Department of Numbers

Methodology Tweak

Posted Thursday, January 07 2010

I recalculated all prices metrics for all New York City geographies last night. So what did I change? First I added 5-6 family homes as well as Converted Dwelling or Rooming Houses to the definition of houses and raised the residential unit filter to 6 (on houses only). I think that makes more of a difference in Brooklyn than in Manhattan, but it's noticeable in all boroughs. One would expect this to raise median prices slightly as we include larger homes in the survey.

The more significant change (at least from an implementation perspective — I don't think it changed the results too much) was consolidating multiple sales records into one. Sometimes you see a property with the same borough, block, lot, address and sale date with multiple entries. I can't think of a good reason why that should represent reality better than a single entry. Luckily a lot of those additional properties had sale prices of zero so I was already filtering them out and they had no effect on the median calculation. Occasionally though you'd get weird ones that had multiple, non-zero sales prices. I'm now consolidating those in a very conservative way (the details would involve too much SQL, but I'm basically using the max price of those multiple records when the max price also equals the sum of all those individual prices). Considering these multiple entry sales were relatively rare, the net effect should be to raise median prices just slightly.

The last change was to increase the price filter to $1,000. It was $1 before. $1 gets rid of the vast majority of unbelievable prices, but you do occasionally see some unrealistic values between $1-$1,000. The impact here is tiny, but it will also raise median prices ever so slightly.

So all of these changes have a tendency to raise median prices but Manhattan's 4Q median didn't actually change. It's still $775,000 (like I calculated yesterday) when I use the new data. When you look at the rolling 12 month median sales prices for Brooklyn and Manhattan though it seems to have generally bumped up prices by about 1-2%. That makes sense to me and I'm glad the impact is small. Still, it's worth keeping this new methodology as I think it gets us marginally closer to reality.

Here's a summary of the process as it stands now.

Department of Numbers Housing Types

Housing Type Residential Units New York City Building Code
Houses <= 6 AO,A1,A2,A3,A4,A5,A6,A7,A9
B1,B2,B3,B9,C0,C2,C3,C5
Cooperatives <= 1 C6,C8,D0,D4
Condos <= 1 R1,R2,R3,R4,R6