins Street does not just affect people with Collins Street destinations; it affects anyone who uses Collins Street trams, including those whose final destinations happen to be in Flinders or Bourke Street. Consider Sue, whose destination is the corner of Queen and Bourke Streets, but who uses the Collins Street tram because she lives near that route. If the Queen Street stop is removed, Sue’s walking distance at the city end increases from 200 metres to 400 metres. If Sue already has to walk this far from home to the tram stop in Victoria Parade, she may well consider driving the car into the city instead. (People like Sue are also worse off if tram stops are moved away from intersections into mid-block locations.)

Sue’s case illustrates the fallacy of reasoning based on the ‘average’ passenger. The median walking distance is 100 metres sounds more acceptable than Half of all potential passengers must walk further than 100 metres, but the meaning is exactly the same.

Melburnians are not a homogeneous bunch of average people; we have specific needs and require transport systems with the flexibility to meet them. This always involves a trade-off between convenience for some passengers and improved travel times for others, but if planners focus on the latter to the complete exclusion of the former (as our privatised system leads them to do), the result can be worse for everyone.

In reality, the optimum spacing of tram stops is a complicated trade-off involving many factors, such as the local urban geography, the speed of the average walker, and the characteristics of the vehicle itself. (This is discussed further in the Appendix.) Even within Melbourne, the spacing of stops varies from one point to another on a route. This reflects the wisdom of tramway planners in the early 20th century who took the above factors into account when locating tram stops. Some of those early planning measures did not survive the postwar decline in tram use: for example, until the 1960s trams would often stop on both sides of large intersections to relieve their passengers of the need to cross the road themselves. Since more people used trams then as use trams now, it seems clear that having ‘too many’ tram stops is no deterrent to passengers!

As the Appendix shows, in central Melbourne the optimal spacing of tram stops is about 200 metres. By happy coincidence, this is also the spacing of the major intersections in the CBD street grid. So the current pattern of one tram stop at each intersection is just about right, and to tamper with it involves a net cost to passengers.

One can of course always argue that people are better off having to walk more, and that removing stops will benefit people by improving their level of fitness. This might be true, were it not for the fact that most people have the option of driving cars. Anyone could, if they wished, improve their fitness by parking their car a block away from home instead of in their own driveway, but of course people don’t actually do this. Similarly, public transport users told to walk further for their own good are as likely as not to abandon public transport and drive into the city instead – particularly if they are elderly or have a disability. But of course those who need most encouragement to use public transport are those who currently find it more convenient to drive their cars, and removing tram stops only makes them less likely to switch.

Technical Appendix: The Optimal Spacing of Tram Stops

A common argument for removing tram stops can be paraphrased thus: Consider a tram carrying 60 passengers, that stops to pick up 6 passengers at a stop just 200 metres from the next one. This (it’s alleged) adds one minute to the travel time for each of the 60 passengers, for 60 minutes total. If the stop were eliminated, then the 6 passengers would have to walk up to 200 metres further. The average person takes 2.5 minutes to walk this distance, so the time cost is 6 times 2.5, or 15 minutes. Conclusion: removing the stop has a benefit-cost ratio of 4 to 1.

This argument may sound convincing, but ignores some important facts about the way people value their time. One minute in one context isn’t necessarily worth as much as one minute in another context. In particular, studies by transport economists have consistently shown that walking time is perceived as having a ‘cost’ two to three times greater than time spent in the vehicle. Making people walk further is not ‘cheap’, and requires a greater operational time saving to be justifiable.

Meanwhile, the passengers who have to walk further will be weighing up the inconvenience against the alternative of driving a car instead, with the result that passengers will inevitably be lost to the system. It is less clear that people will be attracted to public transport by the promise of slightly faster running, than that people will be driven away by having to walk further to catch the tram.

The ‘one minute’ added to the travel time by making the extra stop also needs closer scrutiny. Let us take the argument at face value and suppose that the 6 passengers whose tram stop is removed henceforth walk to the next stop, instead of getting in their cars. Then they will add to the time spent at the next stop by the time required to board 6 extra passengers. If one genuinely does not plan on reduced patronage, the only time saved by removing a stop is the time required to slow to a stop, open and close the doors, and accelerate back to full speed, compared with the time taken when running at full speed throughout.

Now we need a little high-school physics. Suppose a tram slows uniformly from its cruise speed V to a full stop, then immediately accelerates uniformly back to speed V. Let D be the distance covered in the process of decelerating and accelerating (so that deceleration starts at distance D/2 in advance of the stop, and acceleration finishes at distance D/2 past the stop). Then the time taken is exactly twice the time required to cover that distance D at the cruise speed V. (This is most easily seen by sketching the two speed-time graphs and making the areas under the graphs equal.)

It follows that if one measures the time taken by a tram to slow to a stop and speed up again, starting and finishing at the same speed V, only half this time is saved by eliminating the stop. (The other half is still required to cover the distance at full speed.)

Next time you’re on a tram, try counting up the seconds when it makes a stop. We believe you’ll arrive at something like the following:

• 10 seconds to slow from full speed to standstill;
• 1 or 2 seconds to open the doors;
• 5 seconds to close the doors (including the warning tone on the newer trams); and
• 10 seconds to accelerate back to full speed.

These figures give a time loss from stopping equal to about 17 seconds (half of 10+10, plus 2, plus 5). This is much less than a full minute, and when combined with the high cost of walking time is enough to reverse the result above. The time lost to 60 passengers from making the stop is 60 x 17 / 60 = 17 minutes; with the stop removed, the time penalty to the 6 passengers is equivalent to at least 15 x 2 = 30 minutes of in-vehicle time. Suddenly it doesn’t look nearly as attractive to remove the stop any more.

So what is the ideal spacing of stops on a tram route? This is going to depend not only on the factors we’ve just discussed, but also on the way the population is distributed along the route, and the average distance people travel.

Here we’ll cover the simplest case: a single route of length L, with the stops more or less evenly spaced at some distance D, and a uniform population distribution along the route. Let T_s denote the time loss from making a stop; we estimated this as 17 seconds above.

The scenario we are interested in is that where public transport has been improved to world’s best practice, and is therefore used for a healthy proportion of local travel as well as for trips to the city centre. It is therefore reasonable to assume that the length of an (unlinked) trip on this tram varies more or less uniformly between some minimum length L0 and the route length L.

For a single trip of length X, the total travel time is the sum of walking time, waiting time and in-vehicle time. We assume waiting time is both short in duration and roughly the same for all passengers, and can thus be disregarded. The walking time we take as

where D is the stop spacing, and T0 and K are constants. This means that part of the walking time is independent of the spacing between stops (as when the trip origin is a couple of streets away from the route) and the rest is proportional to the spacing, doubling when stops are moved twice as far apart and vice versa. Note that K has the dimensions of velocity: it is roughly equivalent to a person’s average walking speed.

The number of stops on a trip of length X is around X/D, and so the time we expect to lose due to stopping is (X/D) T_s. Proper tram priority means there is very little ‘dead time’ spent waiting for traffic lights. Taking the cruising speed as V, and weighting the walk time by a factor W (around 2 or 3) gives the equivalent travel time ‘cost’ for the trip as

Now take the average over all trips, using the assumption that X varies uniformly between L0 and L. This gives the average total travel time as

This travel time is a function of D, the mean spacing between stops. The ‘optimal’ mean spacing is that which minimises T_av (the ‘aggregate cost’ in economists’ terms). With a little calculus this is shown to be

Note that this result is independent of the cruise speed V (except indirectly through T_s) and the fixed part T_0 of the average walking time. We can simplify the formula by noting that

• D_j = (L + L_0)/2 is the average length of a journey; and
• D_s = K T_s / W is the distance a ‘typical’ passenger can walk in T_s seconds, divided by the weighting factor W.

So we can regard the optimal spacing as the ‘geometric mean’ of two distances:

(The geometric mean of two numbers A and B is G = sqrt(AB); if A and B are the two sides of a rectangle, G is the side length of a square with the same area.)

From the above formula, we can deduce the factors that influence the optimal spacing:

• If the route caters mainly for long-distance trips, then stops should be further apart than on routes catering for short trips. (This helps explain why it’s a good idea for railway stations to be further apart than tram stops.)
• Routes on which vehicles have good acceleration and braking can afford to have more closely-spaced stops than routes with poorer vehicles, as the former lose less time in stopping (within reason!).
• The speed and convenience with which people can walk needs to be taken into account. If infirmity, baggage, street crossings or other factors regularly slow people down, this will affect their requirements of tram services.

In other words, you can’t plan tram services on the assumption that all passengers are young, able-bodied people who always travel from one end of the route to the other and never have heavy loads to carry.

So what’s the bottom line? Let’s take Route 96 between the GPO and East Brunswick as an example. It’s around 6km long, so a reasonable estimate for the average length of a trip is 4km. (This is also the average trip length found in the Metplan studies of the 1980s.) T_s is around 17 seconds, as explained above. At a typical walking speed of 5kph, a person can cover 24 metres in 17 seconds; taking W = 2 we obtain D_s = 12 metres. Then we get

which is close to what the stop spacing is right now on Route 96. On Route 109 in Balwyn, the spacing is closer to 300 metres, which is roughly the optimum given the longer journeys typically made on that route.

Finally, there is at least anecdotal evidence that stopping to pick up and set down passengers does not make a huge difference to running times for trams.

On Monday 24 June 2002 during the morning peak, the author of this piece got to the tram stop at the corner of Nicholson Street and Glenlyon Road, East Brunswick, just in time to miss a tram (‘Tram A’). Approximately 5 minutes later, another tram (‘Tram B’) pulled up, with a third tram (‘Tram C’) immediately behind.

Because Tram C was right behind, the driver of Tram B decided to run express through most of the stops on the way to the city: it in fact made no more than 5 stops out of 15 between Glenlyon Road and Parliament station. Since it nonetheless failed to catch up with Tram A, which had a five minute head start but stopped at every stop, one may conclude that cutting out two-thirds of stops saves at most 3 to 4 minutes on a 20 minute tram journey – and most of that saving is due to not having as many passenger boardings.

It would therefore appear that even if one could remove, say, 50 per cent of tram stops while retaining the same patronage level, the time saving would be marginal. There are much better ways to improve travel time.

Last modified: 22 September 2009