Mastodawn

#People4Bikes #BikeTooter

To start with, let's take a look at how #CambridgeMA and #Somerville scores have changed (graph from https://cityratings.peopleforbikes.org/cities/cambridge-ma)

👀

Somehow when P4B updated their methodology this year, Cambridge and Somerville specifically were massively negatively affected. I can confidently say that both cities have made biking improvements over the last year, so there is something else going on.

FWIW, #Boston, #BrooklineMA, #ArlingtonMA, #MedfordMA, #WalthamMA, #NewtonMA are all basically level

Let's take a look at the #Massachusetts top 10

👀

#1: P-Town - sure, small town that you could bike everywhere in town easily enough

#2: Groton - ??? Take a look at that map and tell me you can reasonably travel throughout town on low-stress streets

#5: Sturbridge - LOL

#6: Cambridge - Yeah, I can confidently say the low-stress network here is much better than basically all those other towns

#Massachusetts #BikeTooter #BikeBoston #People4Bikes

Somerville is down at #32, below such biking havens as Bedford, Hopkinton, Lynnfield, and Acton

To be clear, I actually have a pretty good understanding of how hard a problem P4B are trying to solve. Using #OSM data to estimate biking stress is exactly how our @bikeboston.bsky.social Stress Map (https://labs.bostoncyclistsunion.org/stressmap/) is created.

I know more than most the limitations to this type of project. But I don't think their stress map is off, its how they rate cities after creating the stress map.

Bike Stress Map - BCU Labs

I alluded to it before, but P4B did make a major change to their methodology this year. You can read about the changes and rationale here: https://www.peopleforbikes.org/news/2026-city-ratings-methodology.

2026 City Ratings Methodology Updates: What Is Changing and Why It Matters | PeopleForBikes

Every year, PeopleForBikes uses the Bicycle Network Analysis (BNA) to produce our annual City Ratings, a data-driven measure of how well a city’s bike network connects people to the places they want to go. This program helps cities identify gaps in their bike networks, highlights opportunities to improve, and benchmarks progress over time.

PeopleForBikes

From their description of the methodology updates:

"These updates have steadily improved the accuracy with which City Ratings reflect real-world bike networks."

"In previous versions of City Ratings, the Bicycle Network Analysis included some roads and destinations just beyond city limits — typically within one to two miles — as part of a city’s score. In some cases, that meant scores were influenced by infrastructure outside a city’s control."

https://www.cambridgema.gov/departments/communitydevelopment/2020bikeplanupdate

I am pretty sure the exclusion of routes outside the city lines is the primary culprit for both Somerville and Cambridge ratings tanking.

However, the statement that cities have no control outside their borders isn't fully accurate.

Just taking a look at Cambridge's 2020 Bike Plan you can see that Seomerville's Somerville Ave, Beacon Street, and Community Path (with Extension) are included!

The Community Path is part of 2 major trails (MCRT, Minuteman) that go through Cambridge, and Cambridge has co-signed letters to the state/feds about completing the CPX and MCRT. They have direct influence here!

Neighboring cities also work closely together on projects. Beacon going into Inman had a ton of collaboration. Somerville building better bike lanes on Washington St is pushing Cambridge to make Kirkland better. Huron bike lanes are being extended to Belmont's Grove St as coinciding with their project.

In cities like Cambridge and Boston, paying attention to streets just outside respective borders is critical to understand the bike network! The Beacon/Hampshire route is one of the most important routes for both cities, but P4B analysis changes actually punishes both cities because of that.

How does a network analysis that excludes Brookline understand Boston?

Again, this isn't an idle complaint about looking at networks in isolation by municipal border. The experience of Boston is that it is reasonable to go in a straight line on a commute and experience 5 jurisdictions in 5 miles, including exiting and re-entering the same city (I've personally had this commute).

Viewing the region as a whole is a driving feature of @bikeboston.bsky.social BCU Labs (https://labs.bostoncyclistsunion.org/)

BCU Labs

OK, so P4B should find a compromise solution from all streets within 2 miles of their borders and only looking at what's within a city's borders. But that isn't the only problem that leads to small towns getting outsized ratings, while the cities in the country with the most people biking are systematically under appreciated. Let's dive deep into the methodology!

https://cityratings.peopleforbikes.org/about/methodology

P4B basically tries to see what each census block can reach within 1.67 miles, with categories for People, Jobs/Education, Core Services (Medical/Dental, Supermarkets, Social Services), Recreation, Retail, Transit.

Methodology | PeopleForBikes 2025 City Ratings

Learn how PeopleForBikes establishes a City Rating Score

PeopleForBikes

The tricky decision they have to make is what to do when a town doesn't have any of a given category.

For all cases, they chose to exclude the subcategory if it doesn't exist in a town.

BUT!!! for some things, if you can't bike to it because it's out of town, then the bike network isn't sufficient! If you can't bike to groceries, then THAT is a gap in the network.

This is a major flaw in P4B analysis! They likely are trying to focus on bike lanes, but that is just an aspect of land use that affects bike-ability and biking rates. But if they don't want to wade into land use discussions, then they shouldn't include destinations.

However, towns very much have control over zoning and other land use decisions, so it would be valuable to call this out directly.

P4B are claiming that these ratings are valuable for city planners, so let's talk about what actually gets people on bikes!

"This program helps cities identify gaps in their bike networks, highlights opportunities to improve, and benchmarks progress over time."

If we want to make a national rating system that can help cities improve biking and get more people on bikes, then we should start by looking at who is actually achieving that goal.

Let's start by looking at how well P4B ratings line up with bike mode share. (Named cities are 3σ outliers from the linear trend line.)

(sadly Census data only has commute trips)

Large cities the system seems to actually work pretty reasonably for. A pretty reasonable r^2 of 0.28 (by far the highest) and only 2 outliers.

Good bet that most big cities don't have any excluded categories (and are well mapped in OSM). Just sticking with large cities, I'd probably not really complain about the methodology choices.

Medium cities are interesting! Cambridge, Brookline, and Somerville all are outliers surrounding Boston! Validating the above complaints about the regional nature of Boston

It's interesting that basically all of the outliers are college towns of some sort. What is missing in the analysis to capture what's different in these places? Why aren't these rated higher and why are cities with <5% mode share rated so high?

Hoboken, NJ is the highest rated Mid-Sized city but <5% bike share. If this was an outlier, that'd be reasonable. Over 2x people commute by transit over car, so there is a good argument that transit use is eating into bike share. And back to regional effects, a significant % of commuters are going into NYC, which bike is probably not that easy for most people. (This is a good example of the weakness of commute data, where 80% of trips are ignored)

Rochester Hills, MI and Anchorage, AK are the next highest rated medium cities.

Rochester Hills is a Detroit suburb that definitely has some regional effects (look up its boundary enclosing Rochester), but otherwise I don't know much about it to talk about learnings from its score.

Anchorage is a cool town to bike a little in (saw a moose within a few minutes when I did), but also being suppressed on biking kinda makes sense being in Alaska and all.

Boston Bike Data

Small cities are where this rating system really breaks down. There is basically no relationship between the rating and the biking rate.

P4B shouldn't exclude most of the categories that are excluded in small towns in the current system. I suspect that those lack of destinations are actually what is hurting the biking in these towns and leading to unusually high ratings that I laughed at way up in the thread, even when the stress map is mostly high stress roads.

Back to diving in deep on the methodology, using Cambridge as an example (follow along here https://bna.peopleforbikes.org/cities/United%20States/Massachusetts/Cambridge)

PeopleForBikes Bicycle Network Analysis (BNA) - PeopleForBikes Bicycle Network Analysis (BNA)

Cambridge is rated pretty poorly on transit (35/100), which is interesting because it has some of the best transit service in the US

Looking at the transit coverage view of the map, you can see that Transit=T stops (plus if you zoom into Lechmere the Charles Riverboat "ferry terminal", which is just a sightseeing boat, but that is an issue with OSM tagging)

But there are 2 major things that are negatively affecting the scores, but first let's explain the scoring system:

You get 60 points for being able to get to 1 station, and the other 40 points are the % of other stations within 1.7mi are completely low-stress to get to.

Problem #1: If we draw in the lines for those stations, you notice that most of them are on the same line. If you can bike to any Red Line station, you can get to them all low-stress, but Cambridge is losing points here.

(This is ignoring that there are Somerville stations that maybe you could get to stress free that are ignored)

Problem #2: This is actually a harder decision, but P4B is rigid that if there is any high-stress, then it is inaccessible. But this decision is massively over-simplifying how people choose to bike.

If there is a mostly low-stress route but a block of high-stress, many people will accept that. As that block gets longer, fewer people accept the risk.

Cambridge has been systematically improving its bike network over the past decade, and has seen steady increase in bike usage. But you don't really see that in the ratings. Small high-stress segments isolate in these ratings more than they do in real life.

P4B should learn from this and allow high-stress travel, but where those segments have an added cost. This is standard practice in any routing algorithm.

Year-over-year bike improvements will show up better in the ratings!

Related to P4B simplifying, they chose to compress Level of Traffic Stress from 4 levels to 2 levels. This is a pretty reasonable choice, especially for how they perform their analysis (all high-stress is the same). But, even if they don't show it, calculating the 4 levels and using that to weight each segment and how far people will choose to bike on them would likely give a more accurate version of the suggestion in the previous post.

To recap:

-P4B has a cool and challenging project to rate biking in cities across the US
-Doing this well is HARD
-They made changes this year that don't track with what gets people biking
-Bike networks extend beyond city borders, and cities have influence outside their borders
-Useful destinations are important, ignoring the lack of them hurts accuracy
-People will bike small distances of high-stress, and accounting for that will give a more accurate network analysis