As part of the recent ruling by Consumer Affairs Victoria (CAV) all Victorian estate agents and agents' representatives are obliged to list three comparable properties whenever listing or advertising a vendor’s property.
To support agent compliance, develop the Statement of Information, and allow for a seamless user-experience for both agent listing as well as buyer browsing realestateVIEW.com.au have developed a custom-built 3COMPS platform.
You can explore the legislation, as well as the 3Comps tool, here, but to further clarify the inner workings of the product, the matching technology and data integration realestateVIEW’s Data Product Owner, Kent Lardner, has formulated his thoughts, below:
Firstly, to clarify, there are both similarities and differences between how an algorithm works and how an agent works when selecting the best 3 matching comparable sales and it is because of these differences that agents are integral to the process of matching comparable sales. Let’s start with the algorithm and how comparable sales are typically selected in an automated valuation model (AVM).
How the AVM works
Consistent with the approach of many statistical models, price is generally the ‘dependant variable’ or the thing we are trying to predict. As such, we don't use sale price as one of the ‘independent variables' within the model. We would instead use variables such as land size, bedroom count and bathroom count in a simple model to predict a price, while only sampling sales within a given location and within a defined time-period to build the model.
What are the similarities between how an algorithm and an agent select comparables? Often referred to as an ‘emulation model’, the matching algorithm seeks to emulate the practices used by a professional, such as a valuer or real-estate agent. The match can range from 0 for something totally unsuitable to 100% for something perfectly matched to the subject property. For example, if for a given location the agent always selects comparable sales within 1km radius and on the same side of the suburb, then the model would seek to reproduce this pattern.
Matching variables that are regularly used in Australia include: property type (house or unit), time since sold (typically 12 months), distance (within 1 or 2 km and within the same suburb as an ideal match), street type (main road properties compared to normal suburban roads), land size, bedroom count and bathroom count. The richer the data set, the more variables we can use in the model.
Thinking of the matching algorithm as a scorecard, the ideal matched property would be the same property type, would have sold in the last few months, be within a few metres of the subject property and located on the same street, and have the exact same dimensions as the subject. However, the reality is that few comparables are ever the perfect match, which is further compounded by the differences the model can’t reliably measure.
The matching algorithm will perform reasonably well in homogenous markets, that is, markets that are well-populated, relatively predictable and with plenty of recent sales. However, it will perform poorly where the subject property is unique, where the market is heterogenous (not consistent with the norm) or where few recent sales have been recorded. One of the main limitations of an automatic model is that it rarely reproduces what the human eye can detect in terms of build quality, finishes and aspects such as a view.
A real-estate agent selecting comparable sales has considerable advantages over any algorithm, and as such is vitally important as the final filter comparable sales must pass through before they can be considered compliant with new underquoting laws. Using local market knowledge and an in-built super-computer of a brain, data not accessible to the algorithm can be processed and used instantly by the agent. For example, photos can be instantly interpreted and used to match one property to another in terms of build quality, construction age, finish or views.
Some general guidelines have recently been drafted by the Victorian Government detailing expectations for comparable sales selection. In addition to these guidelines, here are a few extra tips and methods that can help, especially where few comparable sales are available.
- Selecting your suburb:
From a compliance perspective, distance will be the primary measure, so for optimal results, it’s always best to select comparables nearby or at least in a similar profile sub-market (such as property type) within a suburb.
- Street classification:
Each street/road is now classified nationally as a freeway, highway, main road or local road etc. This means that for automated checking, the data does exist that can automatically flag poorly matched comparable sales. As such, if you are listing a property on a main road, selecting comparable sales that are not located on main roads will result in a poor match (and can inflate the price).
- Construction/build quality:
This is often reflected in the sale price and obvious to most agents. If an agent has very few available sales to use and the only comparable sales are inferior (or superior) then they could be outside of the expected price range. If all the comparable sales are superior, the list price will be below the supporting evidence (all other things being equal). In this case, the agent should make a note of all the comparable sales they have reviewed and justify the final selection. While all 3 comparable sales being the ‘same’, in terms of qualitative measures, is the ideal, it is not always achieved. One common technique used by valuers and agents in such situations is to select one better, one worse and one same.
- Land size:
For larger lot sizes, it is harder to find matching comparable sales within a defined and narrow percentage tolerance. For example, metro lots under 1000m2 are much easier to match within a 25% range, whereas blocks above 1500m2 are harder to match. For larger lot sizes, matching based on a m2 rate can often be the best approach.
- Bedrooms and bathrooms:
It is very common to see bedrooms used as the primary matching criteria, with bathrooms overlooked. While matching properties with the same bedroom and same bathroom count is ideal, if few well-matched comparable sales are available then often a larger or smaller sized property is needed. In statistical tests, it is common to find the value of an added bedroom is less than the value of an added bathroom. Therefore, based purely on the variation in value, matching based on bathrooms is every bit as important as matching based on bedrooms.
If few comparable sales are available, adding one larger property and one smaller property can often help where no direct matches are available.
- Date range:
Where 6 months is stipulated as a date range it is unlikely that any market movement would significantly change the price. However, for those regions where up to 18 months is acceptable, market movement can often significantly change the price. In such cases, supporting evidence can be added that highlights the market movement through index measures. Medians as an index are generally less reliable than other standard index measurements, so should be used with caution.
For further information on the product, or to organise a live-demo for your agency, please get in touch with realestateVIEW.com.au through the 3Comps website.