Comments on: No Do-Overs in Pricing

By: Sven

Sven — Thu, 25 Sep 2008 21:43:46 +0000

Kris, I think the real problem with the graphs is just the small sample sizes. Looking at a single zip code and then narrowing in on a single month and you have a huge range of about 8 to 33 (eyeballing your chart) that jumps all over the place. I could calculate the correlation from that chart, but I dont think it would be a good one. Could you redo the charts looking year-over-year instead of month-over-month? Granted, you would only have 4 data items if you went from September 2004-September 2008, but I think the chart would show a clear trend then. Statistically, you need a certain sample size based on how far the deviations are between samples. When you had 30 homes sell in a month, you quite probably had a price range from .25 mil to 2.5mil. You would need a much larger sample size to really make claim from that.

By: Kris Berg

Kris Berg — Mon, 22 Sep 2008 20:29:53 +0000

J – Just don’t over-improve for the neighborhood.

By: Jakob

Jakob — Mon, 22 Sep 2008 20:09:41 +0000

Nice graphs. Scripps Ranch still seems to be hangin’ tough.

Interesting that $/s.f. doesn’t drop more with house size. So if I have a small house in Scripps I could expect to get double my money back with an addition, assuming ~$100/s.f. construction cost. At first blush, it sounds like a no brainer!

By: Kris Berg

Kris Berg — Mon, 22 Sep 2008 17:00:07 +0000

Sven,

Thank you for the lesson in statistics, but I beg to differ. I am not looking for correlation as much as I am looking for trends. Look at what happened in the fall of 2006 versus the fall of 2007, and in the spring/summer of 2006 versus 2007. Our 2008 market is not better than 2007, I believe you will agree, so I can make some conclusions (predictions) about what is coming over the next several months.

You are right that the last chart is nothing new. Just a refresher for those with bigger homes (and the accompanying smaller buyer pool).

By: Sven

Sven — Mon, 22 Sep 2008 16:22:24 +0000

A note I wanted to make on charts and graphing. There’s something you can calculate in statistics called a “correlation”. This can be calculated with a complex formula, or you can simply just look at a chart and make a fair guess how ” strong” the correlation is. The theory is that with one criteria, if the outcome shows a strong “correlation”, the chart is more useful. The more randomly it moves, either your sample base is too small and/or the correlation is smaller.

For an example, the charts you supplied, all of them show no statistical correlation except for the last one. This shows that the median price per square foot goes down the bigger the house is. (which makes sense because you are dwarfing the value of the land by increasing livable square footage)

If you are looking for a chart with a really strong correlation on price trends over time, pull up the case-shiller index that actually tracks homes being resold over time.