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

Since last summer, the Riksbank has put forward a new reason for a higher repo rate (the Riksbank’s policy rate). The Riksbank suggests that households would have too low mortgage-rate expectations. But closer scrutiny shows that there is hardly any basis for the Riksbank’s suggestion. (This aside from the fact that, if households’ mortgage-rate expectations would be a problem, a higher repo rate is hardly the solution.)

In the Monetary Policy Report of October 2013 (page 17), the Riksbank says (a similar statement is in the July report):

A particularly difficult situation would arise if the households’ expectations of future mortgage rates were influenced to too great an extent by the current low level of interest rates. Surveys show that the households’ expectations of mortgage rates five years ahead are lower than is compatible with the Riksbank’s assumptions regarding the repo rate in the long term (see Figure 1:33) [figure 1 below]. Unrealistic mortgage rate expectations could lead to a renewed upward trend in both housing prices and debts.

This leads to two questions: Is it true that the households’ mortgage-rate expectations are too low? And if the expectations would be too low, and this is considered a problem, how should the problem be managed?

If households’ mortgage-rate expectations would be problem, is a higher repo rate really the solution?

Let me not go into any detail on the second question. Let me only note that, if households’ mortgage-rate expectations would be unrealistically low, it does not follow at all that the best way to handle the problem is a higher repo rate. A higher repo rate has large costs, in the form of too low inflation, too high unemployment, and, as far as can be judged, higher real debt. The problem would be an information problem, namely that households have insufficient or biased information about future mortgage rates. It seems pretty obvious that other methods to improve households’ information is better, such as requiring better information from the banks to the lenders, or an information campaign by the Finansinspektionen (the Swedish Financial Supervisory Authority) about realistic long-run levels of mortgage rates. In contrast, the Riksbank seems, without much thinking, always assume that most problems shall be solved by a higher repo rate.

The Riksbank’s assumptions

Let me then deal with the first question, are really households’ mortgage-rate expectations too low? Compared to what? The Riksbank refers to the households’ expectations of the variable mortgage rate 5 years ahead, measured by the survey of the National institute of Economic research, being lower than what is consistent with the Riksbank’s assumption about the long-run level of the repo rate (see figure 1 above). Is this a relevant comparison?

There are three assumptions in the Riksbank’s reasoning and in figure 1: (1) An implicit assumption that the repo rate and the mortgage rate will reach their long-run levels within 5 years. (2) An explicit assumption that the long-run level of the repo rate is in the interval from 3.5 to 4.5 percent. (3) An explicit assumption that the long-run spread between the variable mortgage rate and the repo rate is in the interval 1.7 to 2 percent. All three assumptions may be questioned, but assumption (1) seems particularly questionable.

Normal rates within five years?

Regarding assumption (1), it seems quite reasonable today to assume that it will take more than 5 years before interest rates reach their normal levels. Perhaps households assume this, too. It seems likely that interest rates abroad will be low under a considerably long period. And interest rates in Sweden cannot deviate that much from rates abroad without effects on the exchange rate. From this point of view, it would have been good if the NIER’s survey would have included a question about what households believe about the variable mortgage rate not only in 5 years but also in the longer run.

To this one may add the discussion after Larry Summers’s much noted statement at an IMF conference and in an op-ed article in the Financial Times about the neutral real rate possibly being negative for a considerable period.

One may also note that the average of the repo rate during the 19 years from 1995 until now is only 3.2 percent. As can be seen in figure 2, a 5-year moving average of the repo rate has not exceeded 4 percent from 2001 (13 years ago) and 3.5 percent from 2005 (9 years ago). These facts hardly make it more likely that the policy rate would reach a normal rate between 3.5 and 4.5 percent within the next 5 years.

An important circumstance in this context is that, if it takes longer than 5 years to reach normal levels of interest rates, households’ nominal disposable incomes will likely grow quite a bit during this time. This means that incomes will grow relative to the mortgages taken out today, and relative to the future debt payments on these loans. Assume that real disposable incomes will grow at an average rate of 2 percent of year, and assume that inflation will also be on average 2 percent. (With current monetary policy, this may seem less realistic, but let me not get into that issue here.) Then nominal disposable incomes will grow at an average rate of 4 percent per year. Four percent per year means 22 percent in 5 years and 48 percent in10 years. The debt ratio for any given mortgage taken out today will then fall correspondingly, even if the loan is not amortized. This means that, for a given interest rate, debt-service payments relative to disposable income will fall correspondingly. The longer it takes to reach normal levels of the mortgage rate, the lower the debt service relative to disposable income.

The interval for a long-run repo rate?

Regarding assumption (2), the Riksbank’s assumed interval between 3.5 and 4.5 percent for the long-run level of the repo rate, the low average repo rate since 1995 together with figure 2 possibly indicate that a somewhat lower interval, or at least a smaller lower bound for the interval, perhaps 3 percent, cannot be excluded. But without a more thorough analysis, it is difficult to say anything substantially on this.

The spread between the mortgage rate and the repo rate?

Regarding assumption (3), the Riksbank’s assumed spread of 1.7-2 percentage points between the variable mortgage rate and the repo rate, figure 2 shows that a 5-year moving average of the spread has been below 1.7 percent except very recently. Higher risk weights on mortgages and thereby higher capital requirements on the banks’ mortgage stock may increase the spread at unchanged rate-of-return requirements on equity, but the increase may be less or not materialize, if rate-of-return requirements on equity fall with more capital and thereby less risky banking.

Households’ mortgage-rate expectations, the Riksbank’s repo-rate path, and the market’s repo-rate expectations

But let us look more closely how households’ mortgage-rate expectations relate to the Riksbank’s repo-rate path and market expectations about future repo rates. Figure 3 shows the situation at the Riksbank’s policy meeting in September 2011. The Riksbank decided to keep the repo rate unchanged at 1 percent and postponed rate increases somewhat. The black dashed line shows the Riksbank’s new repo-rate path, the Riksbank’s forecast of the repo rate. The red solid line shows market expectations of future repo rates after the decision had been announced. The black squares show the households’ expectations of future variable mortgage rates. The black solid step-shaped line shows the outcome of the repo rate.

We see that the Riksbank’s repo-rate path on this occasion completely lacked credibility, since market expectations deviated so much from the repo-rate path. We also see that the Riksbank’s repo-rate path was a very bad forecast of the future repo rate. Market expectations were a much better forecast.

We see that households’ mortgage-rate expectations were about 2 percentage points higher than the repo-rate path, except 3 years ahead, for the fall of 2014, when the spread had fallen to about 1.5 percentage points. But the spread to market expectations was then more than 3.5 percentage points. With such a large spread to market expectations, it is difficult to maintain that households’ expectations would have been too low. That the spread to the repo-rate path was only 1.5 percentage points 3 years ahead seems less relevant, since the repo-rate path seemed to have lost contact with reality.

Figure 4 shows the situation at the policy meeting in April 2013 (the last policy meeting I participated in). The spread between households’ expectations and the repo-rate path was then a bit more than 2 percentage points 1 year ahead, for the spring of 2014, but fell to about 1.5 percentage points 3 years ahead, for the spring 2016. Also here, market expectations were a better forecast of future repo rates than the repo-rate path, with the market anticipating a reduction of the repo rate to 0.75 percent later in the fall of 2013. Also on this occasion, it is difficult to maintain that households’ expectations were too low. If anything, households appear to be pretty prudent, in the sense of having a substantial marginal to market expectations. And the market expectations undoubtedly seem more reasonable than the Riksbank’s repo-rate path.

Households’ expectations and banks’ actual mortgage rates

Figure 5 shows the perhaps most relevant comparison, namely the households’ mortgage-rate expectations compared to the actual mortgage lending rates of the banks. The blue line shows the banks’ mortgage rates in October 2013 for different fixation periods, according to the Financial Markets Statistics of Statistics Sweden. The variable and 1-year mortgage rate are both about 2.6 percent, whereas the 5-year rate is about 3.6 percent. Given the households’ expectations of future variable mortgage rates, one can calculate what average variable rates households expect for periods up to 5 years. This average rate (the red line) should be compared to the banks’ actual rates for different fixation periods (the blue line).

According to the households’ expectations, the expected average mortgage rate for a loan length of 2 years is about 3.2 percent. But at the same time, the actual 2-year fixed rate is lower, only 2.8 percent. For a loan length of 5 years, the expected average variable rate and the fixed rate coincide.

Thus, the households do not at all seem to expect lower average rates than the banks. If the Riksbank argues that households’ expectations are too low, it must also argue that the rates and expectations of the banks are too low.

All together, there hardly seems to be any basis for the Riksbank’s suggestion that households’ mortgage-rate expectations would be too low. But it would of course be interesting to have data on households’ expectations of rates further into the future than 5 years. At the same time, the fact that households’ nominal incomes normally grow quite a bit in 5 years and longer in relation to previous mortgages means that interest-rate payments become much less burdensome. But most important in this context is arguably that the banks’ credit reviews are thorough and that, regardless of households’ mortgage-rate expectations, banks use sufficiently high interest-rates in assessing lenders debt-service capacity and resilience to disturbances.