Versatile forward guidance: escaping or switching?

Abstract

We examine forward guidance when an economy faces negative natural real interest rates and subsequent supply shocks. We introduce two versatile designs: escaping and switching. In the former, the central bank escapes low interest-rate commitment when inflation reaches a self-chosen threshold. In the latter, the central bank can switch from interest-rate forecasts to inflation forecasts any time. Central bankers are scrupulous and face intrinsic (or extrinsic) costs when they deviate from their policy announcements in the future. We show that switching forward guidance is preferable over escaping forward guidance if and only if negative real interest rate shocks are moderate. Furthermore, with the polynomial chaos expansion method and Sobol’ Indices, we identify the decisive parameters and show that our findings are globally robust to parameter uncertainty.

Introduction

“[... ] the most logical way to make such commitment achievable and credible is by publicly stating the commitment, in a way that is sufficiently unambiguous to make it embarrassing for policymakers to simply ignore the existence of the commitment when making decisions at a later time.” (Woodford, 2012).

Forward guidance is the attempt by central banks to affect expectations about the path of inflation and output by announcing the interest rate policy they will follow in the near future. Some central banks have provided forward guidance in the form of regular interest rate forecasts without commitment for many years (e.g. Riksbank, Norges Bank, Central Bank of New Zealand). In the last decade, central banks have used forward guidance frequently when attempting to ease the zero lower bound (henceforth ZLB) constraint on interest rates. Generally, three types of forward guidance have been pursued: open-ended announcements, time-contingent announcements, and state-contingent announcements. As to the last type, the period for which the announcement is supposed to hold depends on macroeconomic conditions such as the inflation or the unemployment rate.

Experience with forward guidance reveals three insights. First, forward guidance can influence the market participants’ expectations, as shown e.g. in Woodford (2012) for the Bank of Canada. Second, when announcements about future monetary policies are made in vague terms or indicate measures that are hard to pin down, the central bank can abandon such announcements rather easily, and the impact of such forward guidance may be negligible. An example of “vague” terms is the announcement made by the Bank of England in August 2013, stating that the bank rate would stay at 0.5% at least until the unemployment rate fell below 7%. Simultaneously, three criteria were set that enabled the bank to break this commitment: $a)$ the consumer price index (CPI) inflation eighteen to twenty-four months ahead was, in the Monetary Policy Committee’s view “more likely than not to be” 0.5% above the 2% inflation target, $b)$ inflation expectations became poorly anchored, and $c)$ the policy imposed potential threats on financial stability (Bank of England, 2013). All criteria require interpretation and thus permit discretion as to their application. In May 2014, the Bank of England continued its low bank rate policy, despite the fact that the unemployment rate had fallen below the announced threshold of 7%. Third, when central banks engage in state-contingent forward guidance with objectively measurable criteria such as the unemployment rate, they may change their commitment over time. An example is the Federal Reserve.

In December 2013 the Federal Reserve announced it would keep the federal funds rate low as long as the projected inflation rate stayed below 2%, even if the unemployment rate fell below 6.5%.

The fundamental problem of forward guidance is to make credible, time-consistent announcements while retaining elbow room for reacting to new shocks. While the literature reviewed in Subsection 1.3 examines optimal forward guidance in the presence of a particular type of shock, in this paper we examine how forward guidance can be designed when the economy is hit by a sequence of different shocks, i.e. first a negative natural real interest-rate shock and then a supply shock. We perform this analysis under the assumption that central bankers display a particular degree of scrupulosity. A scrupulous central banker experiences intrinsic (or extrinsic) costs when he deviates from his earlier policy announcements. Since these costs are not infinite, a central banker may deviate to some extent if it is in his interest. Hence, the presence of scrupulous central bankers creates partial commitment, and we use this term to describe the impact of scrupulosity when central bankers engage in forward guidance.

At a more general level, a scrupulous central banker can be viewed as a parallel to the conservative central banker of Rogoff (1985). The difference is that a scrupulous central banker has the same objective per period loss function as the social loss function, but suffers an additional utility loss if he deviates from his policy announcement in the future.

We compare two promising designs for forward guidance: escaping and switching. With escaping forward guidance, the central banker partially commits to low future interest rates after a negative natural real interest-rate shock. At the same time, he announces a threshold inflation rate. As soon as inflation oversteps this threshold, the central banker is freed from his announcement to keep interest rates low and regains flexibility. With switching forward guidance, the central banker switches from interest-rate forecasts to inflation forecasts when the supply shock hits the economy. That is, he switches from a partial commitment to interest rates to a partial commitment to inflation rates.

We examine escaping and switching forward guidance and compare them to discretionary monetary policy, to standard unconditional interest-rate forward guidance, and to each other. We perform these analyses and comparisons in the New Keynesian Framework, with negative natural real interest-rate shocks and a subsequent supply shock. Monetary policy is performed by a scrupulous central banker who faces intrinsic losses if he deviates from his own forecasts.¹

Our main results are as follows: First, a scrupulous central banker only applies standard forward guidance with zero interest-rate forecasts in a severe downturn. Otherwise, he uses either escaping or switching forward guidance.

Second, with escaping forward guidance, announcing zero interest-rate forecasts in the downturn becomes attractive for any negative natural real interest-rate shock. It matches or lowers social losses at any natural real interest-rate shock level, compared to social losses under standard forward guidance or without forward guidance. The reason is that this avoids the risk of having an excessively low interest rate connected with high inflation in a subsequent major boom. The inflation threshold above which the central banker can abandon his announcement without facing costs increases with the severity of the negative natural real interest-rate shock.

Third, switching forward guidance offers a further prospect for decreasing social losses, and it dominates escaping forward guidance for medium-sized negative natural real interest-rate shocks. The reason is that in a medium range, switching forward guidance is better at balancing gains and costs from partially committing to low future interest rates through the switch to inflation forecasts, since such forecasts moderate inflation immediately when positive supply shocks occur. This does not work efficiently for small natural real interest-rate shocks, since the excessive inflation expectations created by the zero interest-rate forecast in downturns will not be lowered sufficiently by inflation forecasts in normal times. This leads to higher expected losses in downturns under switching forward guidance compared to escaping forward guidance. Furthermore, for large natural real interest-rate shocks, switching forward guidance is unable to elevate inflation expectations as strongly as escaping forward guidance due to the moderating effect of inflation forecasts. Under escaping forward guidance, the central banker simply chooses a threshold which signals that he will never escape without cost, which in our setting maximally increases expectations. To sum up, escaping forward guidance provides desirable levels of inflation expectation in downturns for every level of the natural real interest-rate shock, since an adequate inflation threshold will be chosen. Switching forward guidance produces the same inflation expectation in downturns for different natural real interest-rate shocks. This is appropriate only for the medium range of natural real interest-rate shocks. Compared to escaping forward guidance for this range of shocks, switching forward guidance leads to similar losses in downturns but lower losses in normal times due to the use of inflation forecasting in normal times.

It is useful to put the above results into perspective. At first sight, adding a switching possibility, and thus a further possibility to partially commit to inflation forecasts, should lower intertemporal social losses. However, the switching possibility weakens the impact of partial commitment through scrupulosity to interest rate announcements during downturns. For larger real interest rate shocks, adding partial commitment to inflation forecasts is undesirable, since it weakens the possibility to create sufficient inflation expectations in downturns. For low real interest rate shocks, replacing a partial commitment by another partial commitment is not desirable either, since then, inflation expectations become excessive. It is better in this case to choose a possibility to abandon partial commitment to interest rate announcements under escaping forward guidance in order to induce moderate inflation expectations.

In the online appendix to this paper, we use Sobol’ Indices and the polynomial chaos expansion methodology (henceforth PCE) to assess the global robustness of the results with respect to ranges of admissible parameters. This is a method borrowed from natural science (see Sudret (2008) and Harenberg et al. (2017))² and goes beyond standard local sensitivity analysis. Given plausible parameter spaces it enables us to draw a more complete picture of the sensitivity of a model. In particular, PCE helps to efficiently identify those structural parameters that contribute most to the variance of the model’s output, in our setup, expected social losses. The application of this method to the New Keynesian Model with a scrupulous central banker reveals that, typically, the slope of the Phillips Curve turns out to be the parameter to which social losses react the most. Most importantly, the analysis reveals that the areas for which escaping forward guidance and switching forward guidance dominate other monetary policy approaches are robust to parameter uncertainty. That is, escaping forward guidance remains the optimal approach for substantial or small negative natural real interest-rate shocks, while switching forward guidance is preferred for intermediate negative natural real interest-rate shocks under parameter uncertainty. We also show that the principal gains of applying forward guidance will materialize even for central bankers with a low degree of scrupulosity.

The literature on forward guidance and the degree of scrupulosity of central bankers can be divided into three parts. First, there is a considerable body of literature on the pros and cons of forward guidance. In a recent article, Svensson (2014) finds that applying forward guidance in the form of a published policy rate path for the countries Sweden, New Zealand, and the U.S. has had mixed success. Gersbach and Hahn (2011), Woodford (2012), and the survey by Moessner et al. (2016) provide detailed accounts both of what forward guidance can achieve and of its limitations.

Second, the potential and limitations of forward guidance have been analytically and numerically assessed in Eggertsson and Woodford (2003), Rudebusch and Williams (2008), Gersbach and Hahn (2014), Gersbach et al. (2019), with Boneva et al. (2015) and Florez-Jimenez and Parra-Polania (2016) focusing on threshold-based forward guidance or forward guidance with an escape clause. In our paper, we examine how forward guidance should be performed when the economy is hit by a series of different shocks—real-interest rates and supply shocks. How forward guidance should be designed in such circumstances, is the contribution of this paper. For this purpose, we introduce and compare escaping and switching as promising approaches to forward guidance in such circumstances.

Third, the way in which central banks can increase the commitment power of their announcements—or equivalently, deviations from announcements generate material or immaterial costs for central bankers—has been discussed for several approaches to forward guidance. This literature will be reviewed in the next section when we rationalize the central bank loss function.

It is useful to put our results in perspective. As shown by several authors (see the survey of Blinder et al. (2008)) there are two different ways in which agents can interpret the forward guidance announcements of central banks. Under Odyssean forward guidance, forward guidance establishes, or at least increases, the commitment to follow an announced path of future policy rates and not to respond to shocks within the time frame of the announcement. Under Delphic forward guidance, the central bank communicates to the public its forecast of the economic outlook and the expected interest rate policy consistent with the outlook. Thus a Delphic forward guidance does not involve commitment. As comprehensively examined by Moessner et al. (2016), the forward guidance practised by central banks has mostly been of the Delphic style. For instance, on December 12, 2012, the Federal Reserve announced “[...... ] to keep the target range for the federal funds rate at 0 to 1/4 percent and [it] currently anticipates that this exceptionally low range for the federal funds rate will be appropriate at least as long as [...... ] inflation between one and two years ahead is projected to be no more than a half percentage point above the Committee’s 2 percent longer run goal [...... ]” (Federal Reserve, 2012).

Delphic forward guidance has no effect in many theoretical models and not in our model either, since expectations of households are homogeneous rational and central banks have the same information set regarding economic fundamentals. A recent literature has made substantial progress in modelling Delphic forward guidance. Goy et al. (2018) have not only introduced heterogeneous expectations among households but also bounded rationality with a heuristic belief switching model. Moreover, the credibility of central bank is endogenous. Such a framework allows both Delphic and Odyssean forward guidance and the authors show that both types increase the likelihood of recovery from a liquidity trap. Features of this model—e.g. belief switching and heterogeneity—would also introduce a Delphic component, in addition to Odyssean, to forward guidance in our model. The examination of precise consequences of such a model is left to further research.

In this article, we assume that deviating from the forecast leads to costs for the central banker. We will provide a variety of reasons why such costs can occur in the next section. This implies some form of commitment to the forecast which makes the model Odyssean. However, we only have partial commitment (or quasi-commitment), since the costs to deviate are not infinite and thus central bankers will deviate at the margin later if it is in their interest. As we will see, sometimes, there is no deviation and sometimes deviation happens, depending on the severity of the real interest rate shock. In the latter case, the forecast only serves as a partial commitment device, but is not equal in expectation to the actual policy choice.

The paper is organized as follows: The model under standard forward guidance is presented in the next section. In Section 3 we investigate escaping forward guidance. In Section 4 we introduce switching forward guidance and study its welfare implications. In Section 5 we calibrate the model and provide intuition for our results. Finally, a discussion and the conclusion make up Section 6. A global sensitivity analysis is conducted in Appendix E of the online appendix to this paper.