Thursday, July 20, 2023

Most productive bus routes: 2022 edition

Public transport patronage is a good thing. The more people use a bus route the more the community and individual benefits. Monitoring patronage is key to planning networks and services. Especially if you want to optimise usage per amount of service provided (something that's regular practice in Perth but rarer here). Right now there is also interest in how usage patterns have changed post-lockdown as some habits developed during the pandemic, such as white collar working from home, seem to be proving enduring. 

Back in 2019 I looked at the previous year's bus boardings per hour data for weekdays and weekends with a focus on the most productive routes. The most productive weekday routes were either those serving universities or in suburbs with diverse high-bus using populations like Tarneit, Truganina, Point Cook and Craigieburn. On weekends routes to major shopping centres like Chadstone, Box Hill and Doncaster did best. While high productivity is a good problem to have, it can also mean that we are skimping on service and thus forgoing patronage growth, as discussed in detail here.   

Earlier this year I requested and received similar bus productivity data for September - November 2022 from DTP. More on that later. But first, let's discuss the 'elephant in the room'..

Data accuracy

I know what everyone's asking: "How do they know how many people are using the bus when so many people aren't touching on?".

It's a good question. Some say that buses in Melbourne risk becoming like trams were about 20 years ago with regards to fare evasion's prevalence and even social acceptability. Official interest in encouraging compliance is perhaps not what it was, especially after buses stopped doing myki cash top-ups during the pandemic and lower patronage meant less foregone revenue (compared to if everyone touched on) anyway. Also mass rail shutdowns and bus replacements got people used to boarding buses without touching on, which may have weakened reflex behaviours when riding regular buses.

While it wasn't in force in the 2022 period we're looking at, a popular view of proportionality or fairness is that a short trip should cost much less than travel across the state. Hence Labor's now implemented $10 statewide daily fare cap (a 2022 election sweetener) risks undermining the legitimacy of paying for short trips in some minds. Something to keep an eye on as we compare the differences between good policy and 'race to the bottom' election politics. 

Graphs published in the DoT Network Revenue Protection Plan 2022 (p6) show that myki's introduction coincided with improved fare compliance for Metro trains and trams, but not for buses, which spent long periods as the mode with the highest proportion of non-payers. A pre-pandemic recovery in compliance proved short-lived, with the last recorded (late 2021) compliance for bus falling from 96% to 90.6%, the lowest of all modes.

In transport's nonchalant head-office world a non-compliance rate of still less than 10% probably weakens the potential short-run returns from the costs of greater enforcement to the extent that they are willing to forgo some compliance. And 90% would seem pretty good to some. And they know that not everyone who doesn't touch on is necessarily evading a fare, especially if they've just come off a train via a station with barriers. 

However in the 'real world' bus riders on some routes see far more than 10% of passengers not touching on. That's significant as touching on is a learned behaviour that you want everyone to develop into a reflex habit.

When people see others not validating this encourages copycat behaviour that lessens long-run compliance and, important for today's discussion, data accuracy. There are also cases where passengers who wish to touch on cannot due to defective or booting up myki readers at the start of a run.   

How do people not touching on affect the ridership statistics that I'm about to discuss? As it turns out DTP is well aware of people not touching on. Thus they use survey data to boost numbers to account for non touch-ons and improve data quality.

However, given variations in the scale of non-compliance (and thus any data compensation factors), you should still prepare for inaccuracies when reading what follows. The disclaimer with the supplied data warns users of this, saying that 'this methodology provides estimates at the macro level'. 

Hence my use and interpretation of the data will be cautious. And if I think parts are wrong I'll tell you.  

Enough of the caveats. What do the numbers indicate?

Most productive weekday routes in 2022

The top ten are as follows (most productive first): 

601 University shuttle: Huntingdale - Monash
733 Box Hill - Monash - Clayton - Oakleigh
182 Werribee - Tarneit (new route)
703 Brighton - Clayton - Monash - Blackburn
401 University shuttle: North Melbourne - Melbourne University
152 Williams Landing - Tarneit (new route)
486 Sunbury - Rolling Meadows
170 Werribee - Werribee Plaza - Tarneit
907 City - Doncaster - Mitcham
495 Williams Landing - Point Cook South

Like with the 2018 numbers, routes serving universities continue to get high boardings per hour. These include 401, 601, 703 and 733.

Also, like last time, a similar number of top performing routes were in the City of Wyndham (152, 170, 182 and 495). Of special note is that 152 and 182 are new routes that did not exist before 2021. This shows that in the City of Wyndham you can throw just about any bus service out there and people will flock to it, even if it's only every 40 minutes off-peak. The high usage means that these routes would easily justify a 20 or even 10 minute service with longer operating hours. The April 2022 Craigieburn upgrades are a good precedent for this. 

Only two non-university or non-Wyndham routes appears on the list. One of them I can believe but the other I can't. The 907 is a popular route on Doncaster Rd. Other routes overlap parts of it but the 907 is direct, has long operating hours and is more consistently frequent 7 days per week. This makes it close to being a rubber tyred tram with high travel speeds due to its freeway running. Thus it's no surprise that people use it.

Route 486, on the other hand, is a Sunbury local route through an area with fairly low population density and little demographic propensity to use non-school public transport. And it's never been busy on the few times I've ridden it. Maybe the number is right but I'm still surprised that this and some other Sunbury routes appear so highly placed. 

Have any of these routes seen recent service increases to reflect their high productivity? Some have.  152 and 182 are new routes, starting only in 2021. 733 gained frequency upgrades on its busiest portion in May, with funding from the 2022 state budget. And 2021 saw improvements to 907 including higher weekend frequency and 24 hour weekend service. 

601 (the Monash shuttle) recorded over 100 boardings per hour. The rest of the top 10 had between 39 and 49 boardings per hour. However there's some important still productive routes in the next ten that sit just below that with weekday boardings in the 33 to 39 per hour range. These include: 

900 Caulfield - Rowville
481 Sunbury - Mt Lion 
160 Hoppers Crossing - Tarneit
497 Williams Landing - Saltwater Coast
485 Sunbury - Wilson Lane
623 St Kilda - Glen Waverley
737 Monash University - Croydon
246 Elsternwick - Clifton Hill
150 Williams Landing - Tarneit
800 Chadstone - Dandenong

Two of the 'second ten' routes (900 and 737) serve Monash University. Three (150, 160, 497) are in the City of Wyndham. This time big shopping centres come in with three (623, 800 and 900) serving Chadstone and 737 serving The Glen. 246 on Punt Rd is a popular frequent inner city route that has been the main subject of the 'rapid running' trial. 

The remaining two (481 and 485) are short local Sunbury routes whose figures surprised me as noted before. Also route 800 stands out as being, apart from the university shuttles, the only top 20 route without Sunday service and only a skeleton Saturday service including two hour gaps in the afternoon. Indeed you would have to go down to almost the 50th most productive route to find another with such limited service. This makes the 800 an outlier and a demonstration that in Melbourne you can have strong usage but wait decades for timetables to be adjusted accordingly.  

Weekdays - the bigger picture

There are exceptions but bus route numbers in Melbourne are roughly clustered by area, and in some cases, operator history. Hence under 200s are mostly Wyndham area, 200s inner ex Met routes, 300s outer north, 400s west, 500s north, 600s and 700s east, 800s outer south-east and 900s SmartBus. I plotted productivity in route number order below. 

The productive clusters in Wyndham and around Monash University stick out. As does the cluster around Sunbury whose high numbers I questioned before. I'm also wary about some of the known high usage routes like Sunshine's 216 and 220 whose productivity I'd expect would have been much higher.

Genuinely quieter routes are found in parts of Melbourne's outer east and outer north, eg around Lilydale, Belgrave and Mernda whose population density and demographics are weaker for buses than areas around Sunshine (labelled 'West') and Dandenong.

Route 900 - 908 SmartBuses are a mixed bag. 900 and 907 have no really quiet sections, with 907's freeway running being offset by high productivity along almost the whole route. 901, 902 and 903, the very long orbital SmartBuses, have some very busy sections but quieter segments stop them being prominent in route level productivity (as opposed to raw patronage) data.  

How has average productivity changed in the last few years? Weekday buses in Melbourne averaged about 22 boardings per hour pre-pandemic (median measure). The median in 2022 is lower at around 15 boardings per hour. I might discuss this at a later time but will be tempering any conclusions due to concerns over data quality arising from counting or non-validation factors.   

Most productive 10 Saturday routes in 2022

Many familiar route numbers productive on weekdays are also productive on weekends. But there's a couple of new entries. 

900 Caulfield - Rowville
623 St Kilda - Glen Waverley
733 Box Hill - Monash - Clayton - Oakleigh
626 Middle Brighton - Chadstone
279 Box Hill - Doncaster
508 Moonee Ponds - Alphington
907 City - Doncaster Rd - Mitcham
703 Brighton - Clayton - Monash - Blackburn
246 Elsternwick - Clifton Hill
800 Chadstone - Dandenong

Weekend usage appears to be driven by the big shopping centres. No less than 7 of the 10 listed routes serve either Chadstone, Box Hill or Doncaster centres. Four (703, 733, 800, 900) go to or near Monash Clayton, indicating a strong bus using demographic in the area. Inner suburbs across the north and east also find buses useful for cross-suburban trips with the 246 and 508 featuring. 

Service levels remain an issue, with more variation on Saturday than weekdays. For example routes like 246 and 907 enjoy a 15 minute frequency during the day whereas 623 and 626 only run hourly. 800, as discussed before has the least service, with two hour gaps in the afternoon and a finish before 4pm from Chadstone. Routes that have recently had weekend frequency upgrades include 279 and 907 (2021) and 733 (2023). All of the Saturday top 10 operate Sundays except for Route 800 (discussed before). 

Sitting below the top 10 include routes like 631, 767, 742, 506, 495, 170, 402, 182, 903 and 624. Chadstone remains a drawcard, being on routes 624, 742, 767 and 903 with two of these also serving The Glen. Both Wyndham (170 and 495) and the inner north (402 and 506) are represented. So is the Monash area and Southland with the 631. 

Route 767 had a frequency upgrade in May following 2022 budget funding. Route 506 in Brunswick is a lot like the 800 in that it's a busy but neglected route with no Sunday service. It ranks 14th on Saturday. As I'll show later, the most productive bus routes on Sundays are closely correlated with those that are productive on Saturdays. Thus if 506 were to get Sunday service it would likely attract more boardings per hour than about 90% of routes that currently operate on Sundays.  

It's also worth mentioning the 903 which appears in the top 20 but not the top 10. This is a long orbital route with busy and quiet segments. If only the eastern and possibly northern segment was considered it would likely rank in the top 10. Similar comments would apply for the 902 (just out of the to 20), which can be standing room only around Glen Waverley and Springvale along with parts of the 901.  The combination of high usage and low frequency particularly afflicts these orbital routes on weekends. 

The need to boost weekend bus timetables is well understood in senior circles. Including Infrastructure Victoria's Director of Policy and Research and Minister Ben Carroll who recently made the following comments in relation to buses in Ballarat (but also relevant state-wide).    

Most productive 10 Sunday routes in 2022

Mostly like Saturday's list but with some differences, possibly due to potential data quality issues on some routes (*). More later.  

900 Caulfield - Rowville
508 Moonee Ponds - Alphington*
623 St Kilda - Glen Waverley
733 Box Hill - Monash - Clayton - Oakleigh
626 Middle Brighton - Chadstone
555 Epping - Northland*
703 Brighton - Clayton - Monash - Blackburn
556 Epping - Northland*
631 Southland - Waverley Gardens
630 Elwood - Monash University

High productivity is again generally associated with large shopping centres and/or the Monash precinct. Service levels on all routes are generally low. Each route listed runs every 40 to 60 minutes except for the 703 and 900, which run every 30 minutes.

279 and 907 remain productive but drop out of the top 10. So does the 800, but this doesn't get a chance to be ranked because it doesn't run Sundays.

Sunday versus Saturday productivity

A worthwhile graph I did last time was to compare Sunday with Saturday productivity. In a nutshell, the most productive routes on a Saturday were also highly productive on Sunday. Especially where Saturday productivity was high but Sunday productivity was even higher it could identify routes that were underserved on Sunday. The exercise could also be a 'sense check' of data quality to identify data anomalies that are worth questioning before drawing too many conclusions. 

Here is it repeated with the 2022 numbers (click for a clearer view): 

Like last time there remains a high (though apparently less) correlation between Sunday productivity productivity and that on a Saturday. You do need to be aware that service levels vary - some routes operate similar frequencies on Sundays as they do Saturdays while others might be only one-third as frequent (eg every 60 min vs every 20 min). Hence a route can be more productive on a Sunday than a Saturday (ie well above the diagonal line) but ridership remains less due to fewer trips run. This is particularly in safe Labor low car ownership areas where people will crowd onto buses despite unattractive service levels. 

Very productive routes

Routes towards the top right are very productive on both days and are underserved. Route 900 is the stand-out example but the 733, 623 and 626 also show clearly. The latter three all used to run hourly but the 733 got upgraded to operate every 40 minutes in May. 

279 and 907 are a fair distance below the 1:1 line, though Sunday productivity remains at a healthy number in (say) the top quarter of routes. This may be explainable by the 2021 service upgrades which delivered improved frequencies on both. This will likely have attracted some increased usage.  However the survey period was barely a year after the improved services were introduced and short-run elasticity is typically well under 1 (eg 0.5). The result is a bit less productivity, even though it remains above most other routes and overall usage may have grown. 

Routes without Sunday service

Another interesting cluster hugs the x-axis. These are routes that have zero Sunday boardings due to the absence of service. The higher productivity Saturday examples are labelled. They almost always either serve big shopping centres like Highpoint, Northland, Chadstone, Box Hill, Doncaster or high needs, high patronage underserved areas like Greater Dandenong.

It hardly goes without saying that Route 800 on Princes Hwy leads this pack, with it ranking 10th out of over 300 metropolitan routes for Saturday productivity. If you were to prioritise routes to add Sunday service to you'd start with the labelled routes along with others that either don't show due to  lacking Saturday service (eg 802) or are not labelled but have strong social equity cases for a 7 day upgrade (eg 536 and 538 around Broadmeadows). 

Well above the line routes  

The biggest change relative to 2018 is the number of routes well above the diagonal line.

Routes prone to this pattern are those where Sunday service collapses relative to Saturday and which have low income/low car ownership catchments. Prime examples are found in the Footscray / Sunshine / St Albans area involving routes like 406, 408 and 410. These may run every 15-20 minutes on weekdays and Saturdays but collapse to every 40 - 60 minutes on Sundays despite strong all-week demand. The patronage result of this is high crowding on the few Sunday buses that do run.  

However I don't think this is the full story. My judgement is that the shape of the graphed line has changed too much given that bus service levels (and for that matter populations) haven't changed a lot. So another factor may be at play, including ..

Potential data anomalies

The graph showed a similar (if not greater) tendency for a cluster of routes in Melbourne's north to also be above the line. 508 is the stand-out but they also include 555, 556 and numerous others too close to label. At first glance something big appeared to be happening when I compared the 2022 graph with the one I made for the 2018 data.

Unlike the Footscray/Sunshine area routes, whose Sunday service falls off a cliff relative to weekdays and Saturdays, most routes in Melbourne's north had a gentler and sometimes almost no Saturday - Sunday service fall off. For example Wollert and Mernda area routes are typically every 40 minutes off-peak weekdays, Saturdays and Sundays, with operating hours slightly longer on Saturday compared to Sunday.   

My expectation was that if a bus route has a similar number of trips on a Sunday to a Saturday then usage should be fairly similar across the two days, with Sundays a little quieter. But if data shows Sundays dramatically busier than Saturdays then I would query it.

I found 40 routes that had 1.4 times or more the Sunday productivity than they did on Saturdays, mostly in Melbourne's north. This seems extraordinary as Melbourne's traditional working class middle and outer northern suburbs tend to be quieter on Sundays relative to other days than gentrifying inner or lifestyle-oriented bayside or hills suburbs.

Particularly notable was Route 356 in Wollert which had 26 boardings per hour on Sundays versus 13 on Saturday, or a 2:1 difference. Service levels on both days were similar, at every 40 minutes. Other examples were only slightly less extreme. The vast majority (32) of the routes in that top 40 were operated by Dysons. Hence it is possible that there may be an issue with data from particular operators.

Dysons also run the Route 508 discussed above which also appears as an outlier well above the line. However even if it was on the line its high usage would be enough for it to be considered as underserved on weekends, including Sundays. Like 508, data for 555 and 556 indicated high enough Sunday productivity for them to qualify as top 10 routes on a Sunday (despite them being far lower ranked on Saturday). These two (along with others like the 567) are also Dysons, so I'd query data on these too.

Conclusion and implications

Bus usage in late 2022 was still lower than it was pre-pandemic. However of all public transport modes buses are the least exposed to white collar CBD workers, many of whom have continued to work at home for at least some of the week.  

But broad patterns of the productivity data remain unchanged. That is very strong weekday productivity for university and Wyndham area routes. Plus those serving major shopping centres, especially on weekends. 

While I've flagged potential data quality issues for some routes, the patterns are still enduring and robust enough to be able to make a service upgrade priority list with a high confidence of maximising 'bang for buck'. In other words we pretty much know what needs to be done and the costs of doing so are modest where off-peak upgrades are concerned. Thus the main shortfall is not data but political will. 

Commendably some routes that would make a productivity-based priority list (eg 279, 733, 767, 900 & 907) have had some modest service upgrades in the last two years. However attention to many more routes are needed. Notable examples include: 

(i) Better frequency and operating hours for key bus routes in Wyndham
(ii) 7 day service, higher frequency and longer hours for popular routes in Greater Dandenong 
(iii) 7 day service, longer hours and weekend frequency boosts for key shopping centre routes
(iv) 7 day service, longer hours and weekend frequency boosts for inner to middle ring low car ownership areas eg Sunshine, Footscray, Brunswick, Northcote, Box Hill and Clayton. 
Examples of a lot of these are in the Top 40 bus service stocktake with a low-budget incremental approach discussed here

Thanks to DTP for the productivity data that made this analysis possible. 


Damo said...

The routes you are flagging with data anomalies on Sundays need to be considered in a wider network context. The 508 for example is flanked by routes that don't run on Sundays, so naturally the 508 will absorb much of the east-west traffic across the lower northern suburbs.

Also Sunbury routes will be increasing in productivity because the town is growing but the services have not kept up. There are many estates popping up, particularly to the east and south, which explains (at least to some level) why the 486 has risen up the ranks.

Peter Parker said...

Thanks Damo, I did think that could be an issue with the 508 but I compared 2022 with 2019 data (that I didn't write about much here). Saturday productivity of 508 in 2022 was slightly less than 2019 (both in the 43-49 range). But Sundays jumped from 47 to 72. Given that the lack of Sunday service on surrounding routes was also an issue in 2019 and Sunday use closely tracks Saturday use, I don't think any big jump on Sundays only could be anything but a data anomaly.

As for Sunbury I looked back to both 2018 and 2017 data and numbers were high then too. So you might be right there. But I still find it surprising that little routes in middle income / high car owning Sunbury can have higher productivity than the likes of 216 and 220 through Footscray. Though maybe it's more that the Footscray routes were undercounted.

David said...

It would be interesting to see the relationship between weekly patronage and net weekly services; does the data you have allow that chart to be generated in a few minutes, or would you need to manually count services making it an impractical time investment?

Peter Parker said...

@David - This item was about productivity, not patronage. I might look at patronage at another time. Productivity has the advantage of factoring in route length and number of services whereas patronage does not.

Not only might weekly services influence patronage but also the length of the route, eg the SmartBus orbitals have high patronage due to length but are not the network's most productive (although sections of them might be). Hence I really like productivity stats, though they do have the effect of counting a short trip the same as a long distance one.

David said...

Hmm, what's the correlation between weekly patronage and route productivity? Gut instinct is that it should be fairly high, r² of perhaps 0.70 to 0.80? But that leaves open the question of what other factors account for the remainder of the correlation.

Might also be interesting to study productivity vs route length, weekly net services, route span, etc.

Peter Parker said...

@David - If all routes were the same length then there'd be a high correlation between weekly patronage and productivity. But they're not. Hence the very long orbitals have only average productivity but top the patronage chart. And the Wyndham routes are short but despite excellent productivity don't stand out on patronage levels.

Malcolm M said...

DoT don't seem good at advertising increased bus frequency. You gave the example of the 279 and 907 that had recent increases to weekend frequency that had not yet resulted in a corresponding patronage increase. Information is usually restricted to a new timetable on the website and a Ministerial press release. I believe there was previously a program called TravelSmart that had their own internal skills in promoting such services. Each change needs its own communication program, to minimise the lag between imcreased frequency and increased patronage. Methods to look at include
- posters at the stations the route passes, announcing the frequency increase
- direct mail drop in the area serviced
- free travel for the first month, prominantly displaying the date for when touch-on is required
- at the end of a free travel period staff the buses with compliance officers to get users used to touching on again

I believe we should also include a half fare for trips of under 5km, because our PT is quite expensive for short trips

Heihachi_73 said...

The minimum service standard for buses really needs to be upped from its current hourly-until-9PM to half-hourly until 11PM in line with trains and trams.

Sunday and public holiday timetables should have been abolished decades ago (e.g. the 2000s) and replaced with a universal weekend timetable which matches weekdays but lacks the additional peak services. Sunday timetables are anachronistic and are not compatible with a post-1980s Melbourne.

SmartBuses are failing in patronage because the timetables are not keeping up with the times. They are no longer premium services, just mediocre at best. The orbitals in particular haven't seen any real increase in services since day one, with half-hourly services in the mornings, evenings and all weekends and public holidays. Completely unusable outside of working hours.

The 907, which received an upgrade, is still half-hourly at night and still misses the Lilydale/Belgrave train at Mitcham by all of three minutes, forcing a 27-minute wait every single time. Worse still during Night Network, where the wait becomes 57 minutes.