The issue with AI in property prediction

Some data providers are claiming that data driven AI programs enable them to predict the future of property markets with great accuracy, but in my mind there are huge risks when it comes to using AI to make property predictions.

AI works by copying human intelligence using algorithms that analyse large amounts of data, identify data patterns and make decisions based on those patterns.

These tasks can now be accomplished much more quickly than ever with AI, but when we use data driven algorithms and filters to predict property prices, there are some huge issues.

Algorithms give everyone the same prediction

Any solution that relies purely on data to make property market predictions will always produce the same prediction from the same data inputs.

This means that every buyer’s agent or investor using such programs will get the same forecast for any suburb that they research.

This is a huge problem because it means that the AI program itself could cause the outcome it predicts.

Here’s the scenario - the data driven program predicts that a suburb or city is about to boom, and so buyers rush in to take advantage while the market is still warm and not yet hot. This growth in buyer demand then causes prices to soar.

Note: This could work out if the demand was coming from genuine owner occupiers seeking homes, but if the demand comes mainly from investors it can turn into a speculative bubble.

The consequences for investors can be disastrous when the bubble busts and prices crash.

Speculative booms often turn into bubbles and bust

Over a decade ago, property market booms in Queensland and Western Australia became bubbles when investors competed to buy, unsupported by a genuine demand for accommodation.

Rental vacancy rates rose, landlords were unable to secure tenants and as markets were flooded with properties for sale, prices crashed.

The same boom, bubble and bust sequence occurred in the Gold Coast unit market during the period 2002 to 2013, when large numbers of high density, high rise units (with one or two year rental guarantees built in) were sold off the plan to investors.

Speculative investor demand turned warm markets into hot ones, but as soon as the resulting rental supply exceeded the rental demand, rents fell, vacancy rates rose, and as desperate investors tried to sell, prices crashed, falling by 55% in six years.

Data driven predictions can only analyse current trends

Investors need to know where the booms, bubbles and busts of the future are likely to occur as a result of trend changes which have not yet emerged, but AI systems which rely on data points, no matter how many are involved, can only make predictions based on current trends.

This means that if enough investors are motivated to buy in so called warm markets or hot spots, then false booms can occur, leading to price crashes, busts and massive investor losses, none of which were anticipated by the AI generated solution.

About John Lindeman John Lindeman has well over a decade of experience researching the nature and dynamics of various types of assets at major data analysts and is a leading property market researcher, author and commentator. For more information visit Lindeman Reports.