Public transport mode share – according to household travel surveys

Sat 10 April, 2010

[post revised and updated October 2012 with new data from Sydney, Brisbane, and New Zealand]

Arguably the best source of public transport mode share statistics is from household travel surveys that are conducted in most large Australia cities and all of New Zealand (unfortunately some surveys more regularly than others). A common measure is public transport’s share of motorised trips (although public transport will also be competing with unmotorised transport modes).

In household travel survey speak, a linked trip is a journey between two distinct non-travel activities, and may involve several trip legs or unlinked trips. For example, if you walk to a bus stop, catch a bus to the train station, then catch a train to the city, then walk to your workplace, that is one linked trip made up of 4 unlinked trips (walk, bus, train, walk). Similarly if you drive from your home to your workplace, that’s one linked trip made up of one unlinked trip (unless you decide to count walking to and from the car). Hence mode share figures that relate to unlinked motorised trips will always be higher than mode share figures that relate to linked trips.

The data I have been able to obtain for cities is sometimes linked trips, sometimes unlinked trips, and sometimes both. It should be possible to get figures for both for any city, and I hope to obtain such data from state transport agencies in the future.

Here is the data I have for linked trips:

And here are the results for unlinked trips:

The Melbourne and Sydney measures are for weekdays only, whereas the New Zealand data appears to be for all days of the year.

In 2008, Melbourne appeared on track to overtake Sydney on unlinked trip public transport mode share, however the 2009-10 result for Melbourne was lower than predicted. Note that the error bars on the 2007-08 and 2009-10 VISTA survey results for Melbourne indicate the actual mode share might not have actually gone down significantly (similar error bars would apply to the linked trip data points). Over the same period public transport patronage grew by 11% and arterial road traffic grew by around 1.2%.

How reliable is this data?

Given that most household travel surveys interview thousands of households in any one year, the results should be pretty accurate for a high level reported figure such as mode share of trips. Household travel survey techniques have matured over the years, so it is likely they are reasonably reliable (particularly more recent results in larger cities).

The Perth survey data for 2003 to 2006 does not correlate with public transport patronage figures, that show a 12% growth over the same period.

For Brisbane 2003-04 I had to add whole number shares for each mode and divide by the sum of motorised mode shares. So there is some uncertainty about the precise motorised mode share.

The Melbourne official estimates for 2002-2007 were calculated using VicRoads traffic data, and public transport patronage figures.

(For more detail see the end of this post).

Linked or unlinked trips?

Calculating mode share based on linked trips removes the impact of public transport transfers. Cities where the public transport network is structured around feeder services with free transfers (eg bus to train) may have more public transport boardings (unlinked trips) than cities where transfers are “less encouraged” by the network design and fare systems (eg Wellington, Auckland, Sydney).

In fact, here is a chart showing the ratio of unlinked to linked public transport trips for four cities where I have data:

The Perth and Adelaide data is based on patronage figures that are reported as ‘initial boardings’ and ‘all boardings’. Annual reports comment that recent through-routing of bus services through the Adelaide CBD may have reduced the number of transfer boardings. You can see the transfer rate for Perth jumped after the southern suburbs railway opened at the end of 2007 (replacing many CBD bus routes with train feeder bus routes).

The Perth, Adelaide and Melbourne public transport fare systems are dominated by products that allow unlimited transfers within a time window (anywhere from 2 hours to 365 days). So while there may be a time and convenience penalty for transferring between two services, there is no financial penalty. Sydney’s public transport fare system has largely involved tickets for a single trip and/or one mode, such that another fare must be paid to transfer. Sydney’s CBD is also served by seemingly hundreds of bus routes – many of which parallel train lines – which enable people to travel to the city without having to transfer onto trains and pay a higher fare (even if that could provide a faster over journey).

The lower Sydney transfer rate partly explains why Melbourne and Sydney are much closer on mode share of unlinked trips, compared to mode share of linked trips. Network design will probably also have an impact.

There was a slight dip in the trend for Sydney around 2007-08 followed by a rise. I’m not sure what might explain that trend – the revamp of the fare system in April 2010 (introducing more multi-modal and multi-operator tickets) may have had a small impact on the 2009-10 figure.

The difference in these rates suggests that there could be quite substantial change in Sydney public transport use patterns should the fare system be revised to make free transfers the norm. Perhaps this might help ease the bus congestion issues in the CBD and allow higher bus frequencies in the suburbs? (assuming there is capacity to transfer bus passengers onto trains in the suburbs). There is one small area of Sydney where train+bus link tickets are available (no fare penalty for transferring), and the census data reveals a very significant rate of bus+train journeys to work in the Bondi Beach area, much higher than anywhere else in Sydney.

Other measures of public transport mode share

In another post, I looked at BITRE data on estimated passenger kms per mode in Australian cities (presumably calculated using patronage figures and average trip lengths from household travel survey data or elsewhere). That enabled calculation of estimated public transport mode share of motorised passenger kilometres, with continuous time series available for all Australia cities. However there will be many assumptions involved in these estimates.

Another measure is boardings per capita (covered here), although this also has the problem of different transfer rates in different cities.

The quest for a fair measure of public transport use continues!

Household travel survey sources:

Melbourne: Victorian Department of Transport (personal communications), but also available in the Growing Victoria Together Progress Report (page 387), in the 2009-10 Victorian State Budget Papers. Figures until 2001 were from the VATS survey, while the 2008 result is from the VISTA survey.

Sydney Household Travel Survey: Data was supplied by NSW Transport Data Centre by email. Public transport trips are inclusive of trains, buses, ferries, monorail and light rail.

Adelaide Household Travel Survey (AHTS): Adelaide Travel Patterns: an overview (if anyone can tell me about whether more recent surveys have been conducted I would be very appreciative, better still if I can get results data!).

South East Queensland Travel Survey: Brisbane Fast Facts Brochure (unclear dating, but PDF was created in 2006 so I assume the results are for 2003-04. The report does not mention whether these are mode shares for trips or kms, however it seems highly likely they are for trips as the walking mode share was 10% and we know walking trips are generally shorter than motorised trips). I also have results for 2008-09 courtesy of Ian Wallis and Associates. I unfortunately do not yet have results for the 2006-2008 survey.

Perth and Regions Travel Survey (PARTS): Data is from the PARTS Key Findings Report (by Data Analysis Australia). The  2003-2006 results are from PARTS, the 2000 figure is a TravelSmart estimate, and 2001 and 2008 estimates are from unspecified sources.

The New Zealand Household Travel Survey: Because of sample sizes, the figures for the New Zealand cities are two years combined (ie the “2010” figure is for 2008/09 and 2009/10). The Canterbury region includes Christchurch as well as a not insignificant surrounding population. The Auckland region is more similar to the Australian cities statistical divisions. The Wellington figures are for the Wellington Region, but are dominated by metropolitan Wellington.

6 Comments | Australian Cities, Melbourne, Mode share, Mode shift, New Zealand cities, Sydney | Permalink
Posted by chrisloader

Car and transit use per capita in Australian cities (Peak Car)

Fri 8 January, 2010

[post updated in January 2016 with data from the 2015 BITRE Yearbook. For some more recent data see this post published in December 2018. First published January 2010]

Thanks to the Bureau of Infrastructure, Transport and Regional Economics’ Australian Infrastructure Statistics Yearbook 2015, a great set of time series data is available on transport behaviour. This post looks at trends in private and public transport use in Australian cities up until 2013-14.

The first chart shows car passenger kms per capita peaked in 2004 for all cities and has been mostly in decline since then (with the possible exception of Adelaide in recent years).

car pass kms per capita 4

This is clear evidence of “peak car” use per capita in Australian cities.

Canberra has the highest car passenger kms per capita – perhaps due to the low density city, relatively sparse bus services, and general ease of using private transport.

The underlying figures show relatively little growth in total car passenger kms in most cities in the ten years since 2003-04:

index of car passenger kms 2

Perth has shown the greatest increase in total car passenger kms despite the reduced car use per capita, presumably mostly reflecting strong population growth. Canberra, Sydney and Melbourne show a flat-lining in car passenger kms between 2004 and 2009 (a period of time that is becoming very familiar on this blog). I’m not sure why Adelaide has seen declining car passenger kms (it’s population did grow over this period by 10%).

Does this mean we are travelling less? The following chart shows estimated motorised passenger kms per capita, and yes in all cities there was a peak in 2003-04 followed mostly by declines. This might be people taking shorter trips, people taker fewer trips, and/or people substituting motorised transport with non-motorised transport.

motorised pass kms per capita 4

There is another story in the data. Mass transit passenger kms per capita rose significantly in Melbourne and Perth between 2005 and 2011:

mass transit share of pass kms 4

However, more recently there have been slight declines in Brisbane, Perth, Adelaide, and Canberra.

Sydney shows up with many more mass transit kms per capita than any other city (68% of which is on rail). While I am not sure exactly how BITRE sourced their train patronage data for “Sydney”, my experience is that patronage data is only readily available for a very wide definition of Sydney’s railways including lines beyond the Sydney Statistical Division including Newcastle, Nowra, and the Hunter Valley.

You can also see a spike in mass transit use in Sydney in 2000/01 – the year containing the Olympic Games.

Previous strong growth in Brisbane might reflect considerable investment in bus services (BITRE have published three years of bus data showing Brisbane increasing from 53.0 million kms in 2005-06 to 61.2 million kms in 2007-08). More recent stagnation in mode share might reflect above-CPI fare rises and changes to patronage measurement methodology.

The data also allows a calculation of mass transit’s share of motorised passenger kms:

Melbourne and Perth are the stand outs for mass transit mode shift based on these figures, particularly in the years leading up to 2009. Adelaide, Sydney and Brisbane also saw mode shift between around 2004 and 2009, but Adelaide and Brisbane have gone backwards since then.

Note there are other ways to more directly measure transport mode share – primarily household travel surveys, covered in another post.

Finally, here is a stark comparison of total car and mass transit passenger kilometre growth since 2003-04 (with population growth included for reference):

car v pt growth aus large cities 2

Car use has grown more than four times slower than population growth, and almost seven times slower than mass transit use.

Business as usual no longer

In the frequently cited BITRE 2007 report on congestion costs, they made a “business as usual” assumption that mode shares will not change (ref page 8) and that car use per capita would increase (ref page 47), and then forecast that avoidable congestion costs in Australia will more than double between 2005 and 2020. The latest BITRE evidence is that mode shares are not business as usual in most cities and that car use per capita is decreasing. See another post reviewing the forecasts of the costs of congestion.

An updated 2015 BITRE report on the costs of congestion looks at the issues of per capita vehicle user into the future in a bit more detail, including acknowledging researchers highlighting peak car.

Their new high case scenario assumes the recent downturn in per capita vehicle use is essentially all GFC related (and has persisted for 10 years), and that things will now bounce back to old patterns (and grow even faster than the 2007 forecast). Their medium case also assumes a reversal of trend, but that we’ll only return to the previous peak level by 2030, and a low case continues the current trend.

The following chart shows forecasts compared to estimated actual vehicle kilometres per capita.

BITRE vkms per capita estimates

The 2007 forecast was not only trending in the wrong direction, their starting assumption about vehicle km per capita for 2007 was well above the (revised) estimated actual.

The new estimates for the “avoidable social costs of congestion” in 2030 are $48.2b for the high case, $37.3b for the median case, and only $25.1b for the low case. BITRE suggest this median estimate is the upper range of what is plausible, but if current trends continue, the avoidable social costs of congestion would be a third lower.

See also a recent piece on this subject by Professor Peter Newman in The Conversation.

Some disclaimers:

  • I’ve called it “mass transit” as the data does not seem to differentiate public, school and private bus passengers. Passengers on chartered buses aren’t usually considered “public transport”, but they still are a very space efficient way to move people on roads.
  • km figures are reported on financial years, while population figures are for June at the end of that financial year. So the “real” per km figures are slightly less, but I’m really just looking at trends and relative numbers here.

1 Comment | Australian Cities, Melbourne, Mode shift, Road Traffic | Permalink
Posted by chrisloader

Next Entries »