The Bike Network Analysis (BNA) is a data analysis tool that measures how well bike networks connect people with the places they want to go. Because most people are interested in biking only when it's a comfortable experience, our maps recognize only low-stress biking connections.
We compute the BNA score over four steps: data collection, traffic stress analysis, destination access analysis, and score aggregation. Each of these steps is described below.
For U.S. cities, the BNA relies on data from two sources: The U.S. Census and OpenStreetMap (OSM). Census blocks delineated by the U.S. Census Bureau's 2010 Decennial Census serve as the basic unit of analysis for all connectivity measures. The 2010 Decennial Census also supplies block-level population data via the Census of Population and Housing, which the BNA uses to calculate the People score. We obtain block-level data detailing the geographic distribution of jobs from the U.S. Census Bureau's Longitudinal Employer-Household Dynamics (LEHD) data, which underlies the Opportunity Employment score.
For U.S. territories, the BNA again employs U.S. census blocks as the unit of analysis. However, population data is not available at the block level, so we substitute block group data from the 2010 Decennial Census. We omit employment information because no comparable data to the LEHD exists for U.S. territories.
For cities outside of the U.S., we derive the geographic units of analysis, population data, and jobs data from comparable public datasets when available. We create custom geographic units in place of census blocks when no suitable alternative exists. If fine-grained population data is unavailable, we will infer the population distribution at a small scale from broader area population estimates. If we cannot identify comparable jobs data, we exclude it from the analysis. Please contact firstname.lastname@example.org if you have questions about the data sources used for non-U.S. cities.
OSM data is available worldwide, providing a fully-routable network of on- and off-street transportation facilities including details about the types of bicycle facilities on any given street segment. OSM also supplies location and attribute data for all destination types in the analysis except population and jobs. The BNA downloads the most recent OSM data for the area within a city's boundary plus a buffer distance around the boundary equivalent to the default bikeshed distance designated in the tool, 2,680 meters or 1.67 miles. Although OSM data quality varies between cities and countries, anyone can edit OSM to improve the BNA's accuracy.
Traffic Stress Analysis
The BNA relies on the concept of a low-stress bike network. The concept of Traffic Stress has emerged as a useful way to think of bicycle facilities in terms of the types of users who would be comfortable riding on them in a given situation. Since our measures are concerned with low-stress bicycling, our methodology focuses on roadway characteristics that generally translate to a Level of Traffic Stress 1 or 2 rating based on the scale originally developed by the Mineta Transportation Institute. In practical terms, this corresponds with the comfort level of a typical adult with an interest in riding a bicycle but who is concerned about interactions with vehicular traffic.
The OSM data we use to build the bike network employs a system of tags to represent different elements of a roadway. A list of tags for bicycle facilities and destinations here. For a description of how OSM tags relate to on-the-ground bicycle facilities you can refer to these tagging guidelines. Please note that our methodology also accounts for some edge cases involving obsolete or non-standard tagging. For a full review of the logic, we invite you to review the source code.
Once we've built the transportation network, we rate every street segment and intersection for high or low traffic stress. There are several bicycle facility types that the original Traffic Stress methodology did not consider. We have followed the same basic approach but our methodology includes some new facility types. You can follow our logic usingthis analysis logic spreadsheet.
While OSM data gives us a great base on which to build, it can vary in terms of the availability of detailed roadway characteristics. To account for situations where OSM data is not sufficient, we developed default assumptions based on OSM's hierarchy of roads. (The defaults are given in the spreadsheet linked above.) The default assumptions are only used when OSM data is missing.
The BNA evaluates traffic stress for each link in the transportation network by applying the logic outlined in the spreadsheet to the street characteristics documented in OSM. The resulting Stress Network map visualizes the stress rating of every street segment with blue representing low-stress routes and orange representing high-stress routes.
Destination Access Analysis
Once we have established the street segment stress ratings, we evaluate every census block (or for non-U.S. cities, other geographic units) to determine which other census blocks are within biking distance and can be reached on the low-stress network. The BNA assumes a biking distance of 1.67 miles or 2,680 meters as measured along streets or paths, the distance an average rider would travel in ten minutes biking ten miles per hour. No one likes a detour so we also assume that a low-stress route is only available if it doesn't force a person to go out of their way by more than 25% compared to a car trip. We also assume that a census block is connected to any road that either follows its perimeter or serves its interior. In practice, this means you can get to a destination whose front door is on a stressful street if you can get to a low-stress street around the corner. Finally, we assume that two census blocks are connected if and only if there is an unbroken low-stress connection between them. In other words, even a short stretch of stressful biking negates a potential connection. This is consistent with the Traffic Stress concept and also highlights the importance of a continuous network, rather than the patchwork of facilities that is common in many U.S. cities.
We use the transportation network to route from each census block to every other census block within biking distance, noting whether a low-stress connection between the two is possible. We also summarize the number and types of destinations available in each census block. Using this information paired with the knowledge of which census blocks are connected on the low-stress network, we calculate the total number of destinations accessible on the low-stress network and compare that with the total number of destinations that are within biking distance regardless of whether they are accessible via the low-stress network. Destinations outside of the city boundary but within the surrounding buffer area are included in the analysis to enable calculating accessibility from points located on the edge of the city boundary. This means that the quality of the bike network in neighboring cities or unincorporated areas will affect a city's BNA score if there are destinations located within that buffer area.
Points are assigned on a scale of 0 to 100 for each destination type based on the number of destinations available on the low-stress network as well as the ratio of low-stress destinations to all destinations within biking distance. The scoring places higher value on the first few low-stress destinations by assigning points on a stepped scale. Beyond the first few low-stress destinations, points are prorated up to 100 based on the ratio of low-stress to high-stress connections to those destinations. For example, a census block with low-stress access to only one park out of five nearby parks would receive 30 points. A census block with low-stress access to two parks out five would receive 50 points (30 for the first park, 20 for the second). A census block with low-stress access to four parks out of five would receive 85 points (30 for the first, 20 for the second, 20 for the third, and 15 out of the remaining 30 points for connecting one of the remaining two parks).
The BNA's six scoring categories are:
- People: Access to other people in the city based on the resident population distribution
- Opportunity: Access to jobs and educational institutions
- Core Services: Access to critical services such as health care
- Recreation: Access to public recreation outlets
- Retail: Access to shopping areas
- Transit: Access to major transit hubs
Three of the scoring categories are composed of a mix of destination types constituting subcategories. For instance, the Recreation category encompasses the subcategories Community Centers, Parks, and Trails. In these cases, the category score is calculated by combining the scores of each of its member destination type/subcategory scores. Weights for each destination type are used to represent their relative importance within the category. For census blocks where a destination type is not reachable by either high- or low-stress means, that destination type is excluded from the calculations. For example, the Opportunity score within a city with no institute of higher education is produced by excluding the Higher Education destination type so the score is unaffected by its absence. As noted in the Data Collection phase, U.S. territories and non-U.S. cities may lack jobs data comparable to the LEHD, in which case they will not receive an Opportunity Employment score and the overall Opportunity score will only reflect access to educational institutions.
We use the category scores to calculate one overall score for each census block, weighting each category according to its relative importance. The step thresholds, destination scoring, and weighting assumptions are all described in this spreadsheet.
BNA scoring operates at two geographic levels. Up to now, the description has focused on scoring of individual census blocks. Census block scores are visualized on the Census Blocks with Access heat map where blue blocks are relatively well connected and orange blocks are poorly connected.
We use Census block scores to calculate scores for the whole city by weighting each census block according to its population and then averaging destination type (subcategory) scores across the city. We then apply the same category weights used in the block-level calculations to calculate citywide category scores and an overall city score. Like the block-level calculation, the citywide calculation excludes destination types that are not represented anywhere in the city. For example, if a city has no rail stations or bus transfer stations, the transit score is not factored into the overall score.