English translation of Ekonomistas post (in Swedish).

At the latest policy meeting and decision, the Riksbank repeated its statements that a lower policy rate would increase the risks associated with household indebtedness. At the same time these statements are contradicted by the Riksbank’s own analysis. According to this, the cost of a lower policy rate is only about 0.4 percent of the benefit, and hence negligible relative to the benefit. On a direct question about this at the latest press conference, Governor Ingves continued to give evasive answers.

At the latest monetary-policy meeting (April 2014), the Riksbank left the policy rate unchanged at 0.75 percent. It did this in spite of CPI inflation equal to zero for two years and unemployment much above a long-run sustainable rate. Furthermore, after the meeting, Statistics Sweden announced that CPI inflation has fallen further to minus 0.6 percent.

The Riksbank’s justification for not lowering the policy rate is that a lower policy rate would increase the risks associated with household debt:

For some time now, monetary policy has been characterised by a balance between weighing how low the repo rate needs to be for inflation to approach the target soon enough against the increased risks linked to households’ high indebtedness that can stem from a low interest rate.

… A lower repo rate would also increase the risks linked to the high household indebtedness. (Monetary Policy Update April 2014, p. 9)

,

But according to the Riksbank’s own analysis, the policy rate has a negligible effect on risks associated with household indebtedness. This you can see by making precise the cost and benefit of a lower policy rate that follow from the Riksbank’s analysis. Then one can relate the cost to the benefit, in order to clarify the tradeoff and judge whether the cost is larger or smaller than the benefit.

As I have shown in this post (figure 1, but for a *decrease* in the policy rate instead of an increase), the Riksbank’s own analysis shows that a 1 percentage point lower policy rate during 4 quarters would reduce unemployment by 0.5 percentage points the next few years. This is the *benefit* of a lower policy rate, in the form of a lower unemployment rate the next few years.[1]

The *cost* can be separated into two parts. The *first* part arises because higher debt might imply a higher probability of a future crisis. According the Riksbank’s analysis, a 1 percentage point lower policy rate during 4 quarters would increase real household debt by 0.25 percent after 5 years (figure 2, but for a reduction in the policy rate, in this post). This would in turn, according to Schularick and Taylor (2012), increase the probability of a crisis in 5 years by 0.02 percentage points.[2]

This cost can be expressed in terms of increased unemployment, if you, as in this box (figure A10), assume a crisis scenario where the unemployment rate becomes about 5 percent higher in a crises.

If the probability of a crisis increases by 0.02 percentage points, that is, by 0.0002, the expected future unemployment rate will increase by 0.0002*5 = 0.001 percentage points.

This is thus the cost expressed in terms of higher expected future unemployment because of an increased probability of a crisis. It is obviously miniscule relative to the benefit of 0.5 percentage point lower unemployment the next few years.

The *second* part of the cost arises because higher debt might imply a higher increase in unemployment in a crisis. According to the Riksbank’s own analysis, the lower policy rate would imply a lower debt-to-income ratio after 5 years of 0.44 percentage points. (figure 3, but with a reduction of the policy rate, in this post).

According to a post by Deputy Governor Martin Flodén (2014), a 1 percentage point higher debt ratio might imply that the increase in unemployment in a crisis becomes 0.02 percentage point higher. From this follows that the increase in unemployment in a crisis in 5 years would be 0.02*0.44 = 0.009 percentage point larger. If a crisis would occur with such a high probability as 10 percent, the expected larger increase in unemployment in a crisis will still be only 0.1*0.009 = 0.0009 percentage point larger.[3] The second part of the cost is of the same magnitude as the first part discussed above.

Thus, this second part of the cost is the cost of a lower policy rate because of a larger increase in unemployment in a crisis. This cost is also miniscule in relation to the benefit of a 0.5 percentage point lower unemployment rate the next few years.

**The cost of a lower policy rate is only about 0.4 percent of the benefit**

If we add the two parts of the cost, we get a total cost of 0.0019 percentage points, in the form of expected future higher unemployment. This cost of a lower policy rate, 0.0019 percentage point higher expected future unemployment, can then related to the benefit, 0.5 percentage point lower unemployment the next few years. The cost of a lower policy rate as a share of the benefit is then 0.0019/0.5 = 0.0038, that is, only about 0.4 percent of the benefit. Clearly, the cost can be neglected.

The benefit and cost of a lower policy rate is summarized in table 1.

**Q&A at the press conference**

At the press conference on April 9, after the announcement of the policy decision, Governor Stefan Ingves received a direct question about these issues from the journalist Johan Schück at Dagens Nyheter. The question was answered first by Ingves and then by Marianne Nessén, Director the Department of Monetary Policy at the Riksbank. The question comes 29 minutes and 17 seconds into the Riksbank’s web video (in Swedish). (The questions and the answers are translated into English by me.)

**JS:** In the previous Monetary Policy Report, there was a box by the Riksbank, which showed that monetary policy has a very limited effect on household indebtedness. Even if you are worried about the indebtedness, the question arises, how meaningful is it to take household debt into account in monetary policy, as you have done in this latest decision?

**SI:** It is meaningful, because if we make a mistake here in Sweden, we will find ourself confronted with the problem regardless. Then *we* have to do as well as we can, and it is important that also *others* contribute to this. It is very difficult to disregard household indebtedness, when you are doing monetary policy, because we know that, if the debt becomes too large – regardless of whether it is household debt or national debt, which has been a problem in a number of countries – then you also get big problems with monetary policy. Therefore we have to take account of the debt as well as it can be done.

**JS:** But there are two issues here, together. One is whether the debt is a problem; that is a separate thing. The other is whether the policy rate is an effective means to affect the debt.

**SI:** Yes, that is completely correct. Given that more and more people are borrowing at a variable mortgage rate – and then you can discuss whether this is good or not – it is the case that, if you lower the interest rate, people will borrow more. It would be strange if that was not the case.

**MN:** May I just add – this particular box that you are referring to, it is a very special experiment that we are considering. The policy rate is lowered during a limited period and then goes back to some level that you were considering before. This is what you may call a modest monetary-policy experiment, and then we examine what effect it has on debt and other macro variables. It is not the final answer to the question “what is the monetary-policy effect on debt?”; it is just a very special experiment. You lower the policy rate during a limited period and then go back to the old path, and the long-run expectations about where the interest rate is going are not affected. This is difficult material. It is an area of research that is developing fast, so we hope to come back in the future when we have conducted more studies.

It seems clear from this, that Stefan Ingves and Marianne Nessén do not have any good answers to Johan Schück’s important question.

The experiment that is done in the box, and the one Marianne Nessén is referring to, is an obvious experiment to do in order to asses the cost of a lower policy rate relative to the benefit, that is, what the monetary-policy tradeoff is about. From the Riksbank’s analysis so far, it thus follows that the cost, in the form of higher expected future unemployment due to increased risks with household debt, is about 0.4 percent of the benefit, in the form of lower unemployment in the next few years.

Thus, the cost of a lower policy rate because of increased risks associated with household debt is thus completely negligible relative to the benefit. Then it is clear, that it is not meaningful to let risks associated with household debt affect monetary policy now in Sweden.

**References**

Flodén, Martin (2014), ”Shall We Be Concerned about High Household Debt?”, English translation of blog post, February 20, 2014.

Schularick, Moritz and Alan M. Taylor (2012), ”Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870-2008”, *American Economic Review* 102, 1029–1061.

[1] To lower the policy rate by 1 percentage point now, from the current level of 0.75 percent, would imply a new policy rate of -0.25 percent. This is fully possible, and is actually discussed as an option for, for instance, the ECB. But the calculations here aim at making precise the cost *relative to* the benefit of a lower policy rate, so the actual reduction in the policy rate is not important. One can do the calculations by reducing the policy rate by 0.75 percentage points, or by 0.5 percentage points, in which case the cost and benefit will be three quarters or a half, respectively, of those for a 1 percentage point reduction. But this does not affect the relation between the cost and the benefit.

[2] According to Schularick and Taylor (2012, table 3), 5 percent higher real debt in 5 years implies that the probability of a crisis increases by 0.4 percentage points. This means that 0.25 percent higher real debt in 5 years implies that the probability of a crisis increases by 0.25*0.4/5 = 0.02 percentage points.

[3] A probability of 10 percent corresponds to a crisis on average every 10^{th} year. According to Schularick and Taylor (2102), the average probability of a crisis in their data is only 4 percent. That is, it corresponds to a crisis on average every 25^{th} year. With the probability 4 percent, the cost is only 4/10 of the cost for 10 percent.