Two serious mistakes in the Goodfriend and King review of Riksbank monetary policy

Marvin Goodfriend and Mervyn King presented their review of the Riksbank’s monetary policy 2010-2015 on January 19. They provide some severe  and justified criticism of the majority’s policy, but they unfortunately start their evaluation by making two serious mistakes, which are discussed in this post. 

Goodfriend and King avoid doing a thorough evaluation of the dramatic and controversial policy tightening 2010-2011, from a policy rate of 0.25% in June/July 2010 to a rate of 2% in July 2011. On the 2010-2011 hikes, they instead simply state (italics added):

[T]he response of the Riksbank to the rapid recovery of the Swedish economy from the global financial crisis – which entailed raising official interest rates from 0.25% to 2% between June 2010 and July 2011 – was broadly accepted by all members of the Executive Board, and appears not unreasonable in the light of all the information available to the Riksbank at the time.

Although the downturn of the Swedish economy in 2008-09 was similar to that in other industrialised countries, the rebound in the Swedish economy, particularly marked in exports, was more rapid than elsewhere and led to a shared view that it was justified to begin the process of raising rates. (p.86)

“Not unreasonable”?

First, when finding the Riksbank’s response to the “rapid recovery” and rate hikes appearing “not unreasonable”, Goodfriend and King seem to confuse rates of change and levels. It is true that GDP and export had started to grow in 2010 (although the year-over-year growth rates in quarter 1 of 2010 were not exceptionally high, 2.9% for GDP and 4.2% for export). However, the more important levels were not high but quite low.

Goodfriend and King make of point of evaluating the Riksbank in the light of the information available at the time of the decisions (something that I, in a March 2009 speech and a chapter in the John Taylor festschrift, have called ex ante evaluation as distinct from ex post evaluation). So let’s see what information was available to the Riksbank in June 2010, when majority decided to start the hikes.


Figure 1. Swedish GDP, GDP trend, and export in levels (GDP in real-time data). GDP indexed to 100 in 2007Q4; export indexed to 100 in 2008Q1. (Added January 30: Exponential GDP trend 2000Q1-2008Q2). Source: Statistics Sweden.

Figure 1 shows the level of GDP (real time data, although not very different from revised data) and export (which Goodfriend and King explicitly refer to), through quarter 1 of 2010, the latest data available to the Riksbank in June 2010. GDP in quarter 1 of 2010 was a full 5% below the 2007 peak, a full 10 % below trend, and at the same level as in 2006. And export was 13% below its peak in 2008, at the same level as in late 2005.


Figure 2. Unemployment rate and the Riksbank’s estimate a “long-term” unemployment rate. Source: Statistics Sweden and the Riksbank.

Furthermore, as seen in figure 2, the unemployment rate, a more reliable indicator of slack in the economy, was about 9% in quarter 1 of 2010 . At the same time, the Riksbank’s estimate of a “long-term” (sustainable) unemployment rate at the time was 6.5%, making the unemployment gap about 2.5%.[1]

To start a rate hike from 0.25% to 2% in such a situation, with large slack in the economy and inflation not far from its target, is definitely not reasonable. Given the low levels of GDP and export, above-normal growth was necessary for the recovery. Tightening to reduce growth instead slowed down the recovery.[2]


Figure 3. Inflation and unemployment forecasts of the FOMC (solid red) and the Riksbank (dashed blue) in June 2010. Horisontal lines in right panel are FOMC and Riksbank estimates on a long-run sustainable unemployment rate. Outcomes to the left of the vertical line; forecasts to the right. Source: Svensson (2011).

In addition, figure 3 (from figure 1 and 2 in Svensson (2011)) shows the June 2010 inflation and unemployment forecasts of the Riksbank and the Fed’s FOMC (to the right of the vertical line). They are quite similar. Both inflation forecasts are below the inflation target of 2% (explicit for the Riksbank, implicit but generally understood for the FOMC). Both unemployment forecasts are far above the long-run sustainable rate. In such a situation, easier policy shifts the inflation forecast closer to the target and the unemployment forecast closer to the long-run sustainable rate as is therefore unambiguously better. Tightening policy defies all logic, unless the policymaker has other objectives than achieving the inflation target and stabilizing unemployment around its long-run sustainable rate. But in this situation, the Riksbank indeed started to hike the policy rate from 0.25% to 2% a year later, whereas the FOMC kept the policy rate at 0.25% and started to prepare QE2. Are Goodfriend and King suggesting that the FOMC should instead also have hiked the federal funds rate to 2%?[3]

So just stating that the Riksbank rate hike was “not unreasonable,” ignoring the actual GDP and export level data as well as the unemployment rate known at the time, and not providing a thorough evaluation of the hike, is a first serious mistake. This arguably most premature tightening, with its dire consequences for inflation and unemployment, should have been subject to considerable scrutiny. Instead, Goodfriend and King are in effect just repeating weak defenses of the policy made before by Governor Ingves and Deputy Governor Jansson, previously rebutted herehere, and here.

Added January 24: Meanwhile, Paul Krugman has found a great illustration of the error in confusing growth rates and levels, and Simon James Cox has recalled a highly relevant quotation from Mervyn King himself (Quarterly Inflation Report – 12th August 2009, Press Conference Transcripts, p. 3):

It’s levels, stupid. It’s not growth rates, it’s levels that matters here. That’s what matters in terms of thinking about the downward pressure on inflation in the medium term. And after a deep recession you would normally expect a period of robust growth in order to use up that margin of space capacity.

“Broadly accepted by all members”?

I believe it is a second serious mistake of Goodfriend and King to state that the rate hikes were “broadly accepted by all members” and that there was “a shared view that it was justified to begin the process of raising rates.”

I know for sure that I did not share the view, and this is evidenced in my November 2010 speech and repeatedly in the minutes. In this and later speeches, I instead advocated a major redirection of policy. One may also note the headline “Disaster path” (“Katastrofbana”) on the front page of the Swedish newspaper Svenska Dagbladet on September 27, 2010. This referred to my warning about the dire consequences of the majority’s high policy-rate path in an interview (in Swedish) in the newspaper that day. The headline and interview hardly indicates that I “broadly accepted” the rate hikes.[4]

As evidence supporting their conclusion, Goodfriend and King state (italics added):

The dissenters on the Executive Board never voted for a level of the current repo rate more than one quarter of a percentage point below that actually set by the majority, and even the most extreme dissenter, Mr Svensson, having voted for a repo rate of 0.25% in February 2010, was voting one year later for a repo rate of 1.25% and then a few months after that for a rate of 2%. (p. 86)

Since members vote for a level of the repo rate, dissenters who believe that policy is some way off track must make that clear by voting for what they believe to be the appropriate rate. A vote for a slightly lower rate than that adopted by the majority cannot subsequently be used as evidence that a dissenter believed that a very different repo rate was appropriate. (p. 87)

However, a vote for a slightly lower rate than that adopted by the majority cannot subsequently be used as evidence that a dissenter did not believe that a very different repo rate was optimal. Goodfriend and King miss that the repo rate and repo-rate path that I voted for was only the first step in the right direction towards an approximately optimal rate and path; it was not a complete step to such a rate and path.

The principle for setting the policy rate and the policy rate path that I followed is discussed in the June 2010 minutes and described in more detail here. It can be summarized as “Adjust policy step by step until the inflation and unemployment forecasts ‘look good’ ” (where “looking good” means best stabilizing inflation around the target and unemployment around its long-run sustainable rate). This principle implies that, if at a policy meeting  the inflation forecast is below the inflation target and/or the unemployment forecast is above the long-run sustainable unemployment rate, set the new repo rate and repo-rate path lower than at the previous meeting. If the situation is the same at the next meeting, continue to lower the repo rate and repo rate path.

As I have several times said in the minutes, given technical limitations of Riksbank analysis, it was not possible to calculate an optimal policy rate and policy-rate path far from the majority path. But it was clear that a better rate and path were below those of the majority. Even for the rather different paths that I advocated as a first step, the inflation forecast was still normally too low and the unemployment forecast too high to look good. So it was clearly just a first step.

For several reasons, it would not have been appropriate to take a surprising and large step towards an unknown optimal rate and path. Instead I advocated this stepwise procedure towards a rate and path that would eventually make the corresponding forecasts for inflation and unemployment look good. Unfortunately, I never gained a majority for even the first step.

Thus, voting for a rate of 1.25% or 2% didn’t mean that it was the best rate, but that it was better than 1.5% and 2.25%, and a first step to a much lower rate.

Goodfriend and King here actually mainly repeat statements made by Governor Ingves that have previously been rebutted in more detail here.

Finally, a fair evaluation of the performance of Karolina Ekholm and me could have asked the following question: What would the counterfactual outcome for inflation, unemployment and the policy rate had been, if the majority had sided with us? Would the outcome have been a policy rate path only 0.25 percentage points below the actual one? Of course not. We would have postponed any policy-rate increase above 0.25% at least until the unemployment forecast would have undershot the long-run sustainable rate, or the inflation forecast would have overshot the inflation target. And we might have lowered the policy rate further to zero or below, as I voted for from July 2009 through February 2010, if more stimulus would have brought better target achievement.

Added January 25: Furthermore, Goodfriend and King misreport the monetary policy stance of the minority, as discussed here.

Severe criticism of the majority for policy after 2011

Goodfriend and King misreport the monetary policy stance of the minoritySo, in summary, the evaluation unfortunately starts with two serious mistakes, in its attempt to exonerate the majority for the rate hikes 2010-2011. Goodfriend and King do provide a more thorough evaluation of policy after 2011, including some severe criticism of the majority’s policy.

For instance, they state:

One of the problems with the monetary policy discussion was that the majority apparently went along with the overly optimistic inflation forecasts that the model produced in the main scenarios because it suited their desire to raise interest rates to counter the potential consequences of rising house prices and household indebtedness, even if they hadn’t much faith in the inflation forecast when making their judgement about the repo rate. If so, they went along with a forecast which did not really represent their own views. They should either have clarified that they were pursuing an objective other than meeting a target for inflation, or pressed more strongly for a different forecast to be published. (p. 84)

They also state (italics added):

It is clear that by 2012 the majority on the Riksbank Board were sufficiently concerned about developments in house prices and the growth of household credit to set the repo rate at a level higher than was justified by a strict application of targeting inflation two years ahead. There were three problems with such a strategy. First, the concerns over financial stability held by the majority were never explained within a clear conceptual framework. Second, it was not easy to reconcile the objective of “leaning against the wind” with the official mandate of the Riksbank to pursue price stability. Third, no empirical evidence was produced on the magnitude of the costs and benefits of pursuing such a strategy. (p. 92)

Actually, there was already then considerable empirical evidence of the high costs and small benefits of leaning against the wind,[5] but there is certainly more now, for instance, here.

[1] Under the assumption of an Okun coefficient of 2, which number was used by the Riksbank at that time, this translates into an output gap of about minus 5%. (The Okun coefficient is the the ratio of percentage point changes in the output gap to percentage point changes in the unemployment rate.)

[2] My Brookings paper Svensson (2011) and my November 2010 speech discuss and criticize the “growth stabilization argument” and other arguments for the policy tightening 2010-2011. The Brookings paper also argues that the unemployment gap is a more reliable indicator of resource utilization than the output gap. In particular, it contains a comparison of Fed and Riksbank monetary policy 2010-2011, arguing that the Fed did the right thing and the Riksbank the wrong thing.

[3] Furthermore, as my colleague Karolina Ekholm and I consistently argued, the Riksbank’s low inflation forecast was nevertheless biased upward, and the high unemployment forecast was biased downward, because the forecasts were based on an assumption of unrealistically high foreign interest rates. Everything else equal, this resulted in a forecast of a weaker krona, with higher inflation and export demand, and lower unemployment as a consequence. Goodfriend and King agree with this bias in their discussion of policy after 2011, but the same assumption and bias were present also in 2010 and 2011.

[4] My translation of the text below the picture:
DISASTER PATH. The interest-rate forecast “too high”. If the market would believe the Riksbank’s interest-rate path, it would be a disaster. This says Lars E.O. Svensson, one of the six members in the Rikbank’s executive board, who wants a much lower interest rate the next few years than what the majority in the executive board wants. In an interview with Svenska Dagbladet Business, Lars E.O. Svensson says that he on several occasions has asked his colleagues why they don’t agree with him, but that no answer has been forthcoming.
— I am surprised about the disagreement about some issues that should be obvious, for instance, that one should not increase the interest rate when the forecast for inflation and resource utilization is too low, he says.
Translation of the legend in the picture:
The Riksbank’s forecast (Riksbankens prognos)
Market expectations (Marknadens förväntningar).

[5] A study by Claussen, Jonsson and Lagerwall (2011) (included in the Riksbank’s inquiry into the risks in the Swedish housing market) shows that it would have been very costly in terms of both production and inflation to try to prevent the upswing in Swedish housing prices from 2004 onwards using monetary policy. Keeping housing prices at a long-run trend level in 2004-2010 would have required repo-rate increases over several years of up to 5 percentage points in both 2006 and 2007. GDP growth would then have been on average almost 2 percentage points lower up to 2010, which corresponds to an accumulated GDP loss of around 12%. CPIF inflation would have been on average 3 percentage points lower during the same period. The study confirms for Sweden what several other international studies have concluded, such as Assenmacher-Wesche and Gerlach (2010).