OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 85, 1 (2023) 0305-9049

doi: 10.1111/obes.12516

Monetary Policy Uncertainty and Inflation

Expectations*
GABRIEL ARCE-ALFARO†,‡ and BORIS BLAGOV§
†Department of Economics, University Duisburg-Essen, Universitätsstraße 12, Essen, 45117,
Germany (e-mail: [email protected])
‡Ruhr Graduate School in Economics, Hohenzollernstr. 1-3, Essen, 45128, Germany (e-mail:
[email protected])
§Macroeconomics and Public Finance, RWI–Leibniz Institute for Economic Research,
Hohenzollernstr. 1-3, Essen, 45128, Germany

Abstract
Do inflation expectations react to changes in the volatility of monetary policy? They
have, but only until the global financial crisis. This paper investigates whether increasing
the dispersion of monetary policy shocks, which is interpreted as elevated uncertainty
surrounding monetary policy, affects the inflation expectation formation process. Based
on US data since the 1980s and a stochastic volatility-in-mean structural VAR model, we
find that monetary policy uncertainty reduces both inflation expectations and inflation.
However, after the Great Recession this link has disappeared, even when controlling for
the Zero Lower Bound.

I. Introduction

‘‘Inflation targeting, at least in its best-practice form, consists of two parts: a policy framework of
constrained discretion and a communication strategy that attempts to focus expectations and explain
the policy framework to the public. Together, these two elements promote both price stability and
well-anchored inflation expectations.’’ - Bernanke (2003)

The seminal contribution of Lucas (1976) has highlighted the importance of

expectations for the effectiveness of economic policy. Inflation expectations play a
key role for wage and price setting (Clarida, Gali and Gertler, 2000; Svensson, 2000).
As a consequence, over the past 40 years, the conduct of monetary policy has markedly
changed with increasing emphasis on communication and transparency (Bernanke and
Woodford, 1997; Blinder et al., 2008). For example, in 1994, the Federal Open Market

JEL Classification numbers: C11, C32, E52.

*We would like to thank Francesco Zanetti, two anonymous referees, Christoph Hanck, Ludger Linnemann,
Torsten Schmidt, participants at the 35th European Economic Association Annual Congress, 28th Society for
Nonlinear Dynamics and Econometrics Annual Symposium, and 13th RGS-Econ Doctoral Conference for their
helpful comments and suggestions.
70
© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 71

Committee (FOMC) began releasing statements regarding their monetary policy decisions,
and in 2012, the US Federal Reserve (FED) officially adopted an inflation target of 2%.
Since the global financial crisis, the main monetary policy instrument – the federal funds
rate (FED funds) – has been constrained by the zero lower bound (ZLB). Recent literature
has looked at the relevance of the ZLB in light of unconventional monetary policy. While
Swanson and Williams (2014) and Debortoli, Galı́ and Gambetti (2019) suggest that the
ZLB has not been particularly binding due to the effectiveness of forward guidance, Ikeda
et al. (2020) find strong evidence that the constraint has been empirically relevant in the
United States and Japan. Therefore, managing inflation expectations has become even
more important for the transmission mechanism of monetary policy.
The effectiveness of monetary policy relies in part on the ability of economic agents to
anticipate monetary policy movements. Consequently, a large part of monetary economics
has dealt with studying the effects of unanticipated monetary policy shocks. However,
a growing literature has also considered a different take on the relationship – how does
unpredictability of monetary policy affect the economy? This is referred to monetary
policy uncertainty (MPU). Although MPU has been studied not least due to its relevance
for central bank credibility (e.g., Stulz, 1986; Neely, 2005; Swanson, 2006), it has
gained considerable emphasis since the global financial crisis, along with other types of
uncertainty.1
A commonality across the empirical MPU literature findings is that, irrespective of the
chosen proxy, an increase in monetary policy uncertainty suppresses economic activity,
increases unemployment, and leads to a decline in prices. The theoretical underpinnings
of these findings suggest that the consumption channel plays an important role – risk-
averse agents hold back consumption, which creates a decline in demand (Mumtaz and
Zanetti, 2013). Essentially, MPU shocks are a materialization of negative demand shocks
and hence propagate through the expectations of economic agents. However, the focus
of the inflation expectations literature has mostly been on the effects of monetary policy
shocks, not on MPU shocks (Leduc, Sill and Stark, 2007; Canova and Gambetti, 2009;
Leduc and Sill, 2013). Not controlling for the uncertainty component might overlook an
important aspect in the inflation expectation formation process. Hence, in this article, we
investigate empirically the link between MPU and inflation expectations.
To do so, an important question is how to measure MPU. MPU has been typically
defined as some function of the ability of economic agents to forecast monetary policy
instruments, that is, interest rates. One example is measuring surprises to agents via options
and yield curve movements (Swanson, 2006; Bauer, 2012; Chang and Feunou, 2014).
More recently, natural language processing (less formally text analysis) has also been
employed to create MPU proxies, either through newspaper-based articles (Baker, Bloom
and Davis, 2016; Husted et al., 2020) or the FOMC meetings (Hansen et al., 2018).
A further example, popular in structural models, is to postulate specific distributional
assumptions on key target central bank variables, either the money supply (Stulz, 1986)
or, more recently, the interest rates (Mumtaz and Zanetti, 2013; Creal and Wu, 2017;

1 See, for example, Bauer (2012); Kang, Lee and Ratti (2014); Chang and Feunou (2014); Mumtaz and Zanetti (2013);

Istrefi and Piloiu (2014); Mumtaz and Theodoridis (2015); Sinha (2016); Creal and Wu (2017); Kurov and
Stan (2018); Hansen, McMahon and Prat (2018); Istrefi and Mouabbi (2018); Husted, Rogers and Sun (2020);
Bauer, Lakdawala and Mueller (2019); Alessandri and Mumtaz (2019).

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
72 Bulletin

Alessandri and Mumtaz, 2019). In these contributions, MPU is modelled in the second-
order moment, that is, the variance of the central bank policy instrument. An increase
in MPU is then defined as an increase in the variance of monetary policy shocks. The
hypothesis is that larger monetary policy shocks worsen forecasts of economic agents and
thus make it harder to anticipate the correct movements of the target variable.
We incorporate MPU following the latter approach and estimate a structural VAR
(SVAR) with stochastic volatility-in-mean to study the interaction between MPU and
inflation expectations. Capturing the joint dynamics of inflation, inflation expectations,
and economic activity is natural in the framework of SVARs. Our work is related to
the growing literature, which focuses on the effects of economic policy uncertainty on
the economy (Bloom, 2009; Mumtaz and Zanetti, 2013; Baker et al., 2016; Bachmann
et al., 2019). We focus specifically on MPU, departing from recent works that look at
the macroeconomic effects of changes in broader measures of uncertainty (Istrefi and
Piloiu, 2014; Fernández-Villaverde et al., 2015).
In a sample spanning from 1982 to 2019, we find that, on average, in the United States,
short-run inflation expectations do indeed decline following an MPU shock, although not
in the same magnitude as inflation, suggesting that expectations are rigid. On the other
hand, long-run inflation expectations, which are typically found not to react to monetary
policy shocks (Canova and Gambetti, 2009), do not seem to be affected by MPU shocks.
Furthermore, we show that the relationship between MPU and inflation expectations has
not remained stable over time. Since the Great Recession, short-run inflation expectations
have not reacted to MPU shocks, while inflation has. These findings suggest that, while
MPU might have been important in the past for the expectation formation process, this
has not been the case over the past decade, even when we control for the ZLB.
The remainder of this paper is structured as follows. The next section lays out the
methodology used in this article. Section III is devoted to the summary of the data set.
Sections IV and V discuss the results and Section VI concludes.

II. Modelling monetary policy uncertainty

This section presents the adopted empirical strategy for estimating the effects of uncertainty
on inflation expectations. We follow the approach of modelling MPU as a second-moment
shock to the monetary policy instrument (e.g. Stulz, 1986). We adopt the framework
of Mumtaz and Zanetti (2013), estimating a stochastic volatility-in-mean SVAR, which
has already been employed for the analysis of MPU on economic activity (Mumtaz and
Zanetti, 2013; Alessandri and Mumtaz, 2019). In this set-up, MPU is captured by the
time-varying variance of monetary policy shocks, which are extracted from the central
bank interest rate series. In the spirit of Swanson (2006), an increase in the variance lowers
the ability of economic agents to forecast interest rates.2 For the remainder of this paper,
we will use the terms uncertainty and volatility interchangeably.3

2 Thisdefinition of uncertainty is popular in the literature. For example Jurado, Ludvigson and Ng (2015) frame
financial and macroeconomic uncertainty as the variance of a time-varying forecast error from a dynamic factor
model.
3
We do acknowledge that there is a difference between implied and realized volatility. Our results draw on a measure
based on realized interest rate volatility. See Bachmann et al. (2019) for a discussion on the subject.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 73

An appealing property of the model is that it constitutes a one-step procedure in

modelling uncertainty, that is, both the coefficients and the uncertainty measure are
estimated jointly from the data, thus providing an important feedback loop. This is in
contrast to a model where the proxy is constructed first from an exogenous source and
then put into a model as it avoids potential measurement errors.4

SVAR with stochastic volatility in-mean

The framework of Mumtaz and Zanetti (2013) may be summarized by the following key
equations. The VAR is given by:


P 
M
1/2
Zt = c + βj Zt−j + γm h̃t−m + t et , et ∼ N(0, IK ), (1)
j=1 m=0

where Zt is a vector of endogenous variables of dimension K × 1 and the vector

h̃t = [h1t , . . . hKt ] contains the log volatilities of the K structural shocks. The shocks
are extracted from the reduced-form variance–covariance matrix t using an orthogonal

decomposition t = A−1 Ht A−1 and A−1 collects the contemporaneous relationships
between the variable. Ht is a diagonal matrix with exp{h̃t } on its diagonal. The processes
for the log volatilities are assumed to be autoregressive of first order, of the form:

h̃t = h̃t−1 + ηt , ηt ∼ N(0, ξ ), E(et ηt ) = 0, (2)

such that  is a K × K diagonal matrix and ηt = [η1t , . . . , ηKt ] . Furthermore we assume

that the first- and second-moment shocks are uncorrelated.5
Let rt denote the monetary policy instrument in the vector of endogenous variables,
Zt . Then hrt is our MPU measure, which affects the levels of the rest of the variables
through the γ coefficients. The innovations ηt in equation 2 are the uncertainty shocks in
our framework with ηrt being the MPU shock. An increase in ηrt means that the monetary
policy instrument may take a wider range of values as the shock dispersion increases.

Model with time-varying coefficients

This model captures the average response to uncertainty shocks over the full sample.
Throughout the last 40 years, the US economy has experienced significant events,
from the great moderation to the global financial crisis, which could have contributed
to changes between the relationships of economic variables. Furthermore, it has been
shown that inflation expectations depend on the own experience of the economic agents,
that is, people who lived through periods of high inflation expect, on average, higher

4 This is not to say that one-step approaches do not come without drawbacks. For example, model misspecification
would undermine the generated regressors (the uncertainty indicators). The choice of one- versus two-step approaches
presents a trade-off between internal consistency and robustness to misspecification. For discussion of one-step
versus two-step estimations of uncertainty indices, see Bianchi, Kung and Tirskikh (2018).
5
This assumption eases the computational burden but it may be relaxed, as, for example, in Alessandri and
Mumtaz (2019).

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
74 Bulletin

inflation (Malmendier and Nagel, 2016; Draeger and Lamla, 2018). Finally, a large part
of the literature on uncertainty has shown that the relationship between uncertainty and
economic activity might be nonlinear (Caggiano, Castelnuovo and Groshenny, 2014;
Alessandri and Mumtaz, 2019), and specifically monetary policy shocks (Castelnuovo
and Pellegrino, 2018). Therefore, we deem it important to investigate whether the inflation
expectations formation process has evolved over time and how the link between MPU
and expectations has behaved throughout different periods.
To do so, we estimate an extended version of the model by allowing the coefficients to
change over time, along the lines of Primiceri (2005) and Mumtaz and Theodoridis (2020).
Thus, our analysis falls in a broad class of studies analysing the time variation of economic
shocks (Benati and Surico, 2008; Mumtaz and Zanetti, 2015; Liu et al., 2019).
We introduce time-varying parameters (TVP) for the coefficients of the estimated
VAR, β and γ , thus capturing potential changes among the economic variables and how
uncertainty is perceived over time. Equation (1) is modified to incorporate time variation
in the following way:


P 
M
1/2
Zt = ct + βj,t Zt−j + γm,t h̃t−m + t et , et ∼ N(0, IK ), (3)
j=1 m=0

We postulate a random walk process for the coefficients:

Bt = Bt−1 + vt , vt ∼ N(0, Q), (4)

where Bt = [βt , γt ] and Q governs the amount of time variation in the parameters.

Estimation strategy
Due to the presence of the volatility terms in equation (1), the conventional maximum
likelihood approach is not applicable. The model is estimated via Bayesian methods
with Gibbs sampling, that is, drawing the parameters iteratively from their conditional
distributions.
The estimation procedure for the TVP and no-TVP specification is mostly the
same. The parameters may be divided into several blocks based on their distributional
assumptions. The reduced-form coefficients B = [β, γ ], and Bt = [βt , γt ] in the TVP case,
respectively, along with the Q matrix, the stochastic volatility block H = {H1 , . . . , HT },
where H is a diagonal matrix containing the h̃, the parameters  in the equation, and the
contemporaneous responses A.
In order to simplify the exposition, we introduce notation  = {A, B, Q, H, }, which
collects the different blocks of parameters. Let  −i denote the exclusion of the ith block
such that  −B = {A, Q, H, }.
To conduct inference, we draw the ith block from the conditional probability
distribution p(i| −i ), which is derived as a function of a conjugate prior distribution
p(i). The prior for the reduced-form coefficients B and h̃ is based on a GLS estimation on
a training sample. For arbitrary starting values, the estimation proceeds in the following
iterative procedure:

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 75

1. Conditional on H, A may be drawn as a linear regression from p(A| −A ), given the

form in equation (2) in conjunction with the algorithm of Cogley and Sargent (2005).
2. The reduced-form parameters, B, are obtained from a linear regression with
heteroskedasticity and may be estimated via GLS. Following Carter and
Kohn (1994), we introduce stochastic volatility via a matrix Q, employ the
Kalman filter for t = 1, . . . , T to get βT|T and γT|T and draw the parameters
from p(βT|T , γT|T |Q,  −B,−Q ).
3. Conditional on the draws for B, Q may be drawn from an inverse Wishart distribution.
4. For the stochastic volatility estimation, we follow Cogley and Sargent (2005),
where the draws for H may be obtained using a Metropolis–Hastings algorithm
conditional on the model parameters.

The only notable difference when estimating the TVP specification is that the time
variation in the model is governed by the process given in (4), for which the specification
of Q is of importance. The constant parameter version is nested here for extremely
low values of Q, while high values permit larger jumps in the parameters. Given that
the random walk process introduces potential explosiveness, exceptionally high values
could make inference impossible. To deal with this issue, in the seminal contribution
of Primiceri (2005), Q is scaled a-priori via an additional parameter, kQ . This choice
is of specific importance. Our goal is to find the appropriate amount of time variation
without imposing overly strong restrictions on the parameters’ movement nor running
into estimation difficulties. In Primiceri (2005) kQ is chosen using a grid-search with the
aim to maximize the marginal likelihood of the model with an optimal value of kQ = 0.01
for US data. However, our data set is monthly, hence there is no guarantee that this is an
appropriate choice for our application.
We follow recent advancements in Bayesian computation of TVP models and treat
kQ as a hyperparameter to be estimated from the data as in Amir-Ahmadi, Matthes and
Wang (2018) by using a random walk Metropolis–Hastings algorithm with an inverse
gamma distribution as a prior. We plot the posterior densities of kQ in the Appendix,
Section A. The corresponding estimates of kQ lead to a median value of 0.063. We use
this median of the posterior density of kQ to calculate the impulse response functions of
the TVP model.

III. Data and estimation

Our data set consists of five macroeconomic variables: the civilian unemployment rate;
the long-run and short-run inflation expectations; the consumer price index (CPI) inflation
rate; and the monetary policy rate. Given that our measure of uncertainty relies on the
volatility of the variables, the ZLB presents a significant issue. For this reason, we choose
the shadow short rate (SSR) as a proxy for the main monetary policy instrument. Thus,
we also capture the unconventional monetary policy measures, which are reflected in the
SSR. For robustness, we also consider the model with federal funds rate (FED Funds).
The key variables of interest in our study are the inflation expectations. Therefore,
we use the proxies developed in Haubrich, Pennacchi and Ritchken (2012), where
expectations are extracted using a term-structure model of the interest rates. These data

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
76 Bulletin

Figure 1. One-year ahead inflation expectations from the Survey of Professional Forecasters and from
Haubrich et al. (2012). Quarterly frequency.
Source: Federal Reserve Bank of Philadelphia and the Federal Reserve Bank of Cleveland

have several appealing properties. First and foremost, the financial sector is a crucial
channel for the expectations of economic agents; thus, proxies based on interest rate
data should provide timely adjustment of inflation expectations. Second, the expectation
measures are available both for short and long term (1 year ahead and 5 years ahead),
starting in 1982. In contrast, the long-run expectations of the Survey of Professional
Forecasters (SPF) only starts in 2005. Third, the data are at a monthly frequency. This
increases the degrees of freedom in our highly parameterized nonlinear model. Finally,
the term-structure model accounts for liquidity and risk premia. This is an important
difference from other measures of inflation expectations based on interest rate data, like
the treasury inflation-protected securities (TIPS) that do not account for such premia.
However, this comes at a cost, as the proxy is an outcome of a term-structure model, hence
carries model uncertainty, which cannot be incorporated into the analysis. For example,
misspecification could lead to an over- or underestimation of our probability intervals – a
different inflation expectations measure might exhibit different dynamics.6 Figure 1, plots
the quarterly aggregated measure from Haubrich et al. (2012) versus the SPF.
The unemployment rate, CPI inflation, and the FED Funds were obtained from the
Federal Reserve Bank of St. Louis Database. The shadow short rate (SSR) was obtained
from Wu and Xia (2016).7 Finally, we use the estimations from Haubrich et al. (2012) for
the long-run and short-run inflation expectations taken from the Federal Reserve Bank of
Cleveland. The sample ranges from January 1982 to June 2019.

6 We have also explored an alternative specification, where the 1-year ahead SPF data have been interpolated using
a mixed frequency VAR as in Schorfheide and Song (2015). We find that our conclusions remain the same. Results
are available upon request.
7 Our findings are robust to the choice of SSR as we have also considered the alternative by Leo Krippner.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 77

Statistical identification and inference

To determine the optimal number of lags in equation (1), we employed information criteria
tests on the reduced-form specification without stochastic volatility terms.The proposed
lag lengths are 2 and 12 for BIC and AIC, respectively. We choose two lags for the
baseline specification, to prevent parameter proliferation in the time-varying case. In the
constant parameter case, we also test 12 lags and find that the results are robust to that
specification.8
Following the estimation strategy from Section II, we initialize the model estimation
using the first 10 years as training sample. Thus, after accounting for the lag length, the
effective sample begins in March 1992. We use 1,00,000 iterations of the Gibbs sampler,
of which the first 95,000 are discarded as burn-in. We test for the convergence of the
chain with standard procedures and discuss the results in Appendix D.
To identify the monetary policy shocks, we use Cholesky decomposition with a lower
triangular matrix, where following the sticky-information literature (Castelnuovo and
Surico, 2010), we order inflation expectations above inflation. Thus, we assume that they
do not respond within the same month to new information regarding the interest rates or
inflation.9 Apart from that, we follow standard ordering of unemployment, inflation, and
the monetary policy rate (Christiano, Eichenbaum and Evans, 1999). Nevertheless, we test
our results to different specifications and find that they are consistent with the baseline.
This is due to the fact that we are interested in the impulse responses following a shock
to the stochastic volatilities, η. Thus, a major determinant of the effects of uncertainty
shocks are the γ coefficients, and variable ordering plays a secondary role. This holds as
long as the estimated stochastic volatilities h remain the same across orderings. This is an
advantage of the framework over set-ups where the uncertainty measure is an additional
series whose level is shocked (e.g. Istrefi and Piloiu, 2014).

IV. Estimated uncertainty measures

Figure 2 plots the MPU series implied by the two models, the TVP and the constant
parameter specification, respectively, as well as an array of other popular uncertainty
indicators from the literature. The measures from the two models are highly similar, with
the constant parameter specification giving a bit more weight to the MPU at the height of
the financial crisis as well as during 2013 – the year of the ‘‘taper tantrum’’. Other than
that, MPU is estimated to have been particularly high during the early 90s and following
the events of the dot-com bubble. These dynamics mirror closely the estimated monetary
policy volatility of Mumtaz and Zanetti (2013), whose sample ends in 2010Q4.
On the top right of Figure 2, the model implied measure is plotted against the MPU
index from Baker et al. (2016) and Husted et al. (2020). All time series have been
standardized to facilitate comparison. The latter two are based on newspaper data and
are much more volatile, compared with the model estimate. Apart from the differences
in smoothness, the three series appear to follow similar dynamics throughout the 1990s.
However, since the early 2000s, there are notable differences. The model implies that

8
The results are available in the Appendix.
9 This has been found to be true for SPF data, see Coibion and Gorodnichenko (2015).

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
78 Bulletin

Figure 2. Measures of monetary policy uncertainty. Top left: estimates from the time-varying parameter
(TVP) and constant parameter models with probability intervals (shaded). Top right: Model implied MPU
versus the indices MPU indices from Baker et al. (2016) and Husted et al. (2020), standardized. Bottom left:
Macro uncertainty from Jurado et al. (2015), standardized. Bottom right: financial uncertainty from Jurado
et al. (2015), standardized

the highest peak of monetary policy volatility was surrounding the global financial crisis.
The measure of Baker et al. (2016) – during the dot-com bubble and its aftermath, while
the index of Husted et al. (2020) suggests that MPU was highest throughout the first
years of the Trump administration.10 The model estimate is by construction smoother; its
dynamics are given by the autoregressive process in equation (2). It is also based on the
estimated stochastic volatility of the monetary policy shocks identified using our SVAR,
hence tightly related to the underlying SSR series and our identification assumptions.
Since economic shocks are by assumption unforecastable, the measure of Mumtaz
and Zanetti (2013) is in spirit much closer to the approach taken by Jurado et al. (2015)
where uncertainty is related to the second moments of a forecast distribution. We plot
both macroeconomic and financial uncertainty based on Jurado et al. (2015) in the bottom
panels of Figure 2. While all three measures display elevated uncertainty surrounding the
global financial crisis, there appear to be notable differences in the other periods. Macro
uncertainty displays lower but more prolonged peaks around the turn of the millennium
and has even fallen during the ‘taper tantrum’ periods. Similarly, financial uncertainty
also does not fluctuate after the financial crisis and on average appears to follow two long

10
This is valid for our data span, which cuts short right before the index of Husted et al. (2020) reaches its historical
maximum in August 2019 of 407 (long-run mean 110).

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 79

Figure 3. Measures of inflation expectations uncertainty from a time-varying parameter model and a constant
coefficients specification. Median responses (solid and dashed line) and 68% probability intervals (shaded
area)

cycles throughout the sample as opposed to MPU, which is estimated to have peaked
shortly and only during specific events.11
Next, we report the volatility of the shocks to short- and long-run inflation expectations,
which could be interpreted as uncertainty surrounding the inflation expectations. They are
plotted in Figure 3 for each specification. The estimates are overlapping almost perfectly.
Notably, long-run inflation expectations display a low and stable level of uncertainty,
while short-run inflation expectations uncertainty increased during the Great Recession and
remained elevated compared with the period before the 2008 crisis.12 Long-run inflation
expectations appear to be well anchored as their uncertainty level does not fluctuate even
through dramatic events such as the global financial crisis. These estimates may also be
seen through the lens of central bank credibility. Constant uncertainty surrounding the
long-run expectations suggests trust in the monetary authority – a crucial condition for an
efficient transmission of monetary policy. These findings are in line with the inflation gap
literature that establishes low volatility of trend inflation (Cogley and Sbordone, 2008;
Cogley, Primiceri and Sargent, 2010; Coibion and Gorodnichenko, 2011). However, long-
run interest rates have been shown to respond to economic news (Guerkaynak, Sack and
Swanson, 2005), which suggests that higher volatility of long-run inflation expectations
could also be expected. The model, however, recovers the large volatility swings in the
short-run inflation expectations only. Given the similarity between the two series, this
raises the question whether the model always recovers the true peaks in the volatilities.
To test this hypothesis, we conduct a Monte Carlo exercise using artificially generated
data. First, we generate data from a model in which short-run inflation expectations
volatility hSR,t for t = 1, . . . , T is extremely low, while long-run inflation expectations
volatility is high. Then we estimate the model on that data set to see whether it would
recover the volatility series correctly. We mirror this set-up with artificially generated data
where the opposite is true; in the data generating process, short-run inflation expectations

11 We control for financial uncertainty by extending our model with the returns from Standard and Poors’ Index.
We compare the estimated financial uncertainty from that model and report that it matches the Jurado et al (2015)
measure much more closely. See Appendix C
12
The estimated volatilities in the model with FED Funds are almost identical to the ones obtained with the SSR.
Estimates are shown in the Appendix, Figure A2 and A3.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
80 Bulletin

volatility is high and long run is low. We generate the data 200 times, 100 for each set-up
and estimate correspondingly 200 times the model on the artificial data set. We find that
in both cases the model correctly attributes the volatility to the correct scenario. Further
details as well as a plot of the estimates may be found in Appendix, Section B.

V. The effects of monetary policy uncertainty shocks

Next, we will examine the responses of the variables to a one standard deviation MPU
shock. Overall, an MPU shock manifests as a negative demand shock, as documented
by Mumtaz and Zanetti (2013) and Alessandri and Mumtaz (2019). Namely, an increase
in MPU leads to a contraction of the real sector due to risk aversion and consumption
smoothing behaviour and a decline in economic output (Mumtaz and Zanetti, 2013), as
well as a fall in prices and the interest rate. With our monthly data set, we observe the
same effects, unemployment increases while inflation and interest rates decline. However,
when we allow for gradual evolution of the relationships between the variables, we find
interesting developments, especially in relation to inflation expectations.
A concise way to highlight the time variation in the responses is to show the average
responses of the model variables to an MPU shock during four different periods: (i) 1993
as the first full year of impulse responses after accounting for the training sample and
the lag values; (ii) 2003 as a boom period after the recovery of the dot-com crash; (iii)
2008 to look to the effects during the Great Recession; and (iv) 2018 the last full year of
estimates. We do this on Figures 4 through 8. For comparison, we also plot the responses
from the constant parameter model, which by construction would be identical over the
chosen periods.
Figure 4 shows the response of long-run inflation expectations to a monetary
policy uncertainty shock. We find that with the exception of the beginning of the
sample, the response of long-run inflation expectations is close to zero. The 68%
probability intervals bands suggest that this response is not different from zero for the
remaining part of the sample. The median response has evolved from slightly negative
towards showing no link to reacting to monetary policy uncertainty. The constant
parameter specification also captures a negative median response of the long-run inflation
expectations, naturally with a tighter probability intervals due to the higher degrees of
freedom. These findings are of particular interest because long-run inflation expectations
are associated with the level of credibility the central bank maintains. This is especially
important for monetary policy effectiveness in a context where forward guidance and other
unconventional measures have been implemented. Along with the estimated low long-run
volatility, these results reinforce the notion that long-run inflation expectations are well
anchored.
Short-run inflation expectations, depicted in Figure 5, respond somewhat differently.
The constant parameter model shows a negative response with the zero line falling
outside of the probability intervals, in contrast to the long-run expectations case. In
the TVP version, we see that these findings are driven primarily by the response of
the short-run expectations throughout the first half of the sample. The median response
of the TVP case (solid lines) lies below the constant parameter case (dashed lines) in
1993 and 2003, while it is less negative after the global financial crisis. Correspondingly

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 81

the 68% error bands show a decline in the short-run inflation expectations not only in
the beginning of the sample, as long-run expectations do, but also before the financial
crisis. After the beginning of the Great Recession, the negative response to MPU shocks
has dissipated, indicating a distinct pre and postfinancial crisis dynamics of inflation
expectations.
Next, we look at the response of inflation, shown in Figure 6, which is also negative.
It is, however, different than the response of inflation expectations in Figures 4 and 5 in
several key aspects. First, both the TVP and no-TVP model suggest a median estimated
(including the error bands) outside of the zero line throughout almost the entire sample
with the exception of the very last periods – inflation declines following an MPU shock.
Given the lack of a strong response of both expectations series after the financial crisis, it
appears that the transmission mechanism is not through the expectations channel. Second,
the response of inflation has always been several magnitudes stronger than that of inflation
expectations, indicative of a rigidity in the expectation formation process when it comes
to MPU shocks. This is in contrast to the dynamics following an uncertainty shock to
the real variable, unemployment, which is also a negative demand shock. In that case, all
three variables react in a similar magnitude (See Appendix C).
To highlight these points, we present an alternative view of the MPU effects in Figure 7.
It contains a cross section of the median responses of inflation and inflation expectations
at specific horizons following an MPU shock – from 12 to 60 months ahead. The figure

Figure 4. Impulse response of long-run inflation expectations following a monetary policy uncertainty shock,
yearly averages. Solid line denotes the median response and shaded area is the 68% probability intervals for
the time-varying parameter model. The dashed line is the response from a constant parameter case

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
82 Bulletin

Figure 5. Impulse response of short-run inflation expectations following a monetary policy uncertainty shock,
yearly averages. Solid line denotes the median response and shaded area is the 68% probability intervals for
the time-varying parameter model. The dashed line is the response from a constant parameter model

shows the gradual transition towards the flat response throughout the last decade of low
interest rates. Combined with the low estimated volatilities, this evidence points towards
a strong anchoring of inflation expectations.
Figure 8 depicts the responses of unemployment and the monetary policy rate.
Unemployment has always increased following an MPU shock, albeit in different
magnitude. By far, the strongest response is estimated to be during the financial crisis,
a result probably exacerbated through omitted variable bias as additional information
is needed in the model to capture the sharp increase in unemployment during the
crisis. Nevertheless, even in good times, a standard deviation shock in MPU increases
unemployment by about 0.2% points.
Following an MPU shock, the response of the monetary policy rate mimics the
dynamics of inflation – it has declined throughout the first half of the sample. However,
after the financial crisis, interest rates do not seem to react anymore. Thus, it appears that the
importance of MPU shocks has declined in recent years. Mumtaz and Theodoridis (2018)
also show highlighted that uncertainty shocks, namely macro uncertainty, have become
less important over time, specifically towards real and financial variables. However,
they do not find this for prices. Our results suggest that at least MPU has become less
important for inflation, yet only in the recent periods not covered by their sample. A
further distinction is that our real variable, unemployment, reacts stronger to an MPU

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 83

Figure 6. Impulse response of inflation following a monetary policy uncertainty shock, yearly averages.
Solid line denotes the median response and shaded area is the 68% probability intervals for the time-varying
parameter model. The dashed line is the response from a constant parameter model

Long Run Inflation Expectations Short Run Inflation Expectations Inflation

0 0 0
h = 12
-0.1 -0.1 -0.1 h = 24
h = 36
h = 48
-0.2 -0.2 -0.2
h = 60

-0.3 -0.3 -0.3

1990 2000 2010 2020 1990 2000 2010 2020 1990 2000 2010 2020

Figure 7. Cross section of the median impulse responses at specific horizons after the shock – from h = 12
months to h = 60 months

shock during and after the GFC, which is in line with other contributions in the literature,
such as Angelini et al. (2019) and Alessandri and Mumtaz (2019).
These findings may also be related to a large strand of the literature on state dependent
effects of uncertainty (Caggiano et al., 2014; Caggiano, Castelnuovo and Pellegrino, 2017;
Castelnuovo and Pellegrino, 2018; Liu et al., 2019).13 These contributions show that
uncertainty shocks have had different effects throughout time, nonetheless by highlighting
that the impact is more pronounced in recessionary periods. However, we find diminishing
effects of MPU shocks over time (as, for example, also Mumtaz and Theodoridis (2018)
do, albeit with respect to macro uncertainty). In part, this is explained by our model

13 See Castelnuovo (2019) for an overview of the literature.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
84 Bulletin

Figure 8. Impulse responses unemployment and monetary policy rate following a monetary policy uncertainty
shock, yearly averages. Solid line denotes the median response and shaded area is the 68% probability interval
for the time-varying parameter model. The dashed line is the response from a constant parameter model

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 85

choice – we allow for a gradual parameter evolution to capture long-run changes of

the macroeconomic dynamics. The state dependent literature, on the other hand, utilizes
models that allow for large and sudden parameter breaks, such as change-point, threshold,
and smooth-transition approaches.
Furthermore, the VAR structure imposes that all variables are interconnected – as the
monetary policy rate reacts less to MPU shocks over time, the other variables might do so
as well. For example, Liu et al. (2019) show that first-moment demand shocks suppress
prices and interest rates, while unemployment rises.14 However, we show that second-
moment shocks matter also on their own – the median response of the monetary policy
rate is almost flat towards the end of our sample 2018, while inflation and unemployment
continue to react to MPU shocks.
Finally, the model is robust with regard to different specifications. We have explored
robustness with respect to the number of lags, the variables included, and their ordering.
Notably, given the importance of the macro-financial nexus, we test for a possible omitted
variable bias by including a financial variable, namely the S&P500 returns. We find that
controlling for financial uncertainty does not alter the estimated proxy for MPU shocks,
neither it affects the impulse response analysis. Moreover, the model implied financial
uncertainty from the S&P500 appears to match rather well the financial uncertainty
index by Jurado et al. (2015). These findings are reported in the Appendix, Section C.
Furthermore, changing the order of the variables, a prominent issue with SVAR models
with Cholesky decomposition and even more so with TVP-VARs, does not lead to
different conclusions (Figures C4 and C5 in Section C in the Appendix).

VI. Concluding remarks

In this article, we investigate the evolution of monetary policy uncertainty on the
economy, with a special focus on the consequences of unexpected increases of MPU on
inflation expectations. Using monthly data for the United States, we estimate an SVAR
with stochastic volatility-in-mean. Starting with a constant parameter specification, we
find further evidence that MPU shocks act as negative demand shocks with rising
unemployment and decreasing price level. More importantly, we find that inflation
expectations react negatively to an increase in the MPU with a different magnitude for
long and short inflation expectations.
Considering the evidence that inflation expectations differ across age groups and
personal experience, we then estimate a model where we allow the parameters among
the model variables to change over time. We treat the question of how much parameter
evolution is allowed in the model as open and estimate it from the data.
This extension allows us to observe that long-run and short-run inflation expectations
have reacted differently to MPU shocks in the 90s as opposed to today. Long-run inflation
expectations have remained mostly robust and unperturbed by MPU shocks apart from
the beginning of the 90s. On the other hand, short-run inflation expectations have had a
pronounced negative response up to the Great Recession. After the 2008 crisis, short-run

14 The original paper reports the results to a positive demand shock from a change-point VAR model identified via
sign restrictions conditioning on the regime. Thus, the model should preserve its signs following a negative demand
shock.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
86 Bulletin

inflation expectations have become less and less affected by MPU shocks. This finding
is surprising given that most episodes associated with high MPU have occurred after
2008. This distinction between long- and short-run inflation expectations suggests a level
of rigidity surrounding long-run inflation expectations associated with a credible central
bank. Contrary to short-run inflation expectations, inflation reacts negatively throughout
the whole sample and more strongly than the inflation expectations in the short run Jurado
et al. (2015).
Overall we find that MPU shocks have become less important over time. This is true not
only for inflation expectations and inflation but also for unemployment and the interest rate.
These findings suggest that MPU might be less obstructing for the conduct of monetary
policy than perceived. Neither interest rates react nowadays to MPU shocks as they did
in the past, nor inflation expectations appear to respond as much as historically observed.

APPENDIX A: TVP model

The posterior density of the kQ parameter is plotted in Figure A1.
The strong evidence for time variation in the data is also supported in the estimated
stochastic volatilities. As the relationship between the variables changes, the degrees of
freedom for the model to fit the data increase, hence the stochastic volatility declines. We
find that while the most important events of the estimated MPU remain the same across
both specifications, the MPU estimated under the TVP model is markedly lower after the
dot-com bubble (Figure A2).

APPENDIX B: Monte Carlo simulation

The data generating process (DGP) for the Monte Carlo simulations is given by


P 
M
Zt = ĉ + β̂ j Zt−j + γˆm h̃t−m + ut , ut ∼ N(0, 
ˆ t ), (B1)
j=1 m=0

Figure A1. Posterior densities of the time-variation parameter kQ . Left: Shadow short rate model. Right:
Federal funds rate model

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 87

Figure A2. Median estimates of Monetary Policy Uncertainty from a model with time-varying parameters
versus the constant coefficient specification

Figure A3. Log volatilities of short- and long-run inflation expectations based on a model with the SSR (solid
line) and a FED Funds model (blue dashed line) with 68% probability intervals


with  ˆ t = Â−1 Ht Â−1 where H is specified as in the original set-up as being a diagonal
matrix with h̃ on the diagonal. The parameters of the DGP are set at the posterior mode
from the estimation of the model based on real data, which ensures a stable VAR process.
We then proceed to draw a series of shocks for ut for t = 1, . . . , 500, which is in line with
the sample size we have in practice. To draw from N(0,  ˆ t ) we need to specify a series
for h̃t . The shapes of the true volatility series are inspired by the estimates with the real
data – we clone the first 250 observations from our model twice, thus creating a data set
with two large crises. This results in a couple of large peaks in long-run inflation volatility
in the first set-up (MC1), while we force short-run inflation volatility to be extremely
low with a mean at 0.005. Correspondingly in the second set-up (MC2), we adopt a low
long-run volatility series, similar to the one in the real data.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
88 Bulletin

Figure B1. True versus estimated volatility from 200 Monte Carlo simulations. Left: Monte Carlo set-up 1
(MC1) with high long-run inflation volatility and low short-run inflation volatility. Right: Monte Carlo set-up
2 (MC2) with high short-run inflation volatility

We then proceed to draw 200 data sets, 100 for each set-up and estimate the model
200 times. After each estimation, we have an uncertainty measure for long- and short-run
inflation expectations and its probability intervals. In Figure B1, we plot the median
volatility estimates for each data set (100 per set-up) along with the true DGP from which
said data set was created. We find that the model correctly captures the shape of each
special case. Note that each grey line on the graph is the median outcome of one data set;
for ease of readability, we omit the probability intervals surrounding each run.

APPENDIX C: Robustness section

Robustness to the lag specification
Figure C1 plots the impulse responses of the constant parameter models the federal
funds rate with 12 lags. The only quantitative difference is the case of long-run inflation
expectations with the SSR, where the response is significant, while the FED Funds model
shows identical behaviour to the two lag SSR specification.

Extending the model to control for financial uncertainty

Figure C2 indicate that our baseline stochastic volatility estimates, from which our
monetary policy uncertainty measure is derived, are unaffected by including the S&P 500
returns as in Mumtaz and Zanetti (2013) to account for financial uncertainty. Additionally,
one can observe that despite the fact that the monetary policy uncertainty from the
financial model follows very closely the estimates from the baseline one, the estimated
volatility of the S&P 500 returns differs significantly from the monetary policy uncertainty
measure. Moreover, Figure C3 shows the associated financial uncertainty from the model

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 89

Figure C1. Median responses to a monetary policy uncertainty shock. Model with shadow short rate (solid
line) and model with the federal funds rate (dashed line). The shaded areas show the 68% probability interval

Figure C2. Solid blue line depicts the median estimate of the stochastic volatilities of the baseline model.
Dashed red lines are the median estimates of the stochastic volatilities of the alternative financial specification.
Shaded areas show the corresponding 68% probability intervals

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
90 Bulletin

Figure C3. Comparison between financial uncertainty estimated from the model by extending it with S&P
500 returns and plotting the stochastic volatility to its shocks versus Jurado et al. (2015)

the alternative financial model of Figure C2 appear to coincide well with the financial
uncertainty index of Jurado et al. (2015).

Robustness to the variable order

Figure C4. Solid blue line depicts the median responses to a monetary policy uncertainty shock of the baseline
model at different time periods. Dashed red lines are the median responses to a monetary policy uncertainty
shock of the alternative ordering specification. Shaded areas show the corresponding 68% probability intervals

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 91

Impulse responses: Unemployment uncertainty shock

Figure C5. Solid blue line depicts the median responses to a unemployment uncertainty shock of the baseline
model. Shaded areas show the corresponding 68% probability intervals

APPENDIX D: Convergence
To ensure convergence of the Gibbs sampler, we estimate the recursive means of the
retained draws for the time-varying parameters and stochastic volatilities of the baseline
model. Figures D1 and D2 present the recursive means for every 20 draws of the vectorized

Figure D1. Recursive means of the VAR coefficients

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
92 Bulletin

Figure D2. Recursive means of the volatilities

time-varying parameters and stochastic volatilities. Both graphs suggest convergence in

the Gibbs sampler given that along the different draws, the recursive means remain stable
and do not depict any sudden increase among the different draws.

Final Manuscript Received: March 2021

References
Alessandri, P. and Mumtaz, H. (2019). ‘Financial regimes and uncertainty shocks’, Journal of Monetary
Economics, Vol. 101, pp. 31–46.
Amir-Ahmadi, P., Matthes, C. and Wang, M.-C. (2018). ‘Choosing prior hyperparameters: with applications
to time-varying parameter models’, Journal of Business & Economic Statistics, Vol. 38, pp. 1–13.
Angelini, G., Bacchiocchi, E., Caggiano, G. and Fanelli, L. (2019). ‘Uncertainty across volatility regimes’,
Journal of Applied Econometrics, Vol. 34, pp. 437–455.
Bachmann, R., Born, B., Elstner, S. and Grimme, C. (2019). ‘Time-varying business volatility and the price
setting of firms’, Journal of Monetary Economics, Vol. 101, pp. 82–99.
Baker, S. R., Bloom, N. and Davis, S. J. (2016). ‘Measuring economic policy uncertainty’, The Quarterly
Journal of Economics, Vol. 131, pp. 1593–1636.
Bauer, M., Lakdawala, A., and Mueller, P. (2019). Market-based monetary policy uncertainty. Available at
SSRN 3371160.
Bauer, M. D. (2012). ‘Monetary policy and interest rate uncertainty’, FRBSF Economic Letter, Vol. 38, pp.
1–5.
Benati, L. and Surico, P. (2008). ‘Evolving U.S. monetary policy and the decline of inflation predictability’,
Journal of the European Economic Association, Vol. 6, pp. 634–646.
Bernanke, B. S. (2003). A Perspective on Inflation Targeting.
Bernanke, B. S. and Woodford, M. (1997). ‘Inflation forecasts and monetary policy’, Journal of Money,
Credit and Banking, Vol. 29, p. 653.
Bianchi, F., Kung, H., and Tirskikh, M. (2018). The Origins and Effects of Macroeconomic Uncertainty.
Working paper 25386, NBER.
Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J. and Jansen, D. J. (2008). ‘Central bank communication
and monetary policy: a survey of theory and evidence’, Journal of Economic Literature, Vol. 46, pp.
910–945.
Bloom, N. (2009). ‘The Impact of Uncertainty Shocks’, Econometrica, Vol. 77, pp. 623–685.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Monetary Policy Uncertainty and Inflation Expectations 93

Caggiano, G., Castelnuovo, E. and Groshenny, N. (2014). ‘Uncertainty shocks and unemployment dynamics
in U.S. recessions’, Journal of Monetary Economics, Vol. 67, pp. 78–92.
Caggiano, G., Castelnuovo, E. and Pellegrino, G. (2017). ‘Estimating the real effects of uncertainty shocks at
the zero lower bound’, European Economic Review, Vol. 100, pp. 257–272.
Canova, F. and Gambetti, L. (2009). ‘Structural changes in the us economy: Is there a role for monetary
policy?’, Journal of Economic Dynamics and Control, Vol. 33, pp. 477–490.
Carter, C. K. and Kohn, R. (1994). ‘On Gibbs sampling for state space models’, Biometrika, Vol. 81, pp.
541–553.
Castelnuovo, E. (2019). Domestic and Global Uncertainty: A Survey and Some New Results. CAMA Working
Paper, 75/2019.
Castelnuovo, E. and Pellegrino, G. (2018). ‘Uncertainty-dependent effects of monetary policy shocks: a
new-Keynesian interpretation’, Journal of Economic Dynamics and Control, Vol. 93, pp. 277–296.
Castelnuovo, E. and Surico, P. (2010). ‘Monetary policy, inflation expectations and the price puzzle’, The
Economic Journal, Vol. 120, pp. 1262–1283.
Chang, B. Y. and Feunou, B. (2014). ‘Measuring uncertainty in monetary policy using implied volatility and
realized volatility’, Bank of Canada Review, Vol. 1, pp. 32–41.
Christiano, L. J., Eichenbaum, M. and Evans, C. L. (1999). ‘Monetary policy shocks: what have we learned
and to what end?’, in Taylor J. B. and Woodford M. (eds), Handbook of Macroeconomics, Volume I, Part
A, Elsevier, the Netherlands, pp. 65–148.
Clarida, R., Gali, J. and Gertler, M. (2000). ‘Monetary policy rules and macroeconomic stability: evidence
and some theory’, The Quarterly Journal of Economics, Vol. 115, pp. 147–180.
Cogley, T., Primiceri, G. E. and Sargent, T. J. (2010). ‘Inflation-gap persistence in the U. S’, American
Economic Journal: Macroeconomics, Vol. 2, pp. 43–69.
Cogley, T. and Sargent, T. J. (2005). ‘Drifts and volatilities: monetary policies and outcomes in the post
WWII US’, Review of Economic Dynamics, Vol. 8, pp. 262–302.
Cogley, T. and Sbordone, A. M. (2008). ‘Trend inflation, indexation, and inflation persistence in the new
Keynesian Phillips curve’, American Economic Review, Vol. 98, pp. 2101–2126.
Coibion, O. and Gorodnichenko, Y. (2011). ‘Monetary policy, trend inflation, and the great moderation: an
alternative interpretation’, American Economic Review, Vol. 101, pp. 341–370.
Coibion, O. and Gorodnichenko, Y. (2015). ‘Information rigidity and the expectations formation process: a
simple framework and new facts’, The American Economic Review, Vol. 105, pp. 2644–2678.
Creal, D. D. and Wu, J. C. (2017). ‘Monetary policy uncertainty and economic fluctuations’, International
Economic Review, Vol. 58, pp. 1317–1354.
Debortoli, D., Galı́, J. and Gambetti, L. (2019, 2020). On the Empirical (Ir)relevance of the Zero Lower Bound
Constraint Vol. 34, National Bureau of Economic Research, Cambridge, Massachusetts, pp. 141–170.
Draeger, L. and Lamla, M. J. (2018). Is the Anchoring of Consumers’ Inflation Expectations Shaped by
Inflational Experience? Working paper series 7042, CESifo.
Fernández-Villaverde, J., Guerrón-Quintana, P., Kuester, K. and Rubio-Ramı́rez, J. (2015). ‘Fiscal volatility
shocks and economic activity’, American Economic Review, Vol. 105, pp. 3352–3384.
Guerkaynak, R. S., Sack, B. and Swanson, E. (2005). ‘The sensitivity of long-term interest rates to economic
news: evidence and implications for macroeconomic models’, American Economic Review, Vol. 95, pp.
425–436.
Hansen, S., McMahon, M. and Prat, A. (2018). ‘Transparency and deliberation within the FOMC: a
computational linguistics approach’, The Quarterly Journal of Economics, Vol. 133, pp. 801–870.
Haubrich, J., Pennacchi, G. and Ritchken, P. (2012). ‘Inflation expectations, real rates, and risk PREMIA:
Evidence from inflation swaps’, The Review of Financial Studies, Vol. 25, pp. 1588–1629.
Husted, L., Rogers, J. and Sun, B. (2020). ‘Monetary policy uncertainty’, Journal of Monetary Economics,
Vol. 115, pp. 20–36.
Ikeda, D., Li, S., Mavroeidis, S., and Zanetti, F. (2020). Testing the Effectiveness of Unconventional Monetary
Policy in Japan and the United States. IMES Discussion Paper Series 20-E-10, Institute for Monetary and
Economic Studies, Bank of Japan.
Istrefi, K. and Mouabbi, S. (2018). ‘Subjective interest rate uncertainty and the macroeconomy: A cross-country
analysis’, Journal of International Money and Finance, Vol. 88, pp. 296–313.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.
 14680084, 2023, 1, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/obes.12516 by Cochrane Philippines, Wiley Online Library on [27/04/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
94 Bulletin

Istrefi, K. and Piloiu, A. (2014). Economic Policy Uncertainty and Inflation Expectations. Technical Report,
Bank de France.
Jurado, K., Ludvigson, S. C. and Ng, S. (2015). ‘Measuring uncertainty’, American Economic Review, Vol.
105, pp. 1177–1216.
Kang, W., Lee, K. and Ratti, R. A. (2014). ‘Economic policy uncertainty and firm-level investment’, Journal
of Macroeconomics, Vol. 39, pp. 42–53.
Kurov, A. and Stan, R. (2018). ‘Monetary policy uncertainty and the market reaction to macroeconomic
news’, Journal of Banking & Finance, Vol. 86, pp. 127–142.
Leduc, S. and Sill, K. (2013). ‘Expectations and economic fluctuations: an analysis using survey data’, Review
of Economics and Statistics, Vol. 95, pp. 1352–1367.
Leduc, S., Sill, K. and Stark, T. (2007). ‘Self-fulfilling expectations and the inflation of the 1970s: Evidence
from the livingston survey’, Journal of Monetary economics, Vol. 54, pp. 433–459.
Liu, P., Theodoridis, K., Mumtaz, H. and Zanetti, F. (2019). ‘Changing macroeconomic dynamics at the zero
lower bound’, Journal of Business and Economic Statistics, Vol. 37, pp. 391–404.
Lucas, R. E. (1976). Econometric policy evaluation: a critique. Carnegie-Rochester Conference Series on
Public Policy.
Malmendier, U. and Nagel, S. (2016). ‘Learning from inflation experiences’, The Quarterly Journal of
Economics, Vol. 131, pp. 53–87.
Mumtaz, H. and Theodoridis, K. (2015). ‘The international transmission of volatility shocks: an empirical
analysis’, Journal of the European Economic Association, Vol. 13, pp. 512–533.
Mumtaz, H. and Theodoridis, K. (2018). ‘The changing transmission of uncertainty shocks in the U. S’,
Journal of Business & Economic Statistics, Vol. 36, pp. 239–252.
Mumtaz, H. and Theodoridis, K. (2020). ‘Dynamic effects of monetary policy shocks on macroeconomic
volatility’, Journal of Monetary Economics, Vol. 114, pp. 262–282.
Mumtaz, H. and Zanetti, F. (2013). ‘The impact of the volatility of monetary policy shocks’, Journal of
Money, Credit and Banking, Vol. 45, pp. 535–558.
Mumtaz, H. and Zanetti, F. (2015). ‘Labor market dynamics: A time-varying analysis’, Oxford Bulletin of
Economics and Statistics, Vol. 77, pp. 319–338.
Neely, C. J. (2005). ‘Using implied volatility to measure uncertainty about interest rates’, Federal Reserve
Bank of St Louis Review, Vol. 87, pp. 407–426.
Primiceri, G. E. (2005). ‘Time varying structural vector autoregressions and monetary policy’, The Review of
Economic Studies, Vol. 72, pp. 821–852.
Schorfheide, F. and Song, D. (2015). ‘Real-time forecasting with a mixed-frequency var’, Journal of Business
& Economic Statistics, Vol. 33, pp. 366–380.
Sinha, A. (2016). ‘Monetary policy uncertainty and investor expectations’, Journal of Macroeconomics, Vol.
47, pp. 188–199.
Stulz, R. (1986). ‘Interest rates and monetary policy uncertainty’, Journal of Monetary Economics, Vol. 17,
pp. 331–347.
Svensson, L. E. (2000). ‘Open-economy inflation targeting’, Journal of International Economics, Vol. 50, pp.
155–183.
Swanson, E. T. (2006). ‘Have increases in federal reserve transparency improved private sector interest rate
forecasts?’, Journal of Money, Credit and Banking, Vol. 38, pp. 791–819.
Swanson, E. T. and Williams, J. C. (2014). ‘Measuring the effect of the zero lower bound on medium- and
longer-term interest rates’, American Economic Review, Vol. 104, pp. 3154–3185.
Wu, J. C. and Xia, F. D. (2016). ‘Measuring the macroeconomic impact of monetary policy at the zero lower
bound’, Journal of Money, Credit and Banking, Vol. 48, pp. 253–291.

© 2022 The Authors. Oxford Bulletin of Economics and Statistics published by Oxford University and John Wiley & Sons Ltd.