Does Inflation Targeting Affect the Dispersion of Inflation Expectations?

Beginning of article

INFLATION TARGETING (IT) is a monetary policy strategy that has been gaining popularity around the world. Three main benefits, all interrelated, have been associated with IT. First, it successfully lowers inflation and makes it less volatile (Bemanke et al. 1999, Johnson 2002, Levin, Natalucci, and Piger 2004, Vega and Winkelried 2005, Mishkin and Schmidt-Hebbel 2007, Goncalves and Salles 2008). Second, it reduces the real costs of disinflations (Mishkin and Schmidt-Hebbel 2007, Goncalves and Salles 2008). Finally, it anchors long-run inflation expectations at or very close to the inflation target (Bernanke et al. 1999, Johnson 2002, Levin, Natalucci, and Piger 2004, Vega and Winkelried 2005, Gurkaynak, Levin, and Swanson 2006, Mishkin and Schmidt-Hebbel 2007, Goncalves and Salles 2008). Of these, the effect on inflation expectations is, in principle, straightforward, since a key aspect that separates IT from other sensible monetary policies is the public announcement of a numerical target, and the subsequent referral to it in central bank communications. In fact, it is possible that the impact of IT on inflation and on other macroeconomic variables may come through its effect on inflation expectations and on the expectations formation process: e.g., IT could coordinate expectations and, in this way, become the nominal anchor of the economy, or it could be thought of as a commitment mechanism that increases the signal-to-noise ratio in the economy, helping people to make a better informed allocation of resources. For this reason, and in contrast to other investigations that concentrate on the effects of IT on inflation or on macroeconomic variables, we concentrate on the effect of IT on inflation expectations.

By making the inflation target explicit, IT provides a focal point that may anchor inflation expectations. If the central bank does not announce a target and if the performance of the central bank is not evaluated based on a number or a range, then people in the economy need not have the same expectation about the future stance of monetary policy and, therefore, inflation expectations need not be anchored. Indeed, Gurkaynak, Levin, and Swanson (2006), using inflation expectations extracted from market instruments, provide evidence that expectations in Canada, the United Kingdom, and Sweden, all IT countries, seem to be less sensitive to macroeconomic news than inflation expectations in the United States, a non-IT country.

IT may not only affect the level of inflation expectations but also the dispersion of these expectations across economic agents. As an example, take two otherwise identical countries with monetary policies conducive to low and stable inflation, but one with an explicit inflation target (the IT country) and the other with an implicit one. The potential benefit for the IT country is that the target becomes a focal point for the coordination of expectations among agents. In contrast, in the country with an implicit target, economic agents have to estimate the target in order to form their inflation expectations and, therefore, need not have the same inflation expectation. As a result, the dispersion of inflation expectations would be larger in the non-IT country.

In this paper we study how the choice of a particular monetary policy scheme, IT, affects the heterogeneity in inflation expectations. The importance of this heterogeneity for macroeconomic analysis has been emphasized by several authors. Lucas (1973), Phelps (1970), Sims (2003), and Woodford (2002) show, using models of imperfect information, that the real costs of nominal movements in the economy may be related to the dispersion of inflation expectations. Mankiw and Reis (2002) use a model with sticky information and, hence, with dispersion in inflation expectations, to obtain the observed delayed response of inflation to monetary policy shocks. More recently, Mankiw, Reis, and Wolfers (2004, p. 2) go as far as suggesting that "disagreement [about inflation expectations] may be a key to macroeconomic dynamics." Finally, less disperse expectations may enhance the effectiveness of the expectations channel of monetary transmission.

We use a simple macroeconomic model to show that, under IT, the optimal long-run inflation forecast is the target. (1) Since this would be true for each forecaster, under IT the dispersion across forecasters (i.e., the disagreement about inflation expectations) should decrease, eventually collapsing around the target. We test this implication using survey data, collected by the firm Consensus Economics, on inflation forecasts from professional forecasters. We have data per forecaster for 25 countries, of which 12 are industrial countries, 7 are from Latin America, and 6 are from the Asian Pacific region. From the 25 countries, 14 have implemented IT. The data are monthly, with forecast horizons of up to 24 months, and spans 16 years. (2)

Yet, presenting convincing empirical evidence on the effects of IT has proven a difficult task for at least two reasons. First, for what now is a considerable amount of time, favorable conditions worldwide have helped tame inflation around the world (Rogoff 2003, Bernanke 2004, Cecchetti, Flores-Lagunes, and Krause 2006). (3) Among these conditions we have central banks becoming autonomous, fiscal policies more favorable to low inflation (e.g., debt renegotiations and low fiscal deficits), and openness to global trade (e.g., more competitive goods and labor markets). Therefore, in recent times inflation has been under control in most countries. This makes it difficult to identify the specific contribution of IT since, if these conditions are not controlled for, their effects could be erroneously attributed to IT. (4) Second, in particular in emerging countries, IT coincides for some periods with disinflation programs--i.e., a restrictive monetary policy for long periods of time, as well as with other actions such as fiscal retrenchment. Again, if these are not taken into account, their effects could be attributed to IT. In general, the omission of relevant explanatory variables is likely to bias upward (in absolute value) the estimate of the effects of inflation targeting. The problem can be alleviated by adequately controlling for omitted variables such as global inflation and disinflation periods, as well as for other variables that are hard to measure or unobservable, such as the degree of central bank independence.

Our main result is that the dispersion of long-run inflation expectations appears to be smaller under IT regimes than in non-IT ones, after controlling for country-specific events such as the level and the variance of inflation and disinflation periods, and timespecific effects such as global inflation. Thus, we provide evidence that suggests that IT has helped anchor inflation expectations. When we separate the effects between developed and developing countries, we find that the effect is present in the latter and, in line with Johnson (2002), that there seems to be no effect on the dispersion of long-run expectations in the former.

The paper is organized as follows. Section I presents the theoretical model. The data on inflation forecasts are described in Section 2, while Section 3 contains the empirical results. Finally, a discussion and the implications of the analysis are presented in Section 4. Appendix A extends the model to the case of a flexible inflation targeter, while Appendix B extends it to the case of an inflation band targeter.

1. A CANONICAL MACROECONOMIC MODEL

In this section, we use a simple canonical macroeconomic model to define what anchoring of inflation expectations means under IT and to derive the implication that we test in the empirical part.

1.1 Inflation Targeting

Inflation l periods ahead is given by

[[pi].sub.t+l] = [s.sub.t] - [i.sub.t] + [[epsilon].sub.t+l], (1)

where st represents underlying inflationary pressures, [i.sub.t] is the monetary policy instrument, and [[epsilon].sub.t+l] represents unforecastable shocks (with zero mean), s and e are assumed to be independent of the monetary policy action, and the difference between them is that s is realized before the choice of the monetary action while e is realized afterward. Notice that here l represents the control lag. (5) This equation can be derived from a system with an IS and a Phillips curve. In that case, [s.sub.t] would be a vector with variables from both equations.

The central bank is a strict inflation targeter. (6) The central bank's objective in period t is to choose a sequence of current and future instruments [{[i.sub.[tau]]}.sup.[infinity].sub.[[tau] = t] to solve

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (2)

where [delta] is a discount factor, [[pi].sup.T] is the target, and [[OMEGA].sub.t] is the central bank's information set. Since in this simple case the instrument (e.g., the overnight rate) in period t will not affect the inflation rate in period t, but will do so until t + l, we can find the solution to the optimization problem by assigning the instrument in period t to hit, on an expected basis, the inflation target for period t + l, the instrument in t + 1 to hit the inflation target for period t + l + 1, and so on (Svensson 1997). Thus, the central bank can find the optimal instrument in period t as the solution to the simple period-by-period problem

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (3)

The first-order condition to solve (3) is

E [[pi].sub.t+l] | [[OMEGA].sub.t] = [[pi].sup.T] (4)

where the expectation is evaluated at [i.sup.*.sub.t], the …