New Ekonomistas post (in Swedish). Here is an English translation.
Previously, I have reported a so-called counterfactual analysis of the outcome for the Swedish economy from a policy rate from June/July 2010 of 0.25 percent instead of the Riksbank’s rate increases and higher policy-rate path. As far as can be judged, the outcome would have been much better. Inflation would have been much higher and very close to the target, and unemployment would have been much lower and closer to a reasonable long-run sustainable rate. In addition – perhaps somewhat surprising for some – the debt ratio would have been somewhat lower, not higher. This analysis has recently been criticized by Per Jansson, a member of the Riksbank’s Executive Board. But his criticism takes an unusual and unexpected form. Jansson does not question the actual calculation of the effects of a low policy-rate path, and he does not present any new and better analysis. Instead, according to his view, such a low policy rate was “simply not realistic.”
In my world, such a policy [with a policy rate of 0.25 percent from June/July 2010] is simply not realistic, and it is hence excluded as a meaningful comparison.
The new Board member Cecilia Skingsley has on Swedish Television expressed herself in similar terms. But Jansson and Skingsley are guilty of an error in their thinking, by mixing two very different ways of evaluating monetary policy.
One way, so-called ex post analysis, evaluates monetary policy after the outcome, that is, after the event. Such an analysis compares the actual policy and its outcome and target fulfillment with a counterfactual policy and its outcome and target fulfillment. It is such an ex post analysis that I have done.
The other way, so-called ex ante analysis, evaluates the policy, taking into consideration only the information available to the policymaker at the time of the decision. Such an analysis considers whether the Riksbank, given its information at the time of the decision, made reasonable forecasts of inflation, unemployment and other variables, and whether the Riksbank, given these forecasts, made the decision about the policy rate and policy-rate path that seemed to lead to the best target fulfillment. It seems to be such an ex ante analysis that Jansson (and Skingsley) are referring to. Mixing the two kinds of analysis leads to considerable confusion of the issue. Here I try to sort out the issue. It turns out that Jansson’s (and Skingsley’s) reasoning does not withstand scrutiny.
In this Ekonomistas post (English translation) I have reported a counterfactual ex post analysis, a comparison after the event of the actual outcome in Sweden, of inflation, unemployment, and the debt ratio, to the counterfactual outcome, if the Riksbank had kept the policy rate at 0.25 percent from June/July 2010 instead of rapidly increasing it to 2 percent and then slowly lower it to 1 percent. (I write June/July 2001, since the relevant policy meeting was held on June 30, 2010, and the policy decision was announced on July 1.) The result of this comparison is shown in figure 1 (if you right-click the figure, you can open a larger figure in a separate window).
The red curves show the actual outcome for the policy rate, CPIF inflation, unemployment, and the debt ratio. The blue curves show the counterfactual outcome for a policy rate of 0.25 percent from June/July 2010 (that is, estimates of the counterfactual outcome for inflation unemployment, and the debt ratio). We see that, for the low policy rate, CPIF inflation would have stayed pretty stable around 2 percent, instead of falling to 1 percent and lower. Target fulfillment for CPIF inflation would have been as good as is possible. At the same time, unemployment would have fallen to around 7 percent, and thus become about 1.2 percentage points lower. With a labor force of about 5 million people, this means that about 60 000 fewer people would have been unemployed. Target fulfillment for unemployment would have been much better than currently.
We also see, that the lower policy rate would have led to a lower debt ratio, not a higher one. This is because a lower policy rate would have increased the denominator (nominal disposable income) faster than the numerator (nominal debt). Then the debt ratio would have fallen. Instead of 174 percent in the beginning of 2013, it would have been about 3 percentage points lower, 171 percent. (See this paper for details.) If one believes that any risks associated with household debt vary with the debt ratio, any risks would have fallen. However, 171 compared to 174 percent is such a small difference that it would imply an insignificant reduction of any risks. But it is definitely not a matter of a substantial increase in any risks, as the Riksbank seems to have believed.
This analysis has recently been criticized in a speech by Per Jansson, a member of the Riksbank’s Executive Board.
One might think that a natural way to criticize my analysis would be to present a better analysis of the outcome for inflation, unemployment, and the debt ratio of the counterfactual policy rate of 0.25 percent. Especially since Jansson, with respect to the outcome for inflation and unemployment, has access to the same method that I have used and that has been developed at the Riksbank. This would have been constructive and moved the discussion forward. Unfortunately, this is not what Jansson has done.
Instead, Jansson main’s argument is, somewhat unexpectedly, that a policy rate of 0.25 percent from June/July 2010 was “simply not realistic.” Under the heading “A crisis rate for four years is no realistic comparison,” Jansson maintains:
What this argument appears to disregard is that when the Riksbank began to raise the repo rate in the middle of 2010, the situation was such that the Swedish economy had recovered surprisingly quickly from the crisis (see figure 5 [Jansson’s figure 5 is inserted below]). Growth was very fast, around 6 per cent for the first quarter when calculated as an annual rate, with a final outcome of more than 6 per cent for the year as a whole. Almost all indicators of economic activity were pointing upwards and the inflation rate was almost 2 per cent, in terms of the CPIF. The traditional reasons alone, without taking into account risks linked to household debt, gave a strong indication that it was time to begin increasing the repo rate. In addition, although lending to households had declined from double digits prior to the crisis, it was still increasing by 9 per cent, which was much faster than disposable incomes.
…In my world, such a policy is simply not realistic and must be disregarded as a meaningful comparison.
What Jansson is talking about in the above quotation is clearly an ex ante analysis, that is, an analysis of the policy decision given only the information available at the time of the decision. This is something completely different from my ex post analysis, which is analysis after the event. (See this paper for a more detailed discussion of these two kinds of analysis.) Jansson thus seems to mean that, given the information available at the decision, keeping the policy rate constant was “simply not realistic.” But can really a policy that after the event fulfill the objectives so well be disposed of as “not realistic”? Apparently Jansson is not questioning the result that a policy rate at 0.25 percent would have lead to much better target fulfillment for inflation and unemployment. This naturally leads to the question of whether the Riksbank had any excuse for not keeping the policy rate at 0.25 percent.
So, did the Riksbank have any excuse? Was really an unchanged policy rate “simply not realistic” in June/July 2010? Did my then colleague Karolina Ekholm and I dissent in favor an unrealistic policy? In order to assess this, one has to do a substantial ex ante analysis. As discussed in this paper, an ex ante analysis is best done with the help of the forecasts for inflation and unemployment that were done at the time of the decision. Figure 2 (figure 1-3 from this paper) shows how the Riksbank’s forecasts for inflation and unemployment looked like at the decision in June 2010. In the figure, as a comparison, the Federal Reserve’s forecasts in June 2010 for inflation and unemployment in the US are also shown.
The upper right panel shows the Riksbank’s forecast for CPI inflation and the Fed’s (FOMC’s) forecast for PCE and core PCE inflation. The Fed was at the time generally seen has having an unofficial inflation target of 2 percent, thus at the same level as the Riksbank’s official target. The lower right panel shows the forecast for unemployment by the Riksbank and the Fed. The horizontal lines show the Riksbank’s and the Fed’s estimate of the long-run sustainable rate of unemployment.
We see that for both the Riksbank and the Fed, the inflation forecast was clearly below the target and the unemployment forecast was substantially above the long-run sustainable rate. In this situation, the Fed kept the policy rate at its low level and started to prepare QE2, the second round of large-scale asset purchases, which was initiated in the fall of 2010 with the purpose to lower long interest rates. This in order to increase the inflation towards the target and reduce unemployment towards the long-run sustainable rate. Such a policy was evidently perfectly realistic for the Fed.
The Riksbank, in contrast, started a period of policy-rate increases, increasing the policy rate up to 2 percent in July 2011. This in spite of the inflation forecast being too low and the unemployment forecast being too high. As we see in figure 1, the result was CPIF inflation substantially below the target and unemployment far above a reasonable long-run sustainable rate.
Jansson maintains that the low policy rate was unrealistic for two reasons. First, because the Swedish economy had “recovered surprisingly quickly from the crisis,” GDP growth was high, and CPIF inflation was close to 2 percent.
The argument that CPIF inflation was close to 2 percent can be disposed of right away. Since the economy and the inflation respond with a lag to changes in the policy rate, it is the forecast for future inflation that is relevant, not the current inflation at the time of the decision. And as we see in figure 2, the CPIF forecast was below the target.
The argument that the economy had recovered is difficult to maintain, since the unemployment rate was sky-high, close to 9 percent, far above the rate before the crisis. And, as we see in figure 3, the GDP level was still below the pre-crisis level. The high growth rate was also from the very low level that GDP had fallen to in the beginning of 2009. We also see that the GDP level, compared to the pre-crisis level, had only recovered to the same level as in the US, after having fallen deeper in Sweden than in the US. The GDP level was (and is) still very far from a linear trend for 1993-2007 (it was and is even further below an exponential, that is, loglinear, trend).
Thus, the economy had not recovered. After the big fall, a substantial period of above-normal growth would have been required in order to recover the fall in GDP and employment, and undo the rise in unemployment. But, as we see in figure 3, the high growth rate (the steep slope) was just temporary, and GDP leveled off towards the end of 2010, after the policy-rate increases had been initiated. The growth rate fell and was much lower during 2011. We also see in figure 3 that the GDP level in Sweden up to now has followed a path similar to or somewhat weaker than that of GDP in the US. And the path of GDP in the US has been seen as very unsatisfactory there. A lower policy rate in Sweden had clearly lead to a better outcome for GDP, employment, and unemployment here.
Second, Jansson maintains that the low policy-rate path was unrealistic because of high growth of household debt. But regardless of how high that growth was, the relevant thing is how differently household debt would evolve with the lower policy-rate path. Also, as is hopefully by now well known, here it is not sufficient to just look at nominal debt – one has to look at the whole picture, including households’ balance sheets, payment capacity, and resilience against disturbances, and how different that picture would look with a lower policy-rate path. Had Jansson done this, he would have moved the discussion forward.
Furthermore, Jansson disregards, strangely enough, that I have actually done an analysis of the impact on the debt ratio (not because the debt ratio is the most relevant indicator, but because the Riksbank focuses on it). This analysis is further discussed in this Ekonomistas post (English translation) and this paper. As mentioned and shown in figure 1, as far as can be judged, a lower policy-rate path would have lead to a lower debt ratio, not a higher. It seems highly unlikely that a lower policy-rate path would have substantially increased the debt ratio and any risks associated with it, which, however, seems to be Jansson’s starting point.
Thus, Jansson’s arguments why a low policy-rate path was unrealistic – because the Swedish economy had recovered, GDP-growth was high, CPIF inflation was close to 2 percent, and household debt grew – do not withstand scrutiny.
My conclusion from this is that it had been not only realistic, but the right thing to do, in June/July 2010, to direct monetary policy towards fulfilling the inflation target and – in line with the preparatory and explanatory works of the Riksbank Act – also stabilize unemployment around a long-run sustainable rate. With this direction and a realistic and low policy-rate path from June/July 2010, inflation would, as far as can be judged, now be substantially higher and closer the target. Unemployment would also have been much lower and closer to a long-run sustainable rate. In addition, the debt ratio would, as far as can be judged, have been lower, not higher. In any case, the household debt ratio, real debt and loan-to-value ratio would not have been so much higher that it would have any noticeable effect on any risks associated with the debt.
It would have been interesting if Jansson, instead of loose and confusing talk, had presented a substantial and constructive analysis of the outcome for inflation, unemployment and debt for a low policy-rate path from June/July 2010. Absent such an analysis, his discussion carries little weight.