Digitalisation and the real estate industry
Digitalisation and artificial intelligence are intensely discussed in the real estate industry. Talk of disruption is not hype, says Marcelo Cajias of PATRIZIA. “It’s about a fundamental revolution in the way we do business in real estate.”
Interest from institutional players is high. This is a breakthrough in an industry usually slow to adopt new technologies, Cajias, the Head of Data Intelligence, writes in Journal of Property Investment & Finance(14 July 2020).
One driver is that digitalisation and AI promise to increase efficiencies of existing processes and generate value from data. Another is that protechs and fintechs are increasingly conquering parts of the business. They are introducing an arsenal of new technologies, including AI and machine learning, that can change the game. But in adapting instruments from other industries to real estate, they face a steep learning curve.
“As the property cycle progresses – and especially with the ‘Black Swan’ of COVID-19 – there will be consolidation of the new players,” says Cajias. “This is natural, but it also mirrors the earthquake confronting the architecture of existing market players.”
A fundamental revolution
So how is the revolution impacting business? For institutional brokers, understanding properties within markets in the machine-learning era becomes more challenging. For one, clients can easily assess a location online and challenge a broker’s assumptions. But also, a machine-based assessment works with a single artificial target: find the best location and rank the entire market.
“This example illustrates the seismic shakeup around disintermediation, which will impact the cosy industrial organisation of existing players,” says Cajias.
For real estate investment managers, smart meters tracking a building’s energy- and water-consumption, air quality and security could alert property managers to anomalies. Intelligent cameras are already optimising retail sales based on day-to-day footfall estimations, with further innovations underway. And AI will help validate investment decisions. For example, it will enable better rental forecasts prior to signing lease contracts or re-letting portfolio assets.
All this means the mantra ‘location, location, location’ is evolving to ‘data, data, data’. But data only becomes a real value creator if market players know how to analyse it. First, they must say goodbye to two myths: that the more data, the more accurate the insights; and that migrating data to a cloud will deliver all the insights a business needs almost immediately.
More data does not equal better insights
The market now is filled with companies selling real estate data derived from web scraping technologies, social media, pictures and sensors. “Including too many (collinear) variables in a model can inflate the explanatory power of the model without reflecting about the underlying data generation process,” says Cajias.
"The process of creating value out of data is a long journey."
Marcelo Cajias, Head of Data Intelligence at PATRIZIA
Nonetheless, including all this information in a hedonic regression to explain the development of rents in a city has one benefit: it could reveal other factors that are better predictors of rents than the size, location and age of dwellings.
“Machine-learning methods are a powerful tool to focus on an accurate prediction,” says Cajias. “However, they challenge the econometrician to consider statistical-causality and econometric-ethical principles more intensely than ever.”
As to the second myth, there are no short cuts in generating insights. It is a long journey, explains Cajias. Obtaining the technological infrastructure is a relatively easy first step, with large providers offering online solutions that can be accessed within minutes.
The long journey of generating insights
So the real journey starts with the next step: defining data pipelines for collecting, cleaning and organising data. While this sounds simple, it is the biggest drain on human resources and is where many market players currently struggle. The third step is analysing the data. This involves finding “unknown life” in data patterns that intuition says are there, but that have not yet been validated.
But the biggest challenge comes in the last step: the deployment of models.
“The deployment of AI involves trusting computational methods that capture, explain and forecast relationships based on an ‘artificial’ assumption. In other words, there is no assumption with regard to the shape of the relationship, but an algorithm consisting of hundreds of rules.”
All this begs the questions:If AI methods are more accurate, does this mean the econometric understanding of markets we constructed decades ago was wrongly specified? “The short answer is – more often than not – yes,” says Cajias.
However, one must differentiate the overall purpose of an econometric model. The traditional regression approach explicitly chooses the shape of the relationship between rents and covariates.
Two models; two approaches
AI learning methods aim to maximise the explanatory power regardless of the priorities of the econometrician, and with no reference to market experience or real estate theory.
Depending on the investment case, letting the machine find the solution automatically can be a powerful tool, but other cases need traditional methods. AI provides property managers with a powerful model for estimating rents.
“This approach is called forecasting and is used to merely predict and forecast the response of (mostly new) assets,” says Cajias.
However, if the aim of identifying value drivers of rents is to derive an action plan for an investment strategy, then the focus lies elsewhere: on estimating the single price elasticities on every explanatory variable with the highest accuracy possible. This approach is called inference.
"The econometric world we constructed decades ago has an expiry date."
Very soon, evolving technology will enable a marriage of both models. Called Explainable Artificial Intelligence (XAI), it is set to give the econometric world constructed decades ago an expiry date, says Cajias.
To navigate this changing world, real estate players must “jump into” a new analytical system. They must understand functionalities and, through trial and error, apply methods to extract insights. “Until now, leaving the comfort zone and spending money in building up the use case for data intelligence has been accomplished by few companies,” says Cajias. “Yet doing so might be a strong business catalyst after the earthquake.”
This text is based on an article by Marcelo Cajias, Head of Data Intelligence at PATRIZIA’s Department of Investment Strategy and Research in Augsburg Germany, which appeared in the Journal of Property Investment & Finance, (14 July 2020).