Why to use Statistical Modelling in Commercial Real Estate
Statistical modelling and commercial real estate are two areas which seem to have not too much in common until you consider the fact that activities in the commercial real estate sector are exposed to risk and uncertainty, even more so as real estate is naturally a mid- to longterm business with accelerating impact on risk exposures and insecurities in future (market) developments.
Therefore, real estate investors need some tools to cover this uncertainty while being exposed to mid- to longrun equity exposures.
Here, we have a look why it is essential to incorporate statistical modelling into the workflow of commercial real estate operations.
Statistical Modelling vs. Machine Learning
So, why even to bother with statistics when there are much fancier machine learning/ neural network applications available?
Well, it depends what is the task at hand and which kind of material is available!
I guess, anybody prefers to (sometimes) riding a Porsche or Ferrari. But when the task is plowing a cornfield, a Porsche might not be the best choice in that case. The same is true for statistical models and machine learning/ neural network applications.
I found a comprehensive summary (references below) which describes the difference between those two worlds:
- Statistical Model is the use of statistics to build a representation of the data and then conduct analysis to infer any relationships between variables or discover insights. It is a method of mathematically approximating the world. A statistical model will have e.g. sampling, probability spaces, assumptions and diagnostics to make inferences.
- Machine Learning is the use of mathematical and/ or statistical models to obtain a general understanding of the data in order to be able to make classifications and predictions.
One crucial element of distinction is the kind of data material available on which both of the approaches can work with.
Statistical models get already a grip on small data sets in order to distill a suitable approximation of what is going on.
Machine Learning/ Neural Network Applications need a huge amount of data in order to be able to learn from the data, to identify the right patterns in the data and to make sense of this learning effort in terms of, e.g. classification, forecasting, synthetic image generation and similar.
Commercial Real Estate Applications
Talking about digital technical twins in the commercial real estate business, this has enormous potential within the framework of machine learning.
Sensors implemented in all parts of a building deliver millions and millions of data points within a short period of time incomprehensible for a human brain to make sense out of it. Though, this is the finest our for machine learning/ neural network methods which learn on the data flow and distill hidden patterns in the data in no time.
Let’s get to the other end of the spectrum.
Investment management in commercial real estate is at least in part backed by respective market data. Talking market intelligence, two problems unfold here.
- On the one hand, the flow of commercial market data (i.e. its timeline) is by nature very slow. Due to this low frequency of data, machine learning applications are useless.
- On the other hand, there are a lot of market parameters to be taken into account. Combining the information of all this input parameters including their empiric timelines is again incomprehensible for the human mind.
Hence, one of the most important tasks in commercial real estate seems not be overwhelmingly backed by information.
So, what to do? Just checking out the current stage of a market without incorporating empiric information when deciding on investment/ de-investment decisions? Have a (subjective) guess on future market movements and bet on it? All this is more or less of speculative nature while money, i.e. investor’s capital, has to be committed in the mid- to longrun.
One can smell the high level of uncertainty and those unguided risk exposures given that framework!
But there is a way out. Statistical models are a tool to get a grip on the judgement of future market developments. Statistics offers a lot of options to help to reduce risk exposures and to give investors the opportunity to take appropriate market positions in due time before market developments catch them off-guard.
So, implementing statistical models, e.g. in the form of predictive analytics, in the workflow of real estate investment management is the natural next step. It is a crucial support system for decision makers running commercial real estate investments.
A Word of Caution
I think, the benefits of statistical models incorporated in the investment management of commercial real estate investors are profound. Actively managing risk exposures in the context of future market developments, or getting an opinion on the impact of potential disastrous events on one’s own portfolio -to name just two- give an investor a decisive competitive advantage in the market.
It is a great example why to use statistical modelling in commercial real estate. Though, there are no free lunches.
Even within the broad field of statistical models, one has to filter those tools which are suitable for the respective task and appropriate for the given data.
To give an idea of some of the challenges in statistical modelling, here are three examples:
- State-of-the-art ARIMA-GARCH models need at least 800–1,000 data points in a time line. This works excellent with share price quotations but in general is out of scope for real estate market data.
- The incorporation of dependency structures among market parameters could fail when the ratio market parameters/ observations is too stretched. In this case, the structure might fail or deliver biased results.
- Trends in time lines not treated properly while developing the statistical model could lead to biased and therefore wrong results.
So, ask your expert but do not fail in making use out of it!
References
Statistical Modelling vs. Machine Learning by Asel Mendis, KD Nuggets/ August 14, 2019
