Active travel to work in NSW: trends over time and the effect of social advantage

Regular physical activity has been shown to prevent and help manage over 20 chronic conditions, including coronary heart disease, stroke, type 2 diabetes, cancer, obesity, mental illness and musculoskeletal conditions.¹ Australian guidelines suggest that adults should get at least 30 min of moderate-intensity physical activity on most days,² but almost 60% of Australians aged 15 years and over do not meet these guidelines.³ Promoting active commuting (walking or cycling to work) is one way to increase population physical activity and help realise public health benefits.

Governments have started to promote active commuting,^4,5 however overall rates of cycling or walking to work remain low in Australia.^6–8 Examining geographic areas within an urban environment where active commuting is increasing or decreasing may hold clues to the best methods to increase active travel, as does examining variations in active commuting by areas of social advantage/disadvantage. Further, if active travel is not equitably distributed among the population or across regions, there is potential to widen gaps in health outcomes.⁹

The influence of social advantage or disadvantage on active travel is not yet clear. Turrell et al. 2013 found that residents of socioeconomically disadvantaged neighbourhoods of Brisbane were more likely to walk for transport than their counterparts in advantaged neighbourhoods, which may actually contain or reduce neighbourhood inequalities in chronic disease.¹⁰ For cycling a different relationship may exist. For example, registered users of the public bicycle scheme in London were more likely to live in areas of low deprivation, and another health survey in New South Wales (Australia) found that people cycling to work had higher levels of education.^11,12 Similarly, greater income, habitual physical activity, and positive beliefs about physical activity were associated with utility and recreation only cycling among Brisbane residents.¹³ Utility cycling was also associated with some or no vehicle access, part-time employment, and perceived environmental factors (such as little crime, few cul-de-sacs, nearby transport and recreational destinations). Buehler et al. found that there was much less variation in active travel among socioeconomic groups in Germany (a high active travel country) than in the U.S (a low active travel country).¹⁴

The relationship between housing location and active commuting also needs to be explored further. Factors associated with urban environments such as housing density, mixed land use (shops, community buildings and houses in close proximity), distance to destination and cycle paths are all well established environmental correlates of active travel.^15,16 However, Rissel et al. 2013 found that people living in rural and remote areas of NSW were more likely to walk to work than people in major cities.¹² Low car ownership may be directly associated with more active commuting, but it is likely that car ownership also mediates the relationship between active commuting and housing location. In the inner city, space is limited, parking is expensive and traffic congestion is common. Good public transport links and good walking and/or cycling infrastructure further reduce the need to own a car in the inner city.^10,16–19

Previous studies investigating the predictors of active travel have typically analysed data by individuals. Analysis at Local Government Area (LGA) level or city/county level is also appropriate because many active commuting interventions, such as improvements to infrastructure, are implemented at an LGA level,^4,5 and characteristics such as socio-economic status and distance to work are likely to be shared by local residents. Further, intangible predictors such as a ‘culture of cycling’ are often common to, and influenced by, a geographic area. Finally, trends in walking and cycling in Sydney mirror each other, indicating that some of the same environmental influences may be important at the Local Government Area (LGA) level.^6,7

This study has two parts. First, we look at rates of active commuting by LGA in Sydney over time to examine trends in active commuting in different geographic areas. Second, we investigate associations between active commuting and Socio-Economic Index for Areas (SEIFA, an indicator of an area’s socioeconomic status), Accessibility/Remoteness Index of Australia (ARIA, an indicator of housing location) and car ownership across NSW with LGAs as the unit of analysis.

Part 1 of our analysis looks at the Sydney area whereas part 2 includes all of NSW.

Journey to work data from the 2001, 2006 and 2011 Australian Census of Population and Housing were purchased from the Australian Bureau of Statistics (ABS). The dataset included all respondents whose place of enumeration (the location where they spent Census night) was an LGA in NSW, and who answered the question ‘How did the person get to work on xxx date?’ (Census day 2001, 2006 or 2011). Respondents who selected that they worked at home or did not go to work were excluded. For cycling, respondents were able to tick more than one mode of travel, so that a journey to work using bicycle may or may not have also involved another mode of travel, such as train or ferry. For walking, respondents who ticked ‘walked’ only were included in this dataset, thus excluding people who used multiple modes of transport, most notably walking combined with public transport. These results therefore under-represent the total number of people who walk as part of their journey to work. The decision was made to exclude walkers who combined their commute with other modes of transport because the walking ‘leg’ of a multiple-leg trip is poorly recorded and is best captured with a travel diary, which was not an option for this analysis. In contrast, use of a bicycle in Sydney for any part of the journey is more memorable.

In order to avoid release of identifiable Census data, the ABS use a technique that slightly adjusts all cells to prevent any identifiable data being exposed. These adjustments result in small introduced random errors. However, the information provided in the table as a whole is not compromised.²⁰

Active travel data from the 2011 Census were used to calculate journey to work travel mode rates by each LGA in NSW. Population data for each LGA in NSW by age (5 year age groups) and sex from the 2011 Census were derived from ABS table-builder.²¹

Information about SEIFA²² for each LGA in NSW was sourced from the ABS and categorised into population weighted quintiles (2012 data). Similarly, ARIA for each LGA in NSW was also sourced from the ABS and categorised into Major City, Inner Regional, Outer Regional, and Remote (2004 data).²³

A measure of the average number of vehicles per household in each LGA was derived from the 2011 census information on the number of motor vehicles per household with the response categories 0, 1, 2, 3, > = 4. A weighted average was calculated for the entire LGA based on the relative proportion of each category.

Trends in active commuting over time were analysed for three regional groupings; ‘inner Sydney’ comprised LGAs mostly within a 10 km radius of Sydney Central Station (n = 17), representing an easily cyclable distance to the city centre; ‘outer Sydney’ LGAs were the remaining Sydney metropolitan LGAs as defined by the Regional Development Regulation 2012 – Regulation 4²⁴ (n = 23), and the ‘Greater Metropolitan Region’ grouping compromised all the other LGAs in the Greater Metropolitan Region (n = 13) (see Table 1). While there have been minor changes to LGA boundaries between 2001, 2006 and 2011, this analysis was done using current (2011) boundaries.

Table 1.  Characteristics of Local Government Areas (LGAs) of inner Sydney, outer Sydney and the Greater Metropolitan Region and rest of NSW*

The proportion of people actively commuting to work was calculated by dividing the number of journeys to work involving active travel by the total number of journeys to work. Change over time was calculated by dividing the difference at the two time points by the earlier time point. Differences in proportions of respondents reporting walking or cycling between 2001 and 2011 were tested using McNemar’s test.²⁵

The association between active travel to work (number travelled/population for each age and sex category) and SEIFA, ARIA and car ownership of place of enumeration (n = 152 LGAs) were estimated using a negative binomial regression model (initial analysis indicated over-dispersion using a Poisson distribution), with adjustments for age and sex profile of the LGA. Due to the correlation between LGA level variables (e.g. SEIFA, ARIA+ etc.), we estimated these associations both unadjusted (only adjusted for age and sex profile of the LGA), and in a model that mutually adjusts for all of the LGA-level variables. Models were checked by distribution of residuals and against deviance from a fully-specified model. We also inspected the correlation among the parameters in our negative binomial regression models and found reasonably low levels (< 0.5) between parameters of different variables. Before the fully-adjusted model was estimated, the covariates were introduced in a stepwise fashion. The coefficients and variance of coefficients was monitored for any unusual changes when variables were added (such as a large changes in value or estimate:variance ratio); no unusual changes indicative of collinearity problems were detected and all models reached convergence.

Between 2001 and 2011, the proportion of people walking to work in inner Sydney rose by 16% (relative rise), from an absolute rate of 8.71% to 10.12%, but fell by 7% in outer Sydney (from 2.79% to 2.59%), and by 14% in the Greater Metropolitan Region (from 3.44% to 2.94%).

A similar pattern occurred for cycling, with nearly all the increase in inner Sydney (relative rise of 83%, from 1.21% to 2.22%) being offset by declines in the Greater Metropolitan Region (19% relative decrease, from 1.13% to 0.92%).

The trends in walking and cycling over time are strikingly similar, and the pattern of active commuting overall reflects this. Between 2001 and 2011, active commuting increased in inner Sydney (relative increase of 24%, CI 24% to 45%), decreased slightly in outer Sydney (reduced by 5.1%, CI –5.1% to 8.9%) and declined in the Greater Metropolitan Region (declined 15%, CI –15% to –1.9%) (see Fig. 1).

Fig. 1.  Proportion of journeys to work made by walking, bicycle and overall active commuting (walking and cycling) in 53 local government areas in inner Sydney, outer Sydney and the Greater Metropolitan Region (GMR) by Census year.

For the combined Sydney and Greater Metropolitan Region (inner Sydney, outer Sydney plus the Greater Metropolitan Region), increased active commuting in inner Sydney slightly outweighed the declines in the Greater Metropolitan Region, resulting in a small but significant increase in proportion of people actively travelling to work between 2001 and 2011 (6.81% relative increase; CI 6.76–6.85%) (Fig. 2).

Fig. 2.  Proportion of journeys to work made by walking, bicycle and active commuting in the combined Sydney and greater metropolitan region by Census year.

The proportion of people walking to work in the combined Sydney and Greater Metropolitan Region rose just 3% (relative rise), from 4.38% in 2001 to 4.52% in 2011. The proportion of people cycling increased by 25% (relative rise), from 0.85% in 2001 to 1.06% in 2011; however 1.06% of all journeys to work made by cycling is still extremely low compared with other modes of transport and an increase of 0.2 percentage points makes negligible impact.

In unadjusted analysis, there was a relationship between SEIFA and active commuting, such that residents of LGAs with higher SEIFA quintiles (higher SES) cycled to work more and walked to work less than those with a lower SEIFA quintile (Table 2).

Table 2.  Associations between Socio-Economic Index for Areas, number of vehicles per household and Accessibility/Remoteness Index of Australia of Local Government Area (n = 152) and mode of travel to work in New South Wales

Upon adjusting for all other variables, the dose–response relationship between cycling to work and SEIFA band became more pronounced, with residents of those LGAs with the highest SEIFA quintile being over three times more likely to ride to work than those in the lowest quintile. The negative relationship between walking to work and SEIFA was attenuated in all bands except quintile 4 (Table 2).

When unadjusted for either SEIFA or vehicle ownership, it appeared that areas outside a major city were associated with less cycling to work, but more walking to work (Table 2); however after adjusting for these variables, our model showed that residents of LGAs with an ARIA band outside a major city were more likely to cycle and/or walk to work.

In the adjusted model, residents living outside a major city, whether inner regional, outer regional and rural, were around one and a half times more likely to ride to work, however no dose–response gradient between cycling and remoteness was evident (Table 2). The switch in association between ARIA and cycling from negative to positive under the unadjusted and adjusted model indicates that this relationship is heavily confounded by SEIFA and/or car ownership.

A clear dose–response relationship exists, however, between remoteness and walking to work, with residents in very remote areas being around four and a half times more likely to walk to work, those in outer regional areas being around two and a half times more likely to walk and those in inner regional areas around one and a half times more likely to walk to work than those in a major city (Table 2).

In unadjusted analysis, there was a strong positive association between low car ownership and high cycling rates, and a weaker association between car ownership and walking, not being significant until a household had less than one car (the lowest car ownership category) (Table 2).

When fully adjusted, the model showed that residents of LGAs with low car ownership were significantly more likely to both ride and/or walk to work than residents of LGAs with higher car ownership.

A clear dose–response gradient exists between car ownership and riding to work in NSW, with lower car ownership being associated with higher cycling rates. Residents of LGAs where the average car ownership was less than 1.3 cars per household were around 11 times more likely to ride to work, and around two and a half times more likely to walk to work than LGAs with car ownership 1.9 or more (Table 2).

Between 2001 and 2011, active commuting increased in inner Sydney, decreased slightly in outer Sydney and declined in the Greater Metropolitan Region. Overall, the increases in inner Sydney were mitigated by the declines in the Greater Metropolitan Region, so active commuting for the combined Sydney and greater metropolitan area remained stable. Similar patterns were observed for both walking and cycling.

After adjusting for age, sex and all other LGA variables, people living in LGAs with high socioeconomic status, higher rurality and low car ownership were more likely to cycle or walk to work. A clear dose–response relationship was found between cycling and socio-economic status and car ownership, and between walking and rurality.

Our analysis found that active travel to work is low and decreasing in outer Sydney and the Greater Metropolitan Region, despite increases in inner Sydney. This decline should not be considered an inevitable consequence of urban development. In the UK, three towns provide good case studies for successful travel behaviour change programmes in outer areas. Worcester, Peterborough and Sutton implemented a suite of strategies including strong marketing and branding, walking, cycling and public transport strategies, and parking and road network strategies, in conjunction with the overarching travel demand management strategy and lead to reduction in car trips by between 6% and 9%, increases in walking by between 3% and 14% and increases in cycling by between 12% and 75%. These results are achievable in outer and Greater Metropolitan Sydney if a similar suite of programs were implemented,²⁶ and programs must focus on these areas to promote equity.

The relationship between a person’s income and cycling remains unclear, with different countries reporting different relationships;²⁷ however in our study population which is part of a low cycling prevalence country, residents of LGAs with higher socioeconomic status were more likely to cycle to work. This is consistent with the profile of registered users of the public bicycle scheme in London, who tend to live in areas of low deprivation and high cycling prevalence.¹¹ Steinbach et al. argue that the identities of some professional (largely White) men and women could be bolstered by cycling, but that ‘the aesthetic and symbolic goals of cycling were less appealing to those with other class, gendered and ethnic identities’²⁸.

Our findings may be partly explained by living in the inner city, proximity to workplace and car ownership i.e. people who live in the inner city (close to many workplaces) are likely to be well off (housing is expensive) and have fewer cars (car parks are rare, and there is access to good public transport). People living in high socioeconomic status LGAs may also be better educated, both generally and in health-related matters and therefore better understand the importance of physical activity. Education, one component of socioeconomic status, has been shown to be independently associated with active travel on an individual level,¹² and may partly explain the association because it may be a marker of health literacy. The situation may be quite different in In those countries in which cycling is very common, such as Belgium and the Netherlands, where women cycle more and there is little difference in the socioeconomic profile of those cycling.²⁷

The higher level of active commuting in rural and remote areas is consistent with previous Australian research,¹² but is harder to explain. In the Netherlands, small- and medium-sized cities have higher bicycle share than large cities,^29,30 probably as a result of the proximity of the destinations involved,²⁷ and this may also be the case in rural and remote Australian towns. It is also possible that roads in rural areas are less hostile environments for active commuting. Dill and Voros (2007) found that cyclists have a negative perception of roads with high-traffic intensities, but smaller towns are likely to have fewer cars and wider footpaths and may therefore be more attractive to walkers and cyclists.³¹

The Census is a complete enumeration of active commuting behaviour in Australia; however results may be influenced by population-level variations in circumstances, for example bad weather or local events influencing transport choice on Census day. It is well known that active travel is affected by rain and other weather variables.^27,32 The Census is conducted in August, which is winter in NSW, so the timing of this data collection may have underestimated active commuting over the whole year. Similarly, the data report on only one day, which may not be representative of regular modes of travel. Finally, the data do not account for individual level factors associated with active commuting.

These data do not include activities besides walking and cycling, or walking or cycling for recreation, sport or other utility journeys and therefore underestimates total physical activity. Additionally, past research has shown that roughly half of all kilometres walked are associated with a public transport trip.³³ As this dataset excluded all walking journeys with more than one mode of transport, it is likely also to have significantly underestimated total transport walking. Further, each LGA may be quite large, and is unlikely to be homogeneous in terms of SES. Readers should bear this in mind when interpreting findings of this study. Finally, as the census is a cross-sectional survey, our results show associations between active commuting and descriptive factors, but this remains a statistical association, rather than a causal relationship.

The association between SEIFA and ARIA is well known³⁴ and is evident in our adjusted and unadjusted models; however these variables are commonly used together in various models as additional information is gained through their use.^35,36 The switch in association between ARIA and cycling from negative to positive under the unadjusted and adjusted model indicates that this relationship is heavily confounded by SEIFA and/or car ownership.

Increasing active travel is a priority for the NSW Government³⁷ because of the benefits to cities as well as to individuals. Reduced traffic congestion and improved public amenity go hand in hand with health benefits such as lower Body Mass Index¹² and reduced risk of diabetes;³⁸ however our research shows that more needs to be done to ensure that the benefits of active travel are distributed equitably to Australians with lower SES, and policies also include outer Sydney and metropolitan regions. A combined approach to travel management which includes marketing, transport strategies, and parking and road network strategies would be a start. Achieving change clearly requires input from multiple agencies, including roads infrastructure, public transport and health promotion, but successfully normalising active commuting in Australia may well achieve levels of population physical activity that have rarely been reported in purpose-designed physical activity interventions.