The Housing Boom and Household Debt in Norway1
The Norwegian housing market was only moderately affected by the global financial crisis, and the rising trend of house prices resumed shortly after the crisis. In the meantime, household debt reached more than 200 percent of disposable income, and it is expected to grow further. This paper examines the characteristics of household debt and various factors driving the housing boom and debt accumulation, with a particular focus on institutional factors. The paper also examines the vulnerability stemming from the high level of household debt and the potential macroeconomic impact of a possible house price correction.
A. House Prices and Household Debt
1. Norway has seen a long housing boom (Figure 1). House prices have been rising since the early 2000s apart from a short reversal during the global financial crisis. The increasing trend of house prices continued in recent years until 2013 when house prices stabilized with no clear trend and credit to households continues to grow. Estimates suggest that house prices may be overvalued by about 25–60 percent, depending on measures of overvaluation.2, 3
Figure 1.House Prices and Household Balance Sheets
2. Household indebtedness has also risen significantly for the past decades. Household debt was about 140 percent of disposable income in 2002 and now it stands at about 220 percent of disposable income in 2014, which is among the highest in OECD countries. Household debt is heavily skewed toward housing, with about 85 percent in the form of mortgages from banks and mortgage companies.
3. Household assets are mostly illiquid. A large portion of household assets consists of housing wealth and pension assets. Liquid assets such as deposits and shares are relatively small at about 150 percent of disposable income. As a comparison, liquid assets in Sweden and Denmark are about 300 percent and 250 percent of disposable income, respectively (Figure 1).
4. Household net worth is positive but smaller than peer countries. One caveat, however, is that household balance sheets may not fully capture the fact that Norwegian households have large pension assets through the social security system and through occupational pension schemes. Household balance sheets data capture only voluntary individual pension savings, which is relatively modest in size compared with peer countries. On the other hand, social security pension liability was 5.7 trillion NOK, about 412 percent of disposable income in 2013.4 This suggests that Norwegian households’ assets are larger than what is found in data on household balance sheets in practice. Nevertheless, limited liquid buffers could make Norwegian households vulnerable to sharp house price corrections.
5. Household debt is concentrated more among higher income households (Figure 1). Median household debt tends to be higher among higher income families. However, the distribution of debt looks quite different across households depending on the age of main income earners; younger households tend to have more debt, and very high levels of household debt tend to be skewed more toward younger households. These households also tend to have smaller assets, particularly financial assets.5 Young households are thus likely to be more vulnerable to house price corrections or a sharp interest rate hike because they have limited liquid buffers with most of their assets held in illiquid housing. Household vulnerability will be examined further in Section C.
B. Institutional Factors Behind the Housing Boom and Household Indebtedness
6. Various factors can contribute to rising house prices and household indebtedness. These include demand factors such as income growth or population growth and low interest rates and supply factors such as shortage of housing. Other factors are more institutional or structural, such as the size of rental market or rental market regulations and tax incentives including interest deductibility of residential mortgages (IMF, 2015).
7. In the Norwegian context, both demand and supply factors have contributed to the housing boom and rising household debt. Demand factors include high wage/income growth reflecting the oil sector boom, population growth due to the large influx of immigrants, and low interest rates in recent years. Supply factors include constraints due to regulations on land use, minimum unit size, and other construction standards (IMF, 2013). According to an OECD estimate, Norway has a relatively low price responsiveness of housing supply, possibly reflecting both natural land constraints and regulations; the long-run price elasticity of new housing supply is estimated at about 0.5, compared to an OECD average of 0.7.6
8. Structural or institutional factors are also important in explaining household indebtedness in Norway.
- The rental market in Norway is relatively unregulated but limited in size. Norway’s private and public rental combined accounts for about 23 percent of the total dwelling stock, compared to an average of 38 percent for other Nordic countries (Denmark, Finland, and Sweden). Thus, individuals tend to enter the owner-occupied housing market and take mortgages at a relatively younger age.
- The tax system also plays a role as owner-occupied properties receive preferential tax treatment relative to other investment. Home ownership is generally encouraged by tax incentives:
- Mortgage interests payments are fully tax deductible (at a tax rate of 27 percent).
- The imputed rent from home ownership is tax exempt, and a homeowner can rent out part of a property tax free.
- Housing is subject to a lower wealth tax than other assets (25 percent of market value of primary dwellings and 70 percent of market value of secondary dwellings).
- There is no capital gains tax if a house has been owned for more than a year and the owner has used it as their own home for at least 12 out of the last 24 months.
- Saving for house purchase is encouraged by tax deductions on a savings scheme for house purchase by persons under 34 years of age. Under this savings scheme, one can save maximum NOK 200,000 in total and NOK 25,000 per year. 20 percent of what is saved during a year is tax deductable. Savings under this scheme need to be used for purchasing a house or repaying mortgage debt. Otherwise tax advantages will be reversed.
Indeed, Norway ranks among the OECD countries with the highest degrees of tax relief on debt financing cost of homeownership, according to an indicator constructed by the OECD.7
- Supply constraints are likely to push up house prices especially when house completions are running behind the growing number of households. This will create a need to take up a larger mortgage than otherwise. It is likely that supply constraints are partly due to strict planning restrictions that are imposed on new house building.
- The prevalence of variable rate mortgages may also contribute to encouraging household debt accumulation, by reducing interest payments in a low interest rate environment. About 95 percent of mortgages in Norway have variable rates.
Tax Relief for Housing Finance
Sources: OECD and Fund staff calculations.
Sources: OECD and Fund staff calculations.
9. A regression analysis confirms that these factors are important in explaining credit booms in a cross-country context. To analyze the buildup in household leverage, a panel probit model to predict the probability of a household credit boom was estimated using data on 24 OECD countries for 1970Q1–2013Q4.8 Explanatory variables include house prices, mortgage characteristics (e.g., interest type and tax relief for housing finance), financial system characteristics (e.g., loan-to-deposit ratio, pension assets as a share of disposable income), and macroeconomic and financial factors (e.g., unemployment rate, working age population growth, and the short-term interest rate). Regression results suggest that these variables contribute to explaining the occurrence of household credit booms, albeit to varying extents. For example, other things constant, the presence of a house price boom is estimated to increase the probability of a household credit boom by 27 percentage points, whereas countries with predominantly variable mortgage rates are likely to face higher probability of a credit boom by 13 percentage points compared to those with mainly fixed rates.
10. The estimated model suggests that Norway possesses several factors that make a household credit boom more likely to occur. For example, using data for 2010–13, rapidly rising house prices in Norway, by raising the value of collateral available to households, would add about 6 percentage points to the probability of a credit boom relative to average OECD level. Switching to a mortgage system with predominantly variable rates such as one in Norway would add another 8 percentage points to the estimated probability. Norway’s more generous tax system regarding housing finance relative to OECD average adds another 3 percentage points, and the relatively low unemployment rate contributes another percentage point to the credit boom probability.
|Average advanced OECD, 2010-13||0.23||0.03|
|+ NOR probability of house price boom||0.29||0.04|
|+ NOR interest rate type (mainly variable)||0.37||0.05|
|+ NOR tax relief for housing finance||0.40||0.05|
|+ NOR change in unemployment rate||0.41||0.05|
C. Household Vulnerability
11. Despite high debt levels in recent years, Norwegian households do not appear to face significant payment capacity problems under current conditions. Non-performing loans are near historical lows. A Norges Bank study using the 2012 household-level data found that only about 2 percent of household debt is “more vulnerable,” and the proportion of vulnerable households is only about 1 percent.9 These results are based on household debt meeting three risk criteria: (i) debt above five times disposable income; (ii) financial margin (income minus taxes, interest and ordinary living expenses) below one month of annual after-tax income10; and (iii) net debt (debt minus deposits) larger than the value of dwelling. Household debt is considered risky if these indicators are higher than the threshold.
12. However, households appear vulnerable to interest rate hikes. The FSA has analyzed the impact of an interest rate hike on households’ interest burden using Statistics Norway’s micro simulation model based on household-level data.11 The interest burden is defined as interest expenses as a share of income after tax. The analysis finds that households are sensitive to interest rate hikes: an increase of the lending rate by 2 percentage points would double the proportion of households with an interest burden between 20 and 30 percent, from 5.5 percent to 12 percent. At the same time, the proportion of households with an interest burden above 30 percent would more than double from 2.5 percent to 7 percent.
13. To gain insights into households’ vulnerabilities to a change in economic conditions, the Norges Bank approach was expanded to include a set of shocks.12 These shocks are included separately and combined: (i) lending rate increase of 2 percentage points; (ii) real house price drop by 40 percent;13 and (iii) income drop of 20 percent.14 The results are summarized as follows:
Box 1.Sensitivity Analysis
The Norges Bank’s approach applies three criteria. One of the criteria based on “margin” is calculated as income minus taxes, interest, and ordinary living expenses. Ordinary living expenses are based on the Standard Budget complied by National Institute for Consumer Research, which is a standardized estimate of the standard cost of consumption and does not necessarily reflect individual circumstances of households. This measure thus entails some uncertainty while two other risk measures are directly observable. Ordinary living expenses are likely to vary depending on household income levels, and the estimate may also not be capturing other necessary expenses.
To take account of potential overestimation of margin, different definitions of margins are applied. Margins below two months and three months of income were also applied as a sensitivity analysis, and baseline results were recalculated. Using the three combined criteria, the share of vulnerable household debt increases from 2 percent under the baseline with one-month margin to 3.2 percent and 4.4 percent if 2 month-margin and 3 month-margin are applied, respectively. These are still relatively small share, but different assumptions seem to affect different age/income groups somewhat differently. The share of vulnerable debt tends to be higher among lower income and younger households, and the share is 1–4 percentage points higher if 3 month-margin is used instead of 1 month-margin. This variability tends to be more pronounced among income groups 3–5 and age groups between 25–34 years old.
The results for the first decile (D1) and the youngest households (younger than 24 years old) need to be interpreted with caution. Households in the lowest income decile include households that engage in tax planning. Their reported income may be low, but they are likely to have high debt servicing capacity. The age group below 24 tends to hold student loans as part of their debt and has more flexible options regarding principal payments if their income is low. Importantly, both groups hold a very small share of total household debt.
The Share of Vulnerable Household Debt by Income Decile
Sources: Norges Bank and Fund Staff calculations.
The Share of Vulnerable Household Debt by Age
Sources: Norges Bank and Fund staff calculations.
|Interest rate increase||4.6||1.9|
|House price fall||5.5||2.5|
|Fall in income||8.1||3.4|
|Combination of the three shocks||21||8.6|
- The share of vulnerable debt rises from 2 percent in the baseline to about 5 percent, 6 percent, and 8 percent, under each separate scenario, respectively. However, the proportion of vulnerable households rises but remains relatively low, below 3 percent in all three individual scenarios. On the other hand, under the severe scenario of combined shocks, the share of vulnerable debt increases to 21 percent, and the proportion of vulnerable households also rises to about 9 percent.
- The impact varies across different income deciles and age groups. In particular, lower income and younger households are disproportionally more affected by the three combined shocks. For example, roughly 30 percent of the debt held households aged 25–34 years are vulnerable under the severe scenario of combined shocks.15
The Share of Vulnerable Household Debt by Age
Sources: Norges Bank and Fund staff calculations.
14. The aggregate number masks distributional effects. The proportion of vulnerable households remains below 10 percent under the severe scenario, but this share is larger for certain income or age groups. The exercise above thus suggests that household vulnerability could rise under severe stress scenarios, and these effects will be felt unevenly across different income and age groups.
D. Macroeconomic Impact of a House Price Correction
15. What are the possible aggregate impacts of a house price correction? Given that a significant part of Norway’s household debt lies in housing, it is natural to ask how house price corrections in the past have affected household consumption and residential investment. Theory postulates that changes in house prices can have an effect on individual consumption through their impact on household wealth and access to finance via relaxation/tightening of collateral constraints. The household-level analysis in the previous section shows that more households in Norway could become “vulnerable” in the event of a negative shock to house prices; however, whether that translates into more defaults or a cutback in consumption and investment is an empirical question.
House Price Cycles in the Nordics, 1970Q1-2014Q4
Sources: OECD and Fund staff calculations.
16. It is useful to examine past experiences with house price corrections in Norway and other Nordic countries. Since the mid 1980s, the Nordic countries have undergone two major episodes of house price collapse, one around the banking crises of the mid-to late-1980s and one during the 2008–09 global financial crisis. In all four countries, the decline in house prices in cumulative real terms tends to be much larger and more persistent during the first episode, whereas the house price corrections that happened during the most recent crisis seem relatively milder and less long-lasting (perhaps with the exception of Denmark, which had experienced a much more pronounced housing boom previously). In Norway, for example, real house prices declined by a total of 67 percent over almost six years in the late-1980s crisis, compared to only 13 percent over five quarters during the recent global crisis. Better macroeconomic management and bank regulation may have played a role in containing house price overvaluation during the boom and supporting the recovery of the housing market after the bust, resulting in less severe price corrections. However, the decline in real private consumption for every percent of house price decline has increased in the recent episode, possibly reflecting much more developed mortgage markets and thus closer links between the housing sector and the real economy.
|Peak-trough of HP cycle||–||79Q1-85Q3||79Q2-82Q3||73Q4-79Q1|
|Peak-trough of HP cycle||87Q2-93Q1||90Q1-93Q3||86Q2-93Q2||89Q2-93Q2|
|Peak-trough of HP cycle||07Q3-08Q4||07Q4-09Q1||07Q1-12Q2||07Q3-09Q1|
17. A Vector Auto-Regression (VAR) framework is used to assess the average impact of a house price shock. A model is estimated separately for each outcome variable: GDP, private consumption, and residential investment. Following Igan and Loungani (2012), the VAR includes three other variables: CPI, short-term interest rate, and house prices. House price shocks are identified through a Cholesky decomposition of the variance-covariance matrix.16 The model is estimated using quarterly data for all four Nordic countries. Data for Norway span 1986Q1–2014Q4, and thus capturing both historical house price correction episodes discussed above. Estimation results suggest that for Norway, a 10 percent decline in house prices would lead household consumption to fall by almost one percent on impact. The effect is relatively short-lived, reaching cumulative 1.5 percent after two quarters before rebounding. The impact on residential investment is estimated to be larger, lasting about a year with a cumulative decline of 17 percent. Thus, if the past is any indication of the future, the aggregate consumption impact of a potential house price turnaround in Norway would not be severe, perhaps due to the availability of other household assets that can be drawn to avoid a drastic cutback in consumption. Nevertheless, the effect could be large for younger families with more debt and fewer assets, which is not captured in this aggregate analysis.
Response of Private Consumption to a 10 Percent Decline in House Prices
Source: Fund staff calculations.
Response of Residential Investment to a 10 Percent Decline in House Prices
Source: Fund staff calculations.
18. The estimated macroeconomic impact of a house price correction appears milder for Norway compared to Nordic peers. For example, in response to a 10 percent drop in house prices, the maximum decline in private consumption in a given quarter is estimated at 0.9 percent in Norway, compared to 2.1 percent in Sweden and 3½ percent in Denmark. The cross-country heterogeneity in responses is even more pronounced in the case of residential investment. Previous literature (e.g., IMF, 2008; Cardarelli and others, 2009) suggests that mortgage market characteristics defining the ease of access to credit (e.g., typical loan-to-value ratio, availability of mortgage equity withdrawal to finance consumption, prevalence of variable vs. fixed rate mortgages) could be among factors explaining why economic activity in some countries may be more vulnerable to declining house prices than in others.
19. High household debt is an underlying vulnerability for Norway. Risks seem contained so far, but this could change if economic conditions deteriorate significantly. The authorities have implemented several measures, including stricter bank capital requirements in line with Basel III/CRD IV (ahead of schedule) and tightening parameters for risk weights on mortgage lending of IRB models. More recently, the FSA has proposed tighter underlying mortgage loan standards, including a requirement to amortize mortgages, applying a higher stress level for interest rates in assessing borrower’s repayment capacity, and reduction in banks’ scope for deviating from these requirements.
20. A holistic approach will be needed to address risks associated with high house prices and household debt. Macroprudential policy measures play a key role to contain financial stability risks, but other supporting measures will be needed to address the issue more fundamentally. Possible options include the following:
- Reducing tax preferences for owner-occupied housing and mortgage debt: There are several tax incentives that encourage home ownership in Norway including full deductibility of mortgage interest as discussed earlier. Given the current low interest rate environment, which limits the effective benefit of interest deductibility, now seems to be a good time to start reducing mortgage interest deductibility. More recently, the Tax Commission has recommended shifting the tax burden toward indirect taxes and advocated less preferential tax treatment for residential housing relative to other assets. Its recommendations include removing the home savings scheme for young people, repealing the tax exemption on rental income up to 50 percent of the market value of private residences, and increasing the valuation of properties. These measures could be phased in gradually.
- Easing supply constraints: Planning and building requirements could be relaxed to stimulate the supply of new housing units. The Ministry of Local Government and Modernisation is working on various measures to simplify regulations on planning and building matters, with a goal to help keep construction costs down and to increase the pace of planning processes.
- Development of rental market: As noted earlier, the rental market in Norway is relatively small. This is likely to be, at least partially, the result of existing tax incentives that encourage home ownership. The role of rental market could be revisited in light of the growing number of immigrants whose housing need may differ from native Norwegians. Availability of rental housing may also facilitate labor mobility across cities as the Norwegian economy transitions to less oil dependent growth model and goes through structural changes.
Arnold, N., J.Chen, F.Columba, G.Ho, and B.Mircheva,2015, “The Nordic Financial Sector, Household Debt, and the Housing Market,” forthcoming.
Caldera Sanchez, A. and A.Johansson,2011, “The Price Responsiveness of Housing Supply in OECD Countries, OECD Economics Department Working Paper.
Cardarelli, R., T.Monacelli, A.Rebucci, and L.Sala,2009, “Housing Finance, Housing Shocks, and the Business Cycle: Evidence from OECD Countries,” IADB Working Paper.
Igan, D. and P.Loungani,2012, “Global Housing Cycles,” IMF Working Paper 12/217.
International Monetary Fund, 2008, “The Changing Housing Cycle and the Implications for Monetary Policy,” World Economic Outlook Chapter 3 (April).
International Monetary Fund, Norway 2013 Selected Issues, IMF Country Report No. 13/273, 2013.
International Monetary Fund, Multi-Country Report, Housing Recoveries: Cluster Report on Denmark, Ireland, Kingdom of the Netherlands – the Netherlands, and Spain, IMF Country Report No. 15/1, 2015.
Solheim, H. and B. H.Vatne, “Measures of Household Credit Risk,” Economic commentaries, 8/2013.
Lindquist, M., M. D.Riiser, H.Solheim, and B. H.Vatne, “Ten Years of Household Micro Data: What Have We Learned?” Staff Memo 8/2014.
Prepared by Giang Ho and Kazuko Shirono.
These estimates are calculated from deviations in price-to-income ratio and price-to-rent ratio, and also based on a model used in the early warning exercise. See IMF (2013) for more details on the methodology.
The price-to-rent ratio is often used to gauge house price misalignment, but rent series tend to imperfectly capture rent developments in practice. In Norway, the rent series is thought to capture mostly the rent developments of existing rental contracts, which tend to move at the rate of CPI inflation due to regulations. This tendency could lead to an overestimation of house price gaps using the price-to-rent ratio. In addition, the rental market in Norway, being relatively small and very different from the owner-occupied market, provides limited substitutes for the owneroccupied housing market, which makes the price-to-rent ratio an imperfect measure of house price valuation particularly for Norway.
The National Budget 2015.
Non-financial assets from the household level data are significantly larger than non-financial assets reported in aggregate OECD data. The difference is due to different valuations of real assets. Household level data use market values to evaluate real assets.
The indicator takes into account if interest payments on mortgage debt are deductible from taxable income and if there are any limits on the allowed period of deduction or the deductible amount, and if tax credits for loans are available.
See Arnold and others (forthcoming).
See Finanstilsynet, Risk Outlook 2014, and Risk Outlook 2013.
Norges Banks also conducts various sensitivity analyses using this framework. See, for example, Lindquist et al. (2014).
Deposit rates are also assumed to rise by 3 percent at the same time.
Parameters are calibrated based on similar shocks assumed in the bank stress test in the 2015 Norway FSAP exercise.
These results need to be interpreted with caution. About 40 percent of the debt held by household younger than 24 years old is vulnerable under the combined shock scenario, but these households hold less than 3.5 percent of total household debt, and close to 30 percent of their debt is student loans, which are interest free as long as borrowers remain in school. On the other hand, households aged 25-34 years hold more than 20 percent of total household debt. Thus the vulnerability of the latter group of households is likely to be more significant than the very young households.
In particular, the ordering of variables in the VAR is as listed above. An assumption is that macroeconomic variables are affected by monetary policy only with a lag while monetary policy responds contemporaneously to changes in all variables in the system. The house price variable enters last, allowing house prices to respond instantly to macroeconomic variables and monetary policy.