April 8, 2024

Défi i3C: When a predictive data model helps real estate investment

For the second edition of Défi i3C, Ivanhoé Cambridge recognized the work of three students from HEC Montréal and Polytechnique Montréal whose predictive data science model helped identify the asset classes that are most resilient to the rise of hybrid work and, conversely, those most affected by this new work habit.

The spread of remote work has profoundly transformed the world of work and lifestyles. This structural trend, now firmly established in companies, has also had a major impact on the property market, with a fall in office space take-up, higher vacancy rates, migration to the suburbs and a move away from city centres. Against this backdrop, how does hybrid work influence the dynamics of cities of tomorrow? What are the strategic implications for real estate investors and several types of property? These were the questions posed to seven university teams in the second edition of Défi i3C, organized by Ivanhoé Cambridge.

Technology, and more specifically artificial intelligence, is transforming the world and opening up new opportunities for those involved in real estate investment. To encourage research and development in this field, we have launched this academic competition in 2022,” explains Sarah Mjidou, our Senior Director, Digital Transformation, Information Technology, and sponsor of the event. The aim of this initiative is to create links between the academic world and the business world, so that innovative solutions can emerge to meet the needs of the property market.”

The 2023 edition of this intra-university competition crowned the work of two HEC Montréal students – Yacine Ndiaye and Joel Crispin – and one Polytechnique Montréal student – Mady Semega – who developed a predictive data model capable of navigating the new reality of the labor market. “Our mandate was to assess the impact of hybrid work on the real estate market in order to make recommendations to Ivanhoé Cambridge on acquisition or disposition,” recalls Mady Semega. We then developed a predictive model for several major US cities based on various criteria to determine the best cities and the best types of properties to invest in or dispose of over the next two years.

Mady Semega, Yacine Ndiaye and Joel Chrispin during the presentation of their predictive data model

A fall in the value of offices

Specifically, their data model was fed with a sample of data covering 915 different cities, 22 macroeconomic variables and nearly 20 explanatory variables. Initially, Ivanhoé Cambridge gave us access to the Green Street database, which contains a lot of data on commercial properties,” explains Mady Semega. Depending on the city and the type of property, we have three levels of data that we have aggregated to build models by city and by type of property“. At the same time, “we used macroeconomic data, such as GDP (gross domestic product) and the working-age population, as well as specific data on the world of labor and the evolution of hybrid work“, adds Joel Crispin. “These data allowed us to include the impact of remote work on employees in the finance and consulting sectors – two of the sectors most heavily affected by hybrid working – and thus to see the impact on the performance of real estate assets,” adds Joel Crispin.

The model was also trained using historical data to ensure the best results. The three students were thus able to highlight four key metrics to support their recommendations: asset value momentum, property occupancy rate, city desirability rank and, finally, the percentage of remote jobs. Using their model, the three students were able to assess a total of 7 asset classes: 3 types of commercial property (offices, industrial/logistics and shopping centres) and 4 types of residential property (multi-family houses, single-family houses, townhouses and condos). Thus, “in the residential sector, we found that the suburbs will see a sharp rise in property prices“, adds Yacine Ndiaye.

Investment decision support tool

Their model has also made it possible to propose a series of investment recommendations for the period 2024-2025 for four major asset classes: residential, offices, shopping centres and industrial/logistics. “The tool is capable of analyzing a huge amount of data and, as a result, ranking several cities and several types of property,” explains Mady Semega. It can therefore be used to make investment or disinvestment decisions based on pre-defined criteria. It’s a real decision-making tool”. Indeed, “we have identified certain cities where it makes sense to invest in certain asset classes, such as Las Vegas“, points out Yacine Ndiaye.

For example, the model recommends investing in industrial properties, particularly in cities where the capitalization rate are above normal expectations. Against the market, it also recommends exploring investment in shopping centres in certain cities where occupancy rates and net income growth are particularly high. Finally, their tool recommends investing in single-family houses in San José, a city with the best prospects for value appreciation. On the other hand, their model recommends keeping a close eye on office buildings in New York and Chicago for possible disposal.

Mady Semega, Yacine Ndiaye, Joel Chrispin and their mentor Amel Khobzi

We also suggest