Car use is decreasing and has decoupled from GDP growth

13 November 2013 — Dr Jeff Kenworthy, professor in Sustainable Cities at the Curtin University Sustainability Policy Institute, has been researching the relationship between car use and the development of GDP in cities. The following is a summary of his paper “Decoupling Urban Car Use and Metropolitan GDP Growth” published in World Transport Policy and Practice, 19.4, October 2013.

Increasing car use has historically been linked to increases in wealth, and was once considered “virtually a law of transport”. Global agencies dealing with the difficult question of whether the world can adjust to meet the challenge of climate change must find a mechanism to enable GDP and transport to be decoupled. Technological advances can reduce carbon for each kilometre of travel, but this is much more likely to be effective if at the same time the amount of kilometres travelled is also decreasing. Without this their global models will never be able to incorporate a valid way of achieving a future with a maximum of 2°C of temperature rise.

Examining both car use per capita (vehicle kilometres) and GDP per capita for two years (1995/6 and 2005/6) for 42 cities in the USA, Australia, Canada, Europe and Asia, in a single currency (eg, real US dollars pegged to 1995), may enable comparisons between cities in different countries, but allows for the possibility of distortions due to conversions between currencies; thus using real GDP in local currencies is preferable.

The main aim of the analysis here was to compare how the amount of driving per unit of GDP has changed over the 10 year period between 1995 and 2005 in each city, expressed as a percentage.

Table 1, showing the result of the analysis, reveals that the city with the biggest drop in car driving per unit of GDP was Madrid, with 50 per cent.

Table 1: Car kilometres travelled per unit of GDP in real local currencies, 1995-2005

Figure 1 shows the percentage changes in car kilometres per unit of real GDP for each city and reveals the extent to which car use decoupled from GDP over the 10-year period. It shows that thirty-nine out of the 42 cities reduced in their car kilometres per GDP by an average of 24 per cent (across the entire 42 cities the decline was 21 per cent).

These cities have thus been able to grow their economies while experiencing major reductions in the relative amount of car driving associated with this wealth creation (see further section on links to peak car use below). Overall, the data suggest that in the overwhelming majority of cases where real per capita GDP has grown, this wealth creation has decoupled from car use.

Figure 1: Percentage change in car kilometres travelled per unit of real GDP in 42 cities (real local currencies)

Figure 2 shows that in all regions except Asia, the car kilometres driven per dollar of GDP have declined from 1995 to 2005, with the European cities showing on average the least decline. Here we also see that in 2005, US cities experienced by far the most car use for every dollar of generated GDP, Australian cities were a little more efficient in this respect, while Canadian cities were clearly lower than both their US and Australian counterparts. European cities needed less than half the car use to generate the same amount of GDP as in US cities.

These have been the same patterns observed since I and my colleague Prof Peter Newman began collecting city data from the 1960s. The two relatively wealthy Asian cities of Singapore and Hong Kong had about one-fifth of the car use per GDP as the US cities in 2005. Similar patterns are evident in 1995, except that the gap between the Asian cities and the rest of the cities was much greater in 1995 than in 2005 (these Asian cities increased in their car kilometres per dollar of GDP while all the other regions declined). A similar situation is observable between the Canadian and European cities because of the small relative decline in car kilometres per dollar of GDP experienced in European cities.

Figure 2: Car use (vehicle kilometres) per dollar of GDP in US, Australia/New Zealand, Canadian, European and Asian cities, 1995 and 2005 (using real 1995 US dollars)

There may be specific economic circumstances, which could have relevance to interpreting Figure 2. For example, in the USA while national GDP per capita has risen considerably, the bulk of that economic growth has been within already very wealthy population segments. Meanwhile, the period from 1999 to 2009 saw a decline of 5 per cent in real median household income and by 2011 real median household income in the USA was at the lowest level it has been since 1995.

If most economic growth is accruing to very high income people while the bulk of households are not really getting any wealthier in real terms, then one might expect that (a) these already very high income people are not needing to drive more just because they are now even more wealthy, while (b) the average person is roughly as rich in real terms as before and therefore possibly driving about the same as well. The net effect in both cases would be a decline in car kilometres per dollar of real GDP.

Links to peak car use

Peak car use appears to be happening due to a combination of factors relating to the growth of the knowledge/services economy, the urban youth culture and use of social media, an increasing popularity of urban locations with rising urban densities and a revival in the use of transit, especially urban rail. These can all be contributing to the decoupling of car use from GDP. It is possible that GDP and car use could be decoupled, but with both still growing, albeit with GDP growing more strongly to give the effect of lower vehicle kilometres per unit of GDP.

Figure 3 explores this in more detail by examining the actual changes in car vehicle kilometres per capita for each of the 42 cities in this analysis between 1995 and 2005.

Figure 3: Annual per capita car kilometres in 42 cities, 1995 and 2005

A close examination of the chart reveals that 12 out of the 42 cities did in fact not only achieve relative reductions in the amount of car driving associated with their GDP growth, but also absolute declines in per capita car kilometres of driving (Atlanta, Houston, San Francisco, Los Angeles, Oslo, Toronto, Montreal, Zürich, Stockholm, London, Vienna, Graz). The average decrease in per capita car use for these twelve cities was 6.4 per cent.

Furthermore, 16 of the cities experienced per capita increases in car vehicle kilometres of less than 10 per cent over a decade, suggesting a significant slow down in the growth rate of per capita car kilometres. The remaining 14 out of the 42 cities in this analysis grew in car use per capita between 1995 and 2005 by an average of 20.8 per cent. The average decadal percentage increase in car vehicle kilometres per capita in cities from 1960 to 1970 was 42 per cent, from 1970-1980, 26 per cent and 1980 to 1990, 23 per cent. Overall, in this analysis of 42 cities from 1995 to 2005, those that increased in car vehicle kilometres per capita did so by 12.6 per cent, or about half the 1980 to 1990 figure. For the 42 cities as a whole, car use per capita increased on average by only 7.2 per cent, or less than one-third as was typical for the 1980s in similar cities.

If we examine the correlation between annual car kilometres of travel per capita and GDP per capita using the later 2005 data for the 42 wealthy cities here, there appears to be a very weak positive relationship, but with a huge amount of scatter in the data. As a result, GDP per capita, at least in wealthier cities, is an extremely poor predictor of car use per capita.

Total motorised personal mobility and GDP

With the reducing amount of car use (vehicle kilometres) associated with every real unit of GDP, it is interesting to see whether this also applies to the total amount of personal motorised travel that people have to undertake to meet their needs in cities. In other words, is car use simply being replaced by other forms of mobility in generating the GDP of an urban region? In this case we combine annual car, motorcycle and public transport passenger kilometres and express this per unit of real GDP in the local currency. The annual public transport passenger kilometres for each city is for the entire system.

Table 2: Total motorised passenger kilometres per unit of GDP in real local currencies, 1995-2005

Table 2 demonstrates that the large majority of cities needed significantly less total personal motorised mobility in 2005 relative to the amount of real GDP they generated, compared to 1995. There are only seven out of the 42 cities that increased in this factor. Overall, in the thirty-five cities that reduced in this factor, the average reduction was 26 per cent over the ten years, while across the entire sample, including those cities that increased, the overall reduction in motorised mobility per unit of GDP was 20 per cent.

Figure 4: Percentage change in total motorised passenger kilometres travelled per unit of real GDP in 42 cities (local currencies)

Figure 4 shows that the seven cities to have increased in total motorised mobility per unit of GDP from 1995 to 2005 were, Stuttgart, Oslo, Munich, Hong Kong, Frankfurt, Vancouver and Berlin. In the case of Hong Kong the mobility increase per capita was all in transit, with combined car and motorcycle passenger kilometres per capita declining. In the other six cities that increased, all the forms of passenger mobility per capita rose. In fact, only eight cities out of the 42 cities actually declined in per capita total motorised mobility from 1995 to 2005 (Graz, Stockholm, Melbourne, Montreal, Toronto, Atlanta, Houston, and Los Angeles).

In summary, in 83 per cent of the cities in the analysis, total motorised mobility relative to GDP reduced or decoupled between 1995 and 2005.

Grouping these figures into their respective regions reveals that in 1995, Australian cities had the highest level of overall personal mobility per dollar of GDP. This was followed closely by the US cities, then the Canadian cities with a big drop in this factor, then the European cities with another big drop and finally the two Asian cities which were lower again, but by a lesser margin than the other differences. By 2005, all the groups of cities had reduced their personal mobility requirements relative to GDP, except the Asian cities, which went up.

This meant that by 2005 the two Asian cities of Singapore and Hong Kong on average had more personal mobility per dollar of GDP than the European cities and were almost equal to the Canadian cities (but this mobility was mainly transit, not cars). Also the US cities overtook the Australian cities in 2005 by a tiny margin to become the leading cities in this factor, but the remainder of the comparative differences were similar to 1995. In general, in comparing 1995 and 2005 there seems to be a “flattening” process at work so that the differences in motorised personal mobility levels relative to GDP are becoming less pronounced.


Based on research by Kooshian and Winkelman (2011), it is clear that 1995 was essentially the peak in the travel intensity of the US economy and from that year on, notwithstanding some small variations, less vehicle travel was needed for every dollar of GDP generated. Between 1995 and 2005 there was approximately an 8 per cent drop in the Vehicle Miles Travelled (VMT) per dollar of GDP in the USA. By comparison, the data on the ten US cities in this study showed a 26 per cent decline. The data used by Kooshian and Winkelman represent total VMT, not just private passenger vehicles, so it seems that: (a) the decoupling effect is less when freight and commercial traffic is factored in; and, (b) the effect is more pronounced in cities.

From other analyses conducted on the global cities data it is clear that an increase in the relative speed of public transport compared to general traffic is a significant factor in transit’s renaissance in cities and especially in the rise in importance of rail in cities.

In the absence of any area-wide congestion control in any city this is hardly surprising, because buses suffer a worse fate than cars in congested conditions in terms of providing a speed-competitive mid-level public transport option. Yet it is the unrestrained use of cars in peak periods, which causes the congestion that so damages bus competitiveness. An even greater improvement in the relative speed of public transport could be gained by controlling congestion in a systematic way so that buses could compete effectively in speed terms with cars.

When public transport travel rises, it is not simply that one passenger kilometre by public transport replaces one passenger kilometre driven by cars. There is a “transit leverage effect” whereby one passenger kilometre by public transport replaces multiple passenger kilometres by car. It can be suggested then that as public transport gains in relative importance in urban mobility we might expect that the decoupling effect between GDP and car travel should actually strengthen. Essentially, public transport is a much more efficient and cost-effective way of providing urban mobility needs. Even if public transport mobility increases, it is quite possible that both car mobility and total motorised mobility per dollar of GDP can decline due to public transport’s ability to replace multiple kilometres of car travel.

This change in priorities, as cities recognise the greater value in public transport investment as well as the need for congestion control, enables us to see a major policy direction that could be the driving mechanism to enable cities of all kinds to continue their growth in wealth whilst decoupling from car use growth.


The data presented in this paper suggest that in general, personal mobility in relatively wealthy cities, and in particular car vehicle kilometres of travel, have decoupled from GDP growth. This trend is supported by independent research from the USA at both a national and city level. The results also appear to be in line with the idea of peak car use in these more prosperous cities around the world, with evidence that many cities experienced a decrease in car vehicle kilometres per capita from 1995 to 2005, while many others increased by an average of only around 5 per cent in the 10 years, a major reduction on growth rates in previous decades. The amount of total personal motorised mobility (car, motorcycle and public transport passenger kilometres) associated with generating one unit of real GDP appears to be also diminishing in the majority of cases, and the differences between wealthy cities in this factor are also declining. American and Australian cities, however, remain significantly higher than cities in Canada, Europe and Asia. Further research is needed to better understand these observed phenomena and their implications. This would include more detailed examination of Indian and Chinese cities to see if they too may be beginning to decouple transport from GDP as they move to significant investments in rail over roads. The ‘peak car’ phenomenon may well therefore be helping us to imagine a future where wealth can continue to be created globally whilst reducing the use of cars, oil and their damaging global impacts.