Increase time series sample size for home sales data

[tiernanmartin]

The sample should include sales that occurred between 2016 to 2018.

However, given that the sale rate will be calculated over a rolling 3-month period, the sample period should actually begin at October 1, 2015 and end on Dec 31, 2018.

[tiernanmartin]

This will likely require the creation of a new field: SALE_DATE.

[tiernanmartin]

Is a 3-month sample period too short for a census tract size housing market? Perhaps 6 months or a year would provide a more robust sample size?

Counter point: there are example of sales rate per quarter at census tract scale (PolicyMap's home sale data)

[tiernanmartin]

I have decided to keep simplify things a bit:

• count, percent, median indicators: 2016-2018
• change indicators: 2013-2015 to 2016-2018

This will require adding 2013 and 2014 parcel metadata

[tiernanmartin]

Unfortunately, the .gdb archives for 2013 and 2014 do not contain any of the metadata extracts needed.

I have contacted Christie Most and Mindy Tiangson at the KC Assessor to request access to these historical records.

In the meantime, I will just use the 2015 data as the threshold of comparison.

[tiernanmartin]

After connecting with Christie it seems that the Assessor's office does not systematically archive historical "extract" data (i.e., the metadata related to parcels and buildings).

It may not be possible to get a 100% accurate denominator for the SALE_* indicators that include 2013 and 2014.