What's an AVM? No, I'm not talking about an arteriovenous malformation. In this case, it's an Automated Valuation Model. A computer calculates the market value of a home. What could be simpler?
In another lifetime I did a lot of social science statistical analysis. I used a well-known computer program, SPSS, to do the heavy mathematical work. Originally, I used the main frame version. Then I got my own copy when we all got our PCs. Eventually, as I got deep into real estate, SPSS was too seldom used, occupied too much hard drive space and was too expensive to update, so I deleted it, A sad day.
As I got more experience in real estate, I began to miss SPSS. I thought it might be a good idea to do a multi-variate analysis of home prices. If I could build a statistical model that included all the important factors in a home's market value, each weighted properly, why then, I would just have to plug in the information and, behold, a statistically based price would emerge!
I didn't have SPSS at my side any more. Getting the program back would cost a not insignificant sum, so I gave it considerable thought. What are the factors that go into the price of a home? First of all, I realized I would not only have to do it town by town, but area by area. Every town has parts that are more expensive than others, so neighborhood had to be a factor. But figuring out what constituted a neighborhood took a lot of experience and expertise. In some towns, one side of a street literally is valued differently from the other side of the street. So there was going to have to be a lot of human analysis before we got around to the data entry itself.
A family room, for instance, has a value. But that value is different depending on what floor it's on. Some years ago, a family bought a home with a first floor family room. They decided they didn't need it on the first floor, so they set up a family room in the basement and redid the first floor to make better use of the space they didn't need for the family room. But a year later, due to a job change, they had to sell the house. Although the overall market had improved, their house sold for less than they paid for it because by moving the family room and redoing the first floor, they had devalued the house.
A garage has value – even though sometimes it seems most folks use them for storage rather than as shelter for the family car(s). Even so, whether it is a one car or two car garage, or even a three car garage certainly makes a difference. But is that value affected by other features of the house? Is a two car garage with a three bedroom home worth less than a two car garage with a five bedroom home? What role does balance and expectations play when the market places a value on a home.
A lot of questions like those came up. The issue of the effect on value of interacting features seemed to be common.
In any case, even if we kept it as simple and straightforward as possible, the ever-changing market came into play. Values change. How quickly are they changing? The question of what comparable properties are for sale right now certainly affects the marketing strategy, but it also affects the market's perception of the value of a given home.
Incidentally, because of the relatively small number of cases (homes) involved in a given neighborhood, any result would lack statistical significance. I could live with that because, after all, I was used to Social Science analysis where that's a common issue. The question then is, are the cases a sample or are they the whole universe?
Alright, fine! It was going to take a lot of time and effort to build the model. Far worse, I could see I would be spending more time on updating the model than I would on making use of it. So I abandoned the idea and didn't invest in a new edition of SPSS.
Now, lo these many years later, the AVM has made its appearance. It began, evidently, with Zillow and the Zestimate. Here was a site the consumer could visit, and it told you the market value of a home. The Zestimate was key to Zillow's popularity. Few users read the warnings on the site about the possible inaccuracies. They also didn't pay attention to the fact that the big Zestimate number was just part of a rather wide range of (smaller print) numbers.
The Zestimate's popularity meant that other web sites felt pressured to provide AVM estimates of their own. They tried to improve on the Zestimate by being more accurate. RPR (Realtor's Property Resource) came up with a couple of AVMs, available only to Realtors.
There's a home on the market as I write this, with a list price of $850,000. The Zestimate is $719,038, taken from a range of $646,000 to $755,000. The Zestimate is just over half way along the range.
RPR's estimate for the same property is $786,100, over $60,000 higher than the Zestimate, and about $30,000 higher than the high end of the Zestimate range, but considerably lower than the list price. RPR's range is $723,212 to $848,988. RPR says the value has risen by over $20,000 in the past month!
So here we have three notions of the current market value of the home, covering a spread of over $200,000 on a house that will probably sell for less than $900,000. That's a margin of error of about 22%. Strikes me as kind of large.
The list price is set by the seller, who depends considerably on the advice and recommendation of the listing agent. We tend to overvalue things that we already possess, while we undervalue the things we want to possess. So sellers often have an unrealistically high notion of what their home is worth in the fair market, while buyers in that same market tend to undervalue the same home. This gives rise to strategic marketing considerations in setting the list price. It also creates issues for the listing agent who doesn't want to challenge the seller too defiantly on the price, but does want to price it at a level where the property will sell, preferably for the maximum the market will pay. So the accuracy of the list price may depend on the stubbornness of the seller as much as on the pricing skills of the agent. To determine which it is requires some research on the listing agent.
As I've often said, the list price should not be an announcement of how much money the seller wants. It should be a key element in a marketing strategy to get the money the seller wants. So when an offer is being contemplated, don't just look at the list price, or the various AVMs available. The buyers' agent needs to calculate the real fair market value of a home, at the record of the listing agent, at the motivation level of the sellers (not always easy to determine), and the motivation level of the agent's own buyers, at the competition in the market from other homes and other buyers, and try to understand the psychology of all involved.
AVMs don't know the people, and don't really know either the subject property or the comparable properties. The computer can make a guess, based on limited knowledge and understanding. It's a good guess, but it's too broad and inaccurate to be of real use.
Click or call your agent. Try to get a good one.