How Data is Changing the Real Estate Game

Robert Ciemniak2019-03-21

In this article, MioTech speaks with Real Estate Foresight to find out how their startup is making the unpredictable nature of China's property market, predictable with data analysis.

Please tell me about Real Estate Foresight and how you're trying to disrupt the market with data.

At Real Estate Foresight, as an independent research boutique, we focus on helping investors and developers predict and assess the China housing markets, focusing on data-driven predictive analytics, with the models and frameworks we have built over the past 7 years. From the macro level to the project specific due diligence. We have also written software to automate research production and search through local sources of information, that now sits with our sister venture Robotic Online Intelligence. Our angle on disruption? A practical, data-driven foresight in a market seen as problematic in terms of data availability, quality and consistency.


What was the gap you noticed in the market?

The real estate sector is traditionally well known to be highly heterogeneous in nature. Assets are so different, data is sketchy and hard to compare like-for-like. High emphasis on personal relationships and the gut feel, combined with the complexity of the deals or changes in local policy, especially in China, all make real estate less likely to be predictable with data analytics. However, with Real Estate Foresight, we believe there is a real information edge that can be gained with the data-driven approach.


Domestically - not so much a gap as an evolution of the sector. High land prices, greater competition, more discerning buyers-upgraders, tighter credit from banks, and targeted local policy all make it harder to make profitable investment decisions. Risk assessment becomes more important and so greater sensitivity to data and analytics that can reduce the uncertainty.


For foreign investors, unless they are close to China, it can often be quite confusing or outright misleading in terms of what they get to see in the selective stories and some sensationalised analysis. Given the scale of China, aggregate statistics are notoriously hard to interpret and a proper understanding of the drivers underneath, be it regions, city-tiers or types of cities in the case of the housing market, is (or should be) vital.


Where do you get your data?


We have partnered with China Index Academy in Beijing, part of Fang (SouFun), as a major source of extensive underlying data at project and city level that we then feed into our analytics. There are also tons of freely available public data online, ranging from the National Bureau of Statistics to local listings, news and media coverage. Most cities have extensive data available online and publish a range of yearbooks with rich data. The issue there is how to make sense of that data, interpret it correctly, navigate through ambiguous definitions – that’s where one needs to spend time.


How is technology enabling greater insights into the China housing market?


There are three parts to it. First, analytics help discover and interpret the (repetitive) patterns in the data and hence help understand where we are in the cycle, and predict what’s likely next in terms of price growth and sales volume growth overall, or at a city level, which cities are likely to outperform – we like to say that we look at cities as if they were stocks and then ask how would one trade them.


Second, the technology for searching through local sources helps discover important information about the development of new areas or districts, launches of new metro lines and important infrastructure projects, upcoming land auctions and listings, major transactions by developers, or local market sentiment, buyer preferences and conditions including actual mortgage rates offered. This can be valuable in sourcing deals and when conducting due diligence with limited hard data.


Third, more internally, technology can help automate the mundane aspects of research production. The nitty gritty of e.g. getting those city reports with charts, data and text nicely produced. You still need a human analyst but now what that analyst used to spend 10 hours on, and what was just manual processing, can be done by the software in 10 minutes. Speed matters not because real estate is a fast pace trading game but because in practice, when the opportunities arise, people working on deals need the data and insights ready …for yesterday.


Could you cite an interesting case studies?

Picking a winning city

The top city-pick we presented at our annual China Property Outlook conference in January 2017 was Xi’an. We made the case that the cut in land supply, improving inventory clearing, new momentum in sales volumes and prior under-performance, combined with very strong economic fundamentals and favourable policy would make Xi’an an outperformer relative to other cities over the next years.


At that point in time, we presented the following charts, with some internal models in the background:


Subsequently, here’s how Xi’an rose to the top performer rank:

A screenshot of a social media postDescription automatically generatedA screenshot of a cell phoneDescription automatically generated

Here’s how Xi’an relative performance has fared since:

A screenshot of a cell phoneDescription automatically generated

Forecasting the market cycle

A couple of years ago, we were able to incredibly accurately forecast the annual change in the year-on-year house price growth… on a monthly basis a year in advance:



A data secret? Seeing very clear patterns in how M2 money supply growth had been leading the house price growth by approximately 6 months, and how sales volume growth tended to lead house price growth. Back then the housing policy was more uniform, so the tighten-ease-tighten-ease cycle wasn’t as differentiated by city as it is now (that’s a big structural change in the market) and such predictions were possible.


Picking the Winning Districts

We also took it a step further, from districts to cities. For a few years now, we have  been publishing the annual China Districts reports where we look at 200 or so districts across around 20 major cities to identify the next hot spots, based on … the data. The data can point at an interesting location, but then, of course, we would still verify the market dynamics through local sources and primary checks. One example of data indicators that proved highly predictive in terms of relative price growth is a group of districts with strong volume metrics but with relatively weak price growth performance are prime candidates for significant upside in the subsequent year – here’s our picks from 2017 and what happened to them (as expected except one) in 2018:


A screenshot of a cell phoneDescription automatically generatedA close up of a white backgroundDescription automatically generated


How are property developers responding to predictive analytics?


More and more of the major Chinese developers are building up internal big data teams, hiring groups of PhDs to analyse data, especially with the volume of internal proprietary data they have. There is more and more of alternative data, such as from use of mobile phones and apps to track people movements. In general, though, I still think many players may not even have fully structured central research/data departments the way we would define these, based on our conversations with people in the industry.


One challenge with ‘predictive analytics’ for China housing is in the over-promising name… While we have many good examples of actual predictions, the role of such analytics is really to better illuminate what’s going on, to identify patterns, to get that extra bit of information that can make a positive difference to a decision or the overall strategy. The core decisions are still deal-specific and here clearly analytics matter less.


As we know, the China property market is highly policy driven, how do you incorporate this unquantified factor into your data analysis?

What’s fascinating is that these policies often are a response to market conditions expressed in the data… and that in fact would make the policy moves quite predictable with the data, at least in terms of the direction! This works especially well on the overheated markets – these quickly get under policy scrutiny.  In the other direction, it is very ‘easy to ease’ the policy in the current cycle, given that we’ve had a long period of extensive tightening since 2016.

The big change in the current cycle compared to a few years ago is that the policy directive now is towards more city- or even district-specific policies to reflect the particular demand-supply conditions in local markets, as opposed to a blanket tightening or easing. The implication should be a greater divergence of performance and also a greater importance on looking at individual cities rather than the typical clusters such as by region or by tier. It is now also less meaningful to talk about ‘overall policy’, other than the general curbing of speculation, the “housing is for living, not speculation’ and control of the price escalation.


What are some other challenges you face in the analysis of the real estate sector?

As for the data challenges in general, there is no lack of these! Consider the basic question, are house prices rising or accelerating or declining or decelerating? The average of 70 cities new home price indices from the official NBS (National Bureau of Statistics) would draw you into one conclusion, quite different to what you would see in the 100 Cities Index from China Index Academy, and both very different from what you might see in many media headlines! How do you square these up? Actually, there is a perfectly good explanation once you dissect the definitions, scope differences, calculation methods and so on. But if you only look at some select headline figure, you might get it wrong.


Country Garden, one of the largest property developers in China announced to increase investment in AI. How does AI disrupt the traditional property construction from your point of view?


AI can work where there is a lot of data, and the property operations generate lots of information, especially for large developers. Consider the scale of operations if you have around 100 billion US dollars in annual contract sales and hundreds of projects. So many areas for optimization. It’s not our focus, though, as we concentrate on the market dynamics and investment decision support.


What is in store for real estate foresight?


With the blueprint for the Guangdong – Hong Kong – Macau Greater Bay Area (GBA) finally released by the government on February 18, we are putting a special effort into the detailed coverage of the  9 mainland cities in the region, Together with Robotic Online Intelligence, we launched GBA Intelligence where our robo-analysts scan and pick up most actionable or insightful local news and announcements.


We will soon be publishing in depth reports with the focus on discovery of potential opportunities in the GBA for developers and investors.


And we simply continue solving the data puzzle in China - building up more predictive analytics with the same mission - to turn the data into actionable insights, looking at data-driven deals.

Share this to
Share this to