ABSTRACT
On 10 July 2007, the Chairman of the Federal Reserve Board, Ben Bernanke, (Bernanke 2007), gave a major speech at the National Bureau of Economic Research, in Cambridge, Massachusetts. He chose to devote it to the topic of Inflation Expectations. That choice may have been motivated by the fascinating debate on the role that inflation expectations had played in the Great Moderation – the remarkable decline in the average level and variability of inflation that the world’s leading economies have witnessed since the early 1990s. It may also have been prompted by a small but growing sense of alarm that, with recent rises in inflation, this phase might be drawing to a close. The issue of what tethers inflation expectations to a target, declared or implicit, is now uppermost in many central bankers’ and economists’ minds: that, and fears that, in the wake of the explosion of many primary commodity prices from 2004, they might start to get detached from it. The Bernanke speech followed close upon an important conference in Warsaw, organized by the National Bank of Poland in 2006. That conference was devoted, too, to inflation expectations. This volume presents newly finalized versions of papers that were presented there. Why are inflation expectations so important? Part of the answer runs like this. Beyond some pretty modest levels, both unemployment and inflation are invariably thought of as social ills. They are the classic two ingredients of “economic misery”. It is the monetary authorities’ task to try to minimize them – or more properly, in most approaches, the discounted future stream of the weighted sum of the squares of their deviations from optimal or equilibrium values. There is one thing that everyone agrees must increase at least one of these two sources of misery, and often both. That is the expected rate of inflation. Expectations of higher inflation shift up the short-term trade-off (the Phillips curve) between actual inflation on the one side, and unemployment (or some measure of lost output, which will accompany the unemployment: the difference between actual real GDP and its trend value, the output gap, should betray this) on the other. That must entail higher actual inflation at a given rate of unemployment (or higher unemployment at a given rate of inflation). This is the unambiguously sinister side of inflation expectations. Sometimes a rise in them could appear more benign. All else equal, it should lead to higher
consumption and investment spending, in response to the lower expected real interest rate. This will not occur, however, if the rise in expected inflation is more than offset by an accompanying (or staggered) jump in nominal (policy) interest rates. The Taylor Rule2 will advise just that, since anything less can only cause instability. Assume that the central bank is believed to follow the precepts of the Taylor Rule.3 In this case, a rise in near-future inflation expectations will make agents anticipate higher real interest rates; they will then see that that should squeeze output and jobs later on; and the end result will be a fall in the current expectations of inflation at or a little beyond that horizon. When prices are expected to decline, on the other hand, the zero lower bound to nominal policy rates means that real aggregate demand will be boosted (depressed) by expectations of slower (faster) decline in the rate of inflation. Recent Japanese experiences have shown that this threat is no laughing matter. With this one exception, then, we learn two things: first, that higher (lower) inflation expectations have the potential, typically, to create major economic damage (benefit); and second, that the size and character of those effects will depend critically on how agents think the central bank will react. At its best, a fully credible central bank, committed to fighting inflation as a first priority, should succeed in keeping inflation expectations steady, at least on average over time. Departures from any explicit or inferred target should then be brief – and small. So inflation expectations are of enormous practical importance, a point stressed forcefully recently by Mishkin (2007). If expectations were fully rational, all they would do would be to reflect (presumably the latest) information and knowledge. They would simply be consistent with “the” model, whatever that might be.4 They would not really have any life of their own. And presumably we would observe them only indirectly. There would be an element of circularity: it would be rather like using consumption spending to infer the income expectations upon which they were assumed to be based. True, inflation expectations can be inferred from an estimated model, under the assumption of rational expectations. But we have two other potential sources for them, both of them effectively independent of the model in which we formalize their influence upon the course of actual inflation and output. One is financial data, such as those on the prices and relative yields of both indexed and unindexed bonds. The second is surveys. This volume is devoted to the latter. Why study surveys? One practical answer is provided by Ang et al. (2007). US evidence, they find, shows that inflation expectations culled from surveys actually turn out to be better predictors of subsequent actual inflation than either statistical macroeconomic models, or financial asset price data. A second argument turns on the fact that the size and character of macroeconomic models are disputed: model-based expectations require the investigator to preselect the model. Third, there is the point that financial market data may be contaminated by transaction costs, risk premia, tax treatment questions, and, above all, data paucity. In many countries, indexed bonds are comparatively recent, if they exist at all, and the range of maturities on which they are offered, if any, is invariably very limited. Of course one need not confine oneself to any one source: Fu
(2007) shows that there are interesting lessons to be learnt from combining surveys and financial market data. Nonetheless, survey data are remarkably rich and extensive, and surely merit close attention. And, like Mount Everest, they exist.5
