New Ekonomistas post, “What are the effects of a fall in housing prices?” (in Swedish). Here is an English translation.

One of the Riksbank’s reasons for a higher policy rate is that this would reduce the risk of a future fall in housing prices. A future fall in housing prices might lead to lower inflation and higher unemployment. But how large would the effects on inflation and unemployment be of a fall in housing prices of, for instance, 20 percent? There is a fair amount of information about this, including the Riksbank’s own reports and calculations. Most information indicates that the effects of a fall in housing prices of 20 percent would be relatively limited. They are hardly larger than the higher unemployment and lower inflation caused by the current monetary policy. 

One may question if really a higher policy rate would reduce the risk of a future fall in housing prices, especially since a higher policy rate seems to increase household real debt and weaken their balance sheets, as I have discussed in these Ekonomistas post (English translations here and here). But let us here limit the discussion to the possible effects on inflation and unemployment of a future fall in housing prices.

In the Riksbank’s commission of inquiry into risks on the Swedish housing market 2011, there is an excellent paper by Claussen, Jonsson and Lagerwall (p. 67-95), which among other things examines the effects of a fall in housing prices with the help of two different models. The authors find that “a fall in house prices could have a relatively limited effect on the real economy.”

In the Monetary Policy Report of July 1, 2010, there is a box on the effects of a fall in housing prices. In it there is a table, table B2, with a calculation of the effects on inflation and unemployment of a fall in housing prices of 20 percent, under the assumption that monetary policy responds to the fall and the repo rate (the Riksbank’s policy rate) is reduced somewhat (table 1 below). [1]

One may note that the effects on inflation are insignificant. The unemployment rate would in contrast be 0.6-0.8 percentage points higher during three years. At the policy meeting on June 30, (se the minutes, p. 20-22) I argued that the box exaggerates the effects, because some assumptions in the model used are problematic for the question examined, and because the effects can be damped with a better more expansionary monetary policy.

A problematic assumption about the loan-to-value cap binding ex post

All models must rely on some simplifying assumptions to be useful. But the model used in the box (Walentin 2013) relies on some assumptions that are problematic for the particular question here, the effects of a fall in housing prices. For these effects, a crucial issue is how the consumption of borrowers would adjust in the event of a fall in housing prices. In the model, it is assumed that there is a loan-to-value cap of 85 percent of the value of the housing. The households with mortgages are assumed to have no other assets than their housing, and they consume all their disposable income. Every quarter, they adjust their mortgage so that the have exactly a loan-to-value ratio of 85 percent. An crucial assumption is that the loan-to-value cap binds ex post, if housing prices fall. If the price of the housing falls by SEK 100 000, the household has to amortize 85 percent of the fall in the price, that is, SEK 85 000, within a quarter. Since the household has no other assets, it has to reduce its consumption substantially.

In reality, things are quite different in Sweden. Borrowers don’t adjust their mortgages every quarter. Instead only a fraction of the mortgages are refinanced each year, perhaps around a seventh. (This implies that households’ total mortgages become sticky, and is an important starting point when examining the effect of monetary policy on household debt, debt-to-income ratios, and loan-to-value ratios, se here and here.) The loan-to-value-ratio varies a lot across different borrowers (see Finansinspektionen, The Swedish Mortgage Market 2013, diagram 3).

Furthermore, and crucial in the context, the loan-to-value cap of 85 percent is not binding ex post. If a household after a fall in housing prices would have a loan-to-value ratio above 85 percent, the banks would demand a corresponding immediate amortization, as long as the household continues to service the debt in a normal way. In Sweden, mortgages are full recourse, so households are personally responsible for the debt, regardless of the value of the collateral. Because of this and the strict credit assessment done before the mortgage is issued, credit losses on mortgages are very small in Sweden, and it is not in the interest of the banks that borrowers quickly pay back their mortgages. In reality, borrowers does not at all have to adjust their consumption as much as in the model.

The effects can be moderated by a more expansionary monetary policy

But in spite of the unrealistic assumptions in the model used, that the loan-to-value cap binds ex post, the effects on inflation and unemployment of a fall in housing prices can be moderated substantially by a better and more expansionary monetary policy than that assumed in table 1 above. During the preparations before the policy meeting in June 2010, I therefore asked the Riksbank’s Monetary Policy Department to calculate the effects of a more expansionary monetary policy. I wished that the box would be revised to be based on this better monetary policy. Since the majority of the Executive Board rejected this, I felt it necessary to present these calculations at the policy meeting instead. They are therefore included in a table in the minutes, p. 20-22 instead of in the box in the Monetary Policy Report. 

The upper part of the table shows that a slightly lower repo-rate path than in table 1 may neutralize the effect on inflation completely. The lower part shows the effect of a substantially lower repo-rate path that tries to neutralize the effect on GDP (the effects on GDP are not shown). We see that this repo-rate path moderates the effect on unemployment considerably. Unemployment becomes 0.5 percentage points higher in year 1 and only 0.2 percentage points higher in year 2, and it remains unchanged in year 3 (compared to if housing prices had not fallen). At the same time inflation becomes a bit less than 1 percentage point higher in years 2 and 3. This requires a repo rate that is a bit less than 1 percentage point lower in year 1 and 1.5 percentage point lower in year 2.

These calculations show that, in spite of the model’s problematic assumption about immediate amortization of 85 percent of a fall in housing prices, a sufficiently expansionary monetary policy may limit the effects on inflation and unemployment substantially. But what if the repo rate would initially be so low that it cannot be lowered much (however, it can probably be lowered at least to minus 0.25 procent)? Then the same expansionary effect can be achieved by it keeping the repo rate a low level for longer. (See the minutes, p. 20-22, for a more extensive discussion and more details.)

One may also recall that current monetary policy may well have caused as much as 1,2 percentage points higher unemployment the last few years, as discussed in two Ekonomistas post (English translations here and here). If this higher unemployment would be considered to be an insurance premium for reducing the risk of a future damage in the form of future negative outcomes, it seems to be a pretty bad insurance policy. According to table 2, for a fall in housing prices of 20 percent, the increase in unemployment may be so limited as 0.5 percentage points during one year only. Even with the more passive monetary policy and larger effects in table 1, the damage is smaller than the insurance premium. If the damage is smaller than the premium, it is a pretty bad insurance policy, even if the insurance would completely eliminate the risk of a future damage. Furthermore, it may be questioned whether the insurance policy actually reduces the risk at all.

These results indicate that the macroeconomic effects of a future fall in housing prices would be relatively limited and thus hardly a reason for a higher policy rate now. They are hardly larger than the higher unemployment and lower inflation caused by current monetary policy.

———–
[1]. Since housing prices are determined by fundamental factors in the model, some fundamental factor must change for the housing prices to fall. In the box it is assumed that housing prices fall because household preferences change so their demand for housing falls.