Friday, January 10, 2014

Statistical Analysis of Property: Cow Hollow

I’ve always been curious to actually get into the nitty-gritty of the SF real estate data.  We spend a lot of time regurgitating opinions, but for once I actually went to the raw data myself: I got a mathematician from The University of South Carolina involved to conduct a study on property information, covering both the Cow Hollow and Marina Districts.  This is an original piece, so I get to speak with a lot of authority here.

Below are some of the discoveries made on Cow Hollow (click on image to enlarge):



  • The proximity to the median price - the number exactly in the middle - captures over two-thirds of the values of Flats, Single-Family Homes, and Condominiums in Cow Hollow.  Paying attention to the median as a price proxy is, therefore, most important.  It’s also a little better than talking in averages because as you see in the charts above… the data skews nicely.  No surprises there.
  • A predictor model was developed and it was discovered that square feet is a leading predictor of price!  We kind of knew this all along - but now we have the evidence - that's encouraging.
  • After making sure that bedroom and other inputs do not skew the relationship of total square feet to price - we tested for normality.


And that’s exactly what we got: the data we worked with is reliable!  In English: we can trust the results.

Here comes the coolest part of all: we figured out how to model an expected price based on key inputs:
 
Using the above chart as an example of outcomes: what we are able to do is determine relationships between key factors of a property - number of bedrooms, bathrooms, square footage - which are dubbed “residuals” in fancy terms - and how that tentatively translates into price.

Everything is about comparisons, and this is precisely how we maximize the ability to compare and spit out a more reliable listing indicator.  At the end of the day, most of this is behind-the-scenes work, but for once I wanted to put it out there so you could see at least some of the methodology.  Based on the responses to this I would love to get more specific with other sub-districts.  Feel free to email me at dino@dinozuzic.com