Background
Methods
Scoping review of published associations between obesity, PA, walking or cycling and fuel price or taxation
Scoping review of published cross price elasticities of public transport demand for the Australian context
Data selection and cost-effectiveness modelling
Variables | Data source | ||||
Total population estimates (population numbers, mortality rates, BMI distribution) | ABS Census 2011 [38] | ||||
Disease epidemiology, relative risks, disability weights, total years of life lived with disability | GBD 2010 [43] | ||||
Disease healthcare costs | AIHW 2004 [45] | ||||
Transport mortality data | Australian Road Deaths Database [98] | ||||
Transport morbidity data | |||||
Variables | Mean values and 95% UIa (where applicable) | Data source and assumptions | |||
Prevalence of using public transport for commuting purposes |
Males
|
Females
| ABS Census 2011 [38] | ||
18y | 4.5% | 18y | 6.9% | ||
19y | 5.8% | 19y | 8% | ||
20-24y | 8.5% | 20-24y | 11.1% | ||
25-29y | 11.7% | 25-29y | 13.1% | ||
30-34y | 11.1% | 30-34y | 9.9% | ||
35-39y | 9.1% | 35-39y | 6.8% | ||
40-44y | 7.4% | 40-44y | 5.9% | ||
45-49y | 6.3% | 45-49y | 5.7% | ||
50-54y | 5.8% | 50-54y | 5.3% | ||
55-59y | 4.9% | 55-59y | 4.5% | ||
60-64y | 3.3% | 60-64y | 2.9% | ||
Cost of legislation (including RIS process) | AUD1,090,792 (95% UI AUD939,805–1,249,710) | Sampled from a gamma distribution, taken from estimates from Lal et al. [49]. | |||
ABS average weekly earnings | AUD1,241 (95% UI AUD1,126–1361) | ||||
Number of businesses affected | 185,959 (95% UI 172,747–199,317) | Sampled from a pert distribution (most likely = 186,097) quoted from Government source, +/−10% [48]. | |||
Total intervention cost | AUD4,381,691 (95% UI AUD3882,683–4,903,984) |
Results
Results from the scoping review of published associations between obesity, PA, walking or cycling and fuel price or taxation
Study | Location/Population | Study aim | Method | Variable of interest (Outcome) | Relevant findings | QA |
---|---|---|---|---|---|---|
Obesity
| ||||||
Courtemanche 2011 [57] | USA Adults (n = 1,807,266) | To estimate the effect of fuel price on weight and obesity, by looking at its effect on PA, frequency of eating at restaurants and food choices at home. | Cross-sectional | Fuel price (BMI (S)) | USD2004 $1 increase in fuel price reduces BMI by approx. 0.35 units (s.e. 0.050, p < 0.01). A permanent USD1.00 increase in fuel prices would, after 7 years, reduce U.S. overweight prevalence by approx. 7% and obesity by approx. 10%. | 8 |
Rabin et al. 2007 [58] | 24 European countries | To describe obesity patterns and examine macro-environmental factors associated with obesity prevalence. | Ecological, cross-sectional | Fuel price (Prevalence of obesity using BMI (S)) | The price of fuel was associated with obesity prevalence for females (b = −0.096, p-value 0.041) and overall (b = −0.095, p-value 0.0542), but not for men. | 5 |
Sun et al. 2015 [59] | 47 low-middle income countries | To identify CVD risk factors in low-middle income countries. | Ecological, cross-sectional | Fuel price (Prevalence of obesity using BMI (S)) | The price of fuel was not statistically significantly associated with obesity in either men or women. | 5 |
Physical activity
| ||||||
Hou et al. 2011 [60] | Birmingham, Chicago, Minneapolis and Oakland, USA. Young adults 18–30 years at baseline (n = 5115) | To investigate longitudinal associations between fuel price and physical activity. | Longitudinal cohort | Fuel price (Leisure PA - energy units (S)) | A hypothetical USD0.25 increase in fuel price significantly associated with increase in energy expenditure (9.9 energy units (EU), 95% UI: 0.8–19.1, p-value 0.03). Equivalent to an increase in walking per week of 17 min. After controlling for all covariates, an USD0.25 increase in fuel price was associated with 1.3 EU increase in walking score (p = 0.2), equivalent to an additional 3 min of walking per week. Results suggest relatively weak association between fuel price and walking. No significant association for cycling. | 7 |
Sen 2012 [61] | American adults 15 years plus (n = 81,957) | Uses data from the time of fuel price rises due to Hurricane Katrina to explore effect on PA. | Cross-sectional | Fuel price (PA, defined five ways: (1) walking, running, bicycling or rollerblading as part of LTPA, (2) walking or cycling to work or errands, (3) playing with kids, (4) housework of MET > = 3, (5) total time spent on all PA MET > =3. (S)) | Higher fuel prices show some association with increases in LTPA (sig. at p < 0.05). Walking and bicycling to work or errands statistically weak and sensitive to model specification. Only one approach resulted in p < 0.05 with OLS estimate 0.74. Changes in participation and time spent in walking or cycling are not large. No association was found between higher fuel prices and PT use, although may be due to lack of accessibility to PT. | 8 |
Sen et al. 2014 [62] | American high school students grades 9–12 (n = 58,749) | To examine the relationship between fuel price and driving behaviours in teens. | Cross-sectional | Fuel price (moderate PA, defined as: (1) whether participates in PA “that did not lead to sweating or breathing hard”, and (2) whether participates more than five times per week or not (S)). | Higher fuel prices positively associated with higher levels of moderate PA. Higher fuel prices associated with moderate PA at least 1 day of the week for females (ME = 3.25%, t-stat = |2.90|), males (ME = 2.32%, t-stat = |2.36|), other races (ME = 3.01%, t-test = |2.16|), and teens ages 16 years and younger (ME = 3.98%, t-stat =14.70|). Higher fuel prices were associated with frequent moderate PA for females (ME = 1.92%, t-stat = |2.19|), males (ME = 3.63%, t-stat = |4.16|), non-Hispanic whites (ME = 3.88%, t- stat = 12.511), other races (ME = 3.85%, t-stat = |2.27|), and teens ages 16 years and younger (ME = 3.54%, t-stat = |4.54|). | 7 |
Cycling
| ||||||
Buehler & Pucher 2012 [63] | USA, population of 90 cities | To examine the association between levels of cycling and cycle infrastructure. | Cross-sectional | Fuel price (share of workers commuting by cycling (S)) | State fuel prices had a significant positive correlation with cycling levels (correlation coefficient 0.5, sig. at 95%), consistent with the theory that higher fuel prices may lead to more cycling to work. | 5 |
Dill & Carr 2003 [68] | USA, population of 35 large cities | To explore associations between cycling infrastructure and cycling. | Cross-sectional | Fuel price (share of workers commuting by bicycle (S)) | Although results on fuel price were not explicitly reported, authors state that fuel price was not statistically significant. | 4 |
Pucher & Buehler 2006 [67] | USA/Canada Population of 18 cities | To explore higher cycling rates in Canada than US. | Cross-sectional | Fuel price (share of workers commuting by cycling (S)) | Higher fuel prices are associated with higher rates of cycling to work (coefficient 3.040 (s.e. 1.159, significant at 95% level, adjusted R2 0.596). | 6 |
Rashad 2009 [64] | USA, metropolitan area residents (BRFS n = 146,730 NPTS n = 73,903) | To determine the relationship between cycling and fuel price. | Cross-sectional | Fuel price (cycling, defined as (1) cycled for pleasure in past month or (2) cycled in a trip yesterday (S)) | Increasing fuel price by $1 increased the probability of cycling by between 1.6% (t-stat 3.30, p < 0.01) and 4.7% (t-stat 2.17, p < 0.05) for men. Results for women ranged between 1% (t-stat 5.11, p < 0.01) and 3.5% (t-stat 3.05, p < 0.01). | 7 |
Smith & Kauermann 2011 [65] | Residents of Melbourne, Australia | To examine the determinants of cycling, including the cross-price elasticity of cycling. | Cross-sectional | Fuel price (cycling volumes (O)) | Substitution into cycling as a mode of transport observed in response to increase in fuel prices, particularly during peak commuting periods and by commuters originating in wealthy and inner city neighbourhoods. Cross-price elasticities vary depending on loop data and method used and time of day, from approximately 0.18 to 0.48 during peak commuting periods, significant at either 5% or after Bonferroni adjustment. | 7 |
Walking
| ||||||
Ryley 2008 [66] | West Edinburgh, adults living in West Edinburgh who drive (n = 627) | To estimate propensity for motorists to walk for short trips, based on changes to fuel price, journey time, parking costs. | Cross-sectional, discrete choice modelling, using stated preference data from survey | Fuel price (propensity to walk (SP)) | Fuel coefficient − 0.4159, significant at 5% (t-value −6.3). Fuel price had lower relative influence than parking costs, or journey time. | 4 |
Results from the scoping review of published cross price elasticities of PT demand
Source | Type of study | Estimate |
---|---|---|
BITRE database [33], Table 1D03 | Cited from study by Goodwin [101] | 0.34 |
BITRE database [33], Table 2D18 | Cited from studies by Cervero 1990 and Wang & Skinner 1984 (further details not given) | 0.08 to 0.80 |
Currie & Phung 2006 [74] | Review within primary study | 0.07 to 0.8 |
Currie & Phung 2008 [75] | Review within primary study | LR: 0.07 to 0.30 |
Holmgren 2007 [78] | Review Meta-analysis | 0.38 (s.e 0.31) SR: 0.82 (95% UI 0.56 to 1.08) LR: 1.15 (95% UI 0.65 to 1.65) |
Kennedy & Wallis 2007 [76] | Review | 0 to 0.20 |
Litman 2016 [73] | Non-systematic review | SR: 0.05 to 0.15 LR: 0.2 to 0.4 |
Luk & Hepburn 1993 [77] | Review | SR: 0.07 |
Results from cost-effectiveness modelling
Parameter | Mean values and 95% UIa (where applicable) | Sources and assumptions |
---|---|---|
Cross price elasticity for PT with respect to fuel price | 0.07 | Derived increase in the prevalence of PT commuting of 0.61% [38]. Modelled to PA/BMI effect (Appendix 3). Assumed all new PT users were previous car drivers, a reasonable assumption given the high prevalence of driving to work in Australia [38]. |
Average annual retail fuel price (national, metropolitan) (cents per litre) | 125.39 cents (95% UI 124.95–125.86) | Sampled from a gamma distribution, from national metropolitan fuel price [102]. |
Marginal METb value for walking to access PT | 2.5 (95% UI 0.7–6.4) | |
Average distance a person will walk to access PT (metres) | 400 | |
Comfortable gait speed (cm/s) |
Males
18-29y = 139.2 (95% UI 110.5–172) 30-39y = 145.7 (95% UI 128.4–164.2) 40-49y = 145.6 (95% UI 115.6–180.4) 50-59y = 139.5 (95% UI 100.5–192.2) 60-64y = 136.3 (95% UI 100.9–179.7)
Females
18-29y = 140.3 (95% UI 109.3–177.4) 30-39y = 140.8 (95% UI 117.5–166.9) 40-49y = 139.2 (95% UI 111.5–172.1) 50-59y = 139.5 (95% UI 112.2–170.8) 60-64y = 129.6 (95% UI 90.8–172.7) | Sampled from a lognormal distribution, taken from estimates from Bohannon 1997 [106]. Using average distances and gait speeds this results in an average increase in walking to access PT of 18.9 min per day in men and 19.2 min per day in women. This falls within the range summarised by Rissel et al. [30] of 8 to 33 min PA associated with PT use. |
Number of weeks of intervention effect (averaged over year) | 49 (95% UI 46–52) | Sampled from a uniform distribution based on estimate of number of working weeks per year for full-time workers. |
Results per scenario
| Total HALYs saved | Total healthcare cost savings (AUD 2010) | Net cost per HALY saved (with cost offsets) (ICER, AUD 2010) | |
---|---|---|---|---|
Main scenario | Main scenario BMI/PA/injury effect | 237 (95% UI 138–351) | $2,552,925 (95% UI $1,304,017–$3,905,568) | $7702 saved per HALY (95% UI $1366–$22,125) (Probability of cost-effectiveness 99.7%) (Probability of cost-saving 0.8%) |
Main scenario BMI effect only | 195 (95% UI 85–314) | $2,310,366 (95% UI $962,352–$3,762,993) | $10,514 saved per HALY (95% UI $1843–$39,990) (Probability of cost-effectiveness 98.4%) (Probability of cost-saving 0.3%) | |
One-way sensitivity analyses | Cross price elasticity 1 (0.82 from Holmgren [78]) | 2769 (95% UI 1614–4056) | $29,928,506 (95% UI $15,124,893–$45,413,548) | Dominant (Probability of cost-effectiveness 100%) (Probability of cost-saving 99.95%) |
Cross price elasticity 2 (1.15 from Holmgren [78]) | 3882 (95% UI 2233–5714) | $42,000,179 (95% UI $20,713,001–$63,854,358) | Dominant (Probability of cost-effectiveness 100%) (Probability of cost-saving 100%) | |
Distance walked 800 m | 472 (95% UI 258–705) | $5,098,746 (95% UI $2,422,181–$7,810,093) | Dominant (Probability of cost-effectiveness 99.95%) (Probability of cost-saving 71%) | |
“Plausible case” | “Plausible case” scenario – BMI/PA/injury effect | 3181 (95% UI 1797–4633) | $34,239,586 (95% UI $17,433,480–$51,336,591) | Dominant (Probability of cost-effectiveness 100%) (Probability of cost-saving 100%) |
“Plausible case” scenario – BMI only | 2532 (95% UI 1084–4098) | $30,222,697 (95% UI $12,875,579–$47,444,286) | Dominant (Probability of cost-effectiveness 100%) (Probability of cost-saving 99.9%) |
Cost or cost savings per new PT user | Values (AUD) | Source/Estimate |
---|---|---|
Vehicle operating cost (VOC) savings | ||
Annual petrol cost savings per new PT user (out-of-pocket cost savings for fuel saved) | $492.08 | |
Repairs and maintenance cost savings | $197.26 | |
VOC SAVINGS FOR THOSE NEW TO AT a
| $689 | |
Including parking charges of $5 per business dayb
| $1839 | |
Including parking charges of $10 per business dayb
| $2989 | |
Including parking charges of $20 per business dayb
| $5289 |
Discussion
Filter | Summary | Decision points |
---|---|---|
Level of evidence | Quantity and quality of evidence supporting association between fuel price or taxation and AT is limited. May be effective: No Level I or II evidence Modelling based on hypothetical scenario analysis | Weak evidence of effectiveness |
Equity | Equity concerns: Disproportionate effect across low, middle and high-income households. Middle-income households most affected as a proportion of overall weekly household expenditure. High-income households least affected as proportion of overall weekly expenditure. Evidence suggests that public transport is less accessible for persons with disabilities, the elderly, those living in areas not well-serviced by comprehensive networks and those from disadvantaged backgrounds. | Moderate issue |
Acceptability | Would require measures to be put into place to increase acceptability (for instance, revenue reinvestment to deal with potential regressivity and to ensure comprehensive public transport accessibility). | Moderate issue |
Feasibility | The intervention is feasible. The feasibility of modal switch to public transport as a result of the intervention may be limited in rural areas or areas not currently well-serviced by comprehensive public transport networks. A recent Australian survey found that 30% of respondents did not use public transport to work or full-time study due to the fact that no service was available at all, with 5.5% of respondents reporting that services were located too far from home [109]. | Not a major issue |
Sustainability | The sustainability of effect is relatively unknown. Consumers may adjust behaviour to price rises over the longer term. | Weak evidence of sustainability |
Side-effects |
Positive:
Potential for less traffic, pollution, safer environments for pedestrians and cyclists
Negative:
Potential strain on public transport networks | Significant wider positive side-effects |
Policy considerations: The intervention demonstrates potential for cost-effectiveness, but is limited in terms of quality of evidence of effect and sustainability. Concerns around equity and acceptability would need to be addressed. |