How high productivity can wreck a bus route

Last Sunday I wrote about our ten most productive bus routes - that is those that had a high number of boardings per service hour. They were basically concentrated around universities and in the City of Wyndham (which had a major bus network revamp a few years ago).

Productive buses are great. Their usage means that people are finding them useful. They bolster fare revenue while being cheap to run. If you're a little bus company who runs without government subsidy (like most were up to the 1970s) this is just what you want. 

However is there such a thing as bus routes that are too productive for their own good?

I think there is. 

Especially if you want public transport to do all the good things it can only do by maximising its patronage. Like taking cars off the road, free road space for those who need it, lessen parking pressures, support denser retail and jobs in our town centres, enable more space-effective cities, lower overall community expenditure on transport and make our economy less carbon-intensive.  

Certain routes have incredibly high boardings per hour figures on weekends. These are the 900 (over 80 boardings/hour) and the 733 (around 60 boardings/hour). Both serve Monash University Clayton, with 900 roughly east-west from Caulfield and 733 north-south from Box Hill. When you get to very high numbers the service deteriorates, with delays due to boarding/alighting and, in extreme cases, passengers being unable to board.   

Why is productivity so high? Both routes combine very favourable passenger demographics with low service levels. For instance 900 is half-hourly while 733 is hourly on a Sunday. 

You could quadruple 900 to every 7.5 minutes and triple 733 to every 20 minutes on a Sunday and still meet Infrastructure Victoria's 20 boardings per hour productivity threshold. That's guaranteed - ie it assumes that the vastly better frequency will not attract a single extra passenger. 

In the 'real world' things behave differently, especially if the route serves a fertile catchment. The better the service, the better the patronage. Though it's not necessarily a 1:1 relationship. For example you might only get a 40% increase in patronage for a 100% increase in service. 

Although there have been times when we've beaten that. For instance in the few years after serious local bus upgrades started in 2006 we got roughly a 25% increase in patronage for a similar increase in service. And that's largely just with service upgrades, not with network reform that can be more cost-effective. 

Getting back to the 900 Sunday example, if we quadruple service from every 30 to every 7.5 minutes, we increase it by 300%. Even if elasticity is quite low (33%) we will have doubled patronage (100% increase). The productivity will have dropped from over 80 boardings per hour to a still impressive over 40 boardings per hour. There will be some knock-on gains flowing to intersecting routes (as trips involving transfers to the 900 will be easier) along with achieving all the other things associated with high patronage mentioned before.  

Let's say you were to keep increasing service. You will reach a point where frequency is already so good that big improvements won't attract more passengers. Eg the difference between every 3 minutes and every 4 minutes would probably not be noticed. But the 3 minute service requires 33% more buses to run than the 4 minute service. That's hard to justify unless crowding requires the capacity the higher frequency provides.

When frequency gets stupidly high your boardings per hour might fall to make the route unproductive. Unless you can find some other means to justify this (eg wider benefit of everyone arriving by bus and not driving) you would back off, run articulated buses if required for capacity and deploy the drivers on other routes that need a frequency boost. Eventually you'd upgrade your original route with a bus lane, busway or light rail, with the operational benefits of higher productivity.

The graph below might explain this better:

The left scale is productivity in boardings per bus hour (red line). The pink area is below Infrastructure Victoria's 20 boardings per hour while the green area is above it. Our network average is a little higher (faint line).

The leftmost third of the graph has the red line (ie productivity) increasing with service. This will vary with route. Some routes are extremely low service / low productivity. Take for instance the 745 that has four variations with one trip each. That has just 8 passenger boardings per hour on school days and 4 on school holidays. Adding more trips to at least one of these variants (with other bus network reform) would almost certainly increase productivity to somewhere nearer average for buses. With more services, more productivity (but still below average) are routes like 531 and 538 serving residential Campbellfield. Reforms like consolidating these routes into one but lengthening hours and adding 7 day service would likely increase productivity. Straight operating hours upgrades to one or both of these  unreformed routes is beneficial for a low income area and will result in added usage. However the increase won't necessarily be enough to increase productivity although social grounds may exist for these routes to be upgraded. 

Some areas have routes with low frequency but very high productivity. Examples include hourly routes like the 675 in Mooroolbark, the 774 in Frankston South or the 814 in Springvale. The red line on the chart above for those routes would start higher, in the 'Extremely viable' area rather than the 'not viable' area. However service upgrades are still justified for as long as productivity remains above average. This is particularly the type of upgrades that work existing buses harder, eg extended hours and days of operation. Arguably routes like these are amongst the top priority to upgrade since there is already a known and large passenger base who we know would benefit, along with others their neighbours who might not have found the bus useful until it was upgraded.

The right scale is patronage (blue line). Across the bottom is frequency. That's simplified; a more precise measure could be service quantity which also accounts for operating hours (refer to  the Currie/Delbosc ATRF paper as discussed here). However this has its own complications given that separate graphs are desirable for day type (eg weekday, Saturday, Sunday). 

Especially in areas where the catchment is favourable for buses and/or not many people own cars productivity peaks well before patronage. Routes 733 and 900 on weekends are extreme examples with very high productivity despite their infrequent service. Even with modest patronage elasticity you can vastly boost frequency, still have a highly productive route and lift average network performance. But only to a point, as mentioned before. 

Each route will have a different curve shape. And the frequency scale doesn't apply everywhere.  University shuttles (eg 301, 401, 601) combine extremely high productivity with extremely high frequency. In contrast routes in sparsely populated areas may not be viable at any frequency but are run due to a policy decision to retain coverage. And some might not attract many more passengers even with doubled service, meaning a sharp fall in productivity to below viable levels.

Fortunately this is hardly a worry for most of our most productive 40, 50 or even 100 routes that are likely to respond favourably to a service upgrade, especially if efficiency-enhancing network reforms are done simultaneously. If we do those we might be able to increase frequencies on selected routes with the existing bus fleet and a less than proportional increase in driver hours required. 

Below is an attempt to cluster different types of routes according to their productivity and patronage. 

Cluster A: Very productive routes with high occupancy per bus but less overall patronage and service. For example they might not be very frequent and finish at 9pm or earlier. Their catchment is favourable for bus usage and is likely to respond well to service upgrades. 

Weekday examples with at least 40 boardings per hour include 150, 151, 160, 167, 170, 180, 190, 192, 201, 237, 270, 279, 302, 406, 410, 423, 424, 494, 495, 497, 508, 529, 533, 536, 537, 570, 630, 703, 733, 737, 813, 814 and 893. 

Some routes are more crowded (ie productive) on weekends than weekdays. This is because they might drop to half or even quarter frequency even though demand for travel does not drop in the areas they serve (eg major shopping centres). Such routes may qualify for Cluster A on weekends but not weekdays where service is better matched to demand. 

Productive routes with at least 40 boardings per hour on Saturday and/or Sunday include 150, 279, 302, 508, 623, 630, 631, 733, 767, 800, 900, 903 and 907. It is likely that busy segments of longer routes like 216/219, 220, 901 and 902 would also qualify.  

Cluster B: These routes rank amongst Melbourne's best used. They have better service than Cluster A. While productivity might be slightly lower than Cluster A (especially for the more frequent examples) it is still above average.  

Examples include busier ex-Met corridors eg 200/207, 234, 246 and 250/251 which typically have 7-day service until midnight and better frequency than routes in Category A. You would also include weekday services on SmartBus routes. The SmartBus orbitals are a mixed bag as all three have low-productivity segments. 

This is the 'sweet spot'. At least on weekdays all of Melbourne's most used bus routes are in this cluster. Their quality of service is the best on the network and their productivity is above average. You need to bring as many routes as possible into this group if you wish to maximise network patronage and the number of passengers carried with your vehicle fleet. 

Cluster C: We don't actually have any of these in Melbourne, at least with the frequency scale shown. Our two super-frequent routes (401, 601) are university shuttles and their productivity is excellent. Our stronger uni shuttles (301, 401, 601) are really Cluster Bs but at a higher frequency point whereas the low frequency Deakin Uni 201 and 768 are Cluster A ('could do better'). 

If you were to adjust the frequency scale to reflect what would be considered extraordinary for a sparsely populated area, then you would find some Cluster C examples. For example 582 and 695 whose 20 to 30 minute frequencies appear very high for their catchments. The same applies for parts of the 901 and 903 orbitals that provide weekday 15 minute frequencies and service until midnight to some low density industrial, residential and rural areas that don't need it. Efficient planning would transfer service from Cluster C routes to others that need it more.  

Getting from A to B (but not C)

Very high productivity in a bus route isn't everything. 

By not operating routes with favourable catchments as frequently as they should be we are denying ourselves millions of passenger boardings per year and many hours of productive revenue service.

We need to efficiently get our productive high patronage potential routes from A to B by reforming networks and boosting service. If a route is already operating at 40 or more boardings per hour it is possibly under-serviced. 

Just mind the curves above. 

Melbourne on Transit bookshop

Favourably reviewed books about transport and cities. Purchases via these links support this blog and its independent reporting (at no extra cost to you).