How are you handling "support your adjustments" revision requests?

[Scott54]

Getting more of these lately where the reviewer wants paired sales or regression support for basically every line item on the grid. Time/date, fine. But when they want me to prove my GLA rate with a paired sales analysis on a report I'm getting paid $500 for, it feels like the math doesn't work.

For those of you getting these regularly, are you building up a library of paired sales you can pull from? Or are you doing fresh analysis every time and just eating the hours?

Would not library pairs require time and location adjustments? Better to use adjusted pairs from your grid. Fast.

 

[djd09]

waivers are like the mortgage fraud days...nothing will make sense

 

[Terrel L. Shields]

Better to use adjusted pairs from your grid.

In rural property, you would likely have one pair saying $89 a SF and another saying $240/SF. Sensitivity is the way to go and it's quick. Calculate, screen shot, paste.

 

[AppraisingLA123]

Getting more of these lately where the reviewer wants paired sales or regression support for basically every line item on the grid. Time/date, fine. But when they want me to prove my GLA rate with a paired sales analysis on a report I'm getting paid $500 for, it feels like the math doesn't work.

For those of you getting these regularly, are you building up a library of paired sales you can pull from? Or are you doing fresh analysis every time and just eating the hours?

I do new analysis everytime. But i seldom get those request. I am pretty sure you are getting them becaues your adjutsed values vary too much, the adjusted values should be very alligned.

 

[Tom D]

CU can say that some of your adjustments differ from the group thought. Then it's up to the underwriter, or whomever, to let it go or request whatever. It also has to do with your CU score. With a 2.5 or lower score, they have safe harbor. They have no need to bug you. It depends on the sensitivity of the lender. CU never give you a clue as to what it sees, only dings you for not being in the group thought.

And your peer's group thinking isn't always right, but CU only looks at it that way. It's like you didn't do it right, the right way is in invisible ink. You try to guess.

 

[Fritz in NY]

Math is math, if I'm buying Fish at the market, how much per pound, color, age, size, BUT, Current, local, purchaser knowledge beats your graphs all day long. I am hired for my knowledge, not my ability to graph what happened. That is what AI dose. I won't invest in AI Real Estate Securities unless bundled to offset % of loss.

 

[sputnam]

USPAP does not require that you show your work. You have to summarize. Some Clients may insist that you show them in more detail. What you do with a Client that wants details is up to you. It's a business decision. Personally, if it wasn't in the Letter of Engagement that I agreed to when I accepted the assignment, I'd tell them to pound sand.

 

[Digger88]

For the easy ones I seemingly only do here I made this years ago and have never had an issue (once in a blue moon a noob lender wants pretty little graphs). For anything more than a cookie cutter I've done a thousand times or anything private I use Synapse which is ok. I add in regression and paste in the results

zzz, zzz AND zzz ADJUSTMENTS WERE EXTRACTED FROM MARKET DATA. THE ANALYSIS UTILIZED BRACKETING AND SENSITIVITY ANALYSIS. THE ADJUSTMENTS MADE, MOST MINIMIZED THE DIFFERENCES BETWEEN THE COMPARABLES WHEN COMPARED TO THE SUBJECT. MARKET-BASED ADJUSTMENTS ARE MADE TO COMPARABLE SALES IN ORDER TO ESTIMATE THE VALUE OF A PROPERTY. BRACKETING AND SENSITIVITY ANALYSIS ARE TWO METHODS USED TO MAKE THESE ADJUSTMENTS.

BRACKETING INVOLVES SELECTING COMPARABLE SALES THAT BRACKET THE SUBJECT PROPERTY IN TERMS OF SIZE, AGE, LOCATION, AND OTHER RELEVANT FACTORS. THIS HELPS TO ENSURE THAT THE ADJUSTMENTS MADE ARE BASED ON MARKET DATA THAT IS AS SIMILAR AS POSSIBLE TO THE SUBJECT PROPERTY. SENSITIVITY ANALYSIS INVOLVES ANALYZING THE IMPACT OF DIFFERENT ADJUSTMENTS ON THE FINAL VALUE ESTIMATE. THIS HELPS TO ENSURE THAT THE ADJUSTMENTS MADE ARE REASONABLE AND REFLECT THE MARKET’S REACTION TO THE DIFFERENCES BETWEEN THE SUBJECT PROPERTY AND THE COMPARABLE SALES.

THE ADJUSTMENTS MADE TO THE COMPARABLE SALES ARE INTENDED TO MINIMIZE THE DIFFERENCES BETWEEN THE COMPARABLES AND THE SUBJECT PROPERTY. THIS IS DONE BY ANALYZING THE MARKET FOR COMPETITIVE PROPERTIES AND PROVIDING APPROPRIATE MARKET-BASED ADJUSTMENTS WITHOUT REGARD TO ARBITRARY LIMITS ON THE SIZE OF THE ADJUSTMENT.

Then I add the market data page for definitions from Synapse. Yes it's minimally complaint, no I don't care lol

Sales Comparison Adjustment Methods

Allocation
For the allocation method, a certain percentage of the sale price of a property is allocated to each feature. The potential
adjustment is based on that percentage allocated for a particular feature.

Depreciated Cost
This method determines a potential adjustment by subtracting depreciation from the cost to build an improvement with the
result being the value (adjustment) for the feature being measured. The difference between cost and value is depreciation
so if the cost to build an improvement and the depreciation can be determined with relative accuracy then the result is the
potential adjustment for that feature.

Grouped Data
This method involves grouping the data (sales) into two categories based on the feature being measured. The average or
median price of the first group is compared to the average or median price of the second. The difference in those two prices
is the potential adjustment for the feature being measured.

Paired Sales (True)
A method of comparing two properties that are considered to be the same in all features except for one. In theory, the
difference in the sales price of each property is an approximation of the value difference (or adjustment) for the one feature
in which the properties differ. For this analysis, all properties that were analyzed are compared against each other to find all
“pairs” and then the average and median of the results of all of those pairs is found.

Paired Sales (Adjusted)
This is the same as True Paired Sales except that if, in analyzing two properties, they differ in more than one feature (True
Paired Sales requires that only one feature is different) and the appraiser is confident they can adjust for those differing
features such that the result is only one differing feature, then this would be an “Adjusted Pair". Adjusted Pairs will nearly
always have more data points since it allows for more than one differing feature (non-perfect matches).

Sensitivity
This method is based on the theory that the best adjustment is the one that results in the smallest range of adjusted sales
prices for all sales analyzed. It “plugs in" an adjustment and calculates what the sales price would be if that were the
adjustment and it does that for every sale. Then it determines the range (difference between the low and high) of the
adjusted sales prices. It repeats that process to test every possible adjustment. The adjustment that leads to the smallest
range of adjusted prices is the final result.

Survey
In this method, market participants (e.g. appraisers, brokers, real estate agents, etc) are contacted in order to determine
what they believe to be what a typical buyer and seller would agree to as far as the added value for a particular feature
(swimming pool, barn, new roof, addition, etc). Typically the average and/or median of those results is the potential
adjustment based on the survey method.

Ordinary Least Squares Regression
Among the most common of all types of simple regression, this method minimizes the sum of the squares of the differences
between a variable and it's predicted value (called the residual). One of the results of this regression method is the slope of
a line that can be drawn through the data points. That slope is the potential adjustment based on this method.

Theil-Sen Regression
This simple regression method finds the slope of every possible line that can be drawn between every pair of data points if
they were plotted on a chart. It then takes the median of all of the slopes of those lines and that is the potential adjustment
based on this method. Since this method utilizes the median, it does reduce the impact of outliers on the data.

Least Absolute Deviation
This simple regression method determines every line that can be drawn between each pair of data points. For each of those
lines, the distance of the remaining data points to the line is calculated using the absolute value. All of those distances are
then added up and the slope of the particular line that results in the smallest sum of absolute values for the residuals
(deviation) is the potential adjustment result based on this method.

Least Median of Squares
Another form of simple regression that is very similar to Ordinary Least Squares Regression except that instead of taking the
average of the squares of the residuals, this method utilizes the median of the squares of the residuals. As a result this
method tends to be a bit more robust to outliers than Ordinary Least Squares Regression.

Robust Simple Regression
If any of the above Simple Regression methods has the word "Robust" in front of it that means that during the calculations,
when the average of all of the data points is subtracted from the data point in question, instead the median of all data points
is subtracted from the data point in question. This tends to make a particular regression method more "robust" to outliers
(meaning less impacted by outliers).

Modified Quantile Regression
This is a modified type of Robust Least Squares Regression where, instead of subtracting the median (the 50th percentile)
from each data point, 9 different percentiles are tested (from 10% up to 90%) and the result from the one that has the best
(highest) r-squared is the final result. This means that regression is calculated nine times (one time for each percentile
tested) but only the results from the one with the best r-squared score is utilized.

 

[Digger88]

That's interesting, the one page summary for the workfile is basically what I'm after. Something I can point to when a reviewer pushes back instead of scrambling to put together support after the fact.

Does Synapse pull from your own past reports or is it running fresh analysis off MLS data each time?

If you need graphs like OSU said Synapse does an ok job with these type results which take about 10 minutes for your report

 

[Alsie35]

For the easy ones I seemingly only do here I made this years ago and have never had an issue (once in a blue moon a noob lender wants pretty little graphs). For anything more than a cookie cutter I've done a thousand times or anything private I use Synapse which is ok. I add in regression and paste in the results

zzz, zzz AND zzz ADJUSTMENTS WERE EXTRACTED FROM MARKET DATA. THE ANALYSIS UTILIZED BRACKETING AND SENSITIVITY ANALYSIS. THE ADJUSTMENTS MADE, MOST MINIMIZED THE DIFFERENCES BETWEEN THE COMPARABLES WHEN COMPARED TO THE SUBJECT. MARKET-BASED ADJUSTMENTS ARE MADE TO COMPARABLE SALES IN ORDER TO ESTIMATE THE VALUE OF A PROPERTY. BRACKETING AND SENSITIVITY ANALYSIS ARE TWO METHODS USED TO MAKE THESE ADJUSTMENTS.

BRACKETING INVOLVES SELECTING COMPARABLE SALES THAT BRACKET THE SUBJECT PROPERTY IN TERMS OF SIZE, AGE, LOCATION, AND OTHER RELEVANT FACTORS. THIS HELPS TO ENSURE THAT THE ADJUSTMENTS MADE ARE BASED ON MARKET DATA THAT IS AS SIMILAR AS POSSIBLE TO THE SUBJECT PROPERTY. SENSITIVITY ANALYSIS INVOLVES ANALYZING THE IMPACT OF DIFFERENT ADJUSTMENTS ON THE FINAL VALUE ESTIMATE. THIS HELPS TO ENSURE THAT THE ADJUSTMENTS MADE ARE REASONABLE AND REFLECT THE MARKET’S REACTION TO THE DIFFERENCES BETWEEN THE SUBJECT PROPERTY AND THE COMPARABLE SALES.

THE ADJUSTMENTS MADE TO THE COMPARABLE SALES ARE INTENDED TO MINIMIZE THE DIFFERENCES BETWEEN THE COMPARABLES AND THE SUBJECT PROPERTY. THIS IS DONE BY ANALYZING THE MARKET FOR COMPETITIVE PROPERTIES AND PROVIDING APPROPRIATE MARKET-BASED ADJUSTMENTS WITHOUT REGARD TO ARBITRARY LIMITS ON THE SIZE OF THE ADJUSTMENT.

Then I add the market data page for definitions from Synapse. Yes it's minimally complaint, no I don't care lol

Sales Comparison Adjustment Methods

Allocation
For the allocation method, a certain percentage of the sale price of a property is allocated to each feature. The potential
adjustment is based on that percentage allocated for a particular feature.

Depreciated Cost
This method determines a potential adjustment by subtracting depreciation from the cost to build an improvement with the
result being the value (adjustment) for the feature being measured. The difference between cost and value is depreciation
so if the cost to build an improvement and the depreciation can be determined with relative accuracy then the result is the
potential adjustment for that feature.

Grouped Data
This method involves grouping the data (sales) into two categories based on the feature being measured. The average or
median price of the first group is compared to the average or median price of the second. The difference in those two prices
is the potential adjustment for the feature being measured.

Paired Sales (True)
A method of comparing two properties that are considered to be the same in all features except for one. In theory, the
difference in the sales price of each property is an approximation of the value difference (or adjustment) for the one feature
in which the properties differ. For this analysis, all properties that were analyzed are compared against each other to find all
“pairs” and then the average and median of the results of all of those pairs is found.

Paired Sales (Adjusted)
This is the same as True Paired Sales except that if, in analyzing two properties, they differ in more than one feature (True
Paired Sales requires that only one feature is different) and the appraiser is confident they can adjust for those differing
features such that the result is only one differing feature, then this would be an “Adjusted Pair". Adjusted Pairs will nearly
always have more data points since it allows for more than one differing feature (non-perfect matches).

Sensitivity
This method is based on the theory that the best adjustment is the one that results in the smallest range of adjusted sales
prices for all sales analyzed. It “plugs in" an adjustment and calculates what the sales price would be if that were the
adjustment and it does that for every sale. Then it determines the range (difference between the low and high) of the
adjusted sales prices. It repeats that process to test every possible adjustment. The adjustment that leads to the smallest
range of adjusted prices is the final result.

Survey
In this method, market participants (e.g. appraisers, brokers, real estate agents, etc) are contacted in order to determine
what they believe to be what a typical buyer and seller would agree to as far as the added value for a particular feature
(swimming pool, barn, new roof, addition, etc). Typically the average and/or median of those results is the potential
adjustment based on the survey method.

Ordinary Least Squares Regression
Among the most common of all types of simple regression, this method minimizes the sum of the squares of the differences
between a variable and it's predicted value (called the residual). One of the results of this regression method is the slope of
a line that can be drawn through the data points. That slope is the potential adjustment based on this method.

Theil-Sen Regression
This simple regression method finds the slope of every possible line that can be drawn between every pair of data points if
they were plotted on a chart. It then takes the median of all of the slopes of those lines and that is the potential adjustment
based on this method. Since this method utilizes the median, it does reduce the impact of outliers on the data.

Least Absolute Deviation
This simple regression method determines every line that can be drawn between each pair of data points. For each of those
lines, the distance of the remaining data points to the line is calculated using the absolute value. All of those distances are
then added up and the slope of the particular line that results in the smallest sum of absolute values for the residuals
(deviation) is the potential adjustment result based on this method.

Least Median of Squares
Another form of simple regression that is very similar to Ordinary Least Squares Regression except that instead of taking the
average of the squares of the residuals, this method utilizes the median of the squares of the residuals. As a result this
method tends to be a bit more robust to outliers than Ordinary Least Squares Regression.

Robust Simple Regression
If any of the above Simple Regression methods has the word "Robust" in front of it that means that during the calculations,
when the average of all of the data points is subtracted from the data point in question, instead the median of all data points
is subtracted from the data point in question. This tends to make a particular regression method more "robust" to outliers
(meaning less impacted by outliers).

Modified Quantile Regression
This is a modified type of Robust Least Squares Regression where, instead of subtracting the median (the 50th percentile)
from each data point, 9 different percentiles are tested (from 10% up to 90%) and the result from the one that has the best
(highest) r-squared is the final result. This means that regression is calculated nine times (one time for each percentile
tested) but only the results from the one with the best r-squared score is utilized.

AI much? Seriously, AI is actually super useful in generating all of this kind of gobbledygook, after you simply give it the numbers that your analysis indicates and tell it the method you used to derive them.

Saves a heck of a lot of time, and sounds impressive to boot.

The important thing of course is that the data that you're giving it to work with is accurate and is from your files and indicated by your analysis.

I've maybe had to do that two or three times during the last year. Knock on wood, I don't get these types of requests except for once in a blue moon, probably because I only do simple properties but there's a lot of data, and hopefully, because I know what I'm doing.