George Hatch
Elite Member
- Joined
- Jan 15, 2002
- Professional Status
- Certified General Appraiser
- State
- California
I would say that the process is a fair bit harder to verbalize than to demonstrate. You guys will see it in action once with a single dataset and understand how it works.
At least a couple of the formsware programs use similar methodology to suggest adjustment variables. They just program their tools to perform the trial-and-error iterations automatically instead of manually; and the ones I've seen pretty much stick to GLA and date of sale. Once you learn how to write your own you can use the same process for almost any variable: view amenities, paved vs unpaved access, pool vs no pool, guest unit vs no guest unit, etc.
You can also run much wider datasets than you would normally include in an appraisal report. Really, if you have access to more apples that are really apples it's better to do it that way so long as you're qualifying your data.
And I cannot stress enough the importance of data qualification. Whether it's this model or any other model - including the Fannie 3-comp grid - data qualification is the lynchpin of making them work. Raw data that you're not familiar with sux; that's is the primary reason most AVMs don't get any respect from appraisers.
The version I'm sending out right now is kind of down/dirty as opposed to the flexibility I wrote into the 1004MC worksheets. Mostly with respect to handling datasets of different sizes.
If I'm using a spreadsheet for comparables analyses I usually write these on the fly in my own work. It's one of those things where once you learn the syntax in Excel (or any other spreadsheet program) to write the equations they kinda write themselves. Teach a man to fish and all that. Inasmuch as what I'm sending out includes those equations and syntax you guys can just copy those and modify your models to suit.
As I say, this sort of sensitivity analysis is a fair bit simpler to learn, understand, perform and explain than the more sophisticated regression models that are more automated. It's kind of that interim step between the GSE 3-comp manual adjustment grid and the black box regression analyses tools that the quants seem to favor but nobody else understands.
I never really thought about it before but maybe there's some utility in developing this a little more and posting it as a sticky in the general forums for more people to see.
At least a couple of the formsware programs use similar methodology to suggest adjustment variables. They just program their tools to perform the trial-and-error iterations automatically instead of manually; and the ones I've seen pretty much stick to GLA and date of sale. Once you learn how to write your own you can use the same process for almost any variable: view amenities, paved vs unpaved access, pool vs no pool, guest unit vs no guest unit, etc.
You can also run much wider datasets than you would normally include in an appraisal report. Really, if you have access to more apples that are really apples it's better to do it that way so long as you're qualifying your data.
And I cannot stress enough the importance of data qualification. Whether it's this model or any other model - including the Fannie 3-comp grid - data qualification is the lynchpin of making them work. Raw data that you're not familiar with sux; that's is the primary reason most AVMs don't get any respect from appraisers.
The version I'm sending out right now is kind of down/dirty as opposed to the flexibility I wrote into the 1004MC worksheets. Mostly with respect to handling datasets of different sizes.
If I'm using a spreadsheet for comparables analyses I usually write these on the fly in my own work. It's one of those things where once you learn the syntax in Excel (or any other spreadsheet program) to write the equations they kinda write themselves. Teach a man to fish and all that. Inasmuch as what I'm sending out includes those equations and syntax you guys can just copy those and modify your models to suit.
As I say, this sort of sensitivity analysis is a fair bit simpler to learn, understand, perform and explain than the more sophisticated regression models that are more automated. It's kind of that interim step between the GSE 3-comp manual adjustment grid and the black box regression analyses tools that the quants seem to favor but nobody else understands.
I never really thought about it before but maybe there's some utility in developing this a little more and posting it as a sticky in the general forums for more people to see.
