Mapping Neighborhoodness

Neighborhoods: what the heck are they?

Well, they’re a lot of things, defined by a lot of factors. Architecture, landmarks, social connections, spatial experiences — the list goes on. There’s no universal definition of “neighborhood” but most definitions broadly agree that neighborhoods are both geographical and social constructions. Forming out of human experiences in those realms, neighborhoods are almost always subjectively defined. While my neighbor and I probably have similar ideas of what our neighborhood is, they probably aren’t exactly the same.

Although neighborhood extents are fuzzy at best, hard lines are practical and essential for cities. A city must carve itself up somehow in order to provide services efficiently. (Think police and fire, trash collection, parking, &c.) What cities end up with is a bunch of different lines for different purposes, none of which are likely to match people’s collective perceptions perfectly. When mapping or performing research, we often use existing lines drawn for one of those purposes because it’s convenient, even though they weren’t drawn with our questions in mind. The most common example is probably census tracts and zip codes, which often stand in for neighborhoods in all kinds of contexts for which they were never meant.

It’s not that there are no lines. There are too many lines.

Finding Boston’s Neighborhoods

Official neighborhood map, City of Boston.

Tim Wallace and I have an occasional hobby and blog called Bostonography in which we attempt to map everything we find fascinating about Boston. Naturally our interest in the collective neighborhood map comes from this city specifically. Boston, like many American cities, comprises distinctive neighborhoods with strong identities and histories, but lacks a single definitive neighborhood map. While the city does have an official set of neighborhoods, they don’t dare draw lines on the reference map. It would only invite dispute.

Of course, that map is only a front door, and people find plenty of reasons to argue over boundaries as city services and other entities carve out districts in their own way. A fun Boston Globe interactive feature from a few years ago demonstrates just a few of these “tangled boundaries” at any given point in the city, including two different sets of lines from a single agency. Again, such lines are necessary and need to be stable. Consider an example of uncertainty from the Dorchester and Mattapan neighborhoods, in which health workers stress the need for “correct” boundaries:

Bill Walczak, president of Dorchester’s Carney Hospital, says that confusion caused by the shifting boundaries presented a quandary for him when he ran the Codman Square Health Center. It makes it more difficult for educators and health professionals to track trends over decades when the borders keep changing, says Walczak.

Boston’s peculiar layout further confuses boundaries. Much of the area can be thought of as a web of interconnected nodes — usually so-called “squares” which are crazy intersections of multiple major roadways. The squares are major points of local reference and, along with their surroundings, are often thought of as neighborhoods themselves. But official lines tend to follow roads, which frequently puts squares at the edges of official neighborhoods, or at the intersection of several of them.

“Squares” of Boston

And let’s not forget the forces of real estate.

Realtors define neighborhoods in ways that help them describe and drive sales, which may or may not line up with anyone else’s definition. They’re known for nudging boundaries and inventing new names, some of which end up sticking. These departures from the popular map range from amusing to hurtful. The map in the above tweet should have a neighborhood called Roxbury more or less in between Jamaica Plain and the South End. I don’t know the motivation for omitting it, but it’s tempting to believe that dropping predominantly black Roxbury from the map was thought to help sell Jamaica Plain to white yuppies. In its small way this little map further stigmatizes a neighborhood that already tends to get a bad rap.


All the lines are wrong? Fine, then. You draw the map.

Whether they’re serious disputes or casual arguments, we’d seen enough disagreements over neighborhood boundaries in Boston that we wanted to move beyond anecdotes and see the sum of all those different opinions. So we set about asking people to draw their opinions on a map, in the hopes of throwing enough data at the question to begin to see collective patterns emerge.

We made an ugly but functional Google Maps-based survey tool and shoved it onto the internet. Respondents drew shapes on the map for some 21 officially recognized neighborhoods of Boston, or really as many or few of those as they desired. Then they’d submit their drawings to a database from which we could grab everyone’s data.

After a while we ended up with thousands of polygons and hundreds of unique responses (each response could include as many as 21 polygons), enough to start analyzing with GIS. The first interest was in measuring consensus. We’d seen maps before like one of Kevin Lynch’s (below) in which survey answers are aggregated to identify minimum areas of overlap.

Kevin Lynch, Map of City Proper (districts), 1959

It’s still not uncommon to see a method basically like this in research that aims to identify neighborhoods. So we too measured overlap. We did this within a hexagonal grid (hexagons are all the rage in cartography lately), albeit a fairly fine one which you can think of as more or less a raster. To measure consensus for a given neighborhood, we simply count the number of polygons for the neighborhood overlapping a cell, then divide that by the total number of polygons for that neighborhood. We end up with this map:

Some neighborhoods show strong agreement and big cores; others don’t. This map was destined to be fuzzy, but it’s interesting to see where it’s particularly fuzzy. Some neighborhoods have gulfs of uncertainty between them, while others have reasonably hard edges. I’ll avoid in-depth examination here, but anyone who knows Boston will probably recognize a few areas of interest.

That was fun, but from the start we were bothered by several limitations:

  1. It only covered the city of Boston (and not even all of it). Lots of people who live in “Boston” don’t live in Boston.
  2. It only allowed a pre-determined set of neighborhoods, which are not necessarily what people identify with.
  3. We had no way of distinguishing pure guesses from informed responses by longtime residents.
  4. It’s an internet survey; it’s not at all a representative sample.

Meanwhile…

Our project is of course not the only one to approach neighborhood mapping through crowd participation on the web. It’s worth noting other projects, a few of which have inspired us, and a couple of which have been inspired by ours.

The Neighborhood Project by Matt Chisholm and Ross Cohen is a longtime project for San Francisco that determines neighborhoods based on craigslist housing posts and an online form.

The L.A. Times Data Desk has done at least a couple, such as this one for mapping L.A.’s Eastside.

Bill Morris ported our project to Burlington, Vermont.

My company, Axis Maps, has built a neighborhood mapping tool which is currently deployed for Dallas.

DNAinfo has neighborhood mapping surveys for New York and Chicago.

Zetashapes by David Blackman aims to have a set of crowdsourced neighborhood boundaries for pretty much everywhere.

Nick Martinelli developed a mapping tool for Portland, Seattle, and Vancouver. I mention this one last because…


Boston Neighborhood Mapping Redux

Nick’s open-source project was inspired by ours, but he did it way better. It’s more flexible, easy to implement, and fun. A few years after our first maps, we finally implemented a version of this for a new Boston neighborhood mapping tool.

A few differences are noteworthy improvements over the original version, we think. I implemented a few things, but most of this comes directly from what Nick created.

  1. The scope is broad and unrestricted. There’s no limit on geography or neighborhood names. It’s no longer Where is Neighborhood X? but rather What/where is your neighborhood?
  2. There are additional survey questions about how long respondents have lived in the neighborhood/city, if at all, and a prompt for any additional comment or story they wish to leave.
  3. Ease of use is improved, we think. We use free-form tracing to draw polygons, as opposed to clicking vertices. This likely feels more natural in the age of mobile devices — and it has a mobile-friendly design. Just trace with your finger! Conceivably, this would make deliberate field surveys easier too: shove an iPad in people’s faces and ask them to draw their neighborhood.
  4. A lot of feedback and our own thoughts suggested that people don’t necessarily think of neighborhoods in terms of edges, but rather centers or important locations. There are many ways to try to capture this, but the least we could do here was to allow the placement of points in addition to (or instead of) polygons.
  5. The technology is so much nicer! Nick built this on top of Leaflet and CartoDB. The latter especially makes viewing, sharing, and analyzing the data super easy.

Play around with it if you want to see more. But now it’s time for maps! We have close to 1,900 polygons already and are able to do some initial dirty analysis.

Where the data are coming from.
Bounding boxes of submissions
Medial axis: a way to visualize the “heart” of these neighborhoods, and possibly central corridors.
Section of an overall neighborhood map showing the most common neighborhood name applied to any given area.
Crop of a new version of the consensus map.

Takeaways

We still have maps to make, data to look at, and people to reach, we can say a few things after working on a couple of crowdsourced neighborhood mapping projects.

People are passionate. We have received an amazing level of response considering we’ve only pushed out a few tweets and offer nothing in return for people’s participation. People seem very interested in this subject, as it’s of personal relevance to almost everyone. I’ve also spoken about this work at a handful of local events over the last few years, and it never fails to touch off interesting discussions about neighborhoods, change, etc.

People are disagreeable. Maybe this still falls under “passionate,” but no matter how neutral you try to remain in running a project like this and presenting maps from it, people will still think you’re an idiot who’s wrong. And of course they won’t be shy about telling you something doesn’t work or that you should have done it some other way.

Guess what happens when you let people on the internet draw things. Yeah, this:

We’re not alone in this outcome.

LA Times. We warned him!

But hey, not all junk data is rude!

Aww, thanks!

The final word

If you’re a person who lives around Boston, knows Boston, or knows someone who knows Boston, please go to http://bostonography.com/hoods and draw!

This has been the written version of a talk given at NACIS 2015, approximating what was said in person. The actual slides are at https://speakerdeck.com/awoodruff/mapping-neighborhoodness.