Skip to main content

Table 1 Study characteristics of infrastructural interventions to promote cycling

From: A systematic review of the effect of infrastructural interventions to promote cycling: strengthening causal inference from observational data

Reference (country) Infrastructural intervention Controlled comparison Type of comparison Data collection method; time between measurements; time exposed Outcome studied Analytical methodology Confounders Direction of the results; significance;
absolute (A) and relative (R) change
Cycling behavior
Aittasalo [23] (Finland) Environmental improvements made to the main and connecting walking and cycling paths No Employees working at workplaces in the area where new infrastructure was introduced Survey
Time between measurements; 18–24 months
Time exposed; 2 months
Cycling frequency as part of the journey to work (days/week) Difference over time, tested by Wilcoxon Signed Rank Test Unadjusted Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
     Cycling distance as part of the journey to work (km/trip)    Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
     Cycling time as part of the journey to work (min/trip)    Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
     Cycling frequency as part of the journey from work (days/week)    Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
     Cycling distance as part of the journey from work (km/trip)    Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
     Cycling time as part of the journey from work (min/trip)    Not in favor of the intervention, not significant
A: Not applicable
R: Not applicable
Aldred [24] (UK) Infrastructural interventions in 3 neighborhoods, transforming local environments for walking and cycling Yes Residents living in the intervention areas vs control areas Travel diary
Time between measurements; 12 months
Time exposed; not specified but could range from 1 to 12 months
Made a bike trip in the past week (yes-no) Difference-in-difference, tested by regression models Demographic variables, socioeconomic variables, health indicator, and car ownership In favor of the intervention, not significant
A: 3.2%-point
R: 16%
     Cycling time (min/week)    In favor of the intervention, not significant
A: 4 min/week
R: 14%
  Yes Residents living in low-dose or high-dose areas (defined by stakeholders involved in implementation) in the intervention areas vs control areas Travel diary
Time between measurements; 12 months
Time exposed; not specified but could range from 1 to 12 months
Made a bike trip in the past week (yes-no) Difference-in-difference, tested by regression models Demographic variables, socioeconomic variables, health indicator, and car ownership All comparisons in favor of the intervention
Low-dose area: not significant
A: 0.7%-point
R: 10%
High-dose area: significant
A: 7.2%-point
R: 24%
     Cycling time (min/week)    All comparisons in favor of the intervention:
Low-dose area:
not significant
A: 1 min/week
R: 5%
High-dose area: not significant
A: 9 min/week
R: 30%
Brown [25] (US) Complete street intervention including the completion of an incomplete bike lane (10.7 km), connecting the airport to down town districts Yes Residents living near (≤0.8 km) vs far (0.8–2 km) from the new infrastructure GPS and accelerometers
Time between measurements; 12 months
Time exposed; 1–8 months
Made a bike trip on the intervention road (yes-no) Difference-in-difference, but no statistical test conducted Demographic and socioeconomic variables Not in favor of the intervention, significance not tested
A: 0%-point
R: −11%
Brown [26] (US) Same as above No Residents living within 2 km of the new infrastructure GPS and accelerometers
Time between measurements; 12 months
Time exposed; 1–8 months
Cycling time on the intervention road among those who cycled (min/week) Difference tested by paired t-test Unadjusted In favor of the intervention, not significant
A: 7 min/week
R: 38%
     Cycling time off the intervention road among those who cycled (min/week)    In favor of the intervention, not significant
A: 6 min/week
R: 15%
Burbidge and Goulias [27] (US) Installation of a multi-use trail, creating a 4-km loop connecting two currently existing sidewalks, serving as transportation and recreation facility No Residents living within 1.6 km of the new infrastructure Travel diary
Time between measurements; 12 months
Time exposed; 5 months
Total cycling trips (trips/day) Difference tested by fixed effects regression models Not reported In favor of the intervention, not significant
A: 0.01 trips/day
R: 33%
Chowdhury [28] (New Zealand) Introduction of a 3 cycle ways linking suburbs with the central business district, and the associated promotional campaigns No Residents living in the city where the new infrastructure was introduced Survey
Time between measurements; 4 years
Time exposed; 12 months
Cycling at least weekly (yes-no) Difference, but no statistical test conducted Unadjusted In favor of the intervention, significance not tested
A: 10%-point
R: 40%
Crane [29] (Australia) A new cycle way (2.4 km) linking a new urban renewal area with the central business district Yes Residents living in the intervention area (suburbs surrounding the cycle way) vs a control area (matched for demographic characteristics) Survey
Time between measurements; 23–25 months
Time exposed; 15–17 months
Cycling at least weekly (yes-no) Difference-in-difference tested by regression models that included a two-way interaction term between
time and proximity
Demographic variables In favor of the intervention, not significant
A: 44%-point
R: 179%
  Yes Residents living closer (< 1 km, 1–3 km)
vs further (> 3 km) from the new infrastructure
Travel diary
Time between measurements; 23–25 months
Time exposed; 15–17 months
Cycling duration (min/week)    Those living < 1 km of the intervention: not in favor of the intervention, not significant
A: −37 min/week
R: −21%
Those living 1–3 km from the intervention: in favor of the intervention, significant
A: 96 min/week
R: 54%
Deegan [30] (UK) Extension of a city-wide cycling network aiming for 900 km, unfinished No Residents living in 31 intervention areas. Area-wide cycling trends in 2 control areas are presented for comparison Survey (census data)
Time between measurements; 10 years
Time exposed; not specified but could range from 1 to 10 years
Proportion of commuting trips made by bike (%) Difference-in-difference, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average in 31 areas:
A: not reported
R: 87%
Average in 2 control areas:
A: not reported
R: 75%
Dill [31] (US) Installation of 8 bicycle boulevards (1.4 km to 6.7 km long) Yes Residents living within 0.3 km of the 8 intervention streets vs residents living within 0.3 km of the 11 control streets (selected to be similar in urban form and demographic characteristics) GPS and accelerometers
Time between measurements; 12 months
Time exposed; 2–12 months
Cycling at least 10 min a day (yes-no) Difference-in-difference tested by regression models that included a two-way interaction term between treatment and period Demographic variables, weather conditions, distance to downtown, bike attitudes and car safety attitudes In favor of the intervention, not significant
A: 9%-point
R: 22%
     Cycling time (min/day) for those cycling at least 10 min/day    Not in favor of the intervention, significant
A: − 1 min/day
R: − 1%
     Made a bike trip (yes-no)    Not in favor of the intervention, not significant
A: −8%-point
R: −15%
     Number of bike trips (trips/day) for those that made a bike trip    Not in favor of the intervention, not significant
A: −0.4 trips/day
R: −9%
Evenson [32] (US) Extension of an existing trail (4.5 km), along with a spur (3.2 km) passing by schools, shopping areas, apartment buildings, and residential areas No Residents living in census blocks that are crossed by the intervention Telephone interview
Time between measurements; 19–28 months
Time exposed; 2 months
Median cycling time (min/week) Difference tested by Wilcoxon nonparametric test for differences Unadjusted Not in favor of the intervention, not significant
A: 0 min/week
R: 0%
     Median cycling time for transportation (min/month)    Not in favor of the intervention, not significant
A: 0 min/week
R: 0%
Goodman [33] (UK) Construction of new walking and cycling infrastructure and improvement of existing routes in 3 cities plus a modest amount of promotion activities Yes Residents living within 5 km of the new infrastructure using proximity for comparison (per 1 km closer to the intervention) 7-day recall instrument
Time between measurements; 24 months
Time exposed; 7–21 months
Cycling time for transport (min/week) Difference-in-difference tested by regression models Demographic variables, socioeconomic variables, health indicator, and car ownership Not in favor of the intervention, not significant
A: −0.2 min/week
R: not reported
    Survey Cycling time for recreation (min/week)    In favor of the intervention, significant
A: 2.5 min/week
R: not reported
Song [34] (UK) Same as above No Residents living within 5 km of the new infrastructure 7-day recall instrument
Time between measurements; 24 months
Time exposed; 7–21 months
Cycling time for utility purpose (min/week) Difference over time tested by paired sample t-test Unadjusted In favor of the intervention, not significant
A: 0.4 min/week
R: 2%
     Cycling distance for utility purpose (km/week)    In favor of the intervention, not significant
A: 0.4 km/week
R: 7%
Hirsch [35] (US) Expansion of two trails (16.3 km), including a bicycle and pedestrian bridge connecting residential areas to employment centers downtown and at the university No Residents living in 116 areas of the city with the new infrastructure. Historical time trends are presented for comparison Survey (census data) Time between measurements; 10 years
Time exposed; not specified but could range from 3 to 10 years
Proportion of workers who commuted by bike (%) Difference over time, but no statistical test conducted Not reported In favor of the intervention, significance not tested
A: 2.3%-point
R: 130%
Historical trend:
A: 0.1%-point
R: not reported
  Yes Residents living in 116 areas of the city with the new infrastructure using distance to the intervention for comparison (results presented for the 25th, 50th and 75th percentiles)   Proportion of workers who commuted by bike (%) Difference-in-difference tested by regression models that included a two-way interaction term between
time and treatment
Demographic variables, socioeconomic variables, cycling infrastructure characteristics, total work-related trips, proportion of trips that cross the trail system All comparisons in favor of the intervention, and all significant
25th percentile (1.1 km):
A: 2.0%-point
R: 115%
50th percentile (2.8 km):
A: 1.9%-point
R: 107%
75th percentile (5.9 km):
A: 1.6%-point
R: 92%
  Yes Residents living in 116 areas of the city with the new infrastructure using proportion of commuting trips crossing the trail for comparison (results presented for the 25th, 50th and 75th percentiles)   Proportion of workers who commuted by bike (%) Difference-in-difference tested by regression models that included a two-way interaction term between
time and treatment
Demographic variables, socioeconomic variables, cycling infrastructure characteristics, total work-related trips, distance to the trail All comparisons in favor of the intervention, and all significant
25th percentile (11%):
A: 1.0%-point
R: 54%
50th percentile (29%):
A: 1.9%-point
R: 107%
75th percentile (42%):
A: 2.6%-point
R: 146%
  Yes Residents living in 116 areas of the city with the new infrastructure using the joined effect of distance and trips crossing the trail for comparison   Proportion of workers who commuted by bike (%) Difference-in-difference tested by regression models that included a two-way interaction term between
time and treatment
Demographic variables, socioeconomic variables, cycling infrastructure characteristics, total work-related trips, proportion of trips that cross the trail system, distance to the trail In favor of the intervention
The increase in bicycle commuting was restricted to tracts that were close to the intervention, and had a higher proportion of commuting trips that crossed the trails
Krizek [36] (US) Installation of multiple bicycle facilities and major bridge improvements to enhance accessibility to major employment centers No Residents living in areas within 1.6 km of the geographical centroids of a new facility. Area-wide cycling trends are presented for comparison Survey (census data)
Time between measurements; 10 years
Time exposed; not specified but could range from 1 to 10 years
Bicycle mode share (%) Difference tested by regression models Not reported In favor of the intervention, significant
A: 0.2%-point
R: 14%
Whole area:
A: 0.02%-point
R: 5%
   Residents living in areas within 1.6 km of the geographical centroids of a new facility, or within 0.8 km from the endpoints of a facility      In favor of the intervention, significant
A: 0.5%-point
R: 46%
  No Bicycle mode share crossing the river. Cycling trends that remained on the same side of the river are presented for comparison Survey (census data)
Time between measurements; 10 years
Time exposed; not specified but could range from 1 to 10 years
Bicycle mode share crossing the river (%) Difference tested by regression models Not reported In favor of the intervention, significant
Crossing river:
A: 1.6%-point
R: 52%
Average that remained at the same side of the river:
A: 0.6%-point
R: 28%
Lanzendorf [37] (Germany) Cycling infrastructure improvements and marketing campaigns in 4 cities No Residents living in cities with the new infrastructure. Cycling trends in big cities are presented for comparison Survey enriched with regional data
Time between measurements; 6 years
Time exposed; not specified but could range from 1 to 6 years
Cycling frequency (trips/day) Difference over time, tested by Mann-Whitney U-test Not reported In favor of the intervention, significant
Average of 4 cities:
A: 0.07 trips/day
R: 27%
Big cities:
A: 0.09 trips/day
R: 31%
     Bicycle mode share (%) Difference over time, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of 4 cities:
A: 1.8%-point
R: 21%
Big cities:
A: 2.4%-point
R: 24%
Merom [38] (US) The construction of a cycle way (16.5 km) and the associated promotional campaigns Yes Residents living near (< 1.5 km) vs far (1.5–5 km) from the new infrastructure Telephone interviews
Time between measurements; 4 months
Time exposed; 3 months
Cycling time among those who cycled (min/week) Difference-in-difference tested by ANOVA Unadjusted In favor of the intervention, significant
A: 26 min/week
R: 147%
Panter [39] (UK) New bus network and an adjacent traffic-free walking and cycling route (22 km) Yes Residents working in the city with the new infrastructure, and living within ~ 30 km of work using proximity for comparison (results presented comparing those living 4 km from the intervention vs 9 km) 7-day recall instrument
Time between measurements; 3 years
Time exposed; 9–14 months
Likelihood of an increase in cycling time for commuting (yes-no) Difference-in-difference tested by regression models Demographic variables, socioeconomic variables, health indicators, car ownership and work related variables In favor of the intervention, significant
A: 87 min/week (among those who reported more cycling for commuting at follow-up)
R: 34%
    Survey Likelihood of an increase in total cycling time (yes-no)    In favor of the intervention, significant
A: 115 min/week (among those who reported more cycling at follow-up)
R: 32%
Pedroso [40] (US) Infrastructure expansion in bicycle lanes (147 km) and improvements in bicycle signage, parking, and cyclist awareness, and the addition of a bike share program No Residents living in the city with the new infrastructure Survey (census data)
Time between measurements; 9 years
Time exposed; not specified but could range from 1 to 7 years
Proportion of workers who commuted by bike (%) Difference over time tested by regression models Not reported In favor of the intervention, significant
A: 1.5%-point
R: 167%
Smith [41] (US) Bicycle lane expansion (> 160 km), and the introduction of bicycle share programs No Residents living in the city with the new infrastructure Survey (census data)
Time between measurements; 5 years
Time exposed; 4 years
Number of cyclist Difference over time tested by t-test Not reported In favor of the intervention, significant
A: 4388 cyclist
R: 262%
Wilmink and Hartman [42] (The Netherlands) Improvements to an existing cycle route network, creating a comprehensive and interconnected network No Residents living in two neighborhoods with the new infrastructure Home interview
Time between measurements; 3 years
Time exposed; not specified but could range from 1 to 3 years
Proportion of trips made by bike (%) Difference-in-difference, no statistical test conducted Not reported In favor of the intervention, significance not tested
A: 3%-point
R: 7%
  Yes Residents living in two neighborhoods with the new infrastructure vs one control neighborhood without the new infrastructure   Cycling frequency (trips per person per day)    In favor of the intervention, significance not tested
A: not reported
R: 4%
     Cycling distance (distance per person per day)    In favor of the intervention, significance not tested
A: not reported
R: 8%
Usage of the infrastructure
Aittasalo [23] (Finland) Environmental improvements made to the main and connecting walking and cycling paths No 4 locations in the study area Automatic counters
Time between measurements; 24 months
Time exposed; 2 months
Bikes per day during afternoon peak hour Difference, no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of the 4 locations:
A: 367 bikes/peak hour
R: 57%
Barnes [43] (US) Complete street redesign of a gateway to the university to improve the conditions for non-motorized users No 1 location on the study road, for 2 directions of travel Direct observation
Time between measurements; 6 months
Time exposed; not specified but could range from 1 to 6 months
Bikes per hour Difference, no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of the 2 directions:
A: 63 bikes/hour
R: 83%
Crane [29] (Australia) A new cycle way (2.4 km) linking a new urban renewal area with the central business district No 2 locations on the study road. City-wide cycling trends and historic time trends are presented for comparison Automatic counters
Time between measurements; 36 months
Time exposed; 16 months
Bikes per day during peak hours (6 h/day) Difference, no statistical test conducted If adjusted, estimated were adjusted for population growth In favor of the intervention, significance not tested
Average of the 2 locations:
A: 144 bikes/peak hours (unadjusted)
R: 4% (adjusted)
City as a whole:
A: −80 bikes/peak hours (unadjusted)
R: −2% (adjusted)
Historical trend:
A: 300 bikes/peak hours (unadjusted)
R: 126% (unadjusted)
Historical trend, city as a whole:
A: 300 bikes/peak hours (unadjusted)
R: 111% (unadjusted)
Dill [31] (US) Installation of 8 bicycle boulevards (1.4 km to 6.7 km long) No 10 locations on the study roads Method not described
Time between measurements; 3 years
Time exposed; 18 months
Number of bikes Difference, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of the 10 locations:
A: not reported
R: 22%
Fitzhugh [44] (US) Retrofıtting a neighborhood with an urban trail (4.6 km) that enhanced connectivity to retail and school destinations Yes 1 location in the intervention neighborhood vs 2 locations in 2 control neighborhoods (matched along socioeconomic dimensions) Direct observation
Time between measurements; 2 years
Time exposed; 14 months
Median number of bikes per 2 h Difference-in-difference tested by Wilcoxon rank sums test Not reported In favor of the intervention, significant
A: 2.2 bikes/2 h
R: 224%
Goodno [45] (US) The installation of two linked bicycle facilities serving downtown No 4 locations on the study roads. City-wide cycling trends are presented for comparison Methods not described;
Time between measurements; 18–20 months
Time exposed; 7–12 months
Bikes during peak hour Difference, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of 4 locations:
A: 124 bikes/peak hour
R: 438%
City as a whole:
A: 20 bikes/peak hour
R: 32%
Hans [46] (Denmark) Improvements made to two large, interconnected bicycle infrastructures (18 km and 15 km) in city suburbs to enhance connectivity No 2 locations on the study roads Automatic counters, calibrated by visual counts
Time between measurements; 35 months
Time exposed; 16–22 months
Bikes per hour on weekdays during the rush hour in day light Difference over time, but no statistical test conducted Seasonal, weather and temporal variables In favor of the intervention, significance not tested
Average of the 2 locations:
A: 43 bikes/hour
R: 47%
     Bikes per hour on weekdays during the rush hour in dark    In favor of the intervention, significance not tested
Average of the 2 locations:
A: 38 bikes/hour
R: 72%
     Bikes per hour on weekdays during the non-rush hour in day light    In favor of the intervention, significant
Average of the 2 locations:
A: 11 bikes/hour
R: 19%
     Bikes per hour on weekend days in day light    In favor of the intervention, significant
Average of the 2 locations:
A: 10 bikes/hour
R: 29%
Heesch [47] (Australia) The opening of three new segments of a cycling lane (1.4 km, 0.9 km, 2.3 km) connecting the suburbs and the city center No 1 location on the study road before the intervention, 2 locations on the study road after the intervention Direct observation
Time between measurements;
4 years and 1 month
Time exposed; 3–38 months
Bikes per 2.5 h Difference over time, but no statistical test conducted Not reported In favor of the intervention, significance not tested
A: 376 bikes/2.5 h
R: 276%
The opening of the last segment of a cycling lane (2.3 km) connecting the suburbs and the city center Yes GPS tracking information on the study road vs 3 other routes surrounding the intervention Mobile phone application
Time between measurements;
1 year
Time exposed; 6 months
Trend in monthly bike trips on the intervention road Interrupted time-series Seasonal variables In favor of the intervention, significant
A: 225 bike trips/month
R: not applicable
  No GPS tracking information on the major routes between suburbs and city center, including the intervention   Trend in monthly bike trips between suburbs and the city center    In favor of the intervention, significant
A: 90 bike trips/month
R: 102%
Law [48] (UK) The introduction of superhighways for cyclists creating continuous cycling routes in the city center, and a public bike sharing system No 21 locations in the intervention area Direct observations (before) and automatic counters (after)
Time between measurements; 9 years
Time exposed; not specified but could range from 1 to 9 years
Bikes per hour Difference over time, test not described Not reported In favor of the intervention, significant
Average of the 21 locations:
A: 154 bikes/hour
R: 432%
Marques [49] (Spain) Introduction of a cycling network in the city (164 km) No 2000–2005: data from 2006 extrapolated
2006–2010: counts made in the city
2011–2013: algorithm based on count data and the number of rental bikes
Count data, changing methodology over time
Time between measurements; 14 years
Time exposed;
not specified but could range from 1 to 7 years
Million bike trips per year Difference, no statistical test conducted Seasonal variables In favor of the intervention, significance not tested
A: 13.3 million trips/year
R: 435%
McCartney [50] (UK) Construction of a new pedestrian and cyclist bridge across the river towards the city center No 5 locations to enter the city from the side of the bridge. City-wide cycling trends are presented for comparison Direct observation;
Time between measurements; 4 years
Time exposed; 2 years
Bikes counted per 2 days Difference over time, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of the 5 locations:
A: 500 bikes/2 days
R: 62%
Rest of the city:
A: 1700 bikes/2 days
R: 48%
Merom [38] (US) The construction of a cycle way (16.5 km) and the associated promotional campaigns No 4 locations along the new infrastructure Automatic counters
Time between measurements; 5 months
Time exposed; 3 months
Bikes per day Difference, tested by regression models Weather variables, day of the week and holiday season In favor of the intervention, significant
Average of the 4 locations:
A: Not reported
R: 31%
Nguyen [51] (Singapore) Improvement of 20 street segments (4.8 km in total) to complete a well-developed cycling network Yes 20 intervention street segments vs 55 control street segments Direct observation
Time between measurements; 2 years
Time exposed; 12 months
Bikes per hour Difference-in-difference, but no statistical test conducted Not reported In favor of the intervention, significance not tested
Average of the 20 locations:
A: 18 bikes/hour
R: 62%
Parker [52] (US) Introduction of a bike lane (5.0 km) with multiple bus stops, schools, businesses, a police station and private residences located along the intervention No 1 location on the study road Direct observation
Time between measurements; 12 months
Time exposed; 6 months
Bikes per day Difference tested by regression models Not reported In favor of the intervention, significant
A: 53 bikes/day
R: 58%
Parker [53] (US) Introduction of a bike lane (1.6 km) with multiple schools, churches and businesses located along the intervention Yes 1 location on the study road vs 1 location at 2 control streets Direct observation
Time between measurements; 12 months
Time exposed; 3 months
Bikes per day Difference-in-difference, but no statistical test conducted Not reported In favor of the intervention, significant
A: 196 bikes/day
R: 385%
Wilmink and Hartman [42] (The Netherlands) Improvements to an existing cycle route network, creating a comprehensive and interconnected network Yes Counts made along roads in the intervention neighborhoods vs counts made in the control neighborhood Count data, methods not described
Time between measurements; 3 years
Time exposed;
not specified but could range from 1 to 3 years
Bike counts Difference-in-difference, no statistical test conducted Not reported In favor of the intervention, significance not tested
A: Not reported
R: 14%