Article Text

Download PDFPDF

OP49 Impacts of the Paris cycling lane expansion plan on cycling levels: a natural experimental study*
  1. Christina Xiao,
  2. David Ogilvie,
  3. Esther van Sluijs,
  4. Jenna Panter
  1. MRC Epidemiology Unit, University of Cambridge, Cambridge, UK


Background Cities globally have started to seriously invest in more sustainable forms of transportation. Using routinely collected city-level data, we aimed to evaluate whether constructing new cycling infrastructure as part of the Paris Cycling Lane Expansion Plan 2015–2020 affects cycling activity along new routes.

Methods Daily cycle count data from January 2018 to March 2020 were acquired for the city of Paris. Eight newly-built cycling infrastructure improvement projects were identified with pre-post data. Comparison streets were chosen if pre-intervention trends in cycling paralleled those at the intervention sites. Since data collection periods for each street were variable, several comparison streets were chosen for each site as follows: (A) one street for which monitoring data were available for the same one-year period as the intervention street, (B) one street that shared the same six-month pre- and post-monitoring periods as the intervention street. For streets without a full year of data (n=3), all available data were used. The average of all control streets for each method was calculated as an additional comparator. Difference-in-difference (DiD) analysis controlling for a public transportation strike during the study period was performed for all streets. in addition, for streets with at least one year of data, interrupted time series (ITS) analysis was conducted to corroborate DiD results.

Results There was some variation in effects between locations: significant net increases in cycling counts were observed in 4/8 streets (e.g. Boulevard Voltaire, Method A: 894 counts/day; 95% CI: 357, 1431). No significant effects were found for Rue Julia Bartet or streets assessed for only one month post-intervention (3/8). In general, DiD outcomes did not differ between methods for choosing control groups. However, comparisons with individually-matched control streets tended to have greater positive net effect sizes than those using the average of control streets, which were more likely to support the null hypothesis. In general, the ITS results corroborated DiD results in terms of direction of effect, but none of the ITS results besides the level and trend change for the strike were significant.

Discussion Infrastructural improvements were found to be effective for larger arterial streets and those with longer follow-up periods. The use of multiple control streets as well as ITS analysis lends weight to our findings. Further research should investigate why improvements were more effective at increasing cycling levels in certain streets than in others.

  • Built environment
  • Natural experiment
  • Cycling infrastructure

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.