In this post, I test out a number of possible explanations for the trend in Melbourne’s public transport (PT) patronage growth over recent years to try to find out what might be driving growth. Is it population growth, CBD employment, fuel prices, international students, or the widening of CityLink? You’ll have to read on.

The first chart shows estimated financial year public transport patronage in Melbourne. Note: The method of patronage estimation has changed over the years for all modes. I have assumed comparable measurement for trains and trams and applied my own educated adjustments to bus patronage history (although I am less confident about the early 1990s – officially patronage stayed much the same despite significant service cuts).

Patronage was bumpy in the 1990s, followed by modest growth for about 10 years and then a distinct uptick in growth around 2004/05.

I will attempt to find an explanation for this pattern in this analysis (particularly more recent years). Short of a fully comprehensive analysis, I will compare trends in possible drivers with the trend in public transport patronage.

Note due to the nature of available data sources, the years covered in chart will vary – you can spot each year by checking the year range in the chart titles.

Population growth

(note I have estimated 2010-11 population growth at 1.6% based on the trend in state population) If this was a dominant factor then you’d expect to see a straight line on this chart. It does show that as population growth has increased, so has public transport patronage growth, but the overall relationship isn’t very linear.

We know that public transport use is higher closer to the inner city of Melbourne. So is public transport better correlated with inner city population? The following chart compared PT patronage with “inner” population (LGAs of Melbourne, Port Phillip, Stonnington, Yarra, Hobsons Bay, Maribyrnong, Moonee Valley, Moreland, Darebin, Banyule, Boroondara, and Glen Eira).

The correlation appears to be slightly stronger.

Employment

People often use PT to get to work. The next chart compares total employed people in Melbourne to public transport patronage (employment figures average monthly total employed people for each financial year, from ABS Labour Force surveys).

Again, the relationship isn’t very linear – despite a small growth in employed persons in 2008-09, public transport patronage still increased significantly. But then in 2009-10, employed persons grew but patronage didn’t. Likewise PT patronage increased more between 2000/01 and 2001/02, despite little growth in total employment, whereas in the previous year employment grew strongly, but PT patronage didn’t.

This chart also shows kinks in the trend around 2005 and in 2008-09 – so employment doesn’t seem to explain the kink. Note also that journeys to work only make up around 40% of public transport trips in Melbourne (according to VISTA data). And public transport has a very low mode share of journeys to work outside the city centre.

ABS publish figures monthly, and here’s the picture for total persons employed in Melbourne. There was virtually no growth between late 2010 and the end of 2011. There was also a flat patch between the start of 2008 and the middle of 2009 (2008-09 shows substantial patronage growth on public transport).

City population (including visitors)

Another hypothesis suggests that if PT is heavily focussed on the CBD (where it has the highest destination mode share), then if more people need to travel to the CBD, this would probably increase PT patronage. This sounds very plausible, especially for trains and trams. The City of Melbourne has estimated weekday daytime population for 2004 to 2010 calendar years. So I am mixing calendar year visitor data with financial year PT patronage – which is not ideal. Anyway, here is what that relationship looks like:

The year 2005/06 includes the 2006 Commonwealth Games that were held in March 2006 and boosted city visitors considerably. If you take out this anomaly, the other four data points look like they form a very linear pattern (as drawn), suggesting it is quite probably a strong driver. There was weak growth in both public transport patronage and city population in 2009-10, suggesting a strong relationship.

A longer time series of CBD visitation data is available for CBD employment, thanks to the City of Melbourne’s Census of Land Use and Employment. As it hasn’t been an annual survey (red dots are census results), I have made linear interpolations between the years for CBD employment numbers. Unfortunately 2010 figures were still not available at the time of updating this post (February 2012). 

Apart from the early 1990s, the trend looks remarkably linear suggesting a strong relationship. And when CBD employment grew very weakly between 2002 and 2004, so did PT patronage. Looking at census data for 2001 and 2006, we know that PT mode share to the Melbourne CBD for journeys to work (well, technically the inner Melbourne SLA which is much the same) grew only slightly from 59.1% to 60.8%. So it looks fairly safe to assume that the growth in people using PT to get to jobs in the CBD grew at much the same pace as CBD employment itself.

While the CLUE data series only runs until 2010 at present, a more timely and regular dataset that might be related to CBD employment is occupied office floor space, calculated from the Property Council of Australia’s Office Market Reports. While I do not have access to the reports themselves, much of the data is available on the internet in various forms, and I have used that data to reconstruct the data series (there is chance of errors creeping in, particularly for earlier years).

Here is the trend in occupied Melbourne CBD office space:

Slow growth until about 2005, then very strong growth. Does that trend sound familiar?

This charts shows very strong correlation (r-squared = 0.99). Although there are still a few small kinks such as 2009-10.

But the overall strong relationship this confirms the high likelihood of CBD employment being a very significant driver of public transport patronage.

Fuel prices

I have taken the monthly average unleaded fuel prices for Melbourne, adjusted for CPI, and then averaged the months for each financial year, to produce the following chart:

Fuel prices are highly volatile, even on an annual basis. Again, even though fuel prices dropped in 2008-09, PT patronage still increased. There seems to be a lot more at work than fuel prices. That said, since 2004-05, real fuel prices jumped from around 115 cents to over 130 cents and have remained higher since. So fuel might be an explanation for the kick up in PT patronage since 2005, perhaps more as the breaking of a psychological price barrier. Or perhaps people’s responses to fuel prices have longer lag times that wash out short-term fluctuations - as people make major decisions – such as the decision to purchase a new car or not. More on that later.

International students

Another hypothesis is that the recent boom in international student numbers drove public transport patronage, as many international students come from countries where public transport is the “default” mode. And while their finances might stretch to studying in Australia, it might not stretch to owning a car (certainly in the car ownership maps we see low car ownership around many universities).

Unfortunately I’ve only found complete data for financial year 2002/03 onwards, and only at the state level (more detailed data is not freely available).

The boom in international students looks like it really took off in 2007, but fell away sharply in 2009-10 and has been lower since. In 2009-10 patronage grew more slowly, perhaps reflecting the drop in international student numbers. But 2010-11 patronage growth was strong again, despite little growth in international student numbers.

The international student numbers are very small in comparison to the total patronage. However if half of those students averaged 10 trips per week for say 40 weeks a year (purely a guess), that’s 38 million trips. I’ve not got data on what their PT use is actually like (I suspect many live close to their school or university and actually walk). And their boom doesn’t coincide with the boom in public transport patronage that started around 2005. So they might be having an impact – hard to conclude much.

Road congestion

Until 2006-07 there was a fairly linear correlation, but then speeds only slowed slight while public transport patronage increased. In 2009-10 speeds increased and public transport patronage grew slowly. Perhaps congestion wasn’t a driver for patronage growth in 2009-10?

Another point to note is the scale on the X axis – the average speed hasn’t changed by very much. Although the variations in AM peak speeds for particular road segments are likely to have changed more significantly, I somewhat doubt whether the average driver would notice the difference between 35.8 km/h and 34.8 km/h (the change between 2005/06 and 2007/08).

The opening of CityLink in 2001 may have led to a slight increase in AM peak speeds, but this seems to have been quickly eroded the following year (so do new freeways ease congestion?). I’m not sure why traffic sped up in 2003/04, but then dropped again significantly the next year.

Road congestion impacts the majority of the tram network, and essentially all of the bus network. So perhaps only trains are attractive as an alternative to driving in congested traffic. Here’s same chart again but plotted only against train patronage:

The chart looks much the same. So congestion might be a driver of PT patronage growth, but it probably doesn’t explain the growth in tram and bus patronage, and the relationship isn’t nearly as linear as other factors.

Perhaps also at play here is congestion being relieved for non-radial commuting, where PT had a low market share beforehand anyway. Further research might look at congestion on CBD-radial roads only, though even then, many will also cater for some cross-city trips.

Two of the radial freeways that feed inner Melbourne are operated as the CityLink toll roads, and quarterly data is available on average daily transactions. If the CityLink toll roads compete with public transport it is probably mostly with trains for longer distance travel to the inner city. Here is a chart showing growth in CityLink transactions and train patronage:

There was very little train growth in the first few years of CityLink (which started in 2001). But then train patronage grew strongly from 2005 while CityLink transaction growth went flat until 2010. A major upgrade project on the eastern leg of CityLink (M1 upgrade) caused delays between 2007 and early 2010, and there was little traffic growth. After the project was largely completed and the fourth lane opened, traffic growth accelerated over 2010. This happened at much the same time that trains recorded weak patronage growth. Then in 2011, train patronage grew again, while traffic seems to have flattened again.

To take a closer look at the two growth rates, here are financial year growth rates on CityLink and trains:

After most of the works were completed, CityLink transaction growth exceeded train patronage growth in 2009-10 and 2010-11 (note that the flattening evident in the previous chart doesn’t show with annual data). The evidence suggests there could well be a relationship between freeway capacity and train patronage, and that the M1 widening project may have reduced patronage growth on the train network. It has certainly enabled a return to strong growth on CityLink.

Car ownership

People who don’t own cars are probably much more likely to use public transport. The following chart uses cars per 100 persons aged 20-74 (as a proxy for people of car driving age).

This chart shows in the early 2000s that car ownership rose quickly, while public transport patronage growth was slow. Then from 2006-07, car ownership levels peaked and public transport patronage grew quickly. Car ownership dropped in 2008-09 just as public transport patronage surged, but recovered in 2009-10, as public transport stalled. This suggests there may be some relationship between PT patronage and car ownership, but the annual change rates aren’t always consistent.

Service kms

Another potential driver of PT patronage is the amount of service provided. Thankfully, this data is available in Victorian State Budget papers (hidden away in budget paper 3) on the number of timetabled service kms for each mode. As the modes are quite different, I’ve plotted modal charts:

Train patronage doesn’t seem to be very strongly related to timetabled kms. Perhaps this is because the service levels at peak times on most lines are already attractive from a frequency point of view at least. Many of the extra train kms are providing capacity without a substantial jump in frequency (although some of the additional kms have been in off-peak periods).  That’s not to suggest there isn’t a relationship, just that it doesn’t look likely to be the dominant driver. In the early 2000s it seems that there wasn’t a strong response to increased timetable kms (including Sydenham electrification in 2002), while in the mid 2000s patronage grew despite kms staying much the same (other factors must be at work).

Again, not a strong relationship between tram kms and patronage, despite strong growth in timetabled kms in the early 2000s (partly from tram extensions into lower density suburbs in 2003 (Box Hill) and 2005 (Vermont South) – see here for more history). It also looks like some cuts in 2000 (when some city routes had to be joined due to the loss of W class trams) were done in a way that didn’t result in a loss of patronage. Perhaps because service frequencies were still fairly good after the cuts.

There does seem to be a stronger relationship between bus kms and patronage. This is perhaps to be expected as bus service levels are on average very low in Melbourne, so improved service levels are likely to result in existing users travelling more, and better attract new users.

What is unexpected is that patronage grew at much the same rate as kms between 2005-06 and 2009-10 – an average elasticity of around 1, which is much higher than you would normally expect. In 2010-11, the annual elasticity fell to 0.42. One possible explanation for the slightly steeper rate in recent years is that more of the new kms have come in the form of SmartBus kms (with higher frequencies). We know that long run implied service elasticities for SmartBus can be around 2 – which is higher than the textbook expectation of service elasticities of up to 1 in the long run. Bus upgrades in the early 2000s were a little more focussed on providing new low-frequency services to the urban fringe, which would be unlikely to lead to as much patronage growth.

Comparing annual growth/change rates

The following table shows the annual change in Melbourne public transport patronage and a number of potential explanatory factors. I’ve used conditional formatting such that darker green cells indicate values you might expect to contribute to strong PT patronage growth. Rows that have dark green in the same years as PT patronage are potentially stronger at explaining the trends in public transport patronage. I’ve also included the r-squared value for a correlation for each factor compared to PT patronage (based on annual growth values, not actual values). You might need to click to enlarge and make it easier to read.

The table confirms a strong relationship with CBD employment, City of Melbourne visitors (2006 removed due to Commonwealth Games anomaly), international student enrolments, population (particularly inner city), and CityLink volumes.

There is an unexpected positive correlation with car ownership which doesn’t really pass the logic test – other factors must be at play and note the very small changes in car ownership (the table uses growth in cars per population aged 20-74).

Fuel prices don’t show a strong relationship, although it is hard to believe that they would have no impact. If you offset the fuel price changes by one year the correlation rises to 0.3 so there might be some lag involved.

Conclusions

Based on these simple charts, I surmise that City of Melbourne (LGA) visitations is likely to be one of the strongest drivers of overall PT patronage in Melbourne (but certainly not the only driver). And it certainly stands to reason, given PT’s dominant mode share of travel to the CBD.

But international students, radial freeway volumes, population are probably also having an impact. The impact of fuel prices appears to be more complex. Buses probably show less response to growth the inner city travel market (as most do not serve the city centre), so service kms are likely to be the strongest driver of bus patronage.

The PCA’s Office Market Report provides the most timely and frequent data relating to CBD employment growth and reveals much slower growth over calendar 2011 (1.4% in occupied office floor space). We might find this trend reflected in slower patronage growth on the train network as  figures are published.

Patronage is growing strongly in Perth, Melbourne and Auckland. South East Queensland has possibly stalled, although issues in patronage estimation methodology may be masking underlying growth. Adelaide and Wellington are showing more modest growth, while Sydney, Canberra and Hobart continue to be laggards. Christchurch patronage unfortunately collapsed following the terrible earthquake.

For all the details, as well as some 110 year trends in Australian public transport patronage, read the updated post.

In other news, this blog is now available at the easier to remember address chartingtransport.com and there is a new public twitter stream if you want to track updates that way.

In August 2011, BITRE published revised estimates of vehicle kms in Australia (Report 124), derived from fuel sales data (using with fleet/fuel mix and fuel intensities etc).

How are we tracking with forecast traffic volumes?

I don’t have access to the complex model BITRE used to forecast congestion costs, but vehicle kilometres is an obvious major driver of congestion costs, and it is easy to compare the 2007 forecast (Working Paper 71) of vehicle kms in major cities with the most recent estimates of actuals (Report 124):

Consistent with other evidence, the growth in vehicle kilometres appears to be significantly below forecast. In 2007, BITRE assumed that city travel growth would fall to population growth rates, and that mode shares of travel would remain static. They also assumed world oil prices would peak at around US$65 in 2008 and drop to the low US$50s by 2011 (in 2004 dollars). None of these assumptions have played out in reality.

When looking at the components of the vehicle km estimates, the estimated actuals (in Report 124) for 2009-10 appear to be 15% lower than forecasts for cars and light commercial vehicles. For trucks, the 2009-10 estimated actual is around 8% lower than forecast.

To be fair, there was little evidence of the emerging mode shifts available at the time. That said, a BITRE forecast presented at ATRF in September 2011 showed a return to business as usual upwards growth, despite the last 6 years showing little growth.

What cost of congestion might we have avoided?

The relationship between travel volume and congestion costs is not linear. It is usually conceptually represented as an exponential curve. That is, a small reduction in traffic volumes will have a large impact on congestion costs (as evidenced each school holiday period where a claimed 5% reduction in traffic volumes has a significant impact on congestion levels).

While I am not equipped to do a robust calculation, the recent shift away from private car motoring is probably having a significant impact on the avoidable costs of congestion. Estimated actual capital city vehicle kms in 2010 (117.9 billion km) were just under the forecast for 2004 (118.2 billion km). The estimated cost of congestion for forecast 2004 vehicle km levels was $9.1b, while it 2010 it was forecast to be $12.9b. Road capacity has been increased in most cities between 2004 and 2010, which would reduce congestion costs for the same traffic volume, so the difference in 2010 between actual and forecast avoidable congestion costs might be in the order of around $3 billion.

So what is happening with vehicle kms per capita?

In another post, I used BITRE yearbook data on motorised passenger kms per capita. BITRE Report 124 only includes figures on vehicle (not passenger) kms, but they are still interesting figures.

And in response to requests from across the Tasman, I’ve added New Zealand’s one “big” city Auckland (data for ‘Auckland Region’ from their Transport Indicator Monitoring Framework, accessed October 2011).

Total vehicle kms per capita appear to be trending down in all Australian cities since around 2004/2005, with the sharpest drop in Melbourne in 2008-09. Auckland appears to be showing no such trend, with perhaps a flattening at best since 2005-06 (the vehicle km data is marked as under review, as is the public transport data which shows patronage growth of 25% in the four years to 2009-10).

Comparing values for different cities requires caution. The physical size of the urbanised area, and the administrative boundaries used to define cities will have an impact. For example, Adelaide shows up with lower vehicle kms per capita than Melbourne, even though it has much lower public transport mode share. The Adelaide urban area has a smaller footprint and is more constrained than Melbourne, which might explain this difference.

Car vehicle kms per capita appear to have peaked in either 2003-04 or 2004-05 in the five big cities, with Melbourne showing the biggest decline (a 14% decline since 2004-05).

The last two charts showed financial year estimates, but data is actually available at a quarterly level. I’ve created the following chart using simple interpolation of June estimates of residential population for each of the large Australian cities:

The underlying fuel data was actually seasonally adjusted, but there still appears to be some noise in the data (or the world may just be that variable, but I doubt it).

Vehicle use outside the big cities

What about traffic volumes in the rest of Australia? I’ve extracted the five big cities (Sydney, Melbourne, Brisbane, Perth and Adelaide) from the remainder:

The reduction in vehicle use does not appear to be limited to the big cities (most of which have seen strong growth in public transport). The trends for car km per capita outside the five cities are no different to overall vehicle use.

I should note: the report does not actually specify how vehicle kms for each state were split between capital city and other areas (section 8.2, citing unpublished data), but the fractions used were published.

What about total vehicle kms in cities?

While I like to look at per capita transport usage (everything is relative), it is instructive to look at trends in total volume as well. They provide some input into whether increased road capacity might be required, for example.

This charts shows that total vehicle kms in Melbourne, Sydney and Adelaide have been relatively flat since around 2004, while Auckland, Perth and Brisbane have shown continued growth. Perth and Brisbane show a downturn only in more recent times, but have had several years of declining vehicle kms per capita, the difference probably explained by stronger population growth.

How do BITRE Melbourne figures compare with VicRoads’ data?

Here is a chart comparing vehicle km index values for Melbourne from BITRE report 124, and an index created from annual growth figures reported in VicRoads Traffic Systems Performance Monitoring reports (with fully revised history):

A significant gap opens around 2003-04, but this substantially closes from 2008-09. Both datasets show a stabilisation of total traffic volumes, with BITRE data stabilising one year later than for VicRoads. BITRE aimed to estimate total metropolitan traffic, while the VicRoads figures are based on a defined set of monitored roads that might not reflect total traffic, particularly in growth areas on the fringe.

(Note: I did a similar comparison of VicRoads data to BITRE Working Paper 71 estimates of actuals in an earlier post).

In conclusion

• There is strong evidence that “business-as-usual” growth in vehicle kms is just not happening in Australian cities, and thus the 2007 forecast doubling of congestion costs by 2020 is very unlikely to play out.
• The dampened growth in travel demand is probably saving the economy a few billion in avoidable congestion costs, and has implications on the need for multi-billion dollar expansions of road capacity (though changes in demand will not be uniform across road networks).
• I’d also suggest it is important that planners and policy makers understand why travel demand trends have changed so significantly, and apply this understanding to forecasts of future demand.

Following a suggestion in the comments on my last post about car ownership, this post takes a detailed look at car/vehicle occupancy.

What are the trends in car occupancy? (direct measures)

Many state road agencies make direct and regular measurements of vehicle occupancy in capital cities and their data is collated by AustRoads.

Unfortunately only four cities report such data to AustRoads. Brisbane data has several missing years – and the three most recent years’ figures reported are all identical, so I’m inclined not to plot them. That leaves Melbourne, Sydney and Adelaide.

Firstly, all day (weekday) average occupancy:

There doesn’t appear to be much in the way of clear trends as the data seems quite noisy (I’m not sure anyone could explain the year by year variations). Perhaps Melbourne average all day occupancy is trending down?

Data is available for three sub-periods:

Again lots of noise, but maybe a downwards trend in Melbourne.

Noisy again. Possible downwards trend in Melbourne.

This data is remarkably flat for Sydney, while Melbourne again appears to be trending down.

It’s little surprise that AM peak has the lowest occupancy, as it is dominated by journeys to work. More on that soon.

What about trends on different road types?

Looking at Melbourne data in more detail, car occupancy appears to have declined most on freeways and divided arterials:

On freeways, the decline is most evident during business hours:

Notes on the AustRoads/VicRoads data:

Along with the noise in the data, there is some ambiguity in the methodology. The AustRoads website reports “car” occupancy, but the methodology doesn’t seem to filter for cars. Are buses included or not? It says the survey should be undertaken in March/April to avoid school and public holidays. But March and April have heaps of holidays (Easter, Anzac Day and Labour Day in many states).

But the AustRoads data is certainly collected on representative arterial roads, where you might expect lower occupancy because of longer trips that are more likely to be work-related.

What are the trends in car occupancy? (derived measures)

Car occupancy can also be measured as the ratio of car passenger kms to car vehicle kms.

BITRE provide estimates of both, but only for all passenger vehicles in Australia. The long-running Sydney Household Travel Survey provides estimates of both for Sydney. here’s what they look like:

The BITRE figures show a fairly smooth and slow downwards trend from 1.62 in 1990 to 1.59 in 2009. The Sydney figures are a little more noisy, but also show a slight decline – from 1.37 in 1999 to 1.35 in 2010. Presumably non-urban car trips have much higher car occupancy as there’s quite a difference between 1.59 and 1.35.

The Melbourne (VISTA) 2007-08 household travel survey nets an average car occupancy of 1.43 – that being the average over all car kms for trips by residents of the Melbourne Statistical Division (SD) that start and finish within the Melbourne SD. Include trips outside the Melbourne SD and you get 1.49. You could measure it a number of different ways.

The census journey to work question gathers data on how people travelled to work, including car drivers and car passengers. While not a clean measure, it is possible to calculate an implied car occupancy as (car drivers + car passengers) / (car drivers). For the purposes of this calculation, I have only taken “car driver only” and “car passenger only” trips (which excludes park-and-ride and kiss-and-ride public transport trips). I do not have data on trip lengths, and average car passenger trips might be different on average to car driver trips.

There’s a pretty clear downwards trend as fewer people travel to work as passengers. In fact, the data suggests extremely low levels of car pooling., and that over 90% of car journeys to work have no passengers in most cities.

So in summary, there is evidence of gradual declines in car occupancy for all travel, and strong evidence of a decline in car occupancy on the journey to work.

What’s the relationship between car ownership and car occupancy?

You might expect car occupancy to go down as car ownership goes up. In other words: we have more cars and need to share them less.

Here’s what the relationship looks like for Australia as a whole (using car occupancy derived from BITRE data):

There are four quite different periods:

• From 1993 to 1999 (bottom right) car occupancy declined as car ownership increased. As you might expect.
• From 1999 to 2001 car ownership stalled, but car occupancy continued to decline.
• From 2001 to 2005 car ownership rose again, but car occupancy declined more slowly.
• From 2005 to 2010 car occupancy increased slightly, while car ownership had slow growth. This is the period when public transport mode shift took hold in most Australian cities.

The relationship is changing, probably influenced by other factors. BUT it could also be that I’m reading too much into the precision of the car occupancy figures – we are talking about variations in the fourth significant figure only for the last few years. The BITRE figures are estimates themselves. Maybe someone from BITRE would care to comment on the precision?

What about different road types?

Here is a chart showing VicRoads’ 2009/10 car occupancy figures for arterial roads in Melbourne:

You can see car occupancy lowest on freeways, and highest on undivided arterials with trams (all in the inner suburbs). Otherwise very little difference (in 2009/10 at least).

How do Australian cities compare?

To try to take out some of the noise, I’ll take the average of the last four years for the AustRoads data and Sydney and Melbourne household travel survey data:

Melbourne appears to have the lowest occupancy, and Sydney the highest – except when it comes to household travel survey data where Melbourne is much higher. But this might just be differences in methodologies between states.

Factors influencing car/vehicle occupancy (in Melbourne)

Having access to the 2007-08 VISTA data, it’s possible to disaggregate vehicle occupancy on almost any dimension you can imagine. I’ll try to restrict myself to the more interesting dimensions!

For most charts I have used vehicle occupancy rather than car occupancy. Cars and 4WD/SUVs combined accounted for 88% of vehicle kms in the dataset so there shouldn’t be a lot of difference. But I’ll start with looking at..

Vehicle type

Now that’s a surprise: 4WD/SUVs have a much higher average occupancy than cars. Why is that?

Are they used for different purposes?

Not a great deal of difference between cars and 4WD/SUVs, although 4WD/SUVs are slightly more commonly used to pick up or drop off someone.

More likely explanations (from the data) are:

• 4WD/SUV come from larger households on average (3.5 people v 3.1 for cars).
• 4WD/SUVs are also more likely than cars to belong to households that are couples with kids.

Day of the week

Probably not a huge surprise that cars have less occupants on weekdays than weekends. Male drivers are much more likely to have no passengers on weekdays, but an average of one passenger on weekends. Whereas there is much less variation for females.

Is this traditional gender roles in the family? (There is a chart to answer almost any question you know..)

There you go: dads much more likely to drive the family around on weekends, and mums more likely to drive them around on weekdays. And while on the subject…

Household types and sizes

Little surprise that car occupancy increases with household size. It is easier to car pool when you have the same origin.

Note that the sample size of one parent households of size 5 are small (especially for male drivers). But curiously single mothers have much higher occupancies than single fathers.

There is also a small sample of other household structures with 5 people.

Unsurprisingly, people living alone are likely to have the lowest car occupancies. With increasingly prevalence of sole person households, you might expect continuing declines in average car occupancy.

Trip purpose

Again work trips are the least likely to involve passengers, particularly on weekdays (average occupancy 1.07). Driven trips to education are not far behind. Little surprise that accompanying someone, or picking up or dropping off someone averages around 2 or more. Occupancies for personal business, shopping, recreational and social trips are in the middle, but much higher on weekends when householders are probably more likely to travel together to common destinations.

Many people would argue that demand for public transport is lower on the weekend. These figures would support that argument, but lower weekend patronage would also reflect lower service levels.

Note: the sample sizes of weekend education and accompanying someone trips were too small to be meaningful so I left them off.

Time of day

[Note: this chart was corrected on 21/8/2011]

There you go, car occupancy peaks between 8 and 9am and between 3 and 4 pm on school days: parents driving kids to/from school.

But vehicle occupancy is highest on Saturday nights when people are socialising, and interestingly Sundays are well above Saturdays (less personal business on Sundays perhaps?). Non-school weekdays have higher occupancies than school weekdays, possibly with parents also taking time off work and spending time with kids.

Just looking at the school peak more closely, here is a chart showing car driver trip purposes by hour of the day on school weekdays. You’ll almost certainly have to click on this one to read the detail.

The most frightening statistics are in the school peaks. A staggering 40% of car trips between 8 and 9am, and 42% of car trips between 3 and 4pm are to pick up or drop off someone (suggesting a fault in the reported vehicle occupancy for trips picking up somebody). This will almost certainly be dominated by school children. No wonder traffic congestion eases so much in school holidays.

That said, car trips to/from school are shorter than other trip types (as we saw in an earlier post). The data suggests 19% of car kilometres of trips starting between 8-9am are to pick-up/drop-off someone, and for 3-4pm the figure is 24%. That’s still a sizeable chunk of total road traffic. It suggests there are huge congestion relief benefits to be had in getting kids to walk, ride or use public transport to/from school.

Geography

There’s not a lot of difference other than for the inner city, where school day occupancies are lower. For someone in the inner city to drive a car, they are probably heading out of the city and any other members of their household might be less likely to have the same destination and/or would have good public transport options for their travel.

The non-school weekday figures show some variation, and while the sample sizes are all over 250, there are some vehicles with an occupancy of 14 recorded. unfortunately because the underlying data is discrete, medians aren’t an easy way around this issue.

Age

This would suggest traditional gender roles are in play: Average car occupancy is highest for drivers aged 30-45, the most common age groups for parents of pre-driving aged children. And women seem to be doing more ferrying of the kids than men.  In the older age groups men are more likely to be driving with passengers.

Income

Vehicle occupancy seems to go down as we have higher incomes (moreso for females), but there seems to be some noise in the data (eg the spike at 3000 is due to one vehicle with 12 occupants). Females with lower household incomes have higher vehicle occupancies (maybe those without an income but looking after a family).

This trend reflects the fact that car/vehicle ownership goes up as wealth goes up:

The threshold for car ownership is around $1250 per week (equivalised to a single occupant household). As Australians have become increasingly wealthy in real terms, we can afford to own more cars.

Trip distance

While there is probably a little noise in this data, there is a fairly clear pattern. Very short trips and very long trips are likely to have higher occupancies. The median trip distance for non-work trips is around 4kms, while work trips are much longer, which fits with the average occupancies for different trip purposes.

In fact, here is a mode share breakdown by trip distance (for trip legs):

You can see car passenger becomes more common for very long trips (note the X axis scale is not uniform). (Don’t ask me why driving is so popular for distances of 16-16.9 kms! It’s probably a bit of noise)

And if you look at the trip purposes of these very long trips, you’ll longer trips are more likely to be social or personal business:

(note: this chart is by trips, and not trip legs)

Main Activity

Probably little surprise that those “keeping house” have the highest occupancy in general, but that full-time workers have very low occupancy on weekdays, but very high occupancy on weekends.

There you go, possibly more than you ever wanted or needed to know about vehicle occupancy!

Firstly, by car ownership rate I mean the ratio of the number of registered “passenger vehicles” (from the ABS Motor Vehicle Census) to population (also from ABS). Of course some of these cars are owned by companies and not garaged at households, but the data is what it is.

When looking over the last 23 years, it is no surprise to see car ownership rates in Australia have risen considerably:

(click on the chart for a less blurry version)

What is interesting in this chart is the relative rate of car ownership between states and territories. The Northern Territory is consistently the lowest – I’m guessing related to the relatively large indigenous population. I’m not sure for the reasons for other differences. It may be the percentage of the population that is in big cities and the car mode share of those cities. It might also be slight differences in reporting from the state agencies (see ABS’s explanatory notes).

But what about the most recent trends? Here is the same data zoomed in to the last 8 years:

It appears car ownership has more or less levelled-out in a few states:

• Victoria, between 2006 and 2010
• Western Australia, from 2007 onwards
• Queensland, from 2008 onwards (it has actually declined very slightly)

This is fairly consistent with other evidence about declining car use and mode shift to public transport in Australia’s cities.

And the overall rate for Australia appears to have levelled out between 2008 and 2010.

Are cars getting cheaper?

Certainly motor vehicles have been getting cheaper in real terms since around 1996 (relative to overall CPI). They’ve also actually been getting cheaper in nominal dollar terms since 1995.

You might have thought increasing affordability was a strong driver of car ownership rates, but it doesn’t explain growth in car ownership pre 1995, or the slowing of car ownership rates around 2008.

What about usage of each car?

Using data from the BITRE 2011 yearbook, it is possible to calculate an estimated annual kms per passenger car. For this I’m comparing the number of vehicles at the motor vehicle census date with an estimate of total car kms in the previous 12 months (straight line interpolation of BITRE year ending June figures). This isn’t a perfect measure as the number of cars grows throughout the 12 month period where kms are taken, but it is still a guide to the trend.

That’s a pretty clear downwards trend, particularly in recent years.

What might explain this?

• From 1995 to 2005 cars have become more affordable and so we can own more cars and need to share them less.
• From 2005 to 2010 car ownership rates have slowed and we are driving cars less as we travel less and/or shift to other modes.

What about car ownership in cities?

The data available on the ABS website is only available at the state and postcode level (and at postcode level only for recent years). However with some GIS calculations I’ve been able to come up with an estimate of the number of passenger cars with an owner address within the Melbourne Statistical Division (postcode boundaries sometimes do not perfectly align with the Melbourne SD boundary on the fringe, but this is fairly minor and mostly in rural postcodes).

Here is a chart comparing Melbourne, Victoria, and Australia:

So after quite a bit of work extracting and processing all the data, I’ve found very little difference between the Melbourne and Victoria rates or trends. Which does seem a little odd given it is probably easier to live without a car in Melbourne than the rest of the state. It might be that there are many cars with an ownership address in Melbourne, but garaged outside Melbourne. I’m not sure (anyone know more?).

I have covered the spatial variations of car ownership in Melbourne in another post.

What about motorcycles?

Are more people owning motorcycles instead of cars?

You can see motorcycle ownership rates have grown significantly since around 2004 (although still very small).

Does it explain the slowdown in the car ownership rate?

This chart still shows a slow-down after 2008, so it doesn’t look like rising motorcycle ownership explains the car ownership slow-down. Motorcycle ownership took off in 2004, but car ownership slowed in 2008.

What about the ageing population?

Could the data be impacted by a changing age profile? Very old people are probably less likely to drive and hence own a car, so maybe this would lead to a declining car ownership rate per head of population as a greater portion of the population is older.

Suppose most car owners are aged 18 to 80 years. Here’s the percentage of Australia’s population within that age band:

The share has been very steady at around 73 to 74% for all of the last 16 years, which suggests little impact on overall car ownership rates.

Then again, those aged 80 today might be healthier and/or wealthier and more likely to own cars that those aged 80 in 1994. In that case, the rate of car ownership of younger people would have seen less growth, but this is purely speculation and I’m not aware of any available data that could enable a test of this hypothesis.

Notes on the data:

• The ABS Motor Vehicle Census has been taken in different months in different years. State population estimates are only available on a quarterly basis. I have used the nearest quarterly population figure for each motor vehicle census where they do not align (never more than one month out).
• Melbourne population estimates are only available at June each year. I have used straight line interpolation of passenger car figures to produce a June estimate of passenger cars per capita for each year (same as for annual car kms).

Or per capita growth:

I think those charts mostly speak for themselves.

(For the record, the five biggest cities are Sydney, Melbourne, Brisbane, Perth, and Adelaide)

By popular demand, here are charts for each city (plus Canberra):

Note:

• These charts have very different scales on the Y axis. Compare with caution.
• Canberra public transport passenger km (actually just bus passenger kms) is reported as “0.25″ billion passenger kms for five straight years, hence the straight green line.
• While I haven’t drawn the second set of charts for each city, in all cities, car passenger kms per capita have reduced (red lines below blue lines). Public tranpsort passenger kms per capita have increased in all cities except Canberra.

And could you do more to reduce car use by pushing population growth to regional cities instead of the fringe of Melbourne?

I thought I’d take a closer look at travel and trip distances using massive 2007-08 VISTA dataset, and see what factors lead to variations.

In this post I look at travel distances (total and by car) and mode splits across geographies, trip purposes, incomes, ages, and household types. And more.

While the results might not be too surprising, I hope you’ll find the evidence interesting.

How do travel distances vary by geography?

In a previous post I showed that people in the outer suburbs generally have a longer median travel distance:

The patterns were not uniform in the outer suburbs. Nillumbik is the second highest on 35.1 kms per person, while Hume is much lower on 19.6. Factors such as incomes and household types might explain this variation (more on that later).

Most of my analysis will deal with six geographic zones – four rings of Melbourne, Geelong and other regional cities in the VISTA sample combined (Ballarat, Bendigo, Shepparton and the Latrobe Valley). Here’s a map of the Melbourne zones:

Note: I’ve used “city” as shorthand for the central area, and “inner” as shorthand for the inner suburban ring.

Based on those zones, here is a simpler view of daily travel distances (total and by car):

This suggests little difference in total travel distance, but significant differences in car travel distances.

I’ve not used averages because some trips were extremely long (the longest trip by an inner city resident was 833 km) which can skew the averages.

But is median the right measure of travel distance? Probably not, if you look at the following chart of the cumulative distribution of all day travel distances:

How do you read this chart? A point on each line means Y% of people travelled up to X kms per day. Essentially the lower the curve on the chart, the longer distance those people travelled.

You can see differences between distributions are not straight forward:

• The lower half of travel distances were quite similar.
• The differences manifest in the top half of distances. You can see that people in outer Melbourne were much more likely to clock up longer travel distances that those in the inner city. For example, 30% of people in the outer suburbs travelled more than  , while only 15% of people who live in the inner city travelled more than 40 kms.
• In fact, there were more long distance travellers in outer Melbourne than in Geelong or the other regional centres.
• 24% of people in the outer suburbs of Melbourne did not travel at all, while only 15% of inner city residents did not travel on the survey day. This causes the distances to cross around the median.
• There is greater diversity in travel distances of people in the outer suburbs, including about the quarter who did not travel.

Here is the same again for car distance travelled (probably the most important chart in this post):

The differences are much clearer here, with car use and travel distance increasing through Melbourne by distance from the city, and the outer Melbourne suburbs having the longest car travel distances. Distances in Melbourne’s outer suburbs are generally longer than in Geelong and the regional centres.

Interestingly, 48% of inner city residents made no car travel at all, hence the very low median. While the city, inner and middle lines converge at a longer distance, the outer suburbs still had 10% of people doing more than 80 kms in cars.

How does mode share vary by geography?

The distributions on car distance travelled reflect mode splits across the regions. Here is a chart of mode split for trips (using the ‘main’ mode for the trip, which means car+PT trips are counted as PT):

Active and public transport mode shares fell away with distance from the centre of Melbourne. I expect this will be a product of poorer service levels, and a smaller proportion of people travelling to Melbourne’s CBD (the main market where public transport dominates).

But here’s a slightly different take, the mode share of person travel distances:

There is much less variation in public transport mode share of kms travelled. This points to people in the outer suburbs of Melbourne, Geelong and other regional centres making much longer trips when they travelled by public transport. I expect many of these will be long distance rail trips to Melbourne.

The clear difference is that people in the outer suburbs and regional cities did a lot less walking/cycling and lot more travel by car.

What about mode share of very short trips?

Walking is a significant mode in the inner city, and many destinations are within walking distance. You might think that the regional centres are similar, because they are more compact in general.

Well, it appears not. Here is a chart of mode shares of trips under 1km (probably a walkable distance for most people).

Around half the short trips in the outer suburbs , Geelong and regional centres were made by private transport – essentially cars! Why did people drive for such short trips in these areas? Is it a lack of safe places to walk/cycle? Or is it a lack of disincentives to drive?

Digging deeper, even for recreational trips of less than 1km in the outer suburbs, 30% were made by car!

Does the number of trips made vary by geography?

The following chart shows the distribution of the number of trips made. In VISTA, a trip is defined as travel between two activities.

People in the inner city generally made more trips, and those in the middle and outer suburbs made fewer trips. This will also be influencing the total distance travelled per day.

Note: very few people make only 1 trip in a day because it essentially means you start and finish you day in different locations (within the VISTA definitions of a day at least).

How do trip lengths vary?

Here is a distribution chart of lengths of trips (for any purpose):

By almost any measure, those in the outer suburbs of Melbourne made the longest trips. They were followed by people who live in the middle suburbs of Melbourne and Geelong. This means that either people choose to partake in activities that were further away, or (more likely) those activities were further away from home.

What about trip distances for different purposes?

First up, median trip distances by purpose:

Work related trip distances were clearly the longest, especially in the outer suburbs. The “median” person living in the outer suburbs of Melbourne travelled 16 kms for work (note that not all “work related” trips are to/from home).

Here’s a closer look at the distribution of trip lengths between home and work:

The differences when looking only at home to work and work to home trips is much more stark, with the outer suburbs of Melbourne fairing worst by a long way.

The median distances in Geelong and the other regional centres were actually less than the inner suburbs of Melbourne, however they have a long tail with over 10% of trips in Geelong more than 50km.

Back to the previous chart, social trips also get longer as you move to the outer suburbs of Melbourne, which suggests that outer suburbs are not as self-contained for social destinations.

Most other trips purposes had a median around 3-4 kms, although this was more like 2-3 kms in the inner city, and distances increase in the outer suburbs. Chauffeuring trips (pick up or drop off someone) show the least variability (many of these would be taking kids to/from school).

Trips to education were longest in the inner suburbs, possibly reflecting children from wealthy families attending private schools that are further away.

How does travel time vary by trip purpose?

You can see:

• Work trips take the most time in Melbourne, but there isn’t a lot of variation. This supports the hypothesis that people have a commuting travel time budget, and generally find work within that budget.
• Work travel times were highest in the inner suburbs (perhaps related to slower road speeds) and outer suburbs (much longer distances).
• Education trip times were longer in the inner and middle suburbs (perhaps related to congestion and/or longer trips to private schools by children in wealthy families)
• 10 minutes was the most common median trip time – which actually shows up as 9 minutes in the chart, owing to the way I calculate medians in Excel (sorry, not perfect, but Excel doesn’t do medians in pivot tables).

Here’s a closer look at work-home trip time distributions:

You can see big steps at the multiples of five minutes, as people tend to round estimated trip times to the nearest 5 minutes. Median trip times in Melbourne are all around 30 minutes, and much lower in Geelong and regional centres. People in the inner suburbs were least likely to have commute trips less than 20 minutes, while the outer suburbs were most likely to have trip times over 30 minutes.

How does travel speed vary by trip purpose?

As you might expect, trips were faster in the outer suburbs, probably because a combination of less congestion and more roads designed primarily to move vehicles quickly (freeways and divided arterials).

Education trips didn’t speed up as much in the outer suburbs, perhaps because they were more likely to be on public transport. Which brings us to…

How does mode share vary by trip purpose?

Around half of education trip kms were by public transport overall, although this was curiously lower in the inner city and outer suburbs.

Work trips had the next highest public transport mode share, which fell away towards the outer suburbs.

Other trip types mostly had slightly higher public transport mode shares closer to the centre of Melbourne. Note: I have not excluded very long trips from this analysis, so they might throw the figures slightly.

Here is another view, private transport mode share:

You can see more significant trends across Melbourne, as people in the inner city and suburbs were more likely to travel by active transport.

What other factors influence travel distance and mode split?

Different households will have different travel needs, and the distribution of household types across Melbourne is not even:

And median per person travel distance varied by household type:

You can see that the household types more prevalent in the outer suburbs (couples with or without kids) have the highest median car travel distances. So this will be impacting longer travel distances in the outer suburbs. You can also see that couples with kids have the highest car mode share, which is no big surprise!

Here’s a look at the distribution of total travel distance for people living in households that were couples + kids, one of the most common household type:

Couples with kids in the inner city certainly travel less distance, and while the bottom half of people were similar for other regions, the travel distances were much longer in the upper half of such people, suggesting geography still had a big impact.

Equally household incomes were not consistent across Melbourne:

Equivalised household income is a measure that allows income comparisons across different household sizes. It is calculated as household income divided by a measure of householders: the first person is assigned a value of 1.0, subsequent persons over 15 years are 0.5, and any children are 0.3.

Curiously, the inner city has the lowest income profile in Melbourne (note that VISTA 2007 did not include Southbank and Docklands residents), while the wealthy live in the inner suburbs.

It will come as little surprise that household income is a driver of total – and car-based – travel distance:

Do rich people shun public transport?

No, only the very-rich seem to shun public transport. According to the VISTA numbers (which are weighted to census 2006 demographics), only 10% of people live in households with an equivalised income over $2000.

The highest concentration of wealthy people is in the inner suburbs, travel distances are generally shorter (although income might explain longer travel distances in relatively wealthy Nillumbik).

To isolate household income, here is a distribution chart for people in households with an equivalised income of between $500 and $750 per week (the largest $250 bracket overall):

Again the lower half exhibits very little difference, while the outer suburbs of Melbourne has much longer distances in the upper half. (note the inner city line is quite jagged, because the sample size in this instance is only 204).

What about age:

People aged 20-64 certainly travelled longer distances. Looking at the distribution of ages, there were more people aged 20-74 living in the inner city and inner suburbs, compared to middle and outer Melbourne. There is very little difference in the percentage of the population between 25 and 64 across the regions (those with the largest car travel distances).

And yes, public transport mode share is lowest amongst very young children and the middle-aged (the later group often being the decisions makers!):

And finally (without going through all the detail here) people who work full-time tend to travel more, but they become less prevalent as you move away from the centre of Melbourne.

So, should we encourage population growth in regional centres instead of Melbourne’s outer suburbs?

Well, it’s probably the wrong question to ask! People in inner Melbourne do a lot less car travel than anywhere else. This analysis clearly shows that encouraging people to move into inner Melbourne would probably do the most to reduce car travel per capita.

People currently living in the outer suburbs of Melbourne travel more and do more car kms than those in regional cities. The main problem is that their work and social trips are much longer.

The evidence suggests putting people into regional cities would generate less car travel than putting people on the fringe of Melbourne.

However, there are several points worth considering:

• If you can generate jobs in the outer suburbs of Melbourne, you might be able to reduce work travel distances. Easy to say, but it defies agglomeration economies that cause jobs to co-locate in the inner city and suburbs. If Melbourne’s Central Activities Areas (formerly Central Activities Districts (formerly Transit Cities)) can become significant employment destinations then that will certainly help.
• If you do encourage people to settle in regional cities, will they have the same transport profile as existing residents? I would guess that there would be a significant difference between people living in the centre of regional cities, and those living on the fringe. The reduced car travel advantages of regional cities are probably largely eroded on the fringes of the regional cities. However, encouraging higher density in the inner areas of the regional cities would probably generate less car kms.
• If you increase the population in regional cities without also increasing employment opportunities, you’ll create unemployment problems and/or force people to travel further to get to work. This would cancel out some of the benefits of locating people in regional centres. It may also increase demand on long distance V/line commuter trains into Melbourne (which currently consume valuable metropolitan train paths with low passenger density).

It doesn’t seem like there is much difference between the outer suburbs and regional cities. But there is a much bigger difference when you compare these with the inner suburbs of Melbourne.

If we really want to reduce car use, we’ll need to do relatively easy things like:

• Locate people in inner city and suburban areas, where travel distances are short and there is viable high quality public transport (though it will probably require capacity upgrades)
• Increase public transport service levels in existing outer suburbs and regional cities, with a particular focus on efficiently connecting people to employment areas by public transport.
• Break down the barriers to walking and cycling in the outer suburbs and regional cities. Footpaths and safe places to ride would be a good start!

Notes about the data:

• Wherever possible I have used person weightings in VISTA, which are for all week travel and align VISTA data with 2006 census data on demographics.
• I have determined trip purpose by looking at the destination purpose of each trip. If the destination is not home, then I assign the destination purpose as the trip purpose. If the destination purpose is home, then I assign the origin purpose as the trip purpose. This gets around the common problem of nearly half of all trips having “go home” as the trip purpose, which costs you half your data when analysing by trip purpose.

So this post is a guide to five posts I updated over April 2011, with a highlight from each.

Evidence of mode shift in Australian cities (BITRE data)

BITRE have published a 2011 statistics yearbook, and there’s evidence of continued public transport mode shift in 2008-09 in Melbourne, Brisbane, Sydney, Perth, but not in Canberra or Adelaide.

Read the full post.

Trends in transport greenhouse gas emissions

New national emissions data has been published showing transport emissions are increasing, particularly in freight and civil aviation, while car emissions are continuing to decline. I’ve done some new analysis that suggests vehicle efficiency only explains around 40% of the saved car emissions – the rest likely to be a product of travel behaviour change.

Read the full post.

Peak oil

We haven’t yet seen peaking of global oil production (including non-conventionals), but both supply and prices are going up.

Read the full post.

Melbourne urban sprawl and consolidation

Recent ABS population data show that there was a surge in urban sprawl in Melbourne in 2009-10, in both share of growth and numbers of new residents. There was also a lot less urban consolidation in 2009-10, and slower population growth overall (1500 new residents per week in 2009-10).

However, post June 2010  dwelling approvals data suggests a slight decline in sprawl’s share of population growth.

Read the full post.

Traffic volumes on Australian toll roads

There’s been a recent surge in traffic volumes on Citylink in Melbourne, following the completion of major upgrade works:

Read the full post.

In an earlier post, I looked at where employees come from for some major employment destinations around Melbourne. This post does the flip-side: What are the work destinations for people in different parts of Melbourne? How well does the current public transport network connect people to where they work? What implications does this have for the proposed rail lines to Doncaster and Rowville?

I’ll take a detailed look at Rowville, Manningham, and Berwick, and a briefer look at Altona, Broadmeadows, Cranbourne, and Sunshine.

About the maps

I’ve used a dataset of that contains the volumes of people commuting each SLA (Statistical Local Area) to destination zones in Melbourne from the 2006 census (with thanks to the Department of Transport). I’ve then trimmed each destination zone to keep only areas where employment activity would be expected (using ABS mesh blocks). That is, I’ve removed parklands, residential areas, etc.

Then I’ve used dot distribution mapping, where each dot represents 10 or 15 employee destinations (depending on the SLA). I’ve also overlaid each SLA’s percentage share of journey to work destinations from the SLA in question (shaded yellow), to give a broader perspective.

Rowville

The “Knox (C) South” SLA might as well be called Rowville as that suburb dominates the SLA. The following map shows the density of worker destinations in 2006. (As usual, you’ll need to click to zoom in and see the detail).

There are quite a few dense destinations within the SLA, including (hyperlinks are to Melway maps):

• Stud Park Shopping Centre
• the Stamford Hotel
• some small industrial areas
• Caribbean Business Park in Scoresby

Nearby destinations to the west include:

• Monash Medical Centre in Clayton
• The Valley Private Hospital
• Waverley Gardens Shopping Centre
• Mulgrave/Clayton/Notting Hill industrial area
• Tally Ho business park

Major destinations to the north and south include:

• Knox City
• Bayswater industrial area
• Dandenong CAD
• Dandenong Hospital/TAFE
• Cranbourne

Further afield there are concentrations in:

• Chadstone Shopping Centre
• Box Hill CAD
• Coles Headquarters in Tooronga
• Melbourne CBD
• St Kilda Road employment area

You’ll also find that most schools in and near Rowville show up on the map.

Rowville is only served by buses for public transport, but there are now direct connections from Rowville to most of these destinations, particularly following the introduction of SmartBus routes 900 (west to Clayton, Oakleigh, Chadstone and Caulfield) and 901 (north Knox and Ringwood, and south to Dandenong and Frankston). The 900 was a completely new route, and the 901 an upgrade of an existing route, both occurring after the 2006 census.

However neither of those routes operate east of Stud Road, where most Rowville residents are located. Those people need to change buses at Stud Park Shopping Centre, which is not easy as most local bus routes in eastern Rowville run every 30 minutes in the peak (and are highly indirect). You can see the local network on the Metlink public transport map for the Rowville area:

The recent Bus Service Review in the area did propose route 900 be extended east towards Ferntree Gully which would introduce a direct public transport connection to many of the major employment destinations for local residents.

Notable destinations not well connected by public transport include:

• The Caribbean Business Park in Scoresby (marked Caribbean Gardens on the above map): The only access by public transport is the 753 and 693 bus routes at the northern edge. There are around five trips per hour combined in the AM peak, but the headway becomes 30 minutes later in the AM peak. This business park only has one entrance road and is away from major public transport routes making it difficult to service. It is however conveniently located next to Eastlink, and obviously has a high car dependence. They just need to hope oil prices stay cheap. Well, cheap-ish.
• The Notting Hill industrial area around Ferntree Gully Road: People wanting to get to jobs in this area need to change buses. The current network allows you to reach Notting Hill by changing between bus routes that operate every 15 minutes (which could be worse).
• The Tally Ho business park in Burwood East (corner Burwood Highway and Springvale Road): This is at the intersection of a SmartBus route and a tram line, so there is relatively good public transport to the site. However the challenge is the distance involved, and from Rowville a transfer is required between SmartBus routes.
• The Bayswater industrial area: This is only directly connected from Rowville by three extended trips on bus route 691 in the AM peak. In the PM peak, no trips run through from Bayswater to Rowville, so a transfer is required at Boronia.
• The Valley Private Hospital, which is just across Dandenong Creek in Mulgrave but not near any bridges.
• The Coles Headquarters in Tooronga (Glen Iris). This employment centre is only really serviced by one bus route (624), or a lengthy (and hilly) walk from Tooronga train station or the Burke Road tram (72).
• The Box Hill CAD, again some distance away. Can be reached with a bus-bus or bus-train transfer (about 1 hour on PT).
• The St Kilda Road employment area is hard to reach (particularly adjacent to Albert Park). You either have to go to Flinders Street and catch a tram south (an indirect journey), get off a train at Armadale and transfer to a slow tram, or change from train to tram at South Yarra (but this only serves the northern end of the patch). Again, this looks like more evidence to support a new PT route from South Yarra Station to the Albert Park section of St Kilda Road (and possibly beyond). This was recommended as part of the Bus Service Review for the area.

The first two of these destinations are along Ferntree Gully Road, and so it might be tempting to try to run a new bus route from Rowville along Ferntree Gully Road. But rather than trying to pair all origins and destinations with direct bus routes (creating a large confusing network of low-frequency bus routes), maybe the answer lies in better frequency on the 693 (or more even headways between the 693 and 753) to reduce transfer penalties. Another improvement would be to introduce a stop on the four express route 754 trips, should that be deemed acceptable by the powers that be(!). The 754 express bus used to be the fastest way from Rowville to the CBD, but this is no longer the case following the introduction of the 900 SmartBus (in 2006).

The following table shows the total numbers of journeys to work to the top SLA destinations, and the public transport mode share

The City of Melbourne is the only destination with any serious PT mode share, which is important to keep in mind considering the above analysis. Perhaps the mode share for some of the directly connected destinations may have risen since 2006 due to the new SmartBus routes. But then maybe public transport will always struggle to compete with plentiful free employee parking (that employers are paying for).

What does this mean for a rail line to Rowville?

If rail were to run via Wellington Road, certainly it would connect Rowville to a number of its major employment destinations to the west. However if the rail ended at Stud Park Shopping Centre, there would still be a transfer problem for most people who live in the eastern part of Rowville (something that can be more easily fixed by extending existing SmartBus route 900). It would not also be an ideal park and ride station, being an activity centre with limited land space.

1450 people from Rowville commuted to the City of Melbourne, and a further 2550 commuted to the City of Monash, a total of 4000 trips (exactly, as it happens). If the train line achieved a 50% share of all travel to Melbourne and Monash (2000 people), that is 3 trains at comfortable loading from Rowville. Of course there would be additional demand from other areas along the route, and for trips to other destinations. But then again not all destinations in Monash and Melbourne would be well served by the rail line (eg Glen Waverley).

The Rowville Railway Prefeasibility Study 2004 (commissioned by Knox Council), simply assumed 7% of peak period journeys to work in the line’s catchment would use the train (refer page 38), without checking census data about where Rowville residents currently work (though it turns out that many do work along the rail line). They arrived at a figure of 3360 journeys to work by Rowville rail, using a total catchment population of 100,000 (allowing for park and ride), shown in the map below.

They assumed 70% of these would be within one hour and then calculated an “hourly” peak patronage estimate of 2352 trips. They described this figure as the patronage per hour in the peak, but this is a little misleading because the other 30% of journeys to work by train (1008) would be outside the busiest hour of the peak. Assuming a three hour AM peak, it might be around 500 for the first and third peak hours. Not a lot of demand.

Obviously university students and people travelling for other non-work purposes would add to these commuter figures, but I am not sure whether there will be sufficient demand to justify the cost of grade-separated heavy rail to a low-density area of Melbourne that is a not an urban growth area. While the residential densities in the catchment are not the lowest in Melbourne, there are large areas of parkland and other non-residential land use that dilutes the average density of the catchment (an issue that equally applies to the proposed Knox tram extension). A shorter line as far as Monash University might be more viable.

In addition, the study assumed a large park and ride catchment. For many people in this catchment, the car travel time saving of having a park and ride train station slightly closer on the Rowville line would be quite small. I suspect many of these trips will be diverted park and ride from the Dandenong and Glen Waverley lines, rather than trips mode shifted to public transport.

Alan Davies, on his Melbourne Urbanist blog has suggested that perhaps such a rail line should veer to the north to capture more of the Notting Hill employment area (and possibly also Chadstone Shopping Centre). Looking at the current employment destinations, there would appear to be some merit in this idea, as long as it still served Monash University. Although without a reservation in place, it would probably require a very expensive tunnel.

Perhaps in a future post I could look at the destinations of a broader catchment of the proposed Rowville rail line (although my dataset only has origins at the SLA level). Alas, I’m doing this in my own time, so I will have to see how I go.

Hopefully we will get a better feel for the economics in the upcoming study into Rowville rail.

Manningham West

The Manningham West SLA captures the main residential half of Manningham, centred around Doncaster, Templestowe and Donvale. Here is a map showing the employment destination densities for Manningham west residents:

Within the SLA, you can see various pockets of destination density, all of which are either shopping areas or schools. In fact, Manningham lacks any significant industrial areas, large medical facilities, or tertiary education institutions. Only 16% of Manningham west workers went to a job located within Manningham west (but this is not actually very low compared to other SLAs in Melbourne, perhaps the subject of future post).

Nearby major destinations include:

• Heidelberg (in particular the Austin Hospital)
• Ivanhoe
• Box Hill CAD
• the curiously located Hewlett Packard centre in Blackburn North
• The “Golden Mile” along Whitehorse Road between Blackburn and Mitcham Stations
• Ringwood CAD

Major destinations further afield include:

• Kew Junction
• Hawthorn/Camberwell corridor, along Burwood Road/Camberwell Road, including Swinburne University
• inner northern suburbs (Carlton/Fitzroy)
• Melbourne CBD
• St Kilda Road (particularly the Albert Park section)
• Royal Melbourne Hospital and the Children’s Hospital in Parkville

In terms of public transport access to employment, the following destinations are more problematic:

• Access to Heidelberg is only provided by the 903 SmartBus (although at a 7.5 minute frequency in peak periods making transfers from other routes easier). There is no east-west connection into Heidelberg from the Templestowe area (an issue identified in the bus service review).
• There is no direct linkage to Ivanhoe, compounded by limited access to Heidelberg.
• The hospitals in Parkville (which require two bus transfers, or a tram connection in the CBD)
• St Kilda Road requires a tram transfer in the CBD. While this is relatively direct, it’s not particularly fast, and the City of Melbourne is planning to remove the tram stops at the corner of Swanston and Lonsdale Streets, the transfer point for such commuters (this is in fact the busiest tram-bus transfer location in all of Melbourne).

Until recently, the “Golden Mile” along Whitehorse Road had very little public transport coverage, but SmartBus route 901 now connects large parts of Manningham with this area (and Ringwood).

The following table shows the major SLA destinations:

What does this mean for a rail line to Doncaster?

Certainly a railway between Doncaster and the city would connect Manningham West to many of its employment destinations in the inner city area.

The journey to work destination SLAs that would be served by a Doncaster railway (Melbourne Inner, Southbank/Docklands, Yarra (North and Richmond) and maybe half of Melbourne remainder) would cover around 8000 trips on 2006 census figures. If the public transport mode share to these SLAs was the same as for trips from all origins (other than the inner city and Manningham west), then that would be around 3900 journeys to work journeys by public transport, or around five comfortably full trains. Over a three hour peak period, that would require less than two trains per hour. However, adding non-work related trips, you might justify three trains per hour in the peak from Doncaster. This would be much less frequent than most train lines in Melbourne.

Doncaster rail also faces similar challenges to Rowville rail, in that most of the catchment would need to use car or bus to access the train line (assuming a Doncaster train line terminated at Doncaster Hill. Already a large regional shopping centre, there would be little room for park and ride, and commuter parking is probably the last sort of inactive land use you want in a major activity centre anyway.

Which would mean buses would need to be the primary access mode for train passengers. This introduces a transfer to most trips, when compared to current direct to city bus services available to most people in Manningham. Transfers bring walking and waiting times, and inconvenience and risk of missed connections (all up what transport planners call transfer penalties).

The Eddington East West Link Study includes an appendix summarising their assessment of mass transit options for Doncaster (appendix C, pages 272-8). It reports a heavy rail line would provide a journey time of 25-30 minutes from Doncaster Hill to Melbourne Central, and that a bus service with high levels of bus priority could complete the trip in 25-35 minutes. SmartBus route 907 is currently timetabled to take around 39 minutes between Doncaster Hill and Melbourne Central in the AM peak, so it would appear the study team anticipated greater bus priority.

Introducing a bus-train transfer to trips to the city would require maybe 5 minutes of transfer time, plus a transfer penalty to account for the inconvenience and missed transfer risk (particularly if trains only run every 15-20 minutes). The journey time advantage of heavy rail quickly evaporates when you include a bus-train transfer to most journeys. You would also need a large bus terminal capable of holding maybe 10-15 feeder buses all meeting the same train if you want to minimise transfer times.

I suspect the lack of compelling travel time savings, and relatively low transport demand will make it difficult to justify the capital cost of Doncaster rail, especially considering the tunnelling required when the line leaves the Eastern Freeway catchment (Eddington estimated a cost of $1.7-2b). This was the finding of Eddington study, but the detail of their analysis was unfortunately not published. And now we are going to have yet another study.

Another factor limiting demand on the corridor is the lack of specialised destinations (such as hospitals or universities) in Manningham to create demand outside commuter peak flows. For example, only 245 people reported commuting from Melbourne or Yarra to Manningham west in the 2006 census, and only 22 of them did so by public transport.

Of course building a train line would change land use patterns, which would probably increase the travel demand from what it was in 2006. But would this increase be enough to return a favourable benefit-cost ratio? And is the railway being built to meet existing latent demand, or create new demand?

Berwick

The Berwick SLA (part of Casey) is in Melbourne’s outer south eastern suburbs, and is still seeing urban growth.

The major local destinations include:

• shopping centres at Fountain Gate and Berwick Marketplace
• Monash University Berwick campus
• the local Chisholm TAFE campus
• industrial/commercial areas in Beaconsfield and around Vesper Drive

The biggest nearby destinations include:

• Hallam, South Dandenong and (the smaller) west Dandenong industrial areas
• Dandenong CAD
• Dandenong Hospital/TAFE
• Cranbourne

Further afield destinations include:

• Monash Medical Centre in Clayton
• Noble Park North (including RACV headquarters)
• Clayton/Mulgrave/Notting Hill and Braeside industrial areas
• Knox City
• Chadstone Shopping Centre(!)
• Coles Headquarters in Tooronga (again)
• Melbourne CBD
• St Kilda Road

Here is part of the Metlink map showing the Berwick area:

Notable destinations not well connected by public transport include:

• The industrial area of Hallam. Bus route 828 almost reaches the Hallam industrial area, but then deviates north to residential areas (and I’m not advocating a change, by the way). Otherwise Hallam station is the south-east corner of the area.
• The large employment area of Dandenong South. Reopening and providing pedestrian access to General Motors Station might help provide access to some parts (it is now ironically abutted by a large employee car park) (see wikipedia if you are interested in the history of this station). The bus service review for the area advocated new east-west routes from Berwick to Dandenong South, which would obviously greatly assist in connecting employees to workplaces. The bus network in Casey was upgraded in late 2010, but no east-west routes were introduced.
• The Clayton/Mulgrave/Notting Hill industrial area is again a problem area. Some parts are near to train stations (providing a direct connection), but most require a change to a bus. For someone not living near a train station in Berwick, the journey would be bus-train-bus, which would not compete well with the car (especially with free parking). I do wonder if a direct bus service (with express running in between) might be viable.
• Caribbean Gardens Business Park (see Rowville discussion)
• The Coles headquarters(see Rowville discussion)
• St Kilda Road (see Rowville discussion)

The following Metlink map shows the lack of east-west bus services across the rail line between the residential and employment areas:

A number of people travel long distances to get to work from Berwick, where public transport will struggle to compete with the car, due to low average bus speed as much as anything. I’ll look at average travel distances in another post.

Here are the volumes and public transport mode shares for major destination SLAs:

Again, only the CBD shows up with large public transport mode share, although curiously Monash South West is at 5% (almost entirely involving train).

Other SLAs

For interest, I have looked at a few other SLAs around Melbourne. I’ll discuss these briefly in terms of problematic public transport access to employment.

Western

Altona

This map shows some problem areas for public transport access, including:

• Large parts of the industrial areas within the SLA, which have no service at all. Particularly the Toyota factory on Grieve Parade (okay, it does have bus route 232, but that only runs to/from the Melbourne CBD on a few peak period trips).
• Victoria University Newport Campus
• Fishermans Bend industrial area (the bus service review recommended route 232 be re-routed along Lorimer Street to improve connectivity to the main employment area)

Sunshine

Public transport doesn’t provide strong service to:

• The Laverton North/Derrimut industrial area. A new bus route 417 was introduced into this area recently, but it only provides access from Laverton station in the south.
• Moonee Ponds has a surprising concentration of destinations, and currently a transfer is required at Highpoint to a tram.
• Fishermans Bend again shows up, with the only access being via the CBD. The recent relocation of Fishermans Bend bus routes to Southern Cross station will certainly assist these people to use public transport.
• The Altona North industrial area, particularly the Toyota factory on Grieve Parade.
• Melbourne Airport.

Maribyrnong

Maribyrnong is quite well connected by public transport to most employment destinations.

However there are some more difficult destinations:

• Laverton North industrial area (partially connected by bus route 414)
• Altona industrial area (that has almost no public transport)
• Fishermans Bend (need to transfer to bus at Southern Cross Station, much less direct than driving)

Melton East

Melton east includes Caroline Springs, Taylors Hill and Hillside. It represents the north-western fringe of contiguous urban Melbourne.

You can see a wide range of work destinations – many a long distance away, suggesting a lack of nearby employment opportunities is an issue.

There are many popular work destinations difficult to reach by public transport, including:

• Laverton North and Altona industrial areas
• Melbourne Airport
• Tullamarine industrial area around Sharps Road (completion of SmartBus route 902 would assist)
• Somerton industrial area (902 completion would assist, although this is a long distance to travel)

Williamstown

27% of destinations are within the City of Melbourne.

Popular destinations difficult to reach by public transport include:

• The Altona industrial area
• Fishermans Bend (which can be reached much more directly by car)

Other SLAs – North

Broadmeadows

Public transport doesn’t provide strong service to:

• The Tullamarine industrial area around Sharps Road/Airport Drive. The Green orbital SmartBus (902) would close this gap if extended to Werribee as originally planned.
• The Tullamarine industrial area near the airport along Melrose Drive (served by the infrequent 478/479 bus routes, for which an upgrade was promised in May 2010 following the Bus Service Review, but still not delivered as at July 2011)

The recent extension of SmartBus routes 901 and 902 will have greatly improved public transport access to the Somerton/Campbellfield industrial area, and access to Airport West.

You can see a smattering of dots over the land covered by Melbourne Airport and the adjacent industrial area to the south. Because this is all one destination zone, the employement is diluted across the zone, whereas in reality the employment will be concentrated around the terminals and industrial area.

We know that a lot of Melbourne Airport workers come from nearby suburbs, including Broadmeadows. The recent extension of SmartBus route 901 has significantly improved access to Melbourne Airport from the Broadmeadows area. Although its current bus stop at the airport is unfortunately quite a distance from all the terminals!

Sunbury

Sunbury is a satellite urban area north-west of Melbourne, with most of this SLA being rural land and 21% of Sunbury SLA residents work in Sunbury itself.

Around 9% of commuters worked in and around Melbourne Airport (distributed over a large destination zone in the map, refer discussion in Broadmeadows section above). There are bus services between Sunbury and Melbourne Airport but they operate very infrequently (a service upgrade has been promised).

A significant number also work in the Somerton industrial area, which is not directly connected by public transport – and would be difficult to be connected efficiently.

Craigieburn

The vast majority of residents in this SLA live in the suburbs along the eastern edge of the SLA (including the Greenvale area to the east of the “14″ label). The rest of the SLA is mostly rual land, although it includes Melbourne Airport and an adjacent industrial area in Tullamarine.

Work destinations difficult to reach by public transport include:

• Melbourne Airport
• Industrial areas in Tullamarine and Airport West
• Thomastown (although it can be reached via Broadmeadows on SmartBus 902)

Moreland north

This SLA is mostly made up of the suburbs of Glenroy and Hadfield.

The biggest destination is Broadmeadows to the immediate north. Connectivity to the north is limited to the two train lines, and two bus routes (one very infrequent). The northern part of the Somerton industrial area can only be reached by public transport with a transfer between relatively infrequent routes.

Brunswick

A significant proportion of Brunswick residents travelled to the south, including the Melbourne CBD.  There are no significant destinations that are difficult to reach by public transport.

Heidelberg

The only work destination somewhat difficult to reach by public transport is the Carlton/Parkville area, as half of peak period trains do not stop at Victoria Park station, which provides a transfer opportunity to high frequency buses to Carlton.

Other SLAs – eastern and south-eastern

Ringwood

Ringwood has direct public transport links to most destinations, including those along the railway line to the city, and the 742 bus to the Notting Hill area. The Bayswater industrial area and St Kilda Road commercial area are a little more difficult to reach by public transport.

Croydon

A significant destination for Croydon SLA residents is the Bayswater and Bayswater North industrial areas. The lack of a bus service along Colchester Road would make this area difficult to reach by public transport for a number of workers.

A significant number of destinations were along the Lilydale/Ringwood train line, making public tranpsort access relatively easy.

A fair number of people commuted to the Clayton/Notting Hill industrial area, and as it happens, bus route 737 connects Croydon to this area (although travel times would not be short).

Melbourne’s CBD only accounts for 5% of work destinations from Croydon.

Lilydale

There are a number of popular destinations difficult to reach by public transport:

• Just to the north of the station is an industrial area that is beyond walking distance of public transport (around Beresford Road).
• Curiously, the southern most part of the suburb of Kilsyth – a rural area (this might actually mostly be the Boral quarry on Cantebury Road, which is connected by bus 679).
• The Bayswater industrial area.
• The eastern part of Mount Evenlyn and Wandin North (again, most of this land is rural so the actual employment might be in spot concentrations within the destination zone).

Knox – north east

A significant proportion of workers had destinations in Bayswater/Bayswater North industrial areas, Knox City and around Boronia station. Most popular destinations are actually connected by public transport from significant parts of the SLA.

Knox – North west

There is some degree of public transport connection to most destinations, although parts of Bayswater North are more difficult to reach.

Only 6% of residents in this SLA work in the Melbourne CBD. A tram extension to Knox City would connect some employment destinations along the way, but would not be an effective way to reach the Melbourne CBD by public transport given the distance and slow speed (bus+train would be faster).

Cranbourne

The standout issue for this SLA is Dandenong South. There is only one bus route connecting the SLA with just parts of Dandenong South industrial area. From the distribution of dots it appears that around 1 in 5 Cranbourne commuters travelled to Dandenong South.

The recent bus service review recommended much better east-west connectivity, but this was not done in the late 2010 upgrade to Casey bus services (unfortunately you wont find the detail in the executive summary of the bus service review, you have to order the full report from the Department of Transport).

Access to the Clayton/Mulgrave/Notting Hill and Braeside industrial areas, and St Kilda Road employment area is also problematic, for the same reasons as outlined for Rowville.

Unfortunately analysis of census journey to work data was not done in this level of detail in the 2007-2009 Bus Service Reviews across Melbourne. Hopefully my analysis can now provide greater evidence to support public transport planning.

The ABS publication Melbourne.. A Social Atlas has a density map (see pages 12-13) at the Census Collection District (CCD) level, but only has five colour graduations so subtleties are quickly lost.

So I’ve decided to draw one myself.

Arguably the best source of data for housing density is the ABS’s experimental mesh blocks, which are smaller than Census Collection Districts (CCD). Mesh blocks are designed to have more uniform land use, which gets around the problem of a CCD which might contain a mix of residential, parkland and commercial land use showing up as low density. But I’ll come back to this.

So here is a 2006 population density map of Melbourne at the mesh block level:

(I’m using people per square km, which is 100 times larger than people per hectare if you need to convert).

You’ll need to click to zoom in, and you might want to then zoom in again with your favourite image viewer to see the detail.

Some observations:

• Many areas on the very fringe show as low density, but this might be because that area was under development at the time of the census, and only some people had moved in.
• Everyone talks about low density sprawl on the fringe, but even back in 2006 there was evidence of higher density development in the outer suburbs. Have a look at the Craigieburn area in the north or around Narre Warren and you will see many patches of green. New blocks on the urban fringe are now actually quite small in places compared to those in the middle suburbs. Two storey townhouses are actually not uncommon in new estates.
• In the north-west (around Delahey/Sydenham), you can see a north-south divide where there is higher density on the eastern side. This corresponds with the municipal boundary between Brimbank and Melton. Presumably they’ve had different urban development policies.
• The biggest clumps of density are in the inner city, particularly Carlton and Carlton North, Fitzroy, St Kilda, Richmond, and Kensington (the western side of which enjoys a 5½ days per week route 404 bus service).
• Looking at the Central Activities Districts (CADs), there are clumps of density near the Dandenong and Box Hill CADs. But nothing to speak of inside Ringwood, Frankston, or Broadmeadows CADs (in 2006).
• Other curious pockets of density in the suburbs include west of Highpoint Shopping Centre, Sunshine, Glenhuntly/Carnegie, and Glen Iris.
• The lowest density suburbs in Melbourne are found in the middle and outer eastern suburbs (particularly Upwey/Belgrave), and in the north-east around well off areas such as Eltham, Toorak and Eaglemont. North west Reservoir seems to be a problem area – high socio-economic disadvantage and low density (not to mention a bus route that runs 5½ days a week).
• Interesting to see relatively higher densities south of the Dandenong rail line.

For comparison purposes, I’ve also created a version based on larger Census Collection Districts (CCDs):

(note: this map doesn’t show anything outside the Melbourne SD)

What’s the difference you ask? You cannot see a great deal of difference, though the CCD map makes Melbourne look a little less dense.

But if you zoom in you can spot differences in some areas where a CCD is part residential, part not. Here’s an example in the Black Rock/Beaumaris area:

The CCD map on the left shows a few darker red blocks next to the whitespace, but that low density is not visible in the mesh blocks on the right, because the  mesh blocks split the parkland and houses. You can also see that the CCDs run to the shoreline, while the beach area has been split into separate mesh blocks.

The advantage of the mesh block map is that it pretty much shows housing density, as most pieces of land that are not residential have been removed (including suburban parks).

But the advantage of CCD density is that it includes local parkland, which is a measure of open space within and immediately surrounding residential areas.

A better way of looking at the density equation is a cumulative distribution chart, as created by Fedor Manin on his blog We Alone on Earth (also referenced on Human Transit).Rather than having to worry about whether low density areas on the fringe are “urban” or not, you can just look at density by population share, and the fringe areas will quickly tail out anyway. On this basis the problems of using an administrative boundary of a city (which often contains a large areas of rural land) largely go away, but then you don’t get a single number.

I’ve lined up all mesh blocks and CCDs in the Melbourne SD in order of density, and created a cumulative profile of density for each.

You can see a big difference between CCDs and mesh blocks (note the X axis is logarithmic). On a mesh block basis, about half of Melbourne’s population lives at a density of greater than 3200/km2, whereas on a CCD basis, only 30% of Melbourne’s population lives at a density greater than 3200/km2. Take note anyone doing a comparison between cities!

Here’s a chart on the same data showing a population distribution across densities, using mesh blocks and CCDs:

You can see the most common density for mesh blocks is slightly higher than for CCDs. The peak for mesh blocks is between 2818-3162 people/km2 on my intervals. That’s an funny sounding interval because I’ve used logarithmic intervals (if you use even intervals of 100 people/km2, the peak is between 2900 and 3300 people/km2)

So what is the average density of Melbourne?

What is Melbourne? Should we include satellite urban areas around the city? For example, is Sunbury part of Melbourne? It is within the Melbourne SD (Statistical District) but not within the Melbourne “Urban Centre” as defined by ABS. Do you want to include non-residential areas (urban density), or not? (residential density)

Here are six very different measures of the urban density of Melbourne, including some measures that have minimum density threshold to restrict the calculation to “residential” areas. The maps above use 1000 people/km2 as a threshold for colouring, and this appears to include all “residential” areas, except for some very large block estates.

You can quickly see why trying to calculate an average density is a fraught exercise! Though the first two are trying to measure “urban density”, while the later are attempting to measure “residential density” (and note the threshold for residential density makes a big difference).

A density profile chart (as above) is clearly a good way to get around the defined area problem. But you still need to be consistent in the land parcel size you use when comparing cities. Not easy when comparing cities with different statistics agencies.

Land use map of Melbourne

Before I finish up, the other beauty of the mesh block data is that it contains a land use classification for each mesh block.

So it is really easy to produce a land use map of Melbourne (and Geelong for good measure):

What are those two black blobs I hear you ask? Essendon and Moorabbin Airports. Tullamarine and Avalon airports are actually classified agricultural.

And you will see residential areas stretching a fair way east of Frankston, and north of Craigieburn – though these are not actually developed. So it’s not perfect.

In fact, according to the data, there is a mesh block in Melbourne with 358 people living in an area of 420 square metres (852,700 people/km2). That’s 1.17 square metres of land space per person. Really? No, what appears to have happened is that almost every resident of the Burnside Retirement Village was registered to one tiny parcel of land. I suppose that’s census data for you!