Concordia researcher uses AI to predict the future cost of rent
According to projections from a Concordia research partnership, renting a two-bedroom apartment in Montreal in the year 2032 will cost $4,325 per month. In Toronto, it will be $5,600, and in Vancouver, $7,750.
It is well reported that there is a staggering dearth of housing supply in Canada. But Erkan Yönder, associate professor of finance and real estate at the John Molson School of Business, wanted to know exactly how excessive demand could shape the market in the coming years.
His research paper, “AI-Driven Insights into Key Factors Contributing to Rental Growth Across Canada,” leveraged artificial intelligence to create projections that are transparent and free of bias. The system was trained on existing data from the Canada Mortgage and Housing Corporation, Statistics Canada and immigration and population projections from the Government of Canada.
“The idea of this project is to look at the market at a neighbourhood level,” Yönder explains. “By zooming in on 427 census subdivisions, we can see what drives rental prices for very specific demographic groups.”
The model takes into account many factors, but Yönder notes that Canada’s immigration and housing policies are especially out of sync. In 2023, there were more than 1.2 million new Canadians, but only around 200,000 new housing units. “Every year we bring new people to Canada and give a shock to the housing demand. We need to be able to provide an equal shock to the supply to stabilize the cost of rent.”
Currently, annual housing completions are between one and three per cent across Canada, so new dwellings would need to be built at four to 10 times the current rate.
This project was commissioned by the Equiton Research Fund in Real Estate at John Molson. The goal of the partnership between Equiton, a private equity firm, and Yönder is to support innovative research into Canada’s real estate investment landscape.
“There is crucial knowledge and expertise in academia,” explains Aaron Pittman, senior vice-president and head of Canadian institutional investments at Equiton. “We asked Professor Yönder to focus on the critical issues facing our industry, and he drew on his and his team’s academic bent to produce immensely valuable new research in detail.”
AI removes emotion, adds precision
The specific type of AI technology used in this study is a neural network, which generates complex predictions based on a multitude of factors.
There are other linear projections of Canada’s rental housing market, but Yönder notes that using a neural network increases the accuracy of predictions by 30 per cent.
“We’ve not seen this level of detail,” Pittman says. “Real estate in Canada is very emotionally charged. Dr. Yonder’s work stripped away the human bias and emotion. He was able to project well forward while preserving statistical integrity through his modelling.
“This extracted data on rental prices is not political. The numbers are the numbers, and we’ve hit a super saturation point.”
Everyone’s problem
“This is a problem for everyone who lives in Canada,” Yönder explains.
Both he and Pittman agree that these research findings should motivate politicians, the private sector and citizens to dramatically rethink the future of housing across the country.
They say there is urgent work to be done by governments of all levels, investors, pension funds, developers and citizens alike. This work includes getting more housing construction projects completed with less bureaucracy, a healthier workforce and greater community acceptance.
“This is fundamentally a local issue,” Yönder says. “Our research brings the granular data required to understand what needs to be done at the neighbourhood level, which is the first step in making a plan that works for all current and future residents.”
Read the cited research paper: “AI-Driven Insights into Key Factors Influencing Canada’s Rental Market.”
