Sticky problems with mapping historical New York City

Author’s time-lapse of Lower Manhattan’s street network development from 1609-2020

With help from digital and spatial mapping software, urban historians and geographers are examining city growth over time. Time-lapse evolutions are proliferating online of street network development in cities like New York City, Barcelona, London, and Berlin.
In most time-lapse studies, geographers encounter problems with lack of data. The older the city, the less data there exists about pre-modern population densities, demographics, and street networks. This lack of data is a problem when mapping the geographies of older cities.
A way around this problem is to look at street network development as a proxy for population size. The more streets there are built, the more people this city should have, the logic follows. In theory, this seems to work because cities with larger populations require more streets and occupy more built-up area. Knowing how much surface area a city occupies, coupled with knowing the average size and number of occupants in a typical block or building, allows a simple calculation of total population (people/acre x surface area). In addition, more historical data exists about street networks (from maps) than exists about population and demographics (from the census).
The problem with this method of using streets as a proxy for demographics is that cities that occupy more surface area and with more streets do not necessarily have more people. There are several reasons for this:
  • Available land: Some cities are built in harder geographies where acquisition of new land for development is prohibitively difficult to acquire, such as Venice. Manhattan’s high density and land values descend, of course, from a demand for housing that far exceeds supply on an island bordered by water.
    For instance, Oklahoma City covers 621 square miles with a 2018 population of only 650,000. New York City covers 302.6 square miles (half the area of Oklahoma City) and has a 2018 population of 8.4 million (thirteen times the population of Oklahoma City). Despite the major differences between these two cities – in population and surface area – the sum total of all streets if they were lined up end to end to form a continuous road is about the same for both cities. Similarly, the Manhattan grid is identical with the same street widths and block sizes from end to end of the island, even though population density in buildings within this grid varies from zero people per acre to over 200 per acre. Flexible street networks support any variety of housing types and densities, which means that street maps alone cannot reveal all the demographic nuances.
  • Zoning: Some municipalities are stricter than others in enforcing discrete and different land uses for residential, commercial, industrial, and mixed-use. The legal landscape of Manhattan has evolved significantly since the first zoning laws in 1916 restricted building height and density. Since then, city government has more clearly articulated rules about minimum apartment size, ventilation, fire escapes, and water supply. City government has also pulled industrial (and often more polluting) land uses away from residential areas in the name of health and safety.
    Although few of these legal and zoning changes are explicitly imprinted on the street network, they have a tangible and important impact on the quality of urban life. This zoning has largely resulted in lower population density because of restrictions on landlords cramming hundreds of people into the smallest space possible for the maximum profit. Now, over 40% of all buildings on Manhattan could not be built today for violating NYC’s zoning code for at least one reason. For instance, most buildings in neighborhoods like West Village and Lower East Side have not changed in a century; there is limited demolition and reconstruction every year. However, population density has significantly fallen as apartments grow larger and rooms formerly designed for multiple people in one room now only have one or two occupants. Even if the buildings and streets don’t change, the ways they are occupied can and do. In other words, changing in zoning and land use are not necessarily visibly imprinted on the plan of streets, particularly if those streets are rigid grids.
  • Transportation patterns: This is the biggest factor encouraging extensive and rapid street network development with low population density – i.e. sprawl. Before the nineteenth-century inventions of railways and streetcars, and the twentieth-century’s auto-based suburbanization, transportation and commuting were prohibitively difficult. People needed to live near to where they worked in what was largely a pedestrian and walking city on unpaved streets. Transportation challenges caused urban growth to be dense and built-up near to places of employment. As a result, many cities like Paris and London might appear small on old maps and occupy only a few square miles pre-1800, even though their population and economic importance were far larger than their surface area on maps leads one to assume. Again, less developed street networks and a smaller surface area of urban development does not neccessarily mean the corresponding city is less culturally or economically important.
Before the introduction of subways in the early twentieth century, the difficulties of commuting greater distances over land and water drove a denser form of urbanism than today. Manhattan, by 1900, had over 2.3 million residents in comparison to only 1.6 million in 2020. These people were crowded into dense blocks with upward of half a million people per square mile. Over the following century, although Manhattan’s population declined by 700,000 people, the street network today is almost identical to a century ago – no smaller and no larger despite major shifts. These shifts in density and demographics simply do not show up on conventional street maps.
My animation below shows the evolution of Manhattan’s built-up area and population density from 1800 to 2010. Notice the steady upward march of street development versus the sudden spike in population density on the Lower East Side in 1910 at over 300,000 people per square mile (in contrast to less than 90,000 in 2010). For every decade after the construction of subways, there has been more urban sprawl, and the island’s population density has fallen. Notice how fluctuations in population density operate semi-independently of street-network development.


Modified from Shlomo Angel and Patrick Lamson-Hall’s NYU Stern Urbanization Project,
here and here.

The animation on the left tells one story of continuous and north-moving development, while the animation on right tells a more nuanced story of population density. The challenge is to find a graphic representation that tells both stories, as neither visualizes all the nuances of urban history.


Possible Solutions

Discussing this problem of street networks with professor Kenneth Jackson, he suggested looking at building Floor Area Ratio (abbreviated FAR), which is the building height and size relative to the amount of land the building occupies. A skyscraper has high FAR. A trailer park has low FAR.
This method of representing urban growth would, in theory, produce three sets of maps: 1) a map of street network development; 2) a density map of people living per square mile; 3) a map of building height and size. This would complicate things but produce a far more accurate representation of urban growth (how to represent this and if enough data exists is another matter).
These three factors – streets, FAR, and population density – act semi-independently of each other. Different urban typologies will share a different mixture of these three factors. Only through analysis of the relationship between these three factors can one begin to understand the underlying demographic, economic, zoning, and historical differences between neighborhoods. For instance:
  • Downtown commercial district like Lower Manhattan: low population density but high FAR. In this case, FAR operates in inverse proportion to residential population density. Buildings can be dozens of stories but have no residents. There is a high density of tall buildings (i.e. high FAR) but low population density.
  • Slum like South Bronx in the 1980s: extensive (though poorly-maintained) street network development, high density, but low FAR because slum dwellings are typically informal without the construction quality required to build high. Buildings might be fewer than six stories and without elevators, as in the Lower East Side, but can contain hundreds or thousands of residents over the tenement’s lifespan. There is low density of tall buildings (i.e. low FAR) but high population density.
  • Suburb like Forest Hills, Queens: extensive (and well-maintained) street network development, low density, and low FAR. In wealthier suburbs, in particular, FAR is kept prohibitively low. Restrictions on minimum lot size required to build, minimum house size, and legal hurdles on subdividing larger lots into smaller ones all serve to enforce a certain quality and price of residential construction that often prices-out lower-income communities of color. There is low density development with numerous green spaces between free-standing homes (i.e. low FAR) and low population density.


Left: NYC population by day in 2015. Right: NYC population by night in 2015. The population doubles by day.


The two density maps above are one illustration of FAR and help nuance Manhattan’s historical development. Areas with the highest FAR tend to be commercial areas with daytime office workers and commuters. The left map shows the daytime population density of the over two million commuters. The areas with highest worker density neatly map onto the same areas of Lower Manhattan and Midtown with skyscraper clusters (i.e. high FAR). The right map shows nighttime population density of residential areas, which also neatly map onto areas with generally lower building height and density (i.e. low FAR). Notice the gray-colored zones in Lower Manhattan and Midtown with an almost zero nighttime population density, which are incidentally the areas with the highest daytime population density and the tallest buildings.
In twenty-first-century New York City, it is quite easy to examine the relationship between these three factors – street network, population density, and FAR – as the datasets are readily available from NYC Open Data. Yet, this all becomes more difficult – perhaps prohibitively difficult – for historical mapping. Calculating FAR for historical Manhattan is certainly possible through scrutinizing digitized historical Sanborn fire insurance maps that go so far as to specify building footprint, materials, and height. However, at the moment, this data is not easily accessible. Historical building footprints and FAR must be calculated through manually scanning, tracing, and inputting data from the New York Public Library’s collection. This must be done for thousands (even millions of buildings) over hundreds of years.
As technology improves, it may be possible in a few decades to translate historical maps into data files that reveal FAR. If historical maps could be scanned and immediately transformed from image files to geospatial data files, the possibilities of using historical maps to inform contemporary research are endless. If and when there is the data on historical FAR, it may be possible to create a new paradigm for studying urban history.


New York City Population Density in 1900

Author’s illustration based on population per municipal ward from 1900 Federal census

Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *