Many real estate developers dream about making their mark on New York City’s skyline. Building big creates a legacy that can be seen miles away. From a business perspective, height creates value and a sustainable competitive advantage. Views from the clouds command equally stratospheric premiums and only a few buildings can give prospective buyers such views.
The reasons why developers want to build big buildings are obvious, so how do they identify the sites where building big is possible?
In this post, I will examine a model I built, with the help of a few handy tools such as Carto (previously CartoDB) & NYC PLUTO, that outlines the real estate growth potential of New York City and helps identify the biggest possible development sites in the city.
I said I will model New York City, but I am only modeling Manhattan south of Central Park. I understand that New York City extends beyond Manhattan. Likewise, the methodology discussed henceforth extends beyond the data set I used. Please do not hurt me.
The basis of this article is a project I did at NYU Stern for Urban Systems, a real estate class taught by Paul Romer. (A second disclaimer: Paul’s personal blog is the inspiration for my own.) In this class, we were challenged to ask questions about cities, perform quick analyses, and come up with general solutions. That is the methodology I will use here. This model is not intended to be a comprehensive review on any individual property. It is intended to paint broad strokes over the city in regard to real estate development potential.
While I’d love to be able to quickly judge which sites can grow the tallest, height is a function of a number of zoning code components such as tower regulations, setback requirements, air rights, special district privileges, etc. Modeling all of that is too complex for now. A quicker way to get a feel for the development potential of any site is reviewing its Floor-Area Ratio (FAR).
A primer on Floor-Area Ratio (FAR)
Floor-Area Ratio is how a real estate developer determines the total floor space a building can have on any given plot of land.
For example, let’s take a lot that is 100′ wide by 100′ long. That lot has 10,000 SF of area. Assume the zoning code states that lot has an FAR of 5.0. Multiply the lot area by the FAR and you get 50,000 SF of total floor space that can be developed. This total is usually called ZFA, or Zoning Floor Area. It’s your floor space limit.
Simple enough, right?
The other important concept is that the Zoning Floor Area can be molded however you see fit (within the bounds of the zoning code). You can build a building with 5 floors on a 10,000 SF footprint using the entire site, you can build a building with 10 floors on a 5,000 SF footprint using half the site, or you can build any other combination as long as the total floor area of the building does not exceed 50,000 SF. The NYC Zoning Code Glossary has a nice picture to hammer home this concept:
This picture also shows how FAR is related to height. Tall buildings need high FARs because tall buildings have a lot of floor space, but FAR isn’t a perfect metric for height because there are other constraints on height. I would certainly bet that there is a correlation between height and FAR, but I have never done a study. Maybe one day…
But I digress. Now that we’ve gone through Floor-Area Ratio and Zoning Floor Area, we have to figure out which properties have the most potential. For that, it’s best to jump into the model.
Building the Model
Every good model starts with great data. Thanks to New York City’s Open Data Plan, which provides that the city must release a ridiculous amount of data to the public, real estate developers with an eye for data analytics have access to tons of great data.
The specific data we need comes from PLUTO, Property Land Use Tax lot Output, which is aggregated by New York City. PLUTO contains zoning information, square footages, building types, tax values, dimensions, and a host of other fields all the way down at the individual lot level indexed by geocodes. This means we can put this data on an actual map!
My favorite tool for getting a handle on PLUTO Data is Chris Whong’s PLUTO Data Downloader. You can select an area, download the data to a CSV file, and import it right into Excel.
Once the data is in Excel, it’s easy to build in metrics. I focused on a few key fields:
lotarea: Area of the lot
builtfar: Total building area / lotarea
commfar: Maximum allowable commercial FAR
facilfar: Maximum allowable facility FAR
residfar: Maximum allowable residential FAR
New York City provides a handy data dictionary if you want more detailed definitions about the various fields.
Using these fields, I came up with a metric I call “Square Foot Growth Potential,” or SFGP. Conceptually, SFGP is the difference between the square footage of floorspace a building currently has and the square footage of floorspace the lot can support as dictated by the zoning code.
For example, if a 10,000 SF lot with an FAR of 5.0 currently has a building with only 15,000 SF of floorspace–equivalent to a built FAR of 1.5–then the Square Foot Growth Potential is 35,000 SF.
Mathematically, SFGP is defined as such:
SFGP = ( Allowable FAR – Built FAR ) * Lot Area
We already have the Built FAR (builtfar) and Lot Area (lotarea), but we don’t have one field for Allowable FAR. We have three: commfar, facilfar, and residfar. To be conservative, I assume the Allowable FAR is equal to the lowest FAR contained in those three fields.
Note that I’m only running this calculation on properties where Allowable FAR is greater than Built FAR. For this exercise, I don’t care about buildings that are somehow larger than what the zoning code allows (via transfer rights, FAR bonuses, etc.). Additionally, I don’t factor in any FAR bonuses beyond what is already in the data.
That’s all that needs to be done to identify targets with high growth potential, but that alone won’t provide anyone with anything useful. There is too much noise. We need to filter it out.
Filtering Out the Noise
The data from PLUTO includes every tax lot in New York City. That means prime development targets like apartment buildings, vacant lots, and gas stations, but it also means public parks, schools, hospitals, government facilities, and other buildings that would be very difficult to develop. Additionally, some buildings are landmarked or located in historic districts. The following fields help us account for these factors:
bldgclass: Building class codes
histdist: Historic district
ownertype: Tax lot ownership type
Building class codes allow us to distinguish between the different types of buildings. There are over 200 different categories and subcategories ranging from bungalow-style one family dwellings to outdoor pools to orphanages. A full list of the building codes is available in Appendix C of the data dictionary.
I excluded over 80 building codes from my analysis including, but not limited to:
- Luxury hotels and dormitories
- Hospitals, infirmaries, and nursing homes
- Churches, synagogues, and convents
- Asylums, orphanages, and detention centers
- Concert halls, museums, and libraries
- Parks, playgrounds, and public pools
- Airports, piers, and docks
- Utility facilities, transportation infrastructure, and railroads
- Various government-owned land types
- Schools, colleges, and seminaries
- Court houses, post offices, and cemeteries
You can see that the building codes contain a great deal of valuable information that helps us hone our analysis.
Next, I factored out any property that is within a historic district or deemed a landmark.
Lastly, I factored out any lot with an ownertype of ‘O’ or ‘C’. An ownertype of ‘C’ means the lot is owned by the city. An ownertype of ‘O’ means the lot is owned by a public authority, state, or federal government.
All of this checking is an attempt to weed out properties that, in reality, would be significantly more difficult to redevelop than your standard, privately-owned townhouse, apartment, or office building.
The End Result
After downloading the data, building the metrics, and cleaning things up, the dataset is ready for use.
As I mentioned before, this data is geocoded, so we can easily plop it onto a map. That’s where Carto becomes extraordinarily valuable.
Using Carto and geocoded data, we can turn an Excel model into a map and visualize the analysis I outlined above. For your viewing pleasure, I filtered the map to include only lots with over ~20,000 SFGP.
Here’s what the final product looks like:
If you browse around the map, you’ll be sure to notice some of the most well-known recent developments like 432 Park (zoned as 53 East 57th Street), boasting 252k SFGP, and forthcoming 80 Essex (part of Essex Crossing), with 130k SFGP.
You might also discover some sites that you didn’t know were in the works. I didn’t know Silverstein is looking to develop 514 11th Avenue, which we model at over 800k SFGP, into a super-tall tower until this model highlighted that property. (Turns out Silverstein may back out, though.)
What interests me most are the underdeveloped sites in prime locations. For example, our model tabs 420 7th Avenue at 133k SFGP and the site is heavily trafficked next to Madison Square Garden, but the building appears to be a one-story Bank of America/Modell’s according to Google Maps.
I wonder what’s going on with that…
As you can see, PLUTO + Carto can be a very powerful combination for real estate developers. Building this model only took a couple hours of work, but it provides me with the ability to scan the entire city and find opportunities with high growth potential visually. Adjusting Carto filters on SFGP, lot dimensions, and number of floors further enables me to pinpoint buildings of interest.
As I continue exploring real estate and New York City, I’m sure I’ll be diving into things I discovered from this simple tool. I hope it piques your interest as much as it does mine.