The Swedish authorities and international organisations often claim that household debt poses a threat to financial and macroeconomic stability, with reference to a high debt-to-income ratio. This column argues that the debt-to-income ratio is a poor predictor of future crises, and that one should look at aggregate and individual data on household solvency (debt relative to assets) and liquidity (debt-service relative to debt-service capacity) instead. Because solvency and liquidity levels are good, Swedish household debt is not a threat to financial and macroeconomic stability.
Swedish authorities and international organisations such as the European Commission and the European Systemic Risk Board (ESRB) commenting on Swedish economic policy regularly claim that Swedish household debt is too high and poses a threat to financial and macroeconomic stability. For example, the Riksbank stated in a 2019 press release that "households' high debt levels continue to pose the greatest risk to the Swedish economy" (Sveriges Riksbank 2019).
These warnings seldom specify what the threat is. They mostly refer to the debt-to-income ratio (DTI), which looks threatening because it has almost doubled from 95% in 1995 to 180% in 2024 (Figure 1).
Figure 1 Swedish household debt-to-income ratio (%)
Figure 1 also shows DTI ratio thresholds for high risk of a banking and financial crisis of 121% and 98%, respectively, published by the European Commission and the ESRB (European Commission 2024, Bricongne et al. 2020, ESRB 2019, 2022). The current DTI ratio is about 1.5 and 1.8 times those thresholds.2
The thresholds seem to have lost touch with reality.
It is insufficient – and even misleading – to only look at debt and the debt-to-income ratio. Ten years of research have shown that the debt-to-income ratio is a notoriously unreliable and misleading indicator (Broadbent 2019, p. 13, Svensson 2025: Appendix B). Its level has no predictive power regarding future crises or declines in consumption; it lacks scientific support; and looking only at debt is inadequate. As I discuss in a new paper (Svensson (2025b), one must examine the entire balance sheet, including assets and net wealth, as well as the debt service relative to incomes, with both aggregate and individual data.
Three main conditions under which stability is threatened
On closer examination, household debt may pose a threat to financial and macroeconomic stability under three main conditions:
Household debt levels become too high relative to assets – i.e. household solvency (assets relative to debt) becomes too low.
Household debt service (interest and amortisation payments) becomes too high relative to incomes and payment capacity – i.e. household liquidity becomes too low.
Rising housing prices induce households to take out additional mortgages (home equity withdrawals) for mortgage-financed overconsumption of macroeconomic significance. If prices level off or fall, home equity withdrawals stop and consumption decreases sharply. This occurred in Sweden during the 1990s crisis and in Denmark, the UK, and the US during the Great Financial Crisis (Andersen et al. 2016, Broadbent 2019, Mian and Sufi 2018, Svensson 2021b).
Approximately 80% of household debt consists of mortgages, while 6% consists of consumer loans. While consumer loans and over-indebtedness related to these cause significant individual and social problems, they do not threaten financial or macroeconomic stability. The focus here is therefore on mortgages.
It is important not only to analyse newmortgages and borrowers (as is typically the case in Sweden) but also the total mortgage stock and borrowers (following pioneering work by Boije and Hansen 2024). The total mortgage stock and the number of borrowers are much larger and matter much more for financial and macroeconomic stability. The annual flow of new borrowers with loans intended for home purchases is approximately only 9% of the borrowers of the mortgage stock.
Two structural conditions reduce the risks
Two structural conditions reduce the risks associated with household mortgages.
First, mortgages are, in fact, a secure cash cow for banks. They contribute to financial stability rather than threatening it. Mortgages are full recourse. Sweden's enforcement agency (Kronofogden) is effective. Swedish mortgage holders prioritise debt service and avoid bankruptcy. They purchase homes to live in, not to let or to speculate on capital gains. They stay in their homes for an average of approximately 24 years after purchase.
Credit losses on mortgages are therefore negligible. Banks – in a cozy oligopoly – earn secure profits during both economic booms and recessions from the margin between mortgage interest rates and the low rates on covered housing bonds, which have the highest credit rating.
Second, mortgage rates are not determined by independent external shocks but are indirectly determined by the Riksbank's policy rate. The Riksbank has tools that, if necessary, can be (and have been) used to stabilise the margin between mortgage rates and the policy rate. Under flexible inflation targeting, the Riksbank sets the policy rate to maintain macroeconomic stability and contribute to financial stability.
During an economic downturn, the Riksbank typically lowers interest rates, reducing mortgagors' interest expenses and making it easier for them to maintain consumption – a kind of insurance against economic downturns.
When supply shocks raise inflation, as they did in 2022 and 2023, the Riksbank's interest rate hikes trade off lower inflation against lower employment, incomes, consumption, and aggregate demand, subject to the latter not threatening financial and macroeconomic stability.
Solvency levels are good
Regarding condition (1) above and household solvency, households' total assets are significantly larger and have grown much faster than their debt. Net wealth (assets less debt) was twice the size of debt in 1985 and five times the size of debt in 2024. Figure 2 shows assets, debt, and net wealth relative to disposable income. The light-blue line at the bottom is the same DTI ratio as in Figure 1. It looks less threatening here.
Figure 2 Ratios to income of household assets, net wealth, and debt from 1980, and growth rates from 1985 (%)
Source: Svensson (2025b: Figure 4.2). Note: DI refers to disposable income.3
Households’ human capital is large and adds to the picture. Household owner-occupiers, unlike commercial and rental property firms, have a substantial earned income in addition to the (imputed) income from their property. The human capital, the earnings capacity, can be measured by the present value of future after-tax earnings. It is a large off-balance sheet item for households. Figure 3 replicates Figure 2 with human capital included.
Figure 3 Ratios to income of household assets (including human capital), net wealth, and debt from 1980 (%)
Source: Svensson (2025b, Figure 4.6). Note: DI refers to disposable income.4
But the above discussion only refers to aggregate data. So do most warnings about household debt from Swedish authorities and international organisations. Individual data provide more precise information, including about the distribution of assets and debt.
Unfortunately, there are no recent Swedish registry data on individual households’ assets and liabilities. But the Finansinspektionen (FI, the Swedish financial supervisory authority) collects individual data on loan-to-value ratios for new borrowers and all borrowers (i.e. borrowers of the total mortgage stock). The data on new borrowers are published in FI’s annual report on the Swedish mortgage market. The data on all borrowers are not published regularly, despite being more relevant for macroeconomic and financial stability. The annual flow of new borrowers is approximately only about 9% of all borrowers (Svensson 2025b: Section 4.7), so the latter matter much more for financial and macroeconomic stability.
From now on, "new borrowers" refers to new borrowers with loans intended for home purchases – thus excluding home equity withdrawals, treated separately regarding condition (3) – and "borrowers" refers to borrowers of the total mortgage stock.
Figure 4 shows the cumulative distribution of loan-to-value (LTV) ratios among new borrowers, borrowers, owner-occupiers, and all households. The distribution among new borrowers clearly shows the thresholds for the FI’s first amortisation requirement, 50% and 70%, as well as the LTV cap of 85%.5
We also see that 35% of owner-occupiers and 60% of all households have no mortgage.
Figure 4 The cumulative distribution of loan-to-value ratios among new borrowers and among borrowers, owner occupiers, and all households, 2023 (%)
The horizontal difference between the distributions shows that borrowers, owner-occupiers, and all households have much lower LTV ratios (and thus higher home equity-to-value ratios) than new borrowers. For new borrowers, the median LTV ratio is 76%, whereas for borrowers it is only 49% and for owner-occupiers 28%.
A full 78% of borrowers have loan-to-value ratios below 70%, meaning they have home equity exceeding 30%. This percentage exceeds any housing price decline during the past 50 years.
A good solvency assessment requires an assessment of whether assets are reasonably valued, including whether housing is overvalued and there is a risk of a substantial fall in house prices. This would shift the distributions of LTV values in Figure 4 to the right. But Swedish owner-occupied housing is not overvalued. Although the Commission and the ESRB, with arguably misleading indicators, have stated that Swedish owner-occupied housing has been overvalued by 30–50%, in Svensson (2025a) I show – with better indicators – that it has not been overvalued but rather substantially undervalued.7
It follows that not only aggregate solvency levels but also the distribution of solvency levels across mortgage borrowers in Sweden is good. The distribution of solvency levels among all owner-occupiers is even better.
Altogether, household solvency levels are good, and condition (1) is not present. Household solvency is not a threat to financial and macroeconomic stability.
Liquidity levels are good
Regarding condition (2) and household (mortgagors’) liquidity, the new borrowers, with higher LTV ratios, may find debt service (i.e. interest and amortisation payments) to be a significant burden, especially given the FI’s amortisation requirements (Svensson 2020, 2025b). But these borrowers have passed banks' credit assessments, including stress tests with mortgage rates as high as 6–7% (FI 2024, 2025).
Figure 5 shows the cumulative distribution of debt service-to-income (DSTI) ratios for new borrowers, borrowers, owner-occupiers, and all households.
Figure 5 The cumulative distribution of the debt service-to-income ratios among new borrowers and among borrowers, owner-occupiers, and all households, 2023 (%)
Source: Svensson (2025b: Figure 5.16). Note: A mortgage rate of 4% is used.8
The horizontal difference between the cumulative distributions of borrowers and new borrowers shows that borrowers have debt service-to-income (DSTI) ratios about 5 percentage points lower than new borrowers.
At the 75th percentile, new borrowers have a DSTI ratio of 21%, whereas borrowers only have a ratio of 16%. Thus, 75% of borrowers have a DSTI ratio lower than 16% – a rather low ratio. At the median, new borrowers have a DSTI ratio of 15%, whereas borrowers only have a ratio of 10.5%. Owner-occupiers have DSTI ratios an additional 4–6 percentage point slower. These ratios are low relative to the expenditure-to-income ratio of 30% below which the OECD (2021) considers housing to be affordable.9
As mentioned above, the Riksbank sets the policy rate – and indirectly the mortgage rate – to maintain macroeconomic stability and contribute to financial stability. It would not set a policy rate that threatened macroeconomic and financial stability.
Altogether, household liquidity levels are good, and condition (2) is not present.
There is no indication of HEW-financed overconsumption of macroeconomic significance
Regarding condition (3), mortgage-financed overconsumption of macroeconomic significance would appear as large home equity withdrawals, low saving rates, and a high durable-goods share of total consumption.
But, according to the FI’s mortgage reports, home equity withdrawals are not abnormally high (figure 6). Survey data indicate that withdrawals are little used for non-housing expenses.10
Figure 6 New mortgages by intended purpose, 2017–2024 (number of households)
This contrasts with the situation before the 1990s banking crises. Then, the economy was overheating, saving rates were strongly negative, and the share of durable goods was far above its historical average, indicating substantial mortgage-financed overconsumption. The large increase in saving rates in the beginning of the 1990s reflects consumption falling proportionately more than income during the crisis.
Thus, there is at present no indication of any mortgage-financed overconsumption of macroeconomic significance. Condition (3) is not present.
Conclusions
The three proposed conditions under which household debt may pose a threat to financial and macroeconomic stability are low solvency, low liquidity, and overconsumption of macroeconomic significance financed by home equity withdrawals. None of these conditions is present in Sweden. Swedish household debt is neither too high nor a threat to financial and macroeconomic stability. The high debt-to-income ratio is a misleading indicator of such threats.
To assess stability risks from household debt, one should examine whether these three conditions are present. Individual data allow us to assess their distribution among households. Regarding solvency and liquidity, figures such as Figures 2, 4 and 5 above should arguably be standard in financial stability reports from national authorities and international organisations, including a focus on borrowers of the total mortgage stock and not just new borrowers. Regarding mortgage-financed overconsumption, time-series data such as displayed in Figure 6 (including the purpose of howe equity withdrawals, if available) and figures such as Figures 7 and 8 are informative.
FI - Finansinspektionen (2024a), "Fördjupad analys av bolån – 2023 [In-Depth Analysis of Mortgages – 2023]," data collected in the FI’s annual mortgage survey.
Statistics Sweden (see Svensson, 2025b, for all detailed references to Statistics Sweden data). The high-risk thresholds of the European Commission and the ESRB refer to thresholds for a high risk of a banking and financial crisis (European Commission, 2024; Bricongne et al., 2020; ESRB, 2019, 2022).
Swedish authorities and international organisations also regularly state that Swedish household debt stands out internationally among comparable countries (Sveriges Riksbank 2019). But Svensson (2025b, section 3) shows that this is hardly true. Luxembourg, Korea, Denmark, Netherlands, Australia, Switzerland, and Norway all had higher DTI ratios than Sweden in 2022; Canada had only a slightly lower DTI ratio. Sweden does not even stand out among the Nordic countries. Furthermore, countries differ strongly in many relevant structural factors, making superficial international comparisons of household indebtedness uninformative.
Statistics Sweden and own calculations. Income is disposable income. Financial assets exclude tenant-owned apartments (which are considered financial assets by Statistics Sweden). Housing assets include owner-occupied single-family houses, tenant-owned apartments, and seasonal and secondary homes. Total assets are the sum of housing and financial assets. Debt is exclusive of farm mortgages from 2001q4. Quarterly data from 2001q4 or 2003q1.
Human capital is the present value of 30 years of earned disposable income, which is set to 60% of disposable income (Statistics Sweden 2017, Ekonomifakta 2024). The discount and growth rates are assumed to be the same, so the present value is just the sum of 30 years of earned income. Debt is exclusive of farm mortgages from 2001q4.
FI’s first amortisation requirement, introduced in 2016, is a minimum 1% amortisation for LTV ratios above 50% but not above 70% and a minimum 2% amortisation for LTV rates above 70%. The second amortisation requirement, introduced in 2018, is an additional 1% amortisation for loan-to-gross income (LTGI) ratios above 450%. An LTV cap of 85% was introduced in 2010.
The dark-blue histogram for new borrowers: Data for FI (2024b) and own calculations. For a scatterplot of the data, see Svensson (2025, Figure 5.3). The distributions among borrowers, owner-occupiers, and all households: Boije and Hansen (2024), FI (2024a), FI (2024b: Diagram 15), Statistics Sweden, and own calculations.
See Svensson (2025a) for details of and references to the Commission’s and ESRB’s arguments, which rely on price-to-income (PTI) ratios being well above historical averages. Svensson (2025a) shows that the relevant price in this context is not the value of the dwelling but the price of the housing services that the dwelling delivers, the user cost. User-cost-to-income (UCTI) ratios have fallen well below historical averages. Swedish owner-occupied housing is expensive to buy but cheap to live in.
A rate of 4% was the peak in 2024 of average outstanding mortgage rates during the recent hike cycle. In August 2025, average outstanding mortgage were down to about 3%. The distributions among borrowers, owner-occupiers, and all household are only estimated for LTGI rates up to 450%. This is the reason why the cumulative distribution is incomplete and only reaches the 94th percentile for borrowers. The remaining 6% of the borrowers have LTGI ratios above 450% and are subject to the FI’s second amortization requirement of an additional 1% amortization, but they do not affect the cumulative distribution below the 94th percentile. Income is net income, that is, after-tax gross income. See Svensson (2025b) for details. There, instead of DSTI, the notation DSTNI is used for debt service to income, to emphasise that it refers to net income.
OECD (2021) suggests expenditure-to-income ratios, for example, with a threshold of 30% of gross income, below which housing is considered "affordable"'. A related measure is the housing overburden rate, which captures the share of households spending an unacceptably large share of income on housing (that is, above a given threshold); both Eurostat and the OECD set the overburden threshold at 40% of household disposable income (net of housing allowances). Ses Svensson (2025a: Section 5.3.1) for further discussion.
Boije and Hansen (2024: 35) asked 1,100 homeowners with mortgages the following question in December 2023: "Have you, at any time during the past five years, borrowed against your home for purposes that are not housing-related (such as financing a car purchase or travel, etc.)?" Of the respondents, 93% answered "no" and 7% "yes".
Numbers refers to households that took out new mortgages in FI’s two week-long samples, one around 1 September and one around 1 October, for each year.
Statistics Sweden and own calculations. "Net lending" – in Sweden referred to as "financial saving," in direct translation from the Swedish – is the difference between net acquisition of financial assets and net incurrence of liabilities (OECD: 2024a). The saving rates are four-quarter trailing sums of saving divided by 4-quarter trailing sums of income.
Statistics Sweden and own calculations. The shares are four-quarter trailing sums of seasonally adjusted expenditure on durable goods divided by four-quarter trailing sums of total household consumption expenditure, seasonally adjusted at current prices. The blue and red lines use the classification of durable goods according to COICOP 1999 and 2018, respectively. The latter classification results in a somewhat lower share of goods classified as durable. On 30 May 2024, the Statistics Sweden National Accounts started to disseminate household consumption according to COICOP 2018 instead of 1999. Data according to COICOP 1999 are no longer updated by Statistics Sweden.
