Canadian Housing Markets: Bubbles if Necessary but Not Necessarily a Bubble

The debate over whether or not Canada’s housing market is a bubble that is going to crash heated up again this month. Earlier this month, the Financial Times warned that the Canadian housing market was “perched precariously at its peak”. Bank of Canada governor Stephen Poloz said this week that Canada’s housing market was not characterized by a bubble and expected housing prices to stabilize.  Of course, this debate assumes the existence of a “Canadian housing market.”

 

In reality, there are many housing markets in Canada spread out from east to west and local and regional conditions are likely to affect housing prices and by extension, the degree to which those markets are overheated.  As Figure 1 shows, there are some pretty big differences in average MLS prices across Canada as evident from data from the Canada Mortgage and Housing Corporation.  Despite forecast declines for 2013 for some markets, prices are still highest in Vancouver, Toronto and Victoria and lowest in Windsor, Three Rivers and Moncton. 
However, Figure 2 looks at the percent increase in the average MLS prices in Canadian CMAs for the period 2006 to 2013 and the results show the biggest percentage increases in prices have actually been occurring in smaller Canadian centers.  Since 2006, the percentage increases in price have been the largest in Regina, St. John’s and Saskatoon.  Next are Winnipeg, Thunder Bay and Saguenay.  Vancouver and Toronto are decidedly in the middle of the pack when it comes to the percentage increase.

So how can we conclude if some markets are more likely to be characterized by a bubble than others?  Well, I decided to regress the percentage change in the average MLS housing price for the period 2006 to 2013 for thirty-one major Canadian CMAs (avgmlsprice) on percentage increases in population for 2006 to 2012 (population) and percentage increases in median total family income for the period 2006 to 2011 (income). (Kelowna, Abbotsford and Sherebrooke were omitted because of gaps in obtaining some of the variables).  While a very simple demand side model specification, one might expect that increases in the demand for residential housing will be a function of these two key variables.  How much of the increase in housing prices can be explained by demand side factors driven solely by increases in population and income? 

The OLS regression is shown in Figure 3 and the coefficients on population and income are both positive but only income is significant and the regression explains about 60 percent of the percentage change in average MLS housing prices.  What I do next is compare the actual percentage increase in MLS housing prices with what the regression predicts to see if the actual increase exceeds what the model predicts.  I take the actual MLS increase for each of the thirty-one CMAs and subtract the predicted increase and what is left is the growth in excess of what population and income should have driven.  These are ranked in Figure 4 from the highest to lowest.

 

 

Well, the results suggest that Regina, Thunder Bay, Winnipeg, Hamilton and Windsor have seen the greatest positive difference between their actual and predicted housing price growth.  Their prices have grown well beyond what one might have expected based on their income and population growth.  Does this mean these housing markets may be the ones that are most likely facing a housing bubble?  Perhaps, unless there are important factors other local population and income driving housing prices.  In the case of Hamilton, the proximity to the GTA may be a factor driving their prices. 

On the other hand, prices in Moncton, Edmonton, Victoria, Calgary and Halifax have grown by much less than what their income and population growth would have predicted.  Does this mean these markets are undervalued?  Not sure.  These results also suggest Toronto and Vancouver have seen changes in their housing prices quite close to what increase in population and income might predict.  This would suggest that there is no bubble here and given the relative importance of these two housing markets in the Canadian economy, Stephen Poloz is correct in not being too concerned about a Canadian housing bubble.  After all, why would he be especially preoccupied about a potential housing bubble in Regina or Thunder Bay? 

The point remains we should not be thinking of a bubble in the Canadian housing market but perhaps local or regional ones. When it comes to whether or not there is a housing bubble in Canada, one needs to examine specific markets rather than simply treat the Canadian housing market as one entity.  There may indeed be some overheated markets in Canada, but then there may also be some undervalued ones. 

Comments

[Ironman]

You should be able to determine if there is a bubble using just national level data. Start by plotting median house sale prices (vertical axis) against median household income (horizontal axis).

What you should find is that Canada's house prices follow a very distinct linear trend from 1990 (and earlier) up to 2001, when it shifted to a much steeper upward linear trajectory, which has persisted since. If you can explain the reason for that shift in trajectory, you'll understand what's been driving Canada's housing markets since that year. [Hint: It greatly involves China.]

[R I]

That's an interesting way to analyze if there is a bubble. But what if causation runs the other way around? That is, for markets where housing price increases match or exceed income gains, what if the income gains are a result of a strong housing market (as in, the strong housing market is driving new home construction, driving increases in income and employment rates, resulting in increases in family income). There might be a feedback effect, but I imagine the coefficient would be less than 1. In any case, there would have to be a 3rd input driving the increase in housing prices.

[Ebsim]

When I used to live in Saskatoon everyone was waiting for some bubble to burst. And... nothing happened.

There is a bubble in bubble warnings. Even calling them bubbles is a little misleading. There might be a real estate bubble but it seems that people want something to happen, to be done. Which I think will cause unwanted consequences worse than the consequences of falling real estate prices.

[Humble Student of the Markets]

I note that your first chart shows the average absolute MLS prices by city, while your analysis uses the change in MLS prices as the y variable. Why are regressing against the change in price instead of the absolute price level?

When you regress against the change in price, you assume that the initial condition is at equilibrium. Out here in Vancouver, as an example, we know the house price to income ratio is totally out of whack compared to other cities in the world, largely because of the influx of foreign money into the local market.

[Mark]

I'd agree housing markets are a local phenomenon. However, the issue with disaggregating the data is you have to stop somewhere. There have been significant differences within Toronto, say compare Leslieville to Scarborough.

[Livio Di Matteo]

Humble: Good point. I wanted to look at the change in price because smaller centers were seeing much larger percentage increases even if their absolute price level was much lower compared to say Toronto or Vancouver.

[Ironman]

R I: The feedback effect you describe would be a characteristic of bubble economic conditions - since home prices under normal circumstances are predominantly a strong function of household income, such a feedback effect would contribute to prices escalating within a housing market at a much faster pace than incomes grow.

LDM: I'd like to put a twist on a great point you made:

"The point remains we should not be thinking of a bubble in the Canadian housing market but perhaps local or regional ones."

Picking up my earlier point regarding China, what if housing prices in Canada are behaving rationally, maintaining a non-bubble relationship between housing prices and household incomes, but are being driven by a bubble outside the housing market itself? Say a sustained surge in China's demand for natural resources, which began back in 2001? That would lead to higher Canadian incomes, particularly in regions where the economies are closely tied to either natural resource production or international trade, which would in turn drive up housing prices in those areas.

Speaking of which, if that's the case, we should see a similar pattern for home prices in other raw material-producing nations for whom China represents a major export market, such as Australia.

[Livio Di Matteo]

Ironman:
Well the Australians are also seeing talk of a bubble.

http://www.bloomberg.com/news/2013-11-05/bubble-trouble-seen-brewing-in-australia-home-prices-mortgages.html

[Michael Hamilton]

But population and income don't justify real house prices being permanently higher unless supply conditions are so constrained (eg by land use regulation) that supply of new residential land and associated dwellings can never catch up. In the US, for example, Shiller's long-term repeat sales data set suggests that real house prices have risen by only around 25% over 120 years. Mobility among large cities, and the widely differning land use regulatory regimes helps generate that outcome in the US. By contrast, in a country like NZ, a single dominant city, and pervasive restrictions on bringing land into residential use in the face of quite rapid population growth seems to have semi-permanently raised real house prices What is the land use restriction situation like in Canada?

[Livio Di Matteo]

Michael:
in Canada,I think land use is usually at the provincial/municipal level so policies would vary across the country.

[Chris Auld]

Livio, if you can snag rent data easily, showing changes in price to rent ratios would be more compelling, albeit still suggestive, evidence regarding bubbles than price/income ratios. With just price data, as you say, you cannot hope to differentiate between a bubble and changes in prices due to any cause other than incomes or population (and population has no explanatory power, and it's endogenous). Rents and prices should change roughly proportionally in response to many non-bubbly demand shifters but bubbles cause prices and rents to diverge, so it's very useful to stick rents in the model somewhere.

[Livio Di Matteo]

Thanks Chris.

[Jacques René Giguère]

Ironman, Livio: this is exactly what's happening in my area.
Sept-Îles,moved by iron ore exports mostly to China (theres is currently a 300 000 dwt Chinamax in the harbor), has seen a price explosion, helped by constraints on usable lands (most land here is either total granite or coastal swamps) while Baie-Comeau, 200 kms west, with weak aluminum prices and a destroyed paper market, has seen essentially a plunge. And as China weakens, the average lenght for selling a house has gone from 36 days last year to 136 recently.

[Nick Rowe]

Interesting post, Livio.

I was surprised that population didn't have a bigger effect. But then I thought: we might expect the relationship between house prices and population to go either way.

An exogenous increase in population (caused by more jobs, say), would increase the demand for houses and cause house prices to rise.

An exogenous increase in the supply of houses (relaxing zoning rules, say) would cause house prices to go down, and population to go up.

[hosertohoosier]

At least one enabling factor for the US housing bubble lay in persistent current account imbalance between the US and China. Effectively, China's surplus was bankrolling a bubble in American real estate (and obscuring the Fed's inflation targets, keeping interest rates low for too long). Since 2008, Canada swung from a current account surplus of 1-2% to a deficit of 3-4% (without downward pressure on the CAD). Our response was to run deficits, both nationally, provincially, and at the level of Canadian households.

This doesn't look to me like some change in fundamentals. The value of our commodity exports was increasing, not decreasing - surely that should have shored up our [former] position as a surplus country. And if foreign investors were looking to long resource shortages by buying up Canadian real estate, you'd think they'd invest in Edmonton/Calgary real estate particularly.

We've gotten a surge of cheap credit, and we are channeling it to unproductive investments in real estate. That sounds kinda familiar to me (though, at the same time, we regulate the housing and banking sectors differently, so it isn't necessarily going to play out like the subprime crisis did).

[Andy]

Livio,
Interesting analysis. I'd be interested in seeing two further things: one, to deal with the worry that your result is being largely driven by outliers, I'd be interested to see what happens if you throw out the top and bottom three data points; piling on, I'm no Canadian but are those markets relatively small? Second, how does 60% compare to the Canadian aggregate (i.e. income rise over the time period vs house price rise over the time period)?

[Livio Di Matteo]

Andy:
I did run the regression dropping the top and bottom data points - population was still insignificant (but a small negative coefficent rather than a small positive one) while income was positive and significant and almost the same coefficient. The adjusted r-squared was 0.62. Have not run it dropping both the top and bottom three but that would remove nearly 20 percent of the observations. One thing I also did not do and in retrospect might have been useful was weight the variables in the regression to take into account the differences in population across CMAs. After all, there are big differences in the size of say Regina and Toronto and yet both get equal weight in the regression. As for the average increases in these variables across the CMAs, over the period 2006 to 2013, the average MLS price rises about 45 percent, the average CMA population rises 7 percent and average median total family income rose 13 percent.

[paras bhargava]

Can you get reliable data going back further? before buying my house i cajoled the TREB people to give me some files going back to the 40's and it certainly seemed like Real Toronto Average Housing prices did have periods of bubble like overheating then corresponding cooling.

[mobk]

Why start the analysis in 2006 only? Folks that argue that there is a bubble would claim that the bubble started further back (more like 2001 or 2002).

[Tomas]

Consider two other paths to validate this work:

1) Run population, income and house price change for other countries

2) Run change in consumer debt and change in house prices for Canada

[genauer]

maybe some folks find this comparison useful:

http://www.marctomarket.com/2013/11/great-graphic-rent-relative-to-house.html

it also has some comments on Mark Carney, which are more in tune, how many in mainland Europe see him

[genauer]

With the recent comments of Nick Rowe, that he also has difficulties to post in some places,

maybe it helps, that I seem to be able to post here, if I go to the thread page, but not if I go to the comments link

[Mandelbrott]

Can't have a discussion about asset prices within the framework of supply/demand or fundamentals. People with mortgages buy houses hence one has to look at the dynamics in the supply of credit. Any MMTer or MCTer would suggest that since housing is mostly financed by members of the BoC, housing is ultimately directly (via CMHC) and indirectly (supporting the banks) dependent on government support. Good or bad, it's a matter of policy ultimately.