Inflation expectations have become a major indicator to assess the credibility of a central bank’s inflation target. Therefore, the monetary policy and the communication strategy of inflation targeting central banks like the Federal Reserve Bank of the US (Fed) and the European Central Bank (ECB) is concerned with a firm anchoring of inflation expectations at their targets. Yet, long-term inflation expectations did not always coincide with the inflation targets in the recent past. In the aftermath of the 2008 Financial Crisis, inflation expectations in the US and Europe fell short of their respective central banks’ inflation targets. Consequently, unconventional monetary policy measures by the Fed and the ECB sought to re-anchor too low inflation expectations back to the target level. In the 2020s, the challenge faced by central banks is completely reversed: International supply chain disruptions that started to show in mid-2021 as a result of the Covid-19 pandemic and the escalation of the conflict between Russia and Western Europe in 2022 have led to a surge of inflation to levels and an increase in macroeconomic volatility not seen since the beginning of the Great Moderation. The adoption of average inflation targeting by the Fed (see Powell 2020) and the ECB (2021), alongside the slow shift to the challenges of this new regime may have given an impression of higher tolerance for inflation, see e.g. Schnabel (2022). In the course of 2022, longer-term inflation expectations, even those of professional forecasters, have steadily increased to levels well above the inflation targets. This poses a potential threat to the credibility of central bank’s inflation targets and raises the risks of a de-anchoring of inflation expectations.
This dissertation examines the anchoring of inflation expectations from three perspectives. Parts of the following work were conducted within the DFG-project “The Anchoring of Inflation Expectations.” While the papers were motivated by the too-low inflation experience, the recent rise in inflation has brought inflation expectations back into the limelight of current discussions among central bankers and macroeconomists for yet another reason. The findings in this dissertation generally apply when inflation expectations are above target.
The first chapter, which is joint work with Dieter Nautz, investigates the role of long- term inflation expectations for the monetary transmission mechanism and the conduct of monetary policy in a structural VAR framework. It is motivated by preceding work of Hachula and Nautz (2018) and Nautz et al. (2019), who find that long-term inflation expectations react to a macro news shock in bi-variate models of short- and long-term inflation expectations. The macro-news shock identified in bi-variate SVARs is a conglomerate of all structural economic shocks related to short-term macroeconomic developments. Yet, by abstracting from economic key variables, like economic activity, inflation, and interest rates, bi-variate models of short- and long-term inflation expectations cannot account for the various shocks considered by macroeconomic theory. As a result, following Leduc et al. (2007) an ever- growing literature embeds inflation expectations data into empirical models. In line with a re-anchoring channel of monetary policy, Andrade et al. (2016) and Doh and Oksol (2018) find that the central banks’ announcements of asset purchase programs play an active role in steering inflation expectations toward the inflation target. However, the literature does not show how the management of inflation expectations by the central bank affects the dynamics of actual inflation and the transmission of monetary policy shocks.
To fill this gap, we empirically investigate the role of long-term inflation expectations for the monetary transmission mechanism. Following D’Amico and King (2017) and Jarociński and Karadi (2020), we employ a minimal set of uncontroversial sign restrictions to identify a monetary policy shock in a structural VAR including inflation expectations. In contrast to Clark and Davig (2011), our results show that long-term inflation expectations do respond to monetary policy shocks. On impact, monetary policy shocks account for a non-negligible fraction of the variation of long-term inflation expectations.
To shed more light on the role of long-term inflation expectations for the monetary transmission mechanism, we compute two counterfactual scenarios. The first scenario shows that, in line with a re-anchoring channel, the immediate response long-term inflation expectations contributes to the reaction of actual inflation to a monetary policy shock. The second scenario shows that the endogenous reaction of monetary policy to expectations shocks has contributed to stabilizing expectations during the zero lower bound period. These results suggest that monetary policy can act to re-anchor inflation expectations if required.
The second chapter uses a macro model with asymmetric information, unobserved components, and an explicit learning mechanism to strengthen the theoretical link with macroeconomic theory, compared to the SVAR approach of the first chapter. Since the credibility of a central bank’s inflation target should not be assumed a priori to be always perfect, even in the presence of officially announced targets, a key feature of the model is that it allows for an explicit notion time-varying credibility and anchoring of the perceived target by the public to the actual inflation target. This model contributes to the re-emerging literature that assumes imperfectly informed agents learn about the economy and form expectations based on recent data.
While most papers focus on the learning from short-term inflation surprises (see e.g. Carvalho et al. 2022; Jorgensen and Lansing 2022), Chapter two models how the public learns from central bank’s interest rate policy about the imperfectly observed and potentially time-varying inflation target. To that end, the learning mechanism of Kozicki and Tinsley (2005) is extended by allowing for breaks in the learning rate and the shock variances according to the broad US monetary policy regimes. Implications for the optimal rate of learning and the degree of anchoring are derived. The breaks in variances and the learning rate give rise to time-variation, including in the degree of anchoring. To estimate the model efficiently in a Bayesian framework, I extend the precision based framework for unobserved components models of Chan and Eisenstat (2018) to the multivariate case.
The estimated model shows that imperfect credibility is most pronounced after the Volcker Disinflation and to a lesser extent in the aftermath of the Great Financial Crisis. While credibility was poor due a too-high perceived inflation target by the public in the former, the reverse is true for the latter period. The estimated learning rates are lower than the optimal rate in all monetary policy regimes.
The third chapter focuses on yet another dimension of the anchoring of inflation expectations, namely volatility. In particular, the chapter investigates whether there are common factors in the volatilities of macro variables and US inflation expectations. However, the vast majority of the SVAR literature that allows for stochastic volatility (SV) in the shocks assumes that the volatilities evolve independently and, thus, ignores possible common sources of the time-variation in volatilities. Therefore, I suggest a new model that allows for a flexible number of common factors in the time-varying volatilities. This model nests popular alternatives like the common stochastic volatility (CSV) model of Carriero et al. (2016), which allows for just one factor that scales the entire residuals co-variance matrix of a reduced form VAR. To distinguish the new model from existing models, I refer to it as the Common Factor Stochastic volatility (CFSV) model. The model is based on a non-centered parametrization that allows to check the hypothesis of a reduced rank in the stochastic volatilities and linear restrictions on the loadings of the volatility factors via Savage Dickey Density Ratios (SDDR), thus avoiding the computation of the marginal likelihood that is particularly costly in non-linear models. Simulation evidence demonstrates that the model correctly recovers the reduced-rank volatility structure in a recursively identified SVAR. Finally, I use the CFSV model to revisit the application of Clark and Davig (2011), who estimate an SVAR with a set of typical US macro variables and short- and long-term inflation expectations while allowing independent stochastic volatility (SV). Allowing for CFSV reveals that only two factors drive the time-variation of the five shocks. While the shock to long-term inflation expectations has its own volatility factor, the second factor loads onto both actual inflation and short-term inflation expectations. Such volatility spill-overs suggest that analysis based on independent SV might underestimate the adverse effect of increased inflation volatility that occurs via short-term inflation expectations. Thereby, the range of inflation outcomes might be underestimated as well. This could be particularly important in the environment of increased inflation volatility following the Covid-19 pandemic.
