When market forecasting goes wrong, we all suffer

Canada Mortgage and Housing Corp. (CMHC) recently revised its housing market forecast to acknowledge that housing sales and prices are more likely to increase in 2021 than decline, as it had previously projected.

Many wonder why CMHC stayed with its bleak May 2020 forecast for so long, especially when housing markets and the economy started to march to a different tune as early as the following month.

But CMHC was not alone in projecting a pessimistic outlook. The big banks also saw red splattered all over the housing markets.

For example, Canadian Imperial Bank of Commerce forecasted housing prices would drop by five to 10 per cent in 2020 from the previous year. Royal Bank of Canada foresaw a decline of seven per cent from the market highs to the projected lows. And the Bank of Montreal expected a five per cent decline in the quality-adjusted price index produced by the Canadian Real Estate Association (CREA).

Even Moody's Analytics and RPS Real Property Solutions Inc. predicted a seven per cent decline in housing prices for 2021. Moody's gloomy outlook, released in September 2020, was in stark contrast to the housing market resurgence observed in places such as the Greater Toronto Area, where unprecedented monthly sales of more than 10,000 sales were reported in July and August 2020, with the average housing price rising to $951,516 in August from $820,226 in April.

Many believe forecasters follow a herding behaviour where they seek comfort by staying close to or agreeing with forecasts released earlier. If reputed or influential agencies are the first to generate projections, subsequent forecasts by others will likely toe the same line.

Brett Trueman, a professor at the University of California, Berkeley, noted in The Review of Financial Studies that analysts often exhibit herding behaviour “whereby they release forecasts similar to those previously announced by other analysts, even when their information does not justify this.”

Could the same have happened in Canada, with other forecasters deciding to follow CMHC's lead?

In December 2020, CREA was one of the first to break rank by forecasting growth in housing prices and sales. It now predicts housing sales in 2021 to grow by 7.2 per cent and prices by 9.1 per cent.

A question that intrigues many Canadians is why CMHC remained glued to its forecast made in May 2020 for so long. Bob Dugan, CMHC's chief economist and a veteran housing market expert, said that forecast was based on the bleak market conditions observed in April, when millions of workers had been laid off, businesses were shut and economic uncertainty was at its peak.

Back then, almost nine million of the 17 million workers in the labour force had applied for the Canada Emergency Response Benefit (CERB) to replace lost income. Nearly 16 per cent of mortgages were facing deferrals, and household debt levels were already extremely high. CMHC's economic model and assumptions about the future meant it forecasted a drop in housing prices of between nine and 18 per cent.

CMHC did not revise its forecasts when the economy headed in a different direction, because, Dugan said, it does not have the resources to generate more than one housing market forecast per year. In fairness, generating such forecasts is informationand resource-intensive because they are generated for local, national and regional markets.

But in a world of big high-velocity data and ubiquitous computing, an annual forecast does not seem sufficient for either public policy-making or forming a business strategy. Unlike in the United States, where Freddie Mac, Fannie Mae, the Federal Reserve, Housing and Urban Development and numerous other public and private agencies and companies have access to macro-econometric forecasts for housing markets, CMHC is perhaps the only agency in Canada whose forecasts are publicly shared. Hence, there is a need to improve the quality and frequency of its forecasts.

Dugan intends to improve the messaging of those forecasts in the future. CMHC had forecasted a peak-totrough decline of nine to 18 per cent in housing prices, which sounded more extreme than the much lower year-over-year change in average price of just 0.8 per cent. In the future, CMHC intends to report a yearover-year change in housing prices.

Given that CMHC is Canada's premier agency focused on providing mortgage liquidity and improving affordable housing, it should be mandated to provide upto-date information and forecasts. Economic data are released monthly or quarterly, providing an opportunity for the CMHC to generate quarterly housing market forecasts. If resources are the only constraint, incoming CEO Romy Bowers, who takes over April 6, might want to strengthen her forecasting team.

Also, CMHC might want to improve the statistical measures of forecast accuracy. Economists Clive Granger and Hashem Pesaran in the Journal of Forecasting in 2000 “emphasized the need for producing and reporting predictive distribution functions rather than point, or even interval forecasts.”

This implies that instead of providing a point or range forecast, CMHC could report the probability of a change in sales or prices, similar to the way the weather is reported, where the forecast communicates the likelihood of precipitation of a certain magnitude.

Forecasting is, and will always be, associated with errors. After all, there isn't an Nostradamus of economics. What will help people understand and respect forecasts is increased transparency about the assumptions and judgments that drive the forecasts and their timely communication.

FORECASTING IS, AND WILL ALWAYS BE, ASSOCIATED

WITH ERRORS.