MLS statistics

[Mejappz]

How would adding other independent variables to a model that models time and price make that model a better model?

Adding more independent variables to a model that only considers time and price can definitely make it more accurate. Price isn’t just influenced by time—things like location, property features, market conditions, and interest rates all play a role. Condition and quality are huge factors. A well-maintained home with high-quality finishes is going to sell for more than a similar one in poor condition, even if they were built at the same time. If the model only looks at time, it might wrongly attribute price changes to time alone when other factors are driving the difference. By including more relevant variables, you get a clearer picture of what’s really affecting price trends, making the model more reliable and useful. That said, being selective is important—throwing in too many variables can add noise instead of value.

 

[alebrewer]

Adding more independent variables to a model that only considers time and price can definitely make it more accurate. Price isn’t just influenced by time—things like location, property features, market conditions, and interest rates all play a role. Condition and quality are huge factors. A well-maintained home with high-quality finishes is going to sell for more than a similar one in poor condition, even if they were built at the same time. If the model only looks at time, it might wrongly attribute price changes to time alone when other factors are driving the difference. By including more relevant variables, you get a clearer picture of what’s really affecting price trends, making the model more reliable and useful. That said, being selective is important—throwing in too many variables can add noise instead of value.

Think about it this way: if you've already accounted for those differences by ONLY modeling the sales considered comparable alternatives to the subject, then you've addressed - to a large extent - those 'quality and condition' factors. Instead of saying that single variate regressions aren't meaningful, I'd say they're not meaningful unless you're regressing truly comparable data to the subject. Another option is to solve for the other variables and then model the adjusted prices, although its been my experience that doesn't really change the model too much. So long as I'm modeling truly comparable properties, regressing price relative to date works pretty well.

 

[Mejappz]

Think about it this way: if you've already accounted for those differences by ONLY modeling the sales considered comparable alternatives to the subject, then you've addressed - to a large extent - those 'quality and condition' factors. Instead of saying that single variate regressions aren't meaningful, I'd say they're not meaningful unless you're regressing truly comparable data to the subject. Another option is to solve for the other variables and then model the adjusted prices, although its been my experience that doesn't really change the model too much. So long as I'm modeling truly comparable properties, regressing price relative to date works pretty well.

I see what you're saying, but in my markets—where there's a broad range of prices—relying solely on a single-variable regression of price against time is pretty much useless. Even if you’re only selecting comparable sales, there’s still a lot of variability in factors like condition, quality, and lot appeal that can skew the results. In a market with a tight, homogenous price range, I can see how a simple regression might work well, but when price points vary significantly, those other factors start to matter a lot more.

Sure, you can try adjusting for them first and then running the regression, but in my experience, that still doesn’t fully capture the complexity of the market. A broader price range usually means more variation in property characteristics, and a single-variable regression just doesn’t cut it when those differences materially impact price. So while I get the logic of keeping it simple when dealing with truly comparable data, in a diverse market, leaving out key variables makes the model way less reliable.

 

[alebrewer]

I see what you're saying, but in my markets—where there's a broad range of prices—relying solely on a single-variable regression of price against time is pretty much useless. Even if you’re only selecting comparable sales, there’s still a lot of variability in factors like condition, quality, and lot appeal that can skew the results. In a market with a tight, homogenous price range, I can see how a simple regression might work well, but when price points vary significantly, those other factors start to matter a lot more.

Sure, you can try adjusting for them first and then running the regression, but in my experience, that still doesn’t fully capture the complexity of the market. A broader price range usually means more variation in property characteristics, and a single-variable regression just doesn’t cut it when those differences materially impact price. So while I get the logic of keeping it simple when dealing with truly comparable data, in a diverse market, leaving out key variables makes the model way less reliable.

meh - everybody's got to do what they think is right. If multi-variate regression is what does it for you - you'd be remiss to do anything but.

 

[Dublin ohio]

Just ran this one for a rural property I am seeing tomorrow. It is a macro since I only had 4 similar sales in the past year. Did it quarterly since I only had 72 sale within the last year and 61 sales the year before. I use an Excel type program and can change show the same trend lines as Terrell mentions as well as change the type of graph. Can also show the trend calculations and "r" if I choose. What a mess. All of the low points are periods of low sales volume