CUNY MSDS Capstone Project
Taxation Equity for NYC Co-ops
PK O’Flaherty
Stage 8 - wrote data and variables section
2025-06-07
Abstract:
The city’s ordinary property owners, from the most vulnerable to the highly educated, don’t even know they’re being overtaxed because the harm is so abstract and technical.
Introduction:
Initially we supposed data science techniques could verify that there are property tax inequalities in NYC. We could do this by building a model predicting assessed tax values and compare them to actual tax assessments. If wealthier buildings had larger errors than mid and poorer buildings then we would have evidence they are underassessed.
What we found was a rich world of analysis, tradeoffs and historical issues that suggests the NYC property tax system largely shifts a disproportionate tax burden onto the poor and middle classes with a modest burden for wealthier people. Our paper shifted from establishing evidence of inequality to providing a tool for individual buildings to determine if they are overassessed, within the current framework, and have an argument to appeal their taxes.
A Ridiculous Example
NYC property taxes were $33.7 billion in 2025, a $1 billion increase compared to 2024, and is the single largest revenue source for the city, representing 44% of all New York City tax revenue. 45% of those property taxes are levied against multifamily buildings. NYC determines how much in property taxes they need to collect and backs into the tax percent of property value required to meet that objective. Let’s say there were only two multifamily buildings in NYC, yours and your friend’s across the street, and each had a property value of one dollar, the city needs to raise 50 cents and would assess a 25% property tax for the year. Now if your friend had successfully appealed his taxable property value and the city lowered his building’s value to 50 cents, the city still needs to raise 50 cents and so would assess a 33% property tax for the year. You pay 33 cents and your friend pays 17 cents. Over time this preferential tax treatment could allow your friend to save enough, and for you to deplete your savings enough, to where your friend ends up buying your building and starts charging you 60 cents a year in rent, covering both of the building’s property tax and allowing your friend to live with more on less. Not having to work anymore your friend runs for city council and changes the law so rental buildings pay twice as much in taxes, further lowering his personal tax and justifying a 20 cent rent increase for you.
It’s a ridiculous example but widening wealth gaps driven in part by property tax inequities is literally happening in the city. Currently the amount of taxes rental buildings pay is between 4.5 and 6 times what equivalent non rental buildings pay, which is passed through to renters via high rents, so that roughly a third of the rents paid in NYC go to property taxes, exacerbating the difficulty for renters to save enough to buy their homes and reduce their property tax burdens. We are living in a ridiculous example.
A Classist Class System
NYC divides property into four tax classes. Tax Class 1 are 1-3 family homes, think brownstone, and Tax Class 2 are multi-family homes like co-ops and condos. While Class 1 makes up 47% of total market value and Class 2 25% of the total taxable property value in the city, only 15% of the property taxes are collected from Class 1 (first class?) and 45% of the property taxes is collected from Class 2 (second class?). It bears emphasizing that these percentages are fixed and so if the taxable value of all class 2 properties dropped in half, 45% of the property taxes would still be collected from class 2 and all class 2 properties tax liability would stay the same.
Note the remaining 40% of property taxes are collected from utilities (class 3 at 6%) and businesses (class 4 at 34%). Examples of businesses in class 4 being office buildings, stores, hotels, factories and lofts. The last is because lofts were converted from manufacturing, warehousing or storage into high-ceiling residences. Together, class 3 and 4 comprise the remaining 28% of total taxable property in the city but due to tax abatement programs, covid era hotel programs and a disincentive to tax businesses during market downturns, class 4 alone may represent 37% of the market value of the cities properties with class 3 raising that an unknown additional amount shrouded in regulations around utility monopolies.
Where to Go From Here
How to fix the inequalities in the current taxation system in NYC is outside of the scope of this paper. The percentage split between the four classes goes back to when the property tax system was created in 1981 and hasn’t been updated since then to current proportions. Likely class 1 property owners have more political clout and are able to perpetuate their smaller share of the total tax bill. There are additional systemic inequities such as class 2 buildings being valued as if they were a rental building which has the effect of overvaluing cheaper buildings and undervaluing more expensive buildings. This methodology, too, is locked into the 1981 tax reforms and can’t be addressed without major political will and clarity of outcomes.
What we can do is help the roughly 5.75 million residents living in class 2 properties (versus the ~1.98 million living in class 1 properties) determine if they are being treated fairly within the constraints of the current system. We should be able to build a model that predicts the city Department of Finance’s estimate of the value of class 2 properties, and determine if a property is likely under or over valued.
Our original hypothesis was that sophisticated owners would have the resources and education to appeal their tax assessments and win, however the literature reveals this to already be proven, even after factoring out the systemic tax advantages already enjoyed of owning luxury units in class 2 properties.
Now we try to give owners without those same wealth, accounting and legal resources, networks and education, a tool to possibly lower their own tax burden.
Literature Review:
I started out with the premise that sophisticated owners were able to successfully appeal property tax assessments. As I learned about NYC’s Department of Finance’s methodology for calculating taxes I assumed some of the over/under assessment for poor/rich buildings would be due to taking statistical averages. The literature review supported these ideas and suggest that my contribution is in educating the reader on how to understand property taxes in NYC and help them determine if they should appeal their own property taxes.
I restricted articles to those written after 1981 when the current rules for taxation were codified in NYC.
The Invisible Problem
Hayashi (2012) introduces the concept of legal salience, that if people can name how they are injured, then they are more likely to be able to initiate and succeed in getting compensated in court. Property tax assessments have low legal salience, meaning regular property owners can’t easily recognize or articulate when they are being overassessed, especially in an opaque system like NYC’s. As a result all but the most sophisticated property owners are discouraged from the appeals process, reinforcing inequality. With every win by a sophisticated owner, the tax burden is shifted a little more disproportionately onto ordinary owners who are poorer and experience poorer outcomes or complete non-participation in the legal system’s venue for property tax appeals.
Systemic Inequity
In 1975, ‘Hellerstein v. Assessor of Islip’ determined that fractional assessments of full market value for determining property taxes were illegal (Scanlon and Cohen 2011). At the time, tax assessors made determinations individually, which led to idiosyncratic assessments and facile bribery. That court case was about to dramatically shift property tax burden onto commercial properties and to avoid that, the New York State Legislature, to prevent significant disruptions, wrote into law S. 7000A in 1981, which is our current taxation system today. However shifts in population, market value and economic activity in the last 44 years hasn’t been reflected in the unchanged tax burden percentages assigned across the four property classes. For example, Class 1 properties (1-3 unit family homes) have an effective tax rate of 0.67% compared to the 3.31% of Class 2 (co-ops, condos and rentals) or the 3.85% of Class 4 (commercial/industrial). So while Class 1 properties own 49% of property value, they contribute only 15% of tax revenue. Vastly more renters live in Class 2 housing, resulting in renters paying significantly more of passed through property tax than the fewer, wealthier renters living in Class 1 housing. While there are co-op and condo tax abatements to bring Class 2 property taxes closer in line to Class 1, these abatements only benefit non-pied-a-terre owners, not renters. Scanlon and Cohen also confirmed wealthier properties to be more likely to appeal assessments, and more likely to win.
This perpetuation of tax advantages for the housing of largely wealthier residents drives a widening of the wealth gap through sale prices as well. Hodge, Komarek & McAllister (2024) found that overassessed lower value properties sold at a 13% discount and underassessed higher value properties sold for a 10% premium. So not only does the assessment inequities redistribute the tax burden regressively, they reshape actual wealth.
Some of this property tax inequity exists in the NYC Department of Finance’s (DOF) property tax calculation methodology as well. The Furman Center (2013) describes how Section 581 of the city’s Real Property Tax Law values Class 2 properties as if they were rental properties even though they don’t generate rental income. This leads to inaccurate assessments especially for high-value properties because no true rental comparables exist, especially when compared to rent-regulated buildings. For example, in 2012, 50 individual co-op units sold for more than the entire building’s official DOF market value estimate. The Furman Center iterated that the city is clearly a regressive taxation system. In the four-class property tax system the city specifies how much property tax will be assessed from each class and sets the rates accordingly, so when high-value properties are underassessed it leads to larger increases on low and mid-value properties, shifting the tax burden on to poorer people. The city can only pick one rate per class such that when applied to everyone in that class’s assessed property tax value, produces the amount of taxes they need from that class. Again, the tax win of a wealthier property is cent-for-cent distributed across every other property, and lower and mid-value properties are not getting wins.
An additional source of tax inequity may be investor size. Xiao (2022) found nationally that larger investors (100+ units owned) have a tax assessment discount compared to small investors and individual home owners. She found the large investor tax assessment discount was larger in areas with any one of three characteristics: 1) A high tax burden, 2) a high concentration of large investors, and 3) a fairer property tax administration. NYC has both a high tax burden and a large concentration of large investors. It was not clear what made a fairer property tax administration or if New York City was considered one, but it seems like a paradox that a fairer property tax administration would allow for a persistent tax discount for large investors. It may be that large investors are able to get tax abatements in exchange for development promises or are large enough to game the rules in other ways, being ultra sophisticated owners with access to the best resources.
Capitalization Rates’ Role in Horizontal and Vertical Inequities
There are several important, but subtle, concepts: horizontal equity, vertical equity and capitalization rates. The NYC Independent Budget Office (2022) confirms there are horizontal inequities (where similarly priced properties have different assessments) and vertical inequities (where wealthier properties pay proportionally less in property taxes). The DOF uses high capitalization rates (cap rates, rates used to discount future streams of hypothetical rental income to a present day property value) that result in lower market valuations. Since the DOF doesn’t use actual market data this can lead to tax inequities. The DOF can’t change from cap rates to fair market value because they operate under the S7000A law from 1981, however they could lower cap rates for high value properties and raise them for low-value properties to narrow vertical inequities.
An additional example to clarify both the terminology of horizontal and vertical equity, and progressive and regressive tax systems - Horizontal equity is when two people, in two equal homes, pay equal property tax. Vertical equity is when a third person, in a more expensive home, pays more property tax. How much more taxes the person in the more expensive home pays determines what kind of tax system you have. A proportional tax system would have the owner of the more expensive home pay an equal percentage of taxes relative to home value as the first two. Science suggests this still puts a disproportionate tax burden on the first two and a more equal outcome is achieved if the more expensive home pays a greater percentage of property value as tax. This is a progressive tax system and generally leads to better wealth equality and stable social outcomes. The opposite is a regressive tax system where the owner of the wealthier home pays a lower percentage of the home’s value in property tax. New York City’s is a regressive property tax system that exacerbates the wealth-gap between haves and have-nots.
Options to Improve the Current System
The Independent Budget Office (IBO) has two other suggestions to be more fair. Using median instead of average transactions would bring down artificially high capitalization rates. Imagine a long flat trending up curve that spikes incredibly high over a short distance towards the end on the far right. This is a visual representation of property values in the city. By taking the average of the values to determine the capitalization rate the DOF is using a rate high above the curve, influenced by the outlier wealthy values. If the DOF used median values to determine the capitalization rates then it would be on the curve and wealthier properties would be valued higher in proportion to the median properties compared to when taking the average.
The second IBO suggestion for fairness is to adjust cap rates between property types. For example, to use a slightly lower cap rate for office buildings in a prime midtown location compared to a lower-tier office neighborhood. The DOF would be challenged to adjust capitalization rates based soley on property characteristics, such as building type, location and usage. Attempting to tweak the capitalization rates in order to get fair market outcomes could invite lawsuits for violating equal protection under the Fourteenth Amendment, which requires that similarly situated individuals be treated equally under the law, or for violating property rights protected under the Fifth Amendment’s Takings Clause and due process protections.
Inconsistencies in the City’s Processes
Our task of providing a tool to help buildings produce evidence if they are overassessed relative to their peers is further complicated by inconsistencies in the DOF’s processes. Goor (2017) found that the NYC DOF deviates significantly from its publicized process when calculating property taxes and that property taxes are poorly correlated with land, market and assessed values. Since we will be using similar characteristics as inputs in our models we should expect a higher degree of error due to the DOF’s process inconsistencies.
The Burden of Renting
Rachel Michelle Goor’s 2017 paper “Only the little people pay taxes” had several general findings of note such as a lot of Class 2c luxury condos’ nearby rental buildings have rental controls leading to deep undervaluations for luxury condos. She determined that property tax exemptions are granted for people’s ability to organize and lobby the State legislature, not because those tax exemptions were sound tax policy. Her emphasis was on homeowners versus renters and she found homeowners have 46% of the estimated market value but pay 15% of the total property tax burden (coincident to but not the same as the 15% of tax paid by Class 1 properties), while rental properties have 24% of the estimated market value but pay 37% of the property tax. This is 4.73 times more tax paid for market value as a rental building than as a homeowner, though she stated renters in NYC were paying six times as much (presumably on a different basis) in property taxes as homeowners do compared to the average outside of NYC which is closer to renters paying 1.5 times as much property taxes as homeowners
Reform Proposals
A second, highly influential paper for us was Lizzie (Yea Won) Lee’s, 2023, “Evaluating the ‘Road to Reform’ for New York City’s Property Tax System”.
Like Goor used machine learning to identify that the DOF was deviating significantly from its publicly stated processes. It follows that Lee (2023) supports the DOF using machine learning (XGBoost) as a viable tool for improving fairness and transparency in tax assessments by flagging valuation anomalies for further investigation.
Like Goor, Lee found that the DOF’s methodology is not disclosed nor reproducible and so it’s difficult to verify if a DOF valuation is accurate, or how a valuation compares to equivalent properties, or how equitable assessments are between property types. This will directly impact our likelihood of success, but we have to start somewhere, for free, without the need to engage expensive law firms and experts.
Lee characterized NYC’s property tax system as extremely regressive with low-income neighborhoods often facing higher effective tax rates. She agrees with the New York City Advisory Commission on Property Tax Reforms 2021 final report for Class 2 to be split into less than inclusive or greater than 10 units, as well as switching to a sales-based market valuation.
A sophisticated criticism she had of The New York City Advisory Commission on Property Tax Reform is that they used quantiles to group properties together to understand impact but that might mean an $800k home and an $80M luxury penthouse might both be in the same top quantile with drastically different impacts. The Commission didn’t specify their quantiles but it could be treating the top 1% or the top 0.1% the same as the top 25%. This could be used to disingenuously misrepresent the tax changes for different groups, or attribute tax increases of the top 0.1% as being burdened by ordinary homeowners unfairly.
It’s not from the literature review but let’s take a moment to say that switching to fair market value has it’s own drawbacks. Without proper smoothing, gentrification would accelerate as increases in home prices would leverage property taxes higher for the residents who had been there prior to gentrification.
Lee suggests NYC may be reluctant to reform property tax because it’s one third of NYC revenue and fears high-income outmigration away from the city. She encourages NYC to lower income from property tax and discourages carve-out plans which lower revenue hoping to stimulate construction but that have mixed results.
While her analysis of what should be NYC’s path to property tax reform is outside of the scope of this paper, it helps frame what ordinary property owners can do to achieve more equitable property outcomes for themselves.
Summary
The literature strongly supports that NYC’s property tax system is regressive, opaque, and full of systemic and methodological inequities. The 1981 statute prescribes an outdated division of tax burden that punishes poor and middle-class residents in favor of wealthier, low occupancy or luxury buildings. Inconsistencies in processes further reduce the legal saliency, or residents’ ability, to name and confirm the financial injury of being overassessed for property taxes. Our task is not to present a path to reform but to provide a free, data-driven attempt to support an appeal, where justified, for ordinary home owners to reduce their tax assessments to fair levels relative to equivalent neighboring properties.
Referenced Papers
Hayashi, A. T. (2012). The Legal Salience of
Taxation. SSRN Electronic Journal.
DOI Link
Scanlon and Cohen (2011) Distribution of the Burden of New
York City’s Property Tax - The Furman Center for Real Estate & Urban
Policy
PSU
Link
Hodge, T. R., Komarek, T. M., & McAllister, A. (2024).
A Double Negative: Capitalizing on Assessment
Regressivity.
DOI Link
Furman Center Policy Brief (2013) Shifting the Burden:
Examining the Undertaxation of Some of the Most Valuable Properties in
New York City - The Furman Center for Real Estate & Urban
Policy
Furman
Center Link
Xiao, S. W. (2022). Investor Scale and Property
Taxation.
DOI Link
NYC Independent Budget Office (2022). Does NYC’s Method
for Assessing Commercial Property Values Result in
Inequities)
PDF
Link
Goor, R. M. (2017). Only the little people pay
taxes.
URI Link
Lee, Lizzie (Yea Won) (2023) Evaluating the ‘Road to Reform’
for New York City’s Property Tax System
DOI Link
Extended Bibliography
New York City Independent Budget Office (?). The Coop/Condo
Abatement and Residential Property Tax Reform in New York
City
Link
New York City Independent Budget Office (2006). Twenty-Five
Years After S7000A: How Property Tax Burdens Have Shifted in New York
City
Link
New York City Independent Budget Office (2013). The Coop
& Condo Tax Break Has Expired, Giving Albany Chance for
Long-Promised Fix
Link
Shi, Boicourt, Ng, et al. (2024). An Assessment of NYC
Cooperative Housing’s Climate Vulnerability and Barriers to
Adaptation
Link
Cetrino, Benjamin (2014) Classification of Property for
Taxation in New York State
Link
Nadine Brozan (2002) For Co-op Complexes, Complex
Choices
NYT
Archival Link
Berry, C. (2021). An Evaluation of the Residential
Property Tax Equity in New York City
PDF
Link
Research Question:
In NYC, the property tax liability for a building is ultimately based on a single number: the total market value of the tax lot, as assessed by the Department of Finance (DOF). The final tax number is influenced by tax abatements and exemptions that are based on policy, which while maybe not fair or optimized for a given property, are not appealable. Additionally, transitional formulas cause a delay in the full reflection of changes in the market value assessment in any given year. So, while there is a lot of noise in a building’s final tax liability, the signal is the assessed total market value.
This project aims to model this assessed market value using publicly available DOF data, NYC land use files (PLUTO) and averaged sale price information. By comparing our model’s predicted value to the city’s actual assessed market value for each tax lot, we can provide insight to a building as to whether it is likely over-, under- or appropriately assessed. In this way we can support ordinary homeowners who may not otherwise have the means to know if they are overassessed or not.
Our research question is:
Can we build a model that predicts the total market value assessed by the DOF? And in so doing provide a tool for co-op boards and residents, especially those without sophisticated legal or accounting resources, to know whether their building may have a good case for appeal?
We suspect that highly sophisticated owners, with the resources to own high-cost or luxury buildings, are likely able to measure and appeal tax assessments to keep their tax liability fair or underassessed, even in addition to the systemic low-tax benefits already enjoyed by high-cost or luxury buildings in the NYC regressive property tax system. These low assessments would appear as higher predicted assessed market values than what was actually assessed.
There is no need for a traditional hypothesis and null hypothesis, we aren’t running statistical tests on randomized groups. Instead we’re building a predictive model to surface actionable data for buildings considering if they are suffering the invisible harm of being overassessed in property taxes compared to their peers.
Data and Variables:
Our initial data journey led to an ambitious layering of datasets across multiple disciplines and a lot of confusion with attempts to do individual building valuations to support this complex network of data. What we feared when we weren’t doubting ourselves, and which was echoed in the literature review, is that NYC seems opaque and inconsistent with their tax assessments.
The literature review also confirmed our initial hypothesis of there being inequities in the NYC property tax system and sophisticated owners successfully appealing their property taxes.
Because of these two things we shifted our focus to something simpler, a tool to help ordinary owners get a sense of whether they are being fairly assessed.
Overview
We selected the DOF expanded Property Valuation and Assessment Data (PVAD) as the basis for our work, restricted to the 10019 zipcode for computational limitations (and because it includes the co-op your author is the treasurer of). We restricted our work to Class 2 properties as Class 1 properties (1-3 Units, think brownstones) already enjoy enormous property tax protections compared to the majority of resident-owners in co-operative or condo housing.
Ultimately we selected total market value
(FINMKTTOT), a PVAD field, as assessed by the NYC
Department of Finance (DOF) as our target value to predict since it is
the primary signal for property tax assessed, ignoring any distortion
through policies with abatements and exemptions.
We added data from PLUTO and Sales data to try and capture additional information that could offer insight into how the DOL assesses total market value and provide a source of evidence for appealing market value assessments.
Datasets Used
Property Valuation and Assessment Data (PVAD)
The NYC Department of Finance’s (DOF’s) Property Valuation and Assessment Data provides the core features that we’ll use to predict assessed market value. Open Data NYC contains also an expanded PVAD with records up to the present which we used instead of the first one after our initial explorations.
PLUTO (Primary Land Use Tax Lot Output)
The PLUTO data was compiled by the Department of City Planning and largely duplicates PVAD as it pulls from the DOF’s data to produce its own. Some fields of interest are if school districts or police precinct’s have any predictive power of assessed tax value, and if so how much of the that is instead explained by location. PLUTO provides coordinates that we can use later to generate an average local assessed market value (presumably per lot square foot) for use in our regression model for feature analysis, and more directly in our intended final XGBoost model for predicting assessed market value.
Rolling Sales Data
Originally we wanted a flag to identify sophisticated owners. We tried pulling educational attainment data from the US Bureau of Census at the zip code (and then census tract or voting precinct level) but each zip code has thousands of buildings/tax lots. An alternative flag was price per squre foot for the average apartment in the building, where less than $1,200 is low, between $1,200-$1,300 is mid and above $1,300/sqft is high with the possibility of using a decision tree for better cutoffs between the groups, or using quartiles, however we only had building square footage to match to sales and we arrived at simply using the average sales price for a given building or tax lot as our marker for sophistication.
What we found is the Sales data only includes sales going back a year and a number of tax lots experienced no sales in that time frame so we have substantial missingness in this data field.
Datasets Not Used
There are four additional sources of data that could be used to expand scope in future extensions of this project.
Property Tax Exemption Detail
“Government, Not-for-Profit and Commercial Exemptions maintained by NYC Department of Finance. The Department of Finance administers a number of benefits for property owners in the form of exemptions and abatements. Exemptions lower the amount of tax one owes by reducing the property’s assessed value. This data contains exemption information for Government, Not-for-Profit and Commercial Exemptions.”
https://catalog.data.gov/dataset/property-exemption-detail
We are skipping property tax exemption details because we found that exemptions (and abatements) act as a distortion when determining if the city is overassessing your building. Any building should first determine whether they have optimized their exemptions and abatements for their resident shareholders, then confirm their property tax calculations as accurate using historical DOF market value assessments, and then the building can question if those market value assessments are too high or not using something like this model if not traditional methods in the ken of expensive lawyers.
However if a research question were about total impact of tax policy, they would want to reflect exemptions to capture total taxes collected.
Assessment Actions
“Assessment Actions Actions on Applications for Reducing Assessments or Reclassifying Property. Listed here are Tax Commission actions for reducing assessments or reclassifying property. KEY: YR=Assessment year; B=Borough (1=Manhattan, 2=Bronx, 3=Brooklyn, 4=Queens, 5=Staten Island); TC=Tax Class or subclass. Classification claims. Reductions are expressed in total actual assessed value. For condominiums, actions shown are for representative lots only.”
https://catalog.data.gov/dataset/assessment-actions
We are skipping appeal information because the literature had already resoundingly supported that sophisticated ownders were more likely to appeal and win those appeals, shifting tax burden onto less sophisticated home owners. An extension of this project could be to review what evidence is submitted in successful property tax appeals and learning how to duplicate those efforts for buildings identified with our model as potentially being overassessed.
A future researcher could pair the output of our final model against the assessment data. Tax Lots with a successful appeal should also have predicted total market values at or above their actual assessments.
American Community Survey
When the US Bureau of Census data didn’t seem granular enough to distinguish sophisticated buildings, and before we decided average sales price of a unit was the best, simple proxy, we found the following American Community Survey data which could prove a useful additional feature layer for associated research, especially into the impact on taxation on vulnerable communities by family structure, race or immigration status.
MapPluto
MapPLUTO is the spatial version of the PLUTO data using tax lot geometries from the Department of Finance’s Digital Tax Map. This could allow a future researcher to better interface with applications to generate beautiful maps. For our purposes it was sufficient to obtain the coordinate information in the regular PLUTO data for exploring location on assessed tax values.
Domain Context
How the City Calculates Taxes
NYC Finance - Digital Tax Map of Tax Block 1046 in Manhattan
The link above clicks to the DOF’s tax map for tax block 1046 in Manhattan. The block is circumscribed by 55th and 56th Street and 8th and 9th Avenue, then further divided into unique tax lots that are each administered independently. No condos appear to be pictured in this block however they are handled uniquely. Each condo is treated separately as it’s own tax lot and are aggregated together into one combined tax lot for purposes of reporting on a building level.
Each individual lot receives a market value and tax assessment every year. Here is the NYC Department of Finance’s (DOF’s) landing page for lot 23 from 1046, the block linked above: access.nyc.gov/1010460023.
Here we’re able to see some core fields (also in the PVAD) such as building class, lot size and number of stories; as well as historical market values (which are used to smooth out changes in the market value over a five year period) and a progression from estimated market value for the year to a 45% proration of the market value, to a transitional prorated market value to account for not being increasing more than 30% in a given five year period and the sum of past five years’ changes spreadout over the next five years respective. Lastly the exemptions are subtracted the transitional prorated market value to arrive at the assessed taxable value for the property.
NYC adds up all of the taxable values for these Class 2 properties and then decides on the city’s tax rate percentage for class 2 properties, that when multiplied by this total taxable value of Class 2 properties, results in the 45% of total levied property tax that the city collects from Class 2. To learn more, you can visit: http://nyc.gov/assessments.
Assessment Mechanics
The assessment ratio of 0.45 is perceptural. It changes the way the tax rate appears to property owners and to anyone comparing tax rates between classes. If there were no assessment ratio, or effectively an assessment ratio of 1.00, then the city’s tax rate for 2024-2025 would have been 5.757% instead of 12.86%. This suggests intuitively that the assessment ratio exists to make the city’s tax rates seem comparable between classes or across other metrics of taxation such as income or sales tax. Because of this we can ignore the assessment ratio in determining if the assessment is fair.
But besides the assessment ratio making taxable value of building’s look low, the estimated market value of the building’s is low, too. This may be intentional for psychological reasons. People may look at the low estimated market values and even lower assessed values after the application of the 0.45 assessment ratio and think they are getting a deal. They may be scared to inquire because they don’t want a recalculation to go against their interest and be higher. One of the papers in the literature review spoke about the low legal salience of overassessed taxation where it’s hard for someone to understand or quantify the financial injury of overtaxation and so they never appeal their overtaxation in court. But it’s now how low their assessment is, it’s how equal their assessment is relative to equal buildings (horizontal equity), and how low their assessment is relative to more valuable buildings (vertical equity).
Market Value = NOI / Cap Rates
The estimated market value determined by the DOF is not based on sales prices or true market value. Rather the DOF calculates what the building’s Net Operating Income (NOI) would be if it were a rental building and divide by a Capitalization Rate (cap rate) to arrive at the estimated market value.
In the words of the DOF:
> “we use statistical modeling to calculate the typical income and
expenses for properties similar to yours in size, location, age, and
number of units. The process varies depending upon whether your property
has more or less than 10 units.“
Cap Rate
The cap rate isn’t uniform across Class 2. The DOF sets cap rates by asset type, neighborhood and income range. If two equivalent buildings had different cap rates one of them could sue the city for violation of equal protection under the law. And so while this offers a measure of equivalency, the system for cap rate selection is a source of inequality. Two neighboring buildings of similar age and size may have been managed very differently over an 80 year-span and be worth drastically different amounts based on upkeep and underlying indebtedness, yet have identical property tax assessments (vertical inequity). Similarly two identical buildings across the street from each other but across a neighborhood divide, might command similar unit sale prices yet one has a drastically lower taxable value (horizontal inequity) because it’s neighborhood compands a higher cap rate.
So if a building has above-average income for it’s cap rate category, the cap rate will undervalue it. If a building has below-average income for it’s cap rate category, the cap rate will overvalue it. This becomes a mechanism for systemic bias where wealthier properties of the same type will pay less in taxes and poorer properties will pay more.
NOI
Net Operating Income (NOI) also introduces inequality to property taxes. NOI is statistically modeled for the majority of buildings. Since the DOF tends to assume an average income (less expense) for buildings it tends to produce lower numbers for wealthier buildings, lower taxable value and higher numbers for poorer buildings, resulting in higher taxes.
Additionally, since the DOF’s methodology uses statistical averages of nearby similar buildings, as wealthier buildings successfully appeal their taxes, those new low assessmentts get picked up as valid comparables, creating a knock-on effect to other nearby wealthy buildings with similar characteristics. This entrenches wins for the wealthy, further perpetuating vertical tax inequities.
Average Sale Price
While sale prices are not factored into the DOF’s estimate of market value, we’d like to capture the difference between wealthy vs poorer buildings with the same characteristics in our modeling process and see average sale price as the simplest proxy for this.
What we may find is that the model doesn’t see average sale price as statistically significant, nor might it be allowed as a basis for protesting property tax assessments.
On the other hand, it could help approximate the cap rate selection process used by the DOF for each category.
Rationale for Focus on Market Value
We’re modeling total market value (FINMKTTOT) because it
is the only appealable number in the tax formula and the core signal
used to compute tax liability. The final tax bill is both distorted by
policy-based abatements and exemptions but also a market value smoothing
mechanism that clouds the total market value assessments very direct
impact on final taxes.
Our initial approach is to use a linear regression model to predict total market value, primarily to surface insights about the different features’ impact on total market value assessment. This should help us build our final XGBoost model which will give us an indicator, with a predicted market value assessment lower than actual assessment indicating a given building has an argument to protest their taxes.
Variables of Interest
Our target variable is estimated market value (or total market value
depending on the source) TOTMKTVAL. This is the main,
appealable, driver in property tax assessments.
Based on the DOF’s publically expressed methodology for calculating market value, significant fields should be size, location, age of building, number of units and whether there are 10 or more units.
Also cap rates are dependent on the asset type so we expect building class to provide significant input as a feature. For example whether the building class begins with a C or a D determines if the building is a walk up or an elevator apartment building respectively.
A complete data dictionary of the final merged and subset dataset is available in Code Chunk 9, referenced below.
Code References
Loading Data sets
Code Chunk 1: Load PVAD
Code Chunk 2: Load expanded PVAD
Code Chunk 3: Load PLUTO
Code Chunk 4: Load Sales Data
Subsetting Data sets
Code Chunk 5: Subset PVAD
Code Chunk 6: Subset PLUTO
Code Chunk 7: Subset Sales Data
Methods and Models:
This section describes the methods you used to analyze the data.
Goal: design and implement a model or set of models to measure or explore these relationships
Ridge Regression
Try linear regression -> I may know what my features are because of the domain dive but if not I could try lasso regression. For a nonlinear model maybe I could do random forest. How would the residuals from a linear vs nonlinear model be a diagnostic in and of itself?
use test/train split and x-fold cross-validation
Residual Analysis
Analyze residuals across building types.
Visualize anomalies - what we’re expecting to see is greater residuals for wealthy and poor buildings w
Statistical Validity
How sound are the research design and methods
Is the sample of observations selected fro the test reflective of the population are the values of the independnet variables not dependent on each other are there significant confounding or exogenous factors influencing the depenent variable and thus need to be controlled for in the model?
Internal Validity
Do the data sets and variables accurately represent the phenomena being explored
External validity
Can the results of the study be generalized
Limitations Potential concerns: It’s possible that if we train the model on the historically available data and there is the presence of unfairness with richer buildings not paying their fair share, then the model would predict relatively lower property tax assessments than you would expect and so the residuals would be normal for the higher end buildings. If we have a nonlinear model that may capture the unfairness anomaly as an expected part of the model so if we use a linear model, and the actual tax assessment methodology is linear then we will avoid this problem. If we can’t use differences in the residuals to identify the tax anomaly then we may have to find other aspects of the model to look for taxation anomalies. [Later, go through and standardize the nomenclature]
Model Selection
How did the type of relationships among the variables or end result influence which statistical or machine learning model was most appropriate
consider the scikit-learn algorithm cheat-sheet OR https://www.analyticssteps.com/blogs/5-statistical-data-analysis-techniques-statistical-modelling-machine-learning
main classifications are: regression / classifcation / clustering / dimensionality reduction
Model Fit
Discuss my work in relationship to overfitting and underfitting
need to discuss feature importance and partial dependence plots - what other visuals can I interpret?
Data Analysis
Discussion of Results:
One concern is that the model might learn the systemic bias and so maybe that’s another way to look at this. If the model is showing even residuals between poor, mid and rich buildings… I guess that would indicate it’s more systemic unfairness than sophisticated owners being able to appeal.
Also it’s easy to see with linear regression, but if we have some low-interpretable non-linear model then I’m not sure what conclusions I’ll be able to draw.
Conclusion:
Final Presentation
Include a link to the youtube presentation ***
write an article for The Co-operator ***
include link to data repository Github or elsewhere or where on the NYC websites I can find the data or, it should already be processed. maybe I can have links to other RPubs documents where the data scrubbing is visible ***
Simple website where you enter your buildings Block and Lot and are able to see a Tax Anomaly-ometer where: Green - fine/under Yellow - fine Orange - over paying Red - definitely appeal Maybe the website can also produce a small report that a board member can bring back to their board to discuss Share as a marketing tool for Daisy our property management company
make the final presentation off of slides and not the final document. ***