Posted Tuesday, December 15 2009
I thought I'd lay out the methodology for the New York home and apartment sales price data a bit more today. As I've mentioned before, the goal is to have a well defined index for residential real estate prices in New York. In order to do that I think I have to be very clear about how I'm calculating and tracking these metrics. So let's begin.
Background on New York's Property Sales Data
To reiterate, all the aggregate real estate metrics for New York on Department of Numbers are calculated from the city's individual property sales data. The city releases a rolling update every month that includes all property that has sold (and been recorded) within the last twelve months. This is what I use for the monthly price updates. The city usually releases prior month data in the first week of the month (e.g. November data is released the first week of December). In theory, we could estimate the median November sales prices in early December, but in practice it seems like the reported sales for the most recent month are too few to make a reasonable estimate of median prices at that point. At least for now, I'm not reporting aggregate metrics for the month until the following month data has been released (e.g. I report October median prices at the beginning of December).
This brings up another point about the data. All the numbers reported here are preliminary as some sales don't get recorded for a couple of months after the sale of the property. In practice it appears that most sales for the month are fully reported within three months of the sale, but I have seen new records appear far prior months occasionally as well. The idea is that the aggregate measures of price settle down long before the last sale for the month is reported. This seems to be the case from my observations to date. I'd still add the caveat that the data is preliminary and that recent months in particular are subject to revision as more sales are reported.
The Current Methodology for Calculating Median Prices
The rolling update described above covers all kinds of property sales beyond residential, so the first thing we have to do is filter out those property types we're not interested in (at least for this exercise). The city reports so many building class types that it's possible that I've overlooked or misinterpreted something, but I do think I've captured the vast majority of the residential sales. I've summarized the constituent property types that make up the three primary housing types I'm reporting on (houses, co-ops and condos) in the table below.
Housing Types from City Building Codes
|Housing Type||New York City Building Code|
In order to build an index of residential property prices, we also need to make sure we're not including multi-unit property sales when we calculate median prices. Obviously, throwing a 10 unit condo sale into the set will not accurately reflect what an individual would find searching for a singular apartment. Thus, I've limited cooperatives and condos to 1 or less units. Houses in New York are often sold with multiple dwellings, so I've put the residential unit cap for homes at 4 or less.
One question that comes up is why there are so many sales in residential buildings with a residential unit count of zero. I've decided to include these sales in the calculation for individual apartments since I have a feeling that they still represent individual apartment sales. If anyone has any more information on this let me know. I have an email in to the city on that subject as well.
The monthly process then is to filter the rolling property sales data for houses, cooperatives and condos and then filter out multi-unit sales using the definitions above. I then group all of these individual sales by geography as defined by the property records (either city, borough or neighborhood) and calculate aggregate median prices for each. Any suggestions for improvement or comments on things I've missed will be greatly appreciated. Again, it's a work in progress. I'm happy to improve the methodology and revise the data if it leads to a better index for residential apartment prices in the city.