Tuesday, July 8, 2014

The Unemployment Cost of Below-Target Inflation

Recently, inflation in the United States has been consistently below its 2% target. The situation in Sweden is similar, but has lasted much longer. The Swedish Riksbank announced a 2% CPI inflation target in 1993, to apply beginning in 1995. By 1997, the target was credible in the sense that inflation expectations were consistently in line with the target. From 1997 to 2011, however, CPI inflation only averaged 1.4%. In a forthcoming paper in the AEJ: Macroeconomics, Lars Svensson uses the Swedish case to estimate the possible unemployment cost of inflation below a credible target.

Svensson notes that inflation expectations that are statistically and economically higher than inflation for many years do not pass standard tests of rationality. He builds upon the "near-rational" expectations framework of Akerlof, Dickens, and Perry (2000). In Akerlof et al.'s model, when inflation is fairly close to zero, a fraction of people simply neglect inflation, and behave as if inflation were zero. This is not too unreasonable--it saves them the computational trouble of thinking about inflation and isn't too costly if inflation is really low. Thus, at low rates of inflation, prices and wages are consistently lower relative to nominal aggregate demand than they would be at zero inflation, permitting sustained higher output and employment. At higher levels of inflation, fewer people neglect it. This gives the Phillips Curve has a "hump shape"; unemployment is non-monotonic in inflation but is minimized at some low level of inflation.

In the case of Sweden, the near-rational model is modified because people are not behaving as if inflation were zero, but rather as if it were 2%, when in fact it is lower than 2%. Instead of permitting higher output and employment, the reverse happens. The figure below shows Svensson's interpretation of the Swedish economy's location on its hypothetical modified long-run Phillips curve. The encircled section is where Sweden has been for over a decade, with inflation mostly below 2% and unemployment high. If inflation were to increase above 2%, unemployment would actually decline, because people would still behave as if it were 2%. But there is a limit. If inflation gets too high (beyond about 4% in the figure), people no longer behave as if inflation were 2%, and the inflation-unemployment tradeoff changes sign.
Source: Svensson 2014

As Svensson explains,
"Suppose that nominal wages are set in negotiations a year in advance to achieve a particular target real wage next year at the price level expected for that year. If the inflation expectations equal the inflation target, the price level expected for next year is the current price level increased by the inflation target. This together with the target real wage then determines the level of nominal wages set for next year. If actual inflation over the coming year then falls short of the inflation target, the price level next year will be lower than anticipated, and the real wage will be higher than the target real wage. This will lead to lower employment and higher unemployment."
Svensson presents narrative evidence that central wage negotiations in Sweden are indeed influenced by the 2% inflation target rather than by actual inflation. The wage-settlement policy platform of the Industrial Trade Unions states that "[The Riksbank’s inflation target] is an important starting point for the labor-market parties when they negotiate about new wages... In negotiations about new wage settlements, the parties should act as if the Riksbank will attain its inflation target."

The figure below shows the empirical inflation-unemployment relationship in Sweden from 1976 to 2012. The long-run Phillips curve was approximately vertical in the 1970s and 80s. The observations on the far right are the economic crisis of the early 1990s. The points in red are the inflation targeting regime. The downward-sloping black line is the estimated long-run Phillips curve for this regime with average inflation below the credible target. You can see two black dots on the line, at 2% inflation and at 1.4% inflation (the average over the period). The distance between the dots on the unemployment axis is the "excess unemployment" that has resulted from maintaining inflation below target. The unemployment rate would be about 0.8% lower if inflation averaged 2% (and presumable lower still if inflation averaged slightly above 2%).

Source: Svensson 2014
Can this analysis be applied the the United States? Even though the U.S. has only had an official 2% target since January 2012, Fuhrer (2011) notes that inflation expectations have been stabilized around 2% since 2000, since the Fed was presumed to have an implicit 2% target. The figure below plots unemployment and core CPI inflation in the U.S. from 1970 to 2012, with 2000 and later in red. Like in Sweden, the long-run Phillips curve is downward-sloping in the (implicit) inflation-targeting period. Since 2000, however, average U.S. inflation was 2%, so overall there was no unemployment cost of sustained below-target inflation. The downward slope, though, means that if we get into a situation like Sweden's where we consistently undershoot 2%, which could result if the target is treated more like a ceiling than a symmetric target, this would have excess unemployment costs.

Source: Svensson 2014
Svensson concludes with policy implications:
"I believe the main policy conclusion to be that if one wants to avoid the average unemployment cost, it is important to keep average inflation over a longer period in line with the target, a kind of average inflation targeting (Nessén and Vestin 2005). This could also be seen as an additional argument in favor of price-level targeting...On the other hand, in Australia, Canada, and the U.K., and more recently in the euro area and the U.S., the central banks have managed to keep average inflation on or close to the target (the implicit target when it is not explicit) without an explicit price-level targeting framework.  
Should the central bank try to exploit the downward-sloping long-run Phillips curve and secretly, by being more expansionary, try to keep average inflation somewhat above the target, so as to induce lower average unemployment than for average inflation on target?...This would be inconsistent with an open and transparent monetary policy."


Comments appreciated!