Appraisal Statistics: Regression - Do Your Really Know How Much Work It Is?

[Artemis Fowl]

Well, by "sensitivity analysis" I am not sure what you are talking about. As I understand it, it is a way to test a given model's output against variations in the input, for the purpose of prediction, not so much for the purpose of constructing the model in the first place. MARS can and is used for "sensitivity analysis". But I'm guessing you must mean something else. To reiterate, we are talking about constructing value models, i.e. equations, to predict home sale prices based on known inputs such as GLA, date of appraisal, lot size, bath count and so on.

The sales comparison grid IS a valuation model. Whatever the official definition(s): "Sensitivity Analysis", as the term is commonly used on this forum, tests and indicates the adjustment that produces the least variation between predicted sale prices within the sample. Most often used for $/square foot adjustment after all other adjustments are made. A tool integrated into an existing valuation model. No need to reiterate thanks tho.

 

[RCA]

The sales comparison grid IS a valuation model. Whatever the official definition(s): "Sensitivity Analysis", as the term is commonly used on this forum, tests and indicates the adjustment that produces the least variation between predicted sale prices within the sample. Most often used for $/square foot adjustment after all other adjustments are made. A tool integrated into an existing valuation model. No need to reiterate thanks tho.

I would have to disagree. The sales comparison grid is a stage beyond the valuation model, it is the application of a valuation model to a set of comparables and the subject, displaying only the adjustments produced for the differences between the subject and comparables. In fact, it allows for introducing adjustments that a particular valuation model is not capable of producing.

Again, I still don't know where you are at with the definition of "Sensitivity Analysis", in fact I tend to believe it is an incorrect use of the term - with a meaning altered by someone somewhere along the way.

 

[Joe Flacco]

My experience is that analysis of smaller sets of data that are as similar to the subject as possible yields the best results.

 

[Artemis Fowl]

I would have to disagree. The sales comparison grid is a stage beyond the valuation model, it is the application of a valuation model to a set of comparables and the subject, displaying only the adjustments produced for the differences between the subject and comparables. In fact, it allows for introducing adjustments that a particular valuation model is not capable of producing.

Again, I still don't know where you are at with the definition of "Sensitivity Analysis", in fact I tend to believe it is an incorrect use of the term - with a meaning altered by someone somewhere along the way.

Ok, I have to agree with your disagreement. For most, the grid serves as more of a worksheet for the valuation that is generally presentable when complete. I will generally grid many more sales than are presented.

And yes...the term is used improperly but has become common usage here. You'll see it often so that's what is meant by it.

 

[Joe Flacco]

Lets say for example you are trying to determine a adjustment for end unit vs interior unit townhouse. You may include all townhouses built in the decade in multiple subdivisions in the neighborhood. Most subdivisions have end units that have windows on the detached side but some subdivisions have end units that do not have any windows on the detached side. Now does it makes sense to include all of those townhouses weather they have windows or not? The market reaction for a end unit without any windows on the detached side is not going to be the same as a end unit with windows on the detached side. If you only analyze the townhouses in the subdivision then you get a more reliable result even if it is only five or six sales. In a subdivision where the end units do not have windows on the detached side, the market reaction is likely insignificant. In a subdivision where the end units have windows on the detached side, the market reaction may be significant. If you are including a larger data set, you probably end up with something in the middle and end up applying a adjustment where one is not necessary or end up applying a smaller adjustment than what the market's reaction is.

 

[Joe Flacco]

Personally, I'll never use regression or mass data techniques. The clients don't like it then they won't be ordering from me.

 

[RCA]

Lets say for example you are trying to determine a adjustment for end unit vs interior unit townhouse. You may include all townhouses built in the decade in multiple subdivisions in the neighborhood. Most subdivisions have end units that have windows on the detached side but some subdivisions have end units that do not have any windows on the detached side. Now does it makes sense to include all of those townhouses weather they have windows or not? The market reaction for a end unit without any windows on the detached side is not going to be the same as a end unit with windows on the detached side. If you only analyze the townhouses in the subdivision then you get a more reliable result even if it is only five or six sales. In a subdivision where the end units do not have windows on the detached side, the market reaction is likely insignificant. In a subdivision where the end units have windows on the detached side, the market reaction may be significant. If you are including a larger data set, you probably end up with something in the middle and end up applying a adjustment where one is not necessary or end up applying a smaller adjustment than what the market's reaction is.

I would try to get a window count in the input data. Is that possible? If not, what it appears you have in the data, among other fields, is: subdivision, interior vs end unit. So, MARS would create a model with these variables, and you will still be able to compare end and exterior units in one subdivision. The model would include something like this
BF3 = max(0, (subdivision in ('Williams', 'North Side') x (max(0, (EndUnit = 'True')) x $3000; That is, if the subdivision is Williams or North Side and it is an end unit, add $3000, where you have the complete model as: SalePrice = .... + BF3 + ....

 

[Joe Flacco]

I would try to get a window count in the input data. Is that possible? If not, what it appears you have in the data, among other fields, is: subdivision, interior vs end unit. So, MARS would create a model with these variables, and you will still be able to compare end and exterior units in one subdivision. The model would include something like this
BF3 = max(0, (subdivision in ('Williams', 'North Side') x (max(0, (EndUnit = 'True')) x $3000; That is, if the subdivision is Williams or North Side and it is an end unit, add $3000, where you have the complete model as: SalePrice = .... + BF3 + ....

I am just giving that as one of many many examples.

But why would I do that when I can look at the five or six sales and come out with a adjustment? I really don't understand this push for regression and statistics. The issue we have is that people never obtained the skill of developing adjustments. It is not that old school methods of developing adjustments do not work or are not reliable or credible.

 

[RCA]

I am just giving that as one of many many examples.

But why would I do that when I can look at the five or six sales and come out with a adjustment? I really don't understand this push for regression and statistics. The issue we have is that people never obtained the skill of developing adjustments. It is not that old school methods of developing adjustments do not work or are not reliable or credible.

IF the ONLY difference in the sales was Interior vs EndUnit, that is to say, all Interior units sold, market conditions adjusted, for X dollars and all end units sold for Y dollars, then you don't need to go any further. But what if there are other differences? You will need then to extend your data set and things start to get complicated, and then you would be advised to use a regression tool.

 

[Joe Flacco]

The importance of the adjustment is so tiny compared to picking the right comps and writing a good reconciliation to the sales comparison approach.