Biased memo from the Riksbank?

New Ekonomistas post (in Swedish). Here is an English translation.

The Council for Cooperation on Macroprudential Policy set up a joint analysis group in February 2013 to examine a few issues concerning household debt. (The council’s tasks have now been taken over by the newly established Financial Stability Council.)

Finansinspektionen (the Swedish Financial Supervisory Authority) published three memos of very good quality from the group in October, 2013 (unfortunately in Swedish only, although they deserve to be translated into English). They are referred to in this post (Swedish only). Now, somewhat belatedly, the Riksbank has published three more memos, of mixed quality, from the group. One of these, Memo 6 ”Risks to the macroeconomy and financial stability arising from the development in the household’s debt and housing prices,” is unfortunately somewhat biased and possibly even misleading. 

The memo repeats the Riksbank’s previous statements that household’s mortgage-rate expectations may be too low. It should be possible to put these statements to rest after the evidence shown in this and and Harry Flam’s Ekonomistas posts (the links are to the English translations of the posts).

The memo also maintains that a fall in housing prices could have large macroeconomic consequences. It speculates that borrowers after a fall in housing prices might want to quickly restore previous loan-to-value (LTV) ratios, and thereby reduce there consumption substantially in order to amortize their mortgages. This would lead to a substantial increase in unemployment. It also argues that high debt relative to disposable income probably leads to worse consequences of a fall in housing prices.

The memo gives the general impression that current level of household debt implies considerable risks, and, especially, that higher household debt would imply higher risks. But does the memo stand up to scrutiny?

The macroeconomic consequences of a fall in housing prices

The memo warns that the macroeconomic consequences of a fall in housing prices might be large, and that they can be particularly large because of the households’ high debt relative to their disposable income. But household debt is not high but historically normal relative to the households’ real and financial assets, since the households’ assets and net worth has grown at least as fast as their debt. This may actually have increased, rather than decreased, their resilience to disturbances in the form of a fall in housing prices or temporary income loss; see this new paper.

The memo does not discuss what might cause a fall in housing prices. Since Swedish housing prices as far as can be judged are consistent with fundamental factors, some fundamental factor or factors must reasonably change for housing prices to change. It makes sense that the macroeconomic consequences of a fall in housing prices would depend on which precise fundamental factor changes. If housing prices are a bubble and exceed the level motivated by fundamental factors, they might fall fast if the bubble bursts. But if they are consistent with fundamentals, they are less likely to fall fast, since fundamentals normally moves relatively slowly. In model simulations of consequences of a fall in housing prices, the cause of the fall is often assumed to be a change in household preferences, such that households suddenly, all else equal, demand less housing. This need not be a particularly realistic assumption.

Fast amortization and a large fall in consumption from a fall in housing prices?

The memo maintains that households, after a fall in housing prices, might want to quickly restore their previous LTV ratios, which might lead to a substantial fall in consumption, aggregate demand, and employment.

But the statements in the memo about the consequences of a fall in housing prices are contradicted by the Riksbank’s own analysis in its commission of inquiry into risks in the Swedish housing market 2011 and in a box in the Monetary Policy Report of July 2010. There it is shown that the effects on inflation, GDP, and unemployment of a fall of 20 percent in housing prices would be relatively limited. In addition, a more expansionary monetary policy can further limit the consequences of a fall in housing prices, as discussed in the minutes of the monetary policy meeting in June 2010. These issues are extensively discussed in a recent Ekonomistas post (English translation).

The statements in the memo about borrowers quickly amortizing their debt is also contradicted by the excellent analysis in another memo, Memo 4, ”Mortgage prepayment decisions by households”, written by Thomas Jansson. There, it is shown that a household that chooses assets, debt, and consumption in an optimal way, taking into account rates of return, borrowing costs, and risk diversification, prefers a high LTV ratio of over 85 percent. It is simply economically advantageous with a higher LTV ratio that makes it possible to invest in financial assets such as stocks and bonds. Investments in liquid assets also create a liquidity buffer, which is good in unforeseen outcomes. This is probably the main explanation why borrowers these days choose to amortize their debt to only a small extend. Many know that the interest rate on some savings accounts is not much less than the mortgage rate. The Swedish lender SBAB’s savings account for their own borrowers now in January 2013 has an interest rate of 2,20 percent, whereas the 3-month mortgage rate without a discount is 2,67 percent. With a reasonable discount of 0,20 percentage points, the spread is only 0,27 percentage points. This is a small cost of a liquidity buffer.

If borrowers would freely choose their LTV ratio, according to Memo 4 many would choose a higher LTV ration than the current LTV cap of 85 percent. This means that the LTV cap is binding restriction, and may borrowers end up with a lower-than-optimal LTV ratio. After a fall in housing prices, at unchanged mortgages the LTV ratios would increase, and many borrowers might find themselves with actual LTV ratios closer to the one they would freely choose themselves.

Mortgages in Sweden are in practice loans with the household’s repayment capacity and the present value of the household’s disposable income, its “human capital,” as collateral, rather than the housing. This is because Swedish borrowers are personally responsible for serving and repaying the debt; mortgages are full recourse. Mortgages therefor have an important role in household’s lifecycle saving. This means that what determines any amortization on the mortgage is not the value of the housing but how the mortgage and its financing fits into the household’s total economic situation. An important factor for any amortization is then the debt service cost. Since a fall in housing prices with sensible monetary policy will imply low policy rates and thereby low mortgage rates, the household’s incentive to amortize fast in such a situation would be small. In this context, it is a considerable advantage that there is a high share of mortgages with variable rates. This makes monetary policy more potent, and allows it to better counter any fall in consumption. As discussed in this recent Ekonomistas post (English translation), monetary policy can be quite effective in moderating the consequences of a fall in housing prices.

In practice, the LTV ratio for a given amortization-less mortgage is probably determined mainly by the LTV-ratio when it was taken out, how many years that have gone since it was taken out, and what has happened to the value of the housing during these years. This leads to a vintage structure of mortgages with different LTV ratios. It is such a view of mortgages that is behind the analysis in this paper about why “leaning against the wind” is counterproductive as a method to reduce household debt as a share of BNP and disposable income. This is also discussed in a previous Ekonomistas post (English translation).

A bigger fall in consumption and a larger increase in unemployment at higher indebtedness?

The memo maintains that the effects of a fall in housing prices probably is larger with high indebtedness among households and refers to some international studies, includingchapter 3, ”Dealing with household debt”, in IMF’s World Economic Outlook, April 2012. There it is shown that the outcomes in recessions and falls in housing prices (”busts”) are significantly worse in countries with “high debt” than with “low debt.” But by “high debt” and “low debt” is not meant countries with high and low debt, respectively, as one might think. Instead “high debt” means countries with a high increase in household debt relative to GDP during the the three years preceding the bust. In Sweden, household ddebt relative to disposable income has been quite stable since 2010. Sweden is thus a country that would now be classified as a “low debt” country in the study. The conclusion for Sweden from the IMF study is thus that Sweden now belongs to the countries for which the outcome in a bust would be much better than for the “high debt” countries. This is the opposite of how the memo seems to have interpreted the IMF study.

The other international studies referred to in the memo in this context concern the U.S. situation during the housing and financial crisis there. The U.S. housing policy with sub-prime lending implied a collapse of credit assessments, in many cases involving crimes, and a systematic lending with very high LTV ratios to borrowers who could not repay or even service the debt without a substantial and steady increase in housing prices to allow increases in the mortgage. In addition, there are other important differences, including the social safety net and that fact that mortgages in many states are no recourse, that is, without personal responsibility to repay and with the housing as the only collateral. This means that one an hardly draw any reliable conclusions for Sweden from this literature.

It may be more relevant to compare with the actual stress test in real time that the Swedish economy was subject to 2008 and 2009. As we see in an Ekonomistas post (English translation), real housing prices fell by 10 percent for houses and 20 percent for flats by from the fall of 2007 to the end of 2009. Private consumption fell by 3 percent from the third quarter of 1993 to the first quarter of 2009. By the first quarter of 2010 consumption had risen to exceed the pre-crisis level. Export and investment fell much more, but consumption helped keeping up aggregate demand. Borrowers did not amortize their mortgages much.

The most important implicit conclusion from the memo is of course that higher debt, if housing prices would fall, would lead to a larger fall in consumption and a higher increase in unemployment. But data for a number of relevant countries hardly supports this. Figure 1 shows that there does not seem to be any relation between level of household debt as a share of GDP in 2007 and the change in private consumption between 2007 and 2012.


Figure 1. The change in private consumption from 2007 to 20012 and household debt as a share of GDP in 2007 for a number of countries.
Source: OECD.


Figure 2 shows that neither does there seem to be any relation between debt and the increase in the unemployment rate from 2007 to 2012 for these countries. This is consistent with the analysis of chapter 4 of the above-mentioned IMF World Economic Outlook. There it is the increase in debt relative to the GPD during the three years preceding the bust, not the level of debt relative to GDP before the bust that has some explanatory power.


Figure 2. The change in the unemployment rate from 2007 to 2012 and household debt as a share of GDP in 2007 for a number of countries.
Source: OECD.

Altogether, there is hardly any support for the reasoning and conclusions in the meme discussed here. There are other dubious reasoning and conclusions in the memo, but there is not enough space to discuss those in this post.