This appendix provides a technical discussion of the framework used in the Housing Market Assessment. The goal of the report is to provide prompt, accurate, and reliable assessments of Canadian housing markets to the public with the objective of identifying vulnerabilities. It is not a forecasting tool or affordability assessment.
The HMA builds on published economic and financial research taking into account the characteristics of Canadian housing markets, uses different econometric methods and data from various sources, and validates the models with past periods where housing imbalances were observed such as Toronto in the late 1980s and Calgary prior to the global financial crisis of 2007 – 2008.
To obtain an accurate picture of the overall state of the housing market, it is of central importance to consider multiple data points and lines of evidence rather than relying on just one measure or indicator. The Housing Market Assessment undertakes this task by constructing a comprehensive, multidimensional, and integrated framework to assess housing market conditions.
To capture imbalances in housing markets, the framework assesses four factors: (1) overheating when demand outpaces supply; (2) sustained acceleration in house prices; (3) overvaluation of house prices in comparison to levels that can be supported by housing market fundamentals; and, (4) excess inventories when the inventory of available housing units is elevated.
Excess inventories, previously named “overbuilding”, refers to an elevated number of newly built unsold homes or a high rental apartment vacancy rate. Such excess inventories can create financial difficulties for builders and rental companies if they struggle to repay debt used to build or buy these units.
Excess inventories are a risk even in unaffordable markets. In such markets, excess inventories should be cleared in the medium term but can present short-term risks. Centres with unaffordable housing would continue to benefit from long-term increases in housing supply. This would restrain growth in house prices and rents even if there are short-term periods of excess inventories.
As Table 1 shows, several indicators are considered, when possible, in assessing these factors. For instance, excess inventories — measured by the deviation of housing prices from their level consistent with economic, financial and demographic fundamentals — is assessed using five econometric models each of which is using four different price series (see Section 2 for details).
|Acceleration in house prices4||
How does the HMA detect significant imbalances or vulnerable market conditions?
The Housing Market Assessment first uses several indicators to test for the presence of signals of vulnerability in housing markets against some predefined thresholds. Once such signals are detected, the assessment measures their intensity, i.e. to what extent the indicators deviate from their historical average, or how their magnitude compares to what was observed during known or suspected past house price bubble episodes, such as in Toronto in the late 1980s-early1990s or in Calgary prior to the global financial crisis of 2007 – 2008.
It then considers the persistence of imbalances over time. In other words, the framework acknowledges that housing markets can temporarily diverge from their fundamentals and that does not necessarily signal vulnerability if the divergence is mild and short-lived. In some cases the divergence could be caused by temporary factors in markets.
How are the results presented?
The Housing Market Assessment uses a colour scale to characterize our assessment of each of the four factors. Green signifies low evidence of vulnerability, meaning that the imbalance is either absent or too short-lived to suggest that it is vulnerable. Yellow signals moderate evidence of vulnerability and is applied to cases where only one of the indicators shows signs of intensity and persistence of imbalances. The colour scale extends to red only for the factors for which indicators are signaling incidence, intensity and persistence of vulnerability. As a result, only overvaluation and excess inventories can receive a red rating because they are assessed with more than one indicator.
How is the HMA framework validated?
The framework was tested against CMHC’s mortgage insurance claims rate. The results show that the detection of more than one Housing Market Assessment factor is more problematic for insurance claims than the detection of just one factor. Therefore, the individual factors are jointly analysed to provide an overall assessment of the state of a given housing market, which is rated on our three-coloured scale (green, yellow, and red). See text box 1 for the colour scale for the overall assessment.
Text box 1: Colour scale for the overall assessment
The four Housing Market Assessment factors are assessed jointly to provide an overall assessment of a given housing market.
An overall assessment of high vulnerability (i.e. red) reflects a situation where more than one factor of price acceleration, overvaluation or excess inventories exhibits moderate or high evidence of imbalance. As these imbalances are resolved, the overall assessment will be moved from high to moderate to maintain an ongoing monitoring after the detection of a vulnerability. As these inbalances are resolved, the overall assessment will be moved from high to moderate to maintain and ongoing monitoring after the detection of a vulnerability.
An overall assessment of moderate vulnerability (i.e. yellow) can reflect a variety of cases. The first case is when only one of the factors of excess inventories or overvaluation are assessed as showing high evidence. Another case is when only one factor is detected with moderate evidence and we have concerns about another factor that has not technically crossed the threshold for moderate evidence.
Finally, an overall assessment of low vulnerability (i.e. green) applies to all other situations. For example, it can reflect a situation where only one factor shows moderate imbalance.
Housing Market Assessment Factors
As shown in table 1, the framework considers four factors to assess housing market conditions: overheating, acceleration in house prices, overvaluation, and excess inventories. This section provides a detailed description of each of the four factors used in the assessment, including the indicators used to assess each factor and the criteria used to identify whether there is a significant imbalance.
For each indicator our decision rule considers intensity and persistence jointly to decide if a factor is considered as a vulnerability. This approach reduces the risk of misleading assessments that only rely on minor deviations or “blips” in the data. Typically, a threshold for an indicator identifies the lower bound for the zone of imbalances and this zone reflects levels of intensity that were rarely observed over the historical period. The persistence of imbalances is assessed by taking into account not only the most recent observation, but also the recent trend.
There is overheating when housing demand is significantly and persistently outpacing supply. The sales-to-new listings ratio (i.e. the number of existing homes sales divided by the number of new listings entering the market) is used as an indicator to assess possible overheating in the existing home market. When demand is strong relative to supply, house prices typically grow at a faster rate. Until the rebalancing of demand and supply occurs, sustained overheating conditions for the resale market may lead to upward pressures on housing prices for existing and new homes. However, as supply and demand begin to balance out, indicators of overheating would begin to soften and house prices would gradually moderate.
The framework compares the sales-to-new listings ratio to thresholds that indicate evidence of overheating. Since each Census Metropolitan Area market has its own characteristics, estimated thresholds for overheating vary. These thresholds are determined by CMHC’s local market analysts. Typically, taking the Canadian MLS® market as a whole, a sales-to-new-listings ratio above 55 per cent is associated with a sellers’ market while an overheating market corresponds to the upper range of a sellers’ market, with a ratio exceeding 70 per cent.
We evaluate overheating by assessing if the sales-to-new listings ratio exceeds the levels that are seldom seen in historical data. These levels indicate unusual and vulnerable conditions and thereby are defined as thresholds. We detect overheating in a given market if the sales-to-new listings ratio exceeds its threshold for at least two quarters over a period of four consecutive quarters, in the three years preceding the assessment7. This prevents relying on signals that are temporary in nature.
Acceleration in house prices
Expectations of future house price growth can attract investors who want to benefit primarily from short-term capital appreciation, which can then lead to further acceleration in house prices. It has been found that speculative activity, fad-based behaviour or excessive leveraging can cause house prices to accelerate, encouraging additional speculative activity, fad-based behaviour or excessive leveraging, thus propelling prices further upward in a spiral of increasing price growth.
Under balanced market conditions, house prices are expected to increase over time, in line with household’s cost of living (often measured by the Consumer Price Index — the CPI). There is house price acceleration when growth in housing prices starts increasing, and keeps increasing for several quarters. Basically, acceleration involves the level of house prices increasing beyond household’s cost of living and this discrepancy rises over time.
Acceleration in house prices over an extended period can eventually cause house prices to depart from levels warranted by the underlying demographic, economic and financial drivers of housing activity, thus eventually also leading to overvaluation.
The Housing Market Assessment formally deems there is acceleration in house prices if their growth rate exceeds an estimated threshold based on a statistical test that detects explosive behaviour of a time series.10
Acceleration in house prices is detected for test statistics reaching levels above their critical value (or threshold) hence we interpret it as a signal of unsustainable acceleration in prices.
In the short term, house prices will generally tend to fluctuate around levels consistent with the fundamental drivers of housing activity, like income and population growth, and actual and expected financing costs, with alternating periods of slight and non-significant overvaluation while remaining reflective, overall, of evolving market conditions instead of potentially vulnerable developments.
However, sustained overheating and acceleration in house prices can lead to overvaluation in the housing market. Speculative behaviour would push prices above those consistent with economic and demographic fundamental drivers. Overvaluation can also occur when observed prices adjust slowly to deteriorating housing market conditions. Overvaluation is detected when house prices remain significantly above levels warranted by fundamental drivers of housing markets.
The framework uses a combination of several house price measures and models to estimate house price levels warranted by fundamental drivers. Their deviation with observed prices provide the estimated degree of over- or undervaluation.
The Housing Market Assessment uses four different price measures: MLS® average resale price (or the Centris® average price by QPAREB), the New Housing Price Index (NHPI) from Statistics Canada, the Teranet-National Bank House Price Index™) and a CMHC repeat-sales index. It also uses five different models:11 two classic demand driven models, an urban growth model, a borrowing capacity model, and a hybrid model. These four models jointly cover a wide range of economic, demographic, and financial drivers of housing activity and prices. See text box 2 for a description of the four models.
For each price measure used to assess overvaluation, the five econometric models are used to estimate the level of house prices that would be consistent with the fundamental drivers of the housing market. When observed price levels diverge significantly and persistently from their fundamental levels estimated by these models, overvaluation is usually detected. The use of four different price measures and five models improves the reliability of our results.
Combining these five models and up to four price series (depending on the housing markets), generates several potential estimates12 of price levels supported by fundamental factors — and corresponding estimates of over- or undervaluation. The gap between observed and estimated (fundamental) house prices is an indication of the degree of over- or undervaluation.
Text box 2: Description of the models used to assess overvaluation
For each model, observed house prices and fundamental factors (such as personal disposable income and mortgage rates) are adjusted for CPI inflation.
According to the classic demand-driven model,13 the level of house prices is determined by variables such as the level of disposable income per capita, the effective five-year mortgage rate and young adults’ population. Intuitively, the classic demand-driven model supposes that increased household income raises housing demand since it reduces the burden of home ownership costs and facilitates access to credit for households. Mortgage rates are considered since they impact the size of the monthly mortgage payment. The model also supposes that an increase in population, for a fixed stock of housing units, would push house prices upward, with housing units becoming relatively scarcer and, hence, more expensive until supply adjusts through the construction of new housing units.
From the perspective of the urban growth model,14 the level of house prices is affected by the construction costs of housing, land prices, and expected growth in city size. Intuitively, lower construction costs and higher productivity reduces the overall cost of producing new housing units hence putting downward pressures on housing prices. The urban growth model also suggests that an increase in city size will push house prices upward because of land scarcity and commuting costs.
The borrowing capacity model15 focuses on the fact that households have limited borrowing capacity and could face difficulties accessing mortgage credit for a home purchase. This model estimates the price level consistent with fundamentals in relation to a maximum borrowing amount; which is the maximum amount an household can borrow given a 25-year amortization period and the current five-year mortgage interest rate for a mortgage payment equivalent to 30 per cent of the household’s income.
The hybrid model takes into account the factors considered in the three previous models.
The intensity of a house price gap — the magnitude of the difference between observed and fundamental house prices on the market — is determined using a threshold based on statistically significant deviations from historical averages. The threshold is determined as the boundary for the highest decile of estimated historical differences between observed and estimated fundamental levels.
Hence, we detect moderate evidence of overvaluation if the residuals from at least one of the overvaluation models are unusually high, i.e. when the normalized residuals are large (i.e. exceeding 1.29 standard deviations). Strong evidence of overvaluation is assessed if the residuals from several models are unusually high.
Indicators for overvaluation have to remain above their thresholds for at least two quarters over the last four quarters preceding the assessment to be deemed persistent.16
The lack of supply in the resale housing market can push potential buyers to the new housing market to meet their housing needs. Strong demand for housing can lead to higher house prices which, in turn, can lead to a supply increase. The larger supply of housing units will eventually alleviate the upward pressures on house prices and contribute to limit their increase.
However, it is possible that supply may temporarily exceed demand. As a result, this excess supply could put significant downward pressures on house prices. Alternatively, a reduction in demand for existing homes could also result in a condition of excess supply, even in the absence of new construction. The key idea is that a housing market is considered having evidence of excess inventories when supply of new unsold units or the rental market vacancy rate are elevated. In such a context, downward pressures on house prices would occur until excess supply is absorbed.
To assess the possibility of excess inventories conditions in the housing market, the framework uses two indicators that relate to the supply of readily available housing units: the rental vacancy rate, and the inventory of completed and unsold homeowner and condominium housing units per 10,000 population.17
The excess inventories indicators are deemed unusually high if their value lies within the highest decile of their historical values.
We detect moderate evidence of excess inventories if one of the indicators (the rental vacancy rate and/or the number of completed and unsold units relative to the population) is above its threshold for at least two quarters over the last four quarters preceding the assessment.18 We detect strong evidence of excess inventories if both indicators are above their thresholds for at least two quarters over the last four quarters.
- Multiple Listing Service® (MLS®) is a registered trademark owned by the Canadian Real Estate Association.
- QPAREB by the Centris® system
- Taking the Canadian MLS® market as a whole, a sales-to-new-listings ratio below 40 per cent has historically accompanied prices that are rising at a rate that is less than overall inflation, a situation known as a buyers’ market. A sales-to-new-listings ratio above 55 per cent is associated with a sellers’ market. In a sellers’ market, home prices generally rise more rapidly than overall inflation. When the sales-to-new-listings ratio is between these thresholds, the market is said to be balanced.
- The integration of the Teranet-National Bank House Price Index™ and New Housing Price Index (NHPI) as indicators for acceleration remains preliminary at this point, so only the MLS average price measure is currently used to assess acceleration in house prices.
- This refers to the Fall Rental Market Survey average national apartment vacancy rate on the primary purpose-built rental market in structures of 3 units or more.
- The level of inventories discussed here is for urban centres with a population of 50,000 and over. The inventory of housing units is defined as a snapshot of the level of completed and unsold (or unabsorbed) units at a specific time. A unit is defined as “absorbed” when an agreement is made to buy the dwelling.
- The lengths of time periods that are used to evaluate the persistence of signals are based on our analysis of CMHC’s mortgage insurance claims rate. Specifically, we tested the variables underlying our factors (for example, the sales-to-new listings ratio), and the factors using historical data on the claims rate to see if the variables underlying our factors and/or the factors were statistically significant determinants of changes in the claims rate. This analysis showed that combinations of factors were most significant.
- Real house price refers to observed (or nominal) house prices adjusted for CPI inflation.
- See Phillips, Wu and Yu (2008) “Explosive Behaviour in the 1990s NASDAQ: When Did Exuberance Escalate Asset Values?” for further details on the methodology.
- Formally, to detect sustained acceleration in house prices, we use the augmented Dickey-Fuller (ADF) test. The null hypothesis is for a unit root against the alternative of an explosive root (the right-tail). The ADF test rolls over a 3-year period, it moves through time one quarter at a time.
- Econometric models of house prices are estimated using cointegration techniques such as Engle-Granger and Dynamic Seemingly Unrelated Regression (DSUR) estimation.
- Our overvaluation assessment uses exclusively the models that are statistically reliable according to the selected econometric methodology.
- The framework of the demand model was recently used by Gallin (2006) and the IMF for Canada (see Tsounta ). The demand model was originally developed by Muth (1960) and now appears in introductory textbooks such as Mankiw (2001).
- See Kahn (2008) and Mayer and Somerville (2000) for more details.
- See Kiyotaki and Moore (1997) and McQuinn and O’Reilly (2008) for more details.
- The lengths of time periods that are used to evaluate the persistence of signals are based on CMHC’s analysis of mortgage insurance claims rate. See footnote 7 for further details.
- The inventory of completed and unsold units for the national level corresponds to the sum of the inventory in Census Agglomerations and in Census Metropolitan Areas.
- The lengths of time periods that are used to evaluate the persistence of signals are based on CMHC’s analysis of mortgage insurance claims rate. See footnote 7 for further details.