New Ekonomistas post. This is an English translation.
The Riksbank has, in a new Economic Commentary about household debt, presented with a press conference and a speech by executive board member Cecilia Skingsley, announced what seems to be dramatic conclusions and a big news item, which also been much mentioned in the media:
- The problems associated with debt would be both more extensive and more comprehensive than we have previously realized.
- Households would in general be highly indebted in relation to their incomes, above all low and middle-income earners.
However, the conclusions do not stand up to scrutiny. It appears that the media and general public have been misled.
How has the study arrived at its dramatic conclusions?
Well, the study has consequently applied – but misinterpreted – something completely obvious, but something presented as a new insight:
- The aggregate debt ratio underestimates debts among borrowers.
Among households, there are both households with and without debt. The aggregate debt ratio (debt over disposable income) is the average of the debt ratio for both households with debt (a positive debt ratio) and households without debt (a zero debt ratio). The debt ratio for households with debt is obviously higher than the average for households both with and without debt. This is nothing new.
The results presented in the Riksbank’s, Economic Commentary is thus precisely what one would expect. It therefore seems a bit strange that the Riksbank would present them as something surprising and alarming, something that would differ from what is already known. That the average debt ratio for households with debt is about 300 percent, is neither more nor less alarming than that the average debt ratio for all households is about 170 percent. That the average debt ratio for households with debt and low incomes is about the same as the average debt ratio for households with debt and high incomes, is neither more nor less alarming than that the average debt ratio among all households with low incomes is much lower than the average debt ratio among all households with high incomes.
The share of households without debt varies with income and is higher for low incomes
The Riksbank’s dramatic news is thus a tautology and completely obvious, once one understands that the Riksbank simply – and misleadingly – neglects the households without debt. The share of households without debt, and especially that this share varies with income, is of course important for judging potential risks associated with household debt. The share of households without debt is low for low incomes, and high for high incomes. Trying, as the Riksbank does, to neglect this is misleading, and even looks like an attempt of disinformation. When households without debt are taken into account, the Riksbank’s conclusions do not follow.
The Riksbank’s study can be seen as a failed attempt to induce some doubts about the government report that Anna Hedborg presented last fall, SOU 2013:78 ”Överskuldsättning i kreditsamhället?” (“Over-indebtedness in the Credit Society”, unfortunately only available in Swedish, previously commented on by me in Swedish here). It presented and analyzed individual debt data, and the results of the report is of considerable relevance for the ongoing debate about household debt and potential risks associated with it. The results of the report confirmed the picture that previous less comprehensive data had given. The most important result for the debt debate is:
- The large loans are held by groups that can be expected to have the best requirements for managing large loans, such has those with high incomes and high education.
- The largest loans are highly concentrated to those with the highest incomes.
The Hedborg report presented several new and more detailed results, including the distribution of different-size loans across income deciles. Table 2.3a below is a variant of the report’s table 2.3, a variant that shows the distribution of loans within each income deciles.
We see that in the three bottom income groups, about 50 percent or more of the individuals are without debt, whereas in the three top income groups, only about 20 percent are without debt. Higher income deciles also have large loans to a larger extent.
The Hedborg report furthermore shows that the debt ratio for households (including both households with and without debt) is higher in high income deciles (figure 2.4a below, a variant of the report’s figure 2.4 with deciles instead of quintiles), except for the first income decile. The first decile is problematic and difficult to interpret, however, with some households with negative income and large debt (and probably a few well-off people with zero taxable income because of large deductions). It is therefore not shown.
The debt ratio for all households increases from just above 50 percent in decile 2 to almost 200 percent in decile 9. The Riksbank’s study reports that the debt ratio of the indebted households, that is, excluding those without debt, is higher, almost 250 percent, and about the same in the different deciles, except for the problematic decile 1 (chart 5 below).
It is of course obvious and – in spite of the fuss from the Riksbank – neither more nor less alarming that the debt ratio is higher for households with debt than that for all households, including those without debt.
What is more relevant, the debt ratio for those with debt only, or the aggregate debt ratio that includes also the households without debt? From an individual and from a consumer-protection perspective, a focus on those with debt only may be justified. But from a macro risk perspective, the share of households without debt is important, and the aggregate debt ratio would for most macro issues be more relevant.
In contrast, the Riksbank seems to try to maintain that it does not matter for potential risks associated with the indebtedness how large a share in each income decile is without any debt. Thus, it would not matter that 50 percent or more of the low income deciles are without debt, compared to 20 percent without debt of the high income deciles. The Riksbank is probably pretty alone with that view.
In summary, the Riksbank’s first conclusion, that the problems associated with debt seem to be both more extensive and more comprehensive than we have previously realized, does not stand up to scrutiny. It does not hold when households without debt are taken into account. Neither does the Riksbank’s second conclusion, that households in general are highly indebted in relation to their incomes, and that this applies above all to low and middle-income earners, stand up to scrutiny. As the Hedborg report shows, low- and middle-income earners are definitely less indebted than high-income earners, since a higher share are without debt and those with debt might have about the same debt ratio as in higher income deciles.
Furthermore, the debt ratio is not a good risk measure. Just mixing stock and flow measures is problematic. Other measures, such as the interest ratio, the real interest payment, the debt service that keeps the debt ratio constant, the LTV ratio, the net wealth-to-total assets ratio, repayment capacity, and resilience towards disturbances, are much more relevant. This is an issue worth coming back to.
Update: The debt ratio as a risk measure is discussed extensively in this later post: “The debt ratio is an unsuitable risk measure – there are much better ones.”
 The Riksbank’s debt ratios are computed as arithmetic unweighted averages. This is arguably misleading, and income-weighted debt ratios are more relevant, at least if the focus is on macro effects and risks. With unweighted averages borrowers with small debt and even less income have a big impact on the average. This does not make much sense if the focus is on macro effects.
Unweighted debt ratios are normally larger than weighted, and they can be much larger, as the following example shows:
Assume 2 borrowers. Borrower 1 has the debt D1 = 2.7 and the disposable income Y1 = 0.968, that is, a debt ratio of D1/Y1 = 2.7/0.968 = 279 percent.
Borrower 2 has the debt D2 = 0.3 and the disposable income Y2 = 0.032, that is, a debt ratio of D2/Y2 = 0.3/0.032 = 938 percent. The sum of the debt is 3, and the sum of disposable income is 1.
The weighted average debt ratio is (D1/Y1)*(Y1/(Y1+Y2)) + (D2/Y2)*(Y2/(Y1+Y2)) = (2.79*0.968 + 9.38*0.032) = (D1+D2)/(Y1+Y2) = (2.7 + 0.3)/(0.968 + 0.032) = 300 percent.
The unweighted average debt ratio is (D1/Y1)*(1/2) + (D2/Y2)*(1/2) = (2.79 + 9.38)/2 = 610 percent (rounded), much higher than the weighted one.
For most purposes, the weighted debt ratio should be more relevant.
This was discussed in the minutes of the policy meeting in February 2013 (page 26-27). I maintained that the debt ratio of 600 percent for Stockholm that board member Per Jansson had presented (slide 5) at a hearing at the Riksdag’s Committee on Finance was exaggerated and misleading, because it had been calculated as an unweighted average. Its deviation from the debt ratio of 325 percent reported in the mortgage market report 2012 of Finansinspektionen (the Swedish FSA) was largely explained by its calculation as an unweighted average.