Charting Transport

Updated public transport patronage trends

Mon 14 November, 2011

Just a quick note to let you know of an updated post on public transport patronage trends. I’ve got almost all 2010-11 data for Australian and now New Zealand cities.

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.

Leave a Comment » | Uncategorized | Permalink
Posted by chrisloader

Are congestion costs going to double? An analysis of vehicle kms in Australian cities

Tue 25 October, 2011

A frequently cited forecast is that the avoidable costs of congestion in Australia will double in most Australian cities between 2005 and 2020. These BITRE forecasts were published in 2007 (Working Paper 71), assuming continued strong growth in vehicle kms in our cities (“business-as-usual” conditions). But as this blog has demonstrated several times, transport trends have not been business-as-usual in recent years.

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.

I’d like to acknowledge BITRE for conducting the excellent work that went into Report 124 and making the data publicly available, without which this analysis would not have been possible.

Leave a Comment » | Australian Cities, Melbourne, Mode shift, Road Traffic | Tagged: Auckland, BITRE | Permalink
Posted by chrisloader

What’s happening with car occupancy?

Sat 20 August, 2011

Is car occupancy trending down as car ownership goes up? What factors influence car occupancy? What is the impact of parents driving kids to school?

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!

3 Comments | Australian Cities, Car ownership, Melbourne, Mode share, Road Traffic | Tagged: AustRoads, Car Occupancy, Census Journey to Work, VISTA | Permalink
Posted by chrisloader

Trends in car ownership

Sun 7 August, 2011

Is the rate of car ownership still growing in Australia?

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).

3 Comments | Car ownership, Melbourne, Mode shift, Pricing | Tagged: Motorcycles | Permalink
Posted by chrisloader

Charting Transport

Updated public transport patronage trends

Are congestion costs going to double? An analysis of vehicle kms in Australian cities

How are we tracking with forecast traffic volumes?

What cost of congestion might we have avoided?