Canadian housing prices are higher than levels consistent with current fundamentals in some provinces. The empirical estimates suggest that a 10 percent decline in housing prices would lead to a 1¼ percent decline in private consumption. The high level of household leverage and housing prices could prove to be a source of vulnerability. The rebound in debt and housing prices after the crisis largely reflects the resilience of the financial system and the stronger economic recovery in Canada, as well as historically low interest rates.

Abstract

Canadian housing prices are higher than levels consistent with current fundamentals in some provinces. The empirical estimates suggest that a 10 percent decline in housing prices would lead to a 1¼ percent decline in private consumption. The high level of household leverage and housing prices could prove to be a source of vulnerability. The rebound in debt and housing prices after the crisis largely reflects the resilience of the financial system and the stronger economic recovery in Canada, as well as historically low interest rates.

I. House Prices and Household Wealth in Canada1

Like many other advanced economies, Canada experienced an upswing in household debt and house prices in the 2000s. Our estimates suggest that house prices are higher than levels consistent with current fundamentals in some provinces. We study the impact of a potential correction in house prices on consumption through household wealth effects. Our empirical estimates suggest that a ten percent decline in house prices would lead to a 1½ percent decline in private consumption.

A. Introduction

1. Like many other advanced economies, Canada experienced an upswing in household debt and house prices in the 2000s (Figures 1 and 2). The household sector’s debt-to-income ratio climbed to historic highs, reflecting in part a marked increase in homeownership. The increase in household debt was moderate in comparison to many other industrialized countries in the decade preceding the 2008–09 crisis, but in contrast to the sustained price corrections in most other countries, house prices resumed their upward trend by mid-2009 in most Canadian provinces. The growth of household debt has outpaced households’ real estate assets since 2007, but the Canadian household sector as a whole still has a comfortable level of net housing equity at current prices.

Figure 1.
Figure 1.

Homeownership, Household Debt, Owners’ Equity, and House Prices

Citation: IMF Staff Country Reports 2011, 365; 10.5089/9781463929237.002.A001

Sources: Bank for International Settlements; Canadian Real Estate Association: Global Property Guide; Eurostat; Haver Analytics; National Bank Financial; OECD; Statistics Canada: Census 2006, and Fund staff calculations.
Figure 2a.
Figure 2a.

Household Assets in Advanced Economies

Citation: IMF Staff Country Reports 2011, 365; 10.5089/9781463929237.002.A001

Sources: Banca d’ltalia, Banquede France, Board of Governors of the Federal Reserve System, Cabinet Office for the Government of Japan, Deutsche Bundesbank, Economic and Social Research Institute, Haver Analytics, INSEE, OECD, Office for National Statistics of the U.K., Statistics Canada, and Fund staff calculations.
Figure 2b.
Figure 2b.

Household Liabilities in Advanced Economies

Citation: IMF Staff Country Reports 2011, 365; 10.5089/9781463929237.002.A001

Sources: Banca d’Italia, Banquede France, Board of Governors of the Federal Reserve System, Cabinet Office for the Government of Japan, Deutsche Bundesbank, Economic and Social Research Institute, Haver Analytics, INSEE, OECD, Office for National Statistics of the U.K., Statistics Canada, and Fund staff calculations.

2. The high level of household leverage and house prices could prove to be a source of vulnerability. The rebound in debt and house prices after the crisis largely reflects the resilience of the financial system and the stronger economic recovery in Canada, as well as historically low interest rates. However, further increases in leverage could set the stage for a large correction down the road, triggered for instance by an adverse external shock. Against this backdrop, this paper examines regional house prices relative to their equilibrium levels and the sensitivity of consumption to housing wealth in Canada.

B. The Evolution of House Prices

3. House prices in Canada have more than doubled over the past decade, notwithstanding a 11 percent correction after the 2008 crisis.2 British Columbia has witnessed the highest increases, with prices higher by 163 percent relative to the second quarter of 2001. In British Columbia and Ontario, house prices have grown by around 41 and 29 percent since their crisis trough (after falling by 10 and 13 percent after their pre-crisis peaks in Q1 2009 and Q4 2008, respectively). Growth rates of house prices have outpaced those of incomes and rents, leading price-to-income and price-to-rent ratios to historic highs. Given the high urbanization rate in Canada, house prices at the provincial level are mainly driven by the major metropolitan areas in the province. Price-to-rent ratios are elevated in the largest metropolitan areas, particularly in Vancouver (Figure 3).

Figure 3.
Figure 3.

House Prices in Canada, Its Major Provinces, and Major Metropolitan Areas

Citation: IMF Staff Country Reports 2011, 365; 10.5089/9781463929237.002.A001

Sources: Canada Real Estate Association and Fund staff calculations.

4. We estimate models of long-run equilibrium house prices for Canada. We assume that the equilibrium level of house prices in period t, Pt*, is determined by fundamentals: Pt* = f (Xt), where Xt is a vector of variables that affect either the supply or demand of housing. We consider the following variables as the fundamental determinants of house prices: employment, immigration, real income per household, the mortgage interest rate, commodity prices, urbanization, and the borrowing capacity of households.3 Using data from 1980 to 2011 (with data for the first half of 2011 annualized), we estimate various long-run equilibrium house price models for the four major provinces: Alberta, British Columbia, Ontario, and Quebec.

5. The estimates suggest that house prices are associated with income, employment, commodity prices, immigration, and borrowing costs. The regression presented in the first column of Table 1 (Model 0) includes all the explanatory variables we consider. For this specification, the estimated coefficients of employment, income, commodity prices, and immigration are statistically significant; the coefficients on mortgage interest rates, households’ borrowing capacity, and urbanization are not.4 We run regressions with the four explanatory variables that are significant in Model 0 for the pooled sample and for each province individually (Models 1 and 2, respectively). As mortgage interest rates and households’ borrowing capacity are conceptually strongly linked to the demand for owner-occupied housing, and exhibit a significant bi-variate correlation with house prices, we also estimate parsimonious specifications linking house prices to mortgage rates and borrowing capacity (Models 3 and 4, respectively). These regressions suggest that the extraordinarily low level of interest rates is among the factors that have spurred the recent growth in mortgage debt and house prices.

Table 1.

Estimates of House Price Equations

(Dependent variable: log of real house prices; annual observations, 1980-2011)

article image
Sources: Haver Analytics, Statistics Canada, and Fund staff estimates based on data from Canada Mortgage and Housing Corporation.Notes: All variables are in logarithms except for the mortgage rate and urbanization. Real house prices at the provincial level are given by the CREA average existing home price index deflated by the provincial CPI. Real personal income per household at the provincial level is proxied by “average real market income” from Statistics Canada and includes earnings from employment and self-employment, investment income, private retirement income, and items under “Other income”. The commodity price index, produced by the Bank of Canada, is a chain Fisher price index of the spot or transaction U.S. dollar prices of 24 commodities produced in Canada and sold in world markets. Immigration is the the number of new immigrants in a province. Mortgage rate is the average residential mortgage rate from which the inflation rate was substracted. Borrowing capacity estimates the maximum borrowing capacity of an average borrower based on the current amortization limit, the actual mortgage rate (which is sometimes below the posted rate) and real personal income. Urbanization at the provincial level, produced by the Census of Population, is the share of population in cities. *, **, and *** indicate significance at the 10, 5, and 1 percent level, respectively. Model 0 and 1 are fixed-effect panel regressions, Model 2 consists of OLS regressions at the provincal level, Model 3 is a fixed effect panel regression on income and real mortgage rate, and Model 4 is a fixed effect panel regression on maximum borrowing capacity.

6. We use the estimated models to gauge the level of house prices that would be consistent with the long-run determinants of house prices. The residuals from the long-run equilibrium regressions suggest house prices in 2011 to be above the levels consistent with the current levels of fundamentals in British Columbia, with some signs of overvaluation also in Ontario, and to a lesser degree, in Quebec. By contrast, the estimated models suggest house prices to be mildly undervalued in Alberta. A weighted average of our estimates (with weights based on provincial GDP levels) suggest that house prices in Canada are on average ten percent above the level consistent with current fundamentals. We also run estimations on sub-samples to perform out-of-sample forecasts of house prices. Our findings (not shown here) indicate that the earlier the cutoff for the subsample, the larger is the implied overvaluation in British Columbia, Ontario, and Quebec.

uA01fig01

Estimated House Price Overvaluation in Major Provinces

(In percent of equilibrium prices implied by the models)

Citation: IMF Staff Country Reports 2011, 365; 10.5089/9781463929237.002.A001

Source: Fund staff estimates.

C. Housing Wealth and Consumption

7. We update estimates of the sensitivity of private consumption to housing wealth in Canada using data for 1990–2011. Following Pichette and Tremblay (2004) and Lettau and Ludvigson (2001), we first estimate a long run relationship between consumption, disposable income (a proxy for human wealth), net housing wealth, financial wealth, and other non-human wealth (mostly the stock of durable goods). We then estimate a vector-error-correction model that differentiates between the responses of consumption to permanent versus temporary changes in wealth (using the method proposed by Gonzalo and Granger, 1995). The average marginal propensity to consume (MPC) out of each type of wealth (i) is given by:

MPCi=πiΦiT+(1πi)ΦiP

where π is the weight on the wealth variation that is transitory and Φ is the MPC from transitory (T) or permanent (P) changes in wealth. Further details on the data and our estimation method are provided in Appendix 1.

8. Our estimate of the MPC out of average changes in housing wealth is 4.3 cents per dollar. Our findings suggest that consumption responds primarily to permanent changes in wealth; responses to transient changes are negligible. For consumption, disposable income, housing wealth, and other wealth, most of the variation is explained by permanent shocks. In the particular case of housing wealth, permanent shocks represent 97 percent of the variation, implying that π = 0.85. The MPC for permanent changes in housing wealth, Φp, is estimated at 5.4 cents per dollar. The MPC for housing wealth is a weighted average of zero and 5.4, where the weight on the latter is given by π. The sensitivity of consumption to changes in financial wealth is estimated to be lower, around 2.5 cents per dollar on average, in part given the relatively higher importance of transient shocks to financial wealth and the relative importance of real estate assets compared to financial assets in households’ portfolios. The estimates are broadly comparable to those reported in Pichette (2004) for Canada based on a sample for 1965–2003; our estimate of a higher sensitivity to financial wealth relative to Pichette (2004) is likely to be due to the relatively higher importance of financial wealth in the more recent sample period.

9. The housing MPC estimates for Canadian households presented in this paper are similar to estimates based on cross-country datasets but somewhat lower than those recently estimated for the United States. Carroll et al. (2010) find long-run MPCs of 9 and 4 cents per dollar for housing and financial wealth in the United States, respectively. Estimates in IMF (2002) and Case et al (2001) suggest average long-run MPC out of housing wealth of about 4–5 cents per dollar for a group of advanced economies—very close to our estimates for the Canadian economy.

10. Our estimates suggest that a ten percent reduction in housing wealth could be associated with a reduction in private consumption (excluding durables) of 1.1 percent due to wealth effects, corresponding to a ½ percent decline in GDP.5 Such a reduction in housing wealth could be triggered by an external shock—for example, a decline in foreign demand for Canadian exports and weaker commodity prices in the context of increased global risk aversion could lead to higher unemployment in Canada and a downturn in house prices. The results also suggest that around 6.3 percent of the increase in the level of per capita private consumption in the last two years can be explained by increases in housing wealth. Given the significant sensitivity of consumption to house prices in Canada, further adjustments to macro-prudential policies for mortgage standards would be warranted in a scenario of further sustained increases in house prices to reduce the risk of a disruptive adjustment down the road.

D. Conclusion

11. Our econometric findings suggest that house prices are higher than the levels consistent with current fundamentals in a number of Canadian provinces and that a correction in house prices would have measurable effects on consumption and output through wealth effects. As discussed in the staff report, the authorities have appropriately taken macro-prudential measures to curb the growth of household debt. Given the unsettled global economic environment that could trigger adverse shocks on the Canadian economy, the authorities should remain vigilant to the developments affecting household balance sheets; further macro-prudential measures may be needed if the debt build-up continues.

References

  • Case, K.E., J.M. Quigley, and R.J. Shiller. 2001. “Comparing Wealth Effects: The Stock Market Versus the Housing Market.National Bureau of Economic Research. NBER Working Paper Series No. 8606.

    • Search Google Scholar
    • Export Citation
  • Carroll, Christopher D., Misuzu Otsuka, and Jiri Slacalek, 2011. “How Large Are Housing and Financial Wealth Effects? A New Approach,Journal of Money, Credit and Banking, Blackwell Publishing, vol. 43(1), pages 5579, 02.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Cochrane, John H., 1994. “Permanent and Transitory Components of GDP and Stock Prices,Quarterly Journal of Economics, 1994, 109, 241265.

  • Gonzalo, J. and C. Granger, 1995. “Estimation of Common Long-Memory Components in Cointegrated Systems,Journal of Business and Economic Statistics, 1995, 13, 2735.

    • Search Google Scholar
    • Export Citation
  • Igan, Deniz, and Prakash Loungani (2011) “Global Housing Cycles,forthcoming IMF Working Paper.

  • IMF World Economic Outlook, 2002. “Chapter 2: Three essays on how financial markets affect real activity”.

  • Lettau, M. and S. Ludvigson, 2001. “Understanding Trend and Cycle in Asset Values: Bulls, Bears, and the Wealth Effect on Consumption.Centre for Economic Policy Research Discussion Paper No. 3104.

    • Search Google Scholar
    • Export Citation
  • Lettau, M. and S. Ludvigson, 2001. “Consumption, Aggregate Wealth and Expected Stock Returns,Journal of Finance, June 2001, 56(3), 815849.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lettau, M. and S. Ludvigson, 2004. “Understanding Trend and Cycle in Asset Values: Reevaluating the Wealth Effect on Consumption,American Economic Review, 2004, March, 94(1), pp. 276299.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • McQuinn, K. and O’Reilly G., 2008, “Assessing the Role of Income and Interest Rates in Determining House Prices, Empirical Modelling, vol. 25, pp.377390.

    • Search Google Scholar
    • Export Citation
  • Pichette, L., 2004. “Are Wealth Effects Important for Canada?Bank of Canada Review, Spring 2004.

  • Pichette, L. and D. Tremblay, 2003. “Are Wealth Effects Important for Canada?Bank of Canada Working Paper No. 2003–30.

  • Stock, J. and M. Watson, 1993. “A Simple Estimator of Cointegrating Vectors in Higher Order Integrated Systems,Econometrica, 1993, 61, 783820.

Appendix 1. Data Sources and Estimation Method

Data

Data on household wealth are from Statistics Canada National Balance Sheets for the period 1990Q1 to 2011Q2. We divide non-human wealth into three components: housing wealth, financial wealth, and other non-human wealth. Housing is defined as the market value of land plus residential structures, net of mortgage debt. Financial wealth includes all financial wealth such as equity holdings and deposits at market values. The remaining wealth in market value, such as durable goods, is included in other non-human wealth. Disposable income is used as a proxy of human wealth. Data is expressed in per capita, real terms. Consumer expenditure, our dependent variable, is defined as the sum non-durable expenditure components, including services. Durable consumption is not included since the flow of services from durable goods rather than their purchases would be the appropriate measure in our analysis.

Method

We estimate a cointegrating system for nondurables consumption, Ct, income, Yt, housing wealth, Ht, financial wealth, St, and other wealth, Ot. We use lower case letters to denote logarithmic transformations, e.g., ct ≡ In (Ct). Lettau and Ludvigson (2001b) derive an approximate equation for the ratio of log consumption to aggregate wealth (W) using observable variables, which we express, after approximating total non-human wealth (At) as the sum of housing, financial and other wealth, as:

ct − αyyt − αhht − αsst − αootEtΣi=1ρwi((1 − v)γat + i − Δct + i + vΔyt + 1 + i)

where ρwi=1 − exp(c − w¯),, v is the steady state share of human wealth in aggregate wealth and a is the average return of non-human wealth. If labor income follows a random walk and human capital returns are constant, the left hand side is a proxy of the log consumption-wealth ratio. The cointegrating residual on the left-hand side should forecast changes in asset wealth (returns), changes in labor income, changes in consumption growth, or a combination of the three. Lettau and Ludvigson (2001a, 2001b) find this residual to be a strong predictor of excess returns on aggregate stock market indexes but not consumption or labor income growth. The estimated long-run relationship is (following the method of Stock and Watson, 1993) is:

ct =0.12 + 0.51yt + 0.13ht + 0.15st + 0.08ot

This relationship determines the error correction term in the estimation of the reduced-form vector error-correction model (VECM):

Δxt=μt + Σj=1lAjΔxt1 + αβxt1 + εt

Where x is the vector of cointegrated variables, with xt = [ct, yt, ht, st, ot]’. All variables in the vector are first order integrated, I(1), as confirmed by unit root tests. Our tests also reveal the presence of a single cointegrating vector for the five variables, a result we impose on the VECM. Newey-West corrected errors show significance at 1 percent for the parameters estimated. The estimated adjustment coefficients to the error correction term are a^ = (–0.0931, –0.0539, 0.0213, 0.9242, –0.1439). Following Gonzalo and Granger (1995), the permanent and transitory components of the shocks are identified and the fraction of forecast-error variance attributable to permanent and transitory shocks are obtained (σPi2andσTi2 respectively). The relative importance of permanent shocks is for average MPC for a given type of wealth (i) is given by:

Πi=σTi2σTi2 + σPi2
1

Prepared by Oya Celasun, Alejo Costa, and Jihad Dagher.

2

The Canadian Real Estate Association (CREA) national house price index shows an increase of 115 percent between Q2 2001 and Q2 2011; the Teranet national index shows an increase of around 96 percent over the same period. The CREA national and provincial price indices are compiled from statistics of existing homes and properties sold through the Multiple Listing Service; changes in the indices could potentially reflect compositional shifts. By contrast, the Teranet national house price index is constructed using a repeat sales methodology. House price indices based on repeat sales are not available at the provincial level.

3

The data sources are listed in Table 1. McQuinn and O’Reilly (2008) develop a model in which house prices depend on households’ maximum borrowing capacity, which in turn depends on income, the maximum debt service ratio, the maximum allowed amortization period, and mortgage interest rates.

4

Borrowing costs tend not to be significant in house price regressions that also control for employment and commodity prices since the negative co-movement between borrowing costs and house prices is confounded by the Bank of Canada’s policy reaction to the economic cycle (for example, in the 1990’s mortgage interest rates dropped with the monetary easing that followed an episode of house price declines).

5

This finding is very close to the estimate in Igan and Loungani (2011). In contrast, they estimate that a ten percent drop in house prices would lower consumption by 2.8 percent in the United States. Data on the distribution of LTV ratios at current house prices suggest that a ten percent decline in house valuations would put about five percent of mortgage borrowers underwater (based on CMHC data and authors’ calculations).

Canada: Selected Issues Paper
Author: International Monetary Fund
  • View in gallery

    Homeownership, Household Debt, Owners’ Equity, and House Prices

  • View in gallery

    Household Assets in Advanced Economies

  • View in gallery

    Household Liabilities in Advanced Economies

  • View in gallery

    House Prices in Canada, Its Major Provinces, and Major Metropolitan Areas

  • View in gallery

    Estimated House Price Overvaluation in Major Provinces

    (In percent of equilibrium prices implied by the models)