Regression is a mathematical tool used by real estate appraisers to determine the likely value, or adjustment rates, of various property characteristics and ultimately predict sale prices. Rather than relying on human opinion, listings, or previous appraisals, regression analyzes actual sales data to determine adjustment rates, assigning statistical value to characteristics such as GLA square footage, number of garages, acreage, age, and so on.
Regression modeling requires computer software but relies on human logic and commonsense. Here’s how to build a regression model in eight steps.
Our Unlimited Learning Membership includes professional development tools like videos, checklists, and white papers. Learn more.
Step 1: Acquire regression-modeling software
Microsoft Excel is a useful processing tool. If you choose to use Excel, perform an internet search for “Load the analysis tool pack in Excel.” Install the software.
Step 2: Acquire your sales data
You may export sales data from your local MLS or from other sources. Unlike paired-sales analysis, regression modeling works best with a large quantity of data. Focus on obtaining a large pool of sales, ignoring estimates or active listings.
Step 3: Cleanse your data
Review and cleanse your data to ensure it is reliable and usable.
Step 4: Select usable data
Select the data you’re going to use.
Step 5: Filter your data
Filter your data, beginning with basic variables. Select for straightforward traits such as square footage and site size.
Step 6: Narrow down your data even more
Continue narrowing down your data, filtering it by increasingly complex variables. Select for more complicated data, such as recent renovations or the number of bathrooms, to improve your model’s specificity and accuracy.
Step 7: Select variables to account for market forces
To do this, add variables like Days Past Sales or whether a property was sold HUD or REO. Experiment with selecting other variables, such as a certain builder or subdivision.
Step 8: Conduct a back test
Use your regression model to predict the sales price of a home that has already been sold. Then compare your predicted sales price with the actual sales price. If the actual sales price matches the sales price your regression model arrived at, you know you can use your regression model to predict the prices of other homes with a good degree of accuracy.
