Five Boroughs for the 21st Century

In this article we explore what happens when we abandon the century-old five borough partitioning of New York City and remap the city to reflect the realities of 2017.

Topos
Topos
Jun 19, 2017 · 13 min read
Image for post
Image for post
Image for post
Image for post
Map of New York City boroughs, 1898. Source: David Rumsey Map Collection

A Brief History of the New York City Boroughs

Formation of the Boroughs

When New York City was consolidated in 1898, forming a city of 3.4 million, the term ‘borough’ was adopted to describe the five constituent areas brought together under the consolidation. Each borough was represented by a ‘borough president’, an elected officer who sat on the New York City Board of Estimate, an 8-member governing body responsible for budget and land-use decisions which also included the Mayor, the Comptroller and the President of the New York City Council.

Connection of the Boroughs

With the exception of Queens/Brooklyn, all boroughs are separated from one another by water. The implications and limitations of this physical partitioning of land have changed considerably since the initial formation of the boroughs. New York City is now connected by over 2000 bridges and tunnels, the vast majority of which were built after 1898.

Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Verrazano-Narrows Bridge 1964, Bronx-Whitestone Bridge 1939, The Henry Hudson Bridge 1936. Source: MTA Flickr
Image for post
Image for post
City Hall Station: one of the first stations opened by The Interborough Rapid Transit Subway (IRT). Photo by Fred Guenther

Disenfranchisement of the Boroughs as a Unit

In 1989 the Supreme Court unanimously declared the NYC Board of Estimate to be unconstitutional. This decision was made on the grounds that Brooklyn (pop: 2,504,706 according to the 2010 census) had no greater representation than Staten Island (pop: 468,730), thereby violating the Fourteenth Amendment’s Equal Protection Clause. In the wake of this decision, the NYC Board of Estimate was abolished, and most of its governing responsibilities were transferred to the New York City Council, which consists of members drawn from a much more granular partitioning of the city with 51 council districts distributed (unevenly) across the 5 boroughs. Thus, while the constituent council districts of a borough have substantial political power in the New York City government (via their representative council members), the boroughs as a unit have far less political significance.

Using Data and Artificial Intelligence to Understand New York City

We formed Topos earlier this year to advance the understanding of cities through the interconnected lenses of data and artificial intelligence. While there are well-known tools such as the United States Census that use manual techniques to collect information about different locations, using data and AI enables a dynamic, highly granular, and globally scalable understanding of place — an understanding we think is valuable given the rapidly evolving nature of cities and neighborhoods around the world (the US Census, for example, takes place every 10 years, divides the country into 9 regions, and only covers the US).

Image for post
Image for post
Neighborhoods represented on a map (left) and as vectors (right)

From a 2D map to a 65D hyperspace

Mathematically, we can understand this suite of neighborhood features as a high dimensional vector space, where each feature is represented by a unique dimension; creating, in the case of this article, a 65-dimensional space. Each neighborhood[*] becomes a vector in this space, which can now be transformed and analyzed using a wide range of mathematical, statistical and computational techniques.

From 65D to 16D

One of the challenges in constructing a collection of features is understanding the interrelationship between features. 4 dimensions that are tightly correlated reveal much less than 4 completely independent dimensions. This becomes especially important in understanding the ways that entities described by features relate to one another — an understanding that forms the basis of several machine learning applications. For this reason, high dimensional spaces are often transformed through the use of various dimensionality reduction techniques.

Image for post
Image for post
The 16 reduced dimensions arrayed in parallel, with inter-dimensional spacing determined by the amount of variance each dimension explains

Dimensional Exploration

Principal Component Analysis outputs a set of reduced dimensions that are ordered by how much variance they explain; in our case, the first outputted dimension (d1) explains 30% of the variance, the second dimension (d2) explains 14% of the variance, the third dimension (d3) explains 7% of the variance, and so on.

  • via an examination of some of the top correlates in our original 65 dimensional feature space
Image for post
Image for post
d1 Choropleth
Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Image Recognition technology allows us to detect the visible presence of nature within neighborhoods
Image for post
Image for post
d2 Choropleth

Division by 5

Having a linearly independent, dimensionally reduced vector space is a powerful starting point for several machine learning applications. In particular, such a space allows the application of clustering algorithms, which group entities (neighborhoods) together in various ways.

Image for post
Image for post
The Topos platform allows clustering using a range of values for K. It also allows user to change the set of inputs that power the clustering.

The New 5 Boroughs

Looking at our new 5 borough mapping of New York, one thing is immediately clear: geographic boundaries and proximities are much less important than they were in the original boroughs. With the exception of the green cluster, all other clusters are spread across a minimum of 3 of the original, geographically defined boroughs, and divided by at least one body of water.

Image for post
Image for post
  • [2] Purple cluster: 4.34 miles
  • [3] Blue cluster: 7.92 miles
  • [4] Red cluster: 7.86 miles
  • [5] Yellow cluster: 11.09 miles

Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Typical Minhattan photos
Image for post
Image for post

Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Typical photos taken from The Ring
Image for post
Image for post

Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Typical North Bend photos
Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Typical South Bend photos
Image for post
Image for post
Image for post
Image for post
Red — ‘North Bend’ and Blue—‘South Bend’

Image for post
Image for post
Image for post
Image for post
Image for post
Image for post
Typical photos taken from The Meadows — nature and suburban style homes are frequent subjects
Image for post
Image for post

Why Divide New York City?

As with many new technologies, data and AI based partitioning can be used to both positive and negative ends. Enabling a new level of strategic gerrymandering and malapportionment is an undeniable risk. We also believe the cultural, ecological, institutional and architectural diversity of New York City is an incredible strength and should be celebrated.


topos.ai

topos.ai