How we calculated property-tax takes in London and New York
By THE DATA TEAM
THE current issue of The Economist includes an article and leader comparing the property-tax systems in New York and London. Compared with, say, income taxes, property taxes are a dizzyingly complex and woefully under-studied area of public policy, and precious little information about them is available to the public in a digestible form. A handful of researchers and organisations—notably Max Galka of Metrocosm, New York’s Independent Budget Office and the Institute for Fiscal Studies in London—have done pioneering work on this subject. (A full reading list is set out at the end of this post.) But the only way to conduct an apples-to-apples comparison of the real-life property-tax rates paid in each city on homes in distinct price bands was to crunch the numbers ourselves. This post explains how we went about the task.
The raw material for our analysis of the New York market resides primarily on the website of the city’s Department of Finance. It hosts lists of both every home sale—there were over 50,000 in 2014—and the annual property-tax bill paid by each of the 1.1m households in the Big Apple. The two datasets are compiled separately. We merged them, and calculated the effective annual property-tax rate (after accounting for all listed abatements and exemptions), across four separate home types and 11 price brackets, using a script written in R, an open-source statistical programming language. We have uploaded both the raw data and the script’s output file to GitHub, a code-sharing website.
On top of around $20 billion a year in annual property taxes, New York city and state together collect a further $3 billion or so from taxing real-estate sales. Although the governments do not release records of these levies for individual transactions, they do report the total revenues that the taxes generate. We allocated these proceeds to each price band of properties based on their share of the total sales value in the city in 2014, as reported by the Department of Finance.
Reconciling these “bottom-up” numbers with the overall tax totals announced by the government presented another challenge. Adding up all 1.1m individual property-tax bills yields a sum of revenue that is 12% higher than the city and state’s actual collections. The discrepancy comes primarily from tax abatements not listed in the public record and homeowners who are behind on their payments. Since we could not assign these differences to individual properties, we reduced the tax paid on all properties in order to make the figures line up.
After deriving the aggregate property taxes paid in each price band, the final step to arrive at a tax rate was to calculate its denominator: the total market value of homes in each group. This figure is by definition an estimate, since no one knows exactly how much a property would sell for until it changes hands. We based our valuation on the actual transactions in 2014 listed on the Department of Finance website. We divvied up the total number of units in New York in proportion to the share of 2014 sales represented by each value band and property class, and multiplied the result by the average price for that group.
Our methodology on the other side of the Atlantic was broadly similar. However, working with different data sets required different techniques. The bulk of London’s property-tax revenue comes from a monthly “council tax”, which is collected by the city’s 33 constituent boroughs. The total number of dwellings and rates charged in each borough for each band of home value (as determined by an assessment in 1991) are published on a state statistics website. Homes owned by companies are subject to a separate annual tax charged by Britain’s national government, which announces how much revenue it collects from properties in distinct value groups and how much comes from London.
The final levy in London is stamp duty, which is applied to home sales. Britain’s Land Registry offers a database of every property transaction in the country. Because these do not distinguish London from elsewhere, we wrote another R script to segment out residences in London based on their postcodes, apply current stamp-duty rates to each sale and apportion them by price band.
As with New York, the most difficult step is joining up these distinct records. The numerator of the effective tax rate was simply the sum of the proceeds from council tax, stamp duty and the corporate-owned-residence levy in each price group and borough. As for the denominator, for homes worth less than £500,000 ($780,000) we estimated the average current market value of each council-tax band by applying a borough-specific house-price appreciation average to the bands set in 1991. We could not apply this methodology for properties worth more than £500,000, because the highest council-tax band for some boroughs includes homes worth as little as £750,000 and as much as tens of millions of pounds. In order to distinguish between the residences of plutocrats and those of the merely rich, we made the same assumption we did in New York, and used 2014 sales data to apportion all homes above £500,000 into price buckets.
As in New York, our totals derived from the estimated taxes and resale value of each property in London failed to square up with the higher official numbers on stamp duty receipts. Multiple factors probably account for this discrepancy: transfers of properties under special circumstances, for example, and transactions that have not yet been registered with the agency. So we applied an increase to every home, as we did in New York, to get the aggregates to match up.
Because valuing homes is an inexact science and the data sources are incomplete, the figures in our story should be interpreted as estimates rather than gospel. Different assumptions or techniques can lead to different results. But we are confident that the overall conclusions of our study—that New York’s property taxes are higher, less dependent on transactions, more regressive and less consistent than London’s—are so large that they would easily withstand tweaks in our approach.
