RDP 2008-05: Understanding the Flattening Phillips Curve 1. Introduction

In recent years, inflation appears to have become less responsive to fluctuations in output and unemployment – that is, the Phillips curve has become ‘flatter’. This has been documented for the United States by Roberts (2006), among others, and a similar phenomenon seems to have occurred in other countries as well (for example, see Beaudry and Doyle 2000 for Canada).

A decline in the output-inflation trade-off, if it has occurred, would have consequences for monetary policy. As discussed in Bean (2006) and Mishkin (2007b), a benefit is that higher levels of the output gap and lower levels of unemployment would be less inflationary. The problem is that inflation, once established, would be harder to bring down.

While the stylised fact of a flatter Phillips curve has been reasonably well established, the precise reasons for this change are not well understood. Firmer anchoring of inflation expectations is one possibility, advanced by Roberts (2006), Williams (2006) and Mishkin (2007b), among others. (This line of reasoning has tended to emphasise the effects of anchoring on inflation persistence rather than the responsiveness of inflation to fluctuations in real activity.) Others, such as Borio and Filardo (2007) and Razin and Binyamini (2007), cite the effects of globalisation.

The purpose of this paper is, therefore, to understand why the Phillips curve seems to have become flatter, using insights from new-Keynesian macroeconomic theory to dissect the linkages between real activity and inflation. Variants of the new-Keynesian framework are extensively used in macroeconomic models at central banks worldwide (for example, at the Riksbank – Adolfson et al 2007, and the Bank of Canada – Murchison and Rennison 2007). According to this perspective, a fruitful way to think about the reduced-form output-inflation nexus is in two stages: first, as a relationship between the output gap and costs; and second, in terms of the linkage between costs (or more precisely, the current and expected future costs) and inflation. A reduction in the overall sensitivity of inflation to output may result from a change in either one of those two stages.

The paper proceeds as follows. Section 2 (briefly) documents the change in the reduced-form Phillips curve in the United States and a small open economy, namely Australia. Section 3 reviews the new-Keynesian inflation model and discusses why a change in the reduced form need not imply a change in firms' price-setting behaviour. In an effort to determine whether there may also have been a change in the structural inflation equation, in Section 4 we estimate the new-Keynesian Phillips curve, finding that there does appear to have been a reduction in the responsiveness of inflation to marginal costs. Section 5 considers several possible explanations for these findings, none of which is entirely satisfactory.