Today's consumer is tech-savvy and, with unprecedented access to information, can be educated and discerning about even mundane transactions. For large transactions, like those involving real property, sellers and buyers are going to do their research, and generally that's a good thing. The problem is that their first stop will probably be an online real estate portal like Zillow, Redfin, Trulia, etc. and the information supplied by these third party providers can be largely inaccurate (which we will cover in the following sections). If sellers are thus misinformed before visiting a real estate agent the discrepancy in estimated value makes it difficult to trust the agent actually representing their interests.
The proprietary algorithm used by Zillow to generate home value assessments is an automated valuation model (AVM), a system that provides market analysis through mathematical modelling. Most AVMs calculate a property's probable value by comparing the values of similar properties at the same point in time. An industry-wide technology, this type of model is used mainly on properties not actually for sale (e.g. for portfolio valuation or second mortgage transaction) and is favored by many lending institutions. AVMs are touted for their speed and low cost but estimates generated this way have a significant margin of error. While this can be a useful overview its inaccuracy means it should not be relied upon as a final determinant of value.
To understand the limitations of AVMs let's go a bit further into how they work. AVMs typically look at factors like property location and characteristics, tax assessments, comparable sales, and price. This sounds pretty thorough, but without proper controls to validate and contextualize data it's easily misrepresented. An automated algorithm can't conduct inspections, so that property information doesn't account for actual, physical condition. If things like upgrades, construction quality, and neighborhood boundaries can't be assessed, it's questionable how accurately an AVM can establish which properties are, in fact, comparable.
Data sources can also be problematic. AVMs draw from public record, for which there exists no national standard to ensure quality. Information like square footage totals, taken from the county assessor's office, is often flat-out wrong. If prominence is given to MLS list price, where maximal pricing serves the interest of parties creating that data, the result will indicate bias more than actual market value. Additionally, any transactional data may lag up to six months, failing to reflect current market conditions. Biased or inaccurate data isn't the only way to impair performance—sometimes the necessary data isn't on record at all.
This last point is the real coup de grâce for AVM credibility. Texas is what we call a non-disclosure state, one of the several states where there is no law requiring that sales price be provided to third parties as part of a real estate transaction. There is also a legal provision that implicitly bars such disclosure outside of specific circumstances for the purpose of facilitating a sale. As sales price is considered information confidential to the transaction, to share this information would constitute a breach of responsibility to the client and put an agent at risk of being held liable. Where sales prices are not reported any model dependent on such record faces a serious shortfall in data. Bottom line: non-disclosure renders AVMs effectively useless.
So where can interested sellers find the information they need? We’ll get into this in Part II.