Full original article can be found hereAppraisers and real estate agents often ask what adjustments I use and/or how I support my adjustments.  The answer is that most properties require a different adjustment that is specific to its market (e.g. size, location, condition, etc.) and there are many different ways to support any individual adjustment.  No one method for supporting adjustments is perfect.  Appraisers should select the method or methods that will produce credible results for the given assignment and available data.

1. Paired Sales – Paired sales are a cornerstone of textbook appraisals, but textbook cases of paired sales rarely occur in practice. In a common textbook scenario, paired sales are two sales that are the same in every way except the one factor for which the appraiser is trying to estimate an adjustment. For this reason, it is easy for appraisers to forget that a paired sale can have other differences (although it is important that the differences are minimal and that adjustments for the differences can be supported). In this assignment, my grid included four sales that had very little difference from one another except for GLA. After adjusting for a couple of minor factors, the paired sales all suggested an adjustment of $51 and $60 per square foot for GLA.
2. Simple Linear Regression – I’ve blogged in the past about supporting adjustments, particularly GLA, using simple linear regression. Linear regression is basically analyzing trends in data.  For this assignment, simple linear regression suggests $53 per square foot when comparing sales price to GLA. Significant variation exists among the data of this sample, but the datum points are spread evenly along the entire regression line suggesting that the indicator is not being skewed by a small subset of outliers. It is okay if the properties in the sample have differences, however it is important to make sure to filter out differences that would skew toward one end of the range or the other. For example, if a larger site size also tends to include a larger home, then it would be important to make sure that the homes in the sample all have similar site sizes or the adjustment could be falsely overstated. Also, it is helpful to the outcome of the regression analysis that the subject property is in similar condition to the majority of the sales in the sample. The following chart shows the linear regression outcome in this appraisal.
3. Grouped Data Analysis – This method is closely related to simple linear regression and is essentially many paired sales representing a fast way to estimate an adjustment simply by sorting comparable sales. This can be done using quick searches on the local multiple listing service or using data exported to a spreadsheet. But remember that the same factors that can skew linear regression will also skew grouped data analysis. For best results, it is important to sort out all of the features that might distort the results without sorting to the point where the sample sizes are small and wildly varied. For this assignment, I filtered out all ranch sales in the past two years with a lot size of 7,000 to 9,999 square feet, that feature two baths and three bedrooms, and that were built within ten years of the subject. Sales of homes meeting these criteria between 1,000 and 1,199 square feet have an average of 1,128 square feet and an average sale price of $212,637. Sales of homes meeting these criteria between 1,200 square feet and 1,299 square feet have an average of 1,253 square feet and an average sale price of $220,055. The difference between the average of these two sets is $7,418 and 125 square feet or $59 per square foot. The median could also be compared as well to provide another indicator that is less likely to be skewed by outliers.
4. Depreciated Cost – The cost approach value in this assignment is consistent with values suggested by recent comparable sales. This suggests that the cost approach is likely valid and could be used as a way to test reasonableness or support adjustments. The subject’s original cost is estimated at $108 per square foot and the depreciated cost is estimated at $81 per square foot. A simple depreciated cost adjustment might not be a good adjustment to apply to comparable sales. This is because the depreciated cost is a straight-line measure from zero square feet all the way to the total area including the kitchen, bath, mechanical, and everything else in the house. For this adjustment, we are just looking for the value difference from a similar-sized comparable to the subject. To obtain this adjustment using the cost approach, I ran a cost estimate for the smallest comparable sale and another cost estimate for the largest comparable sale with no physical changes for anything other than living area (e.g. room count, garage, quality, and all other factors kept equal). The original cost difference between the low and the high came out to $79.53 per square foot. If this number is depreciated based on the cost approach in the appraisal, a reasonable adjustment of $60 per square foot of GLA is estimated.
5. Income Approach – The income approach was not performed for this appraisal assignment, but if it had been, the income approach could have been used to support another indicator for the GLA adjustment. One way the income approach could be used to support a GLA adjustment is by taking the estimated loss or gain in rent from an additional square foot of living area (can be estimated using any of the above approaches except for cost) and apply a Gross Rent Multiplier (GRM). Critical to this approach is that the multiplier and rent estimates are market derived and that rent might be a consideration for the typical buyer.
6. Sensitivity Analysis – This method is closely related to paired sales and I think it works best for secondary or tertiary support for an adjustment or helping to reconcile what adjustment is most effective. However, this method is not very useful if adjustments for other comparable sale differences are not accurate. Once all of the comparable sales have been placed side-by-side in a comparison grid and adjusted for all other factors using market derived adjustments, the appraiser can test different GLA adjustments to see what adjustment produces the tightest range of adjusted value indicators. If the appraiser is unsure by simply looking at the data, the Coefficient of Variation (CV) can be applied to each set of adjusted indicators to mathematically test what adjustment is producing the tightest range. The lower the CV, the better the adjustment is working within this sample of sales. Here is a link to a free CV calculator. Just enter your adjusted indicators separated by commas and press calculate. Then test another adjustment and repeat with the calculator. An appraiser could also set up a formula using the Worksheet function in a la mode Total to instantly provide the Coefficient of Variation. For this appraisal, sensitivity analysis helped me reconcile that the simple linear regression adjustment is most well-supported adjustment because it has the lowest CV as seen in the following table.

None of the above methods for supporting an adjustment are without limitations and there are many more ways an appraiser could support an adjustment.  Although this is an example where data sets are particularly plentiful, the example shows that information does exist outside of textbooks for supporting adjustments; and when multiple approaches are combined and reconciled, a strong case for the appraiser’s conclusion can be made.  An appraiser won’t always need to go this far to support one adjustment, but if that one adjustment is crucial to the outcome of the appraisal or the appraiser believes they will be challenged on this adjustment, then the appraiser should expand and explore multiple methods for support.

By Gary F. Kristensen, SRA, IFA, AGA

Full original article can be found here