I put this together as my final project for Bostonography: The City through Data, Texts, Maps, and Networks at Northeastern University.
The inspiration for this comes from a proposal by Anne Hidalgo, the incumbent Mayor of Paris, France. The proposal, called Ville du Quart d’Heure, or the Fifteen-Minute City, here reported on in great detail by CityLab, is an urban design plan with the intent of designing a city in which all of the basic needs of life for its residents can be satisfied within a fifteen-minute radius, by foot or by bicycle. Such a design not only promotes sustainability in the era of climate change, but also enhances the well-being of its residents.1
Paris, as it turns out, is not the only city that has presented such a sweeping urban design plan. As CityLab reports, Barcelona already has
Superblocks -- "40-acre, tic-tac-toe sections of the street grid that the city has transformed into pedestrian-first environments"2. London is promoting it's
Every One Every Day urban design plan.3 And in America, Portland, Oregon is promoting a plan of 20-minute neighborhoods for 90% of it's residents.4
But what would such a plan look like for Boston? The Boston Planning & Development Agency has not laid out any specific 15- or 20-minute city plan, although, in the Imagine Boston 2020 planning initiative, there are specific references to improving the mixed-use downtown and expanding job growth in the downtown area. To specifically look at how different areas of Boston fit into a fifteen-minute city design, let's first define the specific parameters of a fulfilling daily life for a citizen of Boston.
As quoted in the CityLab Article--Carlos Moreno, a professor at Paris-Sorbonne University, defines the basic needs as such: “There are six things that make an urbanite happy” he told Liberation. “Dwelling in dignity, working in proper conditions, [being able to gain] provisions, well-being, education and leisure."1
We can define these into several categories, with expected data points for each:
Since the major metric in this analysis concerns determining access in a "fifteen-minute" radius, it's important to define exactly how we determine this radius. Per Healthline, the average human walking speed is 3-4 miles per hour -- for ages to 20-49 the average speed is somewhere between 3.0 and 3.2 miles per hour.5 Taking the lower bound of this range (3 mph), a person walking would cover around 0.75 miles. This radius will be the baseline for our analysis.
For the sake of simplicity, we're calculating the majority of these distances from the midpoint of the neighborhood/census tract, using distance as the crow flies. It's important to note that this does not show the exact distance from point A to point B -- further analysis could use a tool like the Google Maps Directions API which would calculate walking time based on an ideal route on walking-friendly roads. Additionally, unless otherwise stated, these distances are calculated from the center point of the census tract, again for the sake of simplicity.
The fifteen-minute plan in Paris also included biking as a mode of transit within the fifteen-minute radius. While bike infrastructure is an important transportation question, it too is subject to a lot of variability based off of the infrastructure in a given neighborhood -- a bike commute along the Southwest Corridor, for example, is significantly different than a bike commute along Massachusetts Avenue or Huntington Avenue, both in terms of speed and in terms of safety. Given that are likely a number of people who may not be comfortable with the latter, and the general fact that biking is not as accessible of a mode of transit, this would be an inexact measure. We are therefore only going to look at walking as the mode of transport for this project.
There are, however, a number of good fields of analysis for biking accessibility, such as the Level of Traffic Stress criteria developed by Peter Furth of the Northeastern College of Engineering, that could be used to expand on this analysis.
Our basic unit of analysis for the city will be census tracts, one unit of magnitude below neighborhoods. Shown below is each census tract outlined, color-coded by neighborhood name. Each census tract has some basic information available, sourced from 2010 Census Data compiled by the Boston Area Research Initiative (BARI).6
The purpose of this project was to create a set of tools that could "heatmap" the availability of different services to each census tract within a fifteen-minute radius. This has a broad range of potential users. For example, this analysis and visualization could be used by city planners when determing where to increase density of housing, or to determine how to rezone underserved areas of the city. Potential renters or home buyers could use this tool to make decisions about which parts of the city are most ideal to live in. These visualizations offer a number of potential avenues of further research as well -- do underserved areas of the city typically have different racial breakdowns? Different median household incomes? This line of analysis can therefore answer questions about equity in the city.
The first category we'll look at is probably the most critical factor in social well-being in a neighborhood -- commute times. Studies have shown that commute times are often directly correlated with an individual's mental health -- higher commute times, especially when traveling in traffic by vehicle, correlate with higher emotional stress levels.7
The simplest way to get a clear picture of the commute of the average individual in an area of the city is by using self-reported census data. Here, we're using commute time data again from BARI.6
Data Analysis Background
The census data is offered to us per census tract as a proportion, broken down into five categories of commute time:
less than 10 minutes, 10-30 minutes, 30-60 minutes, 60-90 minutes, more than 90 minutes. Using
midpoint coding,, we find the midpoint of each of these categories (treating the midpoint of > 90 min as 105 min)
-- so, respectively, 5, 20, 45, 75, and 105 minutes for each category.8 We then sum category_proportion * total_population * midpoint for each category, and divide by the total total_population
to find the average time.
Commute times are currently displayed on a sliding color scale -- green is <= 10 minutes, red is >= 75 minutes, and everything in between is an average color value.
However, commute time isn't the only metric we should measure -- also important is commute type, data that BARI also offers us. Rather than use a red-yellow-green scale for this, we'll mark each commute type with it's own color.
So how does this fit into the "15-minute city" model? From a glance, it seems not well. Almost all of the commutes are in the 30-45 minute range, with some even longer. What's especially interesting to note is that even the "downtown" neighborhoods that one might expect to have short commute times (for example, Back Bay) still report commutes of 25-30 minutes -- a short commute, by most people's standards, but almost double the "15-minute" metric we're looking at. Let's look at the best-performing neighborhoods more specifically.
From looking at this, it is clear that the neighborhoods that score the best do so with primarily walking commutes. Additionally, they're all close to the downtown center -- clustered primarily in Fenway/Kenmore and in the neighborhoods of Government Center, Chinatown, the West End, and the Financial District, typically considered the center of businesses and workplaces in the town, but not in Dorchester, West Roxbury, or Hyde Park/Roslindale. The Allston/Brighton tract, additionally, is right along the section of Commonwealth Avenue that is home to Boston University.
The next major component of our analysis concerns how accessible grocery stores, specialty food markets, and pharmacies are to each area of Boston. We're using data of grocery store locations from the Metropolitan Area Planning Council10, found here. This data has been broken down into several categories of establishment -- Supermarkets/Other Grocery stores, Convenience Stores, Pharmacies & Drug Stores, Fruit & Vegetable Markets, Fish & Seafood Markets, Other Specialty Food Stores, Meat Markets, Farmers Markets, Winter Markets, and Warehouse Clubs. We'll start by taking a look at just the supermarkets, which should provide a wide range of options. A disclaimer: this data is from 2016, so there may be markets that have changed names since, or have closed since.
As it turns out, basically all of the neighborhoods of Boston do fit our criteria in regards to local grocery stores -- however, we can break this down into more specific categories if we choose. Let's now look at specialty markets -- Fish & Seafood Markets, Fruit & Vegetable Markets, and Meat Markets to get a sense of what the access breakdown is.
The access breakdown is similar -- almost all areas in Boston have relatively good access to specialty food stores.
We'll now turn to locations of pharmacies in Boston.
Thanks to the prevalence of chain pharmacies across Boston -- chiefy, CVS and Walgreens, most neighborhoods do quite well on this metric.
The Massachusetts public school system is continually ranked #1 among all US states, and Boston has one of the highest college student populations in the country.11 How do these rankings match up with their accessibility to residents? To do this, we'll break education down into three categories:
Boston's data portal offers us a data set with the location of each library in the Boston Public Library system.12 This map uses this data to show the
distance from the given census tract to the closest library. For the sake of simplicity, the distance is calculated as the crow-flies distance from the center of the tract (rather than
the distance if one was traveling along roads). Again, a green to red scale was used here -- green representing a distance of <= 0.75 miles (as discussed above), and red indicating a distance
of >= 3 miles.
As is clear from the map coloring -- save for a few tracts which lie towards the edges of the city, Boston generally has readily available library access, which slides in quite neatly to our "15-minute" metric.
It's unfortunately hard to get a perfect image of public school access in Boston due to the complex network of school choice, an subject that the Boston Area Research Initiative (BARI) has studied.
To start, we can take a high-level look at the Boston Public Schools Student Assignment Policy as it is stated on their website. The Boston Public School system offers families a choice of where to send their students for grades K-8 from a list of schools tailored to their home address, on the following criteria: "every school within a one-mile radius of their home plus, as needed, nearby high-quality schools based on BPS's School Quality Framework (SQF) system of measurement".13
Per this policy, every household should be able to choose a K-8 school that is within 15-20 minutes of their household (0.8-1 mile, roughly). High schools remain, per the website, citywide options. Working off of this, we can look at how many K-8 schools every census tract has access to in a one-mile radius. We are using again data from the Boston Data Portal on public schools in the city.14 The dataset lists the type of school in regards to what grades it offers -- "ES" (elementary school), "MS" (middle school), "HS" (high school), "K-8", "6/7-12", "Special", "K-12". For this mapping, given that this policy is for K-8 schools only, we have excluded any schools that exclusively offer the high school grades (9-12) as well as special education schools, as these are not offered as options to every household.
Per the BPS website: "Every family has a choice of at least six schools; most will have between 10 and 14 choices." With that metric in mind, we color the map so with red at 1 school and green at 10 schools in a one-mile radius.
What does the map show us? Chiefly, that there are a number of census tracts with few schools that meet the 1 mile radius criteria. The areas that do are mostly clustered in Roxbury and Dorchester, areas that do worse in other areas of our analysis, such as commute time. On the other hand, the neighborhoods that performed the best for commute times and types -- the Back Bay/Financial District/Government Center areas, do much worse on this metric.
Finally, let's look at the higher education distribution in the city. We're again using data from the Boston Data Portal, a dataset on the locations of colleges and universities in the city.15 We have again omitted parts of this data for clarity -- several universities in the dataset have since been listed as closed online, and there were several double locations of other universities in the data. Similarly to the library analyis, we're going to take a look at what tracts have a university in walking distance (calculating the distance off the centroid of the tract).
What tracts are in "college neighborhoods" and which ones aren't? Again, it's clear that the downtown tracts are very dense with college students, while tracts further out in the city, and especially in neighborhoods like Jamaica Plain and Roslindale, are much less walkable.
We'll now turn to the well-being metrics in a city -- chiefly, hospitals, health centers, and urgent care. We're using data from MassGIS data sources for acute care hospitals within the city, which "contain a majority of medical-surgical, pediatric, obstetric, and maternity beds, as defined by the Massachusetts Department of Public Health (DPH)".16
We see a similar distribution of hospitals within the city -- the downtown neighborhoods, and predictably, those in the Longwood Medical Area, all have readily accessible hospitals in a walking radius. The neighborhoods further out in the city -- West Roxbury, Hyde Park, East Boston, have further to travel for medical care. It is also worth noting that in emergency situations, this may not be the most ideal metric, as one will presumably not be walking to medical care, but since acute care hospitals provide a large range of non-emergency services, it is still a valid metric to study.
We'll turn next to MassGIS data on community health centers, which cover a broad range of other non-acute medical services.17 MassGIS uses the definition from the Massachusetts League of Community Health Centers, which "defines a community health center as a non-profit community-based organization that offers comprehensive primary and preventive health care, including medical, social and/or mental health services, to anyone in need regardless of their medical status, ability to pay, culture or ethnicity"17
Community health centers have a better distribution overall than hospitals, and the further-out neighborhoods that are further from acute-care hospitals do slightly better than in the previous map. Still, there are a number of areas with poor access -- Hyde Park especially, where the closest health center is almost three miles away.
The final point of our analysis is leisure, and in general, quality of life within the neighborhoods of Boston. The Boston Data Portal offers us data of all the public spaces within Boston.18
As it turns out, basically every neighborhood in Boston has very good access to open space. Let's look more specifically at the "greenness" of a certain tract. To do that, we'll use the Trees dataset from the Boston Data Portal.19 Here, we've parsed the tree data from the city to determine how many trees are in each tract. On the map, green is > 2000 trees, and brown is < 100 trees. The tract with the most trees had over 6000 trees, the tract with the least had a little over 50.
The neighborhoods that do the best are predictably ones covered mostly by public parks -- Franklin Park and Arnold Arboretum for example. While the downtown areas perform better on other metrics, they do quite poorly here, due to the density of buildings in these areas.
How does Boston measure up as far as fifteen-minute city proposals go? The answer is -- in some metrics, surprisingly well, in other metrics, quite poorly.
On these metrics, Boston did best when looking at specifically businesses -- most neighborhoods have relatively decent grocery store and pharmacy access, within walking distance. Educationally, it does better in some areas than others -- public libraries are very evenly distributed, but public schools are not -- more central areas of the city have more schools within walking distance than other areas. And as a city with some of the best hospitals in the country, there is a relatively even distribution of care across the city.
Perhaps not surprisingly, for anyone who has commuted in the city, workplace access was the metric that fared the worst on the 15-minute test. Almost none of Boston's tracts met this criteria -- with commute times hovering around 30-40+ minutes in most areas. And much of this commuting, especially with neighborhoods further from the downtown area of the city, was done by car.
There is of course another project's worth of analysis on how these metrics can be improved. Certainly, none of these problems can be solved quickly. Boston's "Go Boston 2030" plan addresses a lot of these concerns, with plans to put all city residents within a 10-minute walk of bikeshare, transit, and carshare services, and plans to reduce commute times by 10%.20 These incremental efforts will undoubtedly make a large difference in city life for both residents and those who commute in, and will perhaps one day improve the metrics shown in this project.
The full source code for this project can be found here
This project was made possible with a number of different open-source libraries and tools, offered generously under the MIT license by their developers.
The site was built with GatsbyJS, a static site generator, with ReactJS handling the frontend and GraphQL handling data queries.
This repository is built off of the Gatsby Leaflet Starter offered under the MIT license by Colby Fayock.
All mapping was done with the LeafletJS library, offered as React components by React-Leaflet. All base maps are provided by OpenStreetMap. React-Boostrap provided all non-mapping frontend components.
A number of smaller open-source packages were invaluable for the visualizations. They include Rainbowvis.js which provided the color gradient for the mapping, Gatsby-Source-Geo, a Gatsby plugin that handles GeoJSON and Shapefile data, and proj4js which converted GIS northings and eastings into latitude and longitude.
Zotero was used to collect citations, and MDX was used to lay out content. Loadable Components allowed mapping components that required access to the browser window to be included in .mdx files (see the LoadableComponent file for details).
A simple Python script was used to parse tree data with census tracts (a 13-minute operation that was impractical to do every time the site was built). This code exists in python-utils and made use of the Shapely and pyshp packages.
This site is built and hosted by Netlify.