Long-Run Ination Uncertainty

Stefan Nagel
University of Michigan, NBER, and CEPR

In this commentary I argue that option price data oer

useful insights into the long-run macroeconomic uncertainty
perceived by investors. Data on ination options in the United
States show substantial dispersion in the risk-neutral distribu-
tion of long-run ination rates. This may indicate that substan-
tial uncertainty about the ination target still exists. However,
I argue that a high dispersion in the risk-neutral distribution
could also reect disagreement among investors who are con-
dent in their own forecasts and do not necessarily perceive
a high degree of subjective uncertainty. Disagreement could
potentially reconcile the relative stability of ination in recent
years with the substantial dispersion in the risk-neutral dis-
tribution of long-run ination and in survey forecasts of long
ination.
JEL Codes: E31, E44, G13.

1. Introduction

The topic of the session that I have been asked to comment on is

expectations. In any evaluation of the challenges to achieving price
stabilitythe overall theme of this conferencemacroeconomic
expectations should play an important role. Decisions of price set-
ters and investors today depend on their forward-looking assess-
ment of the macroeconomic situation. Many instruments in central
bankers toolkits, and forward guidance in particular, work through
their eects on the publics expectations. However, macroecono-
mists understanding of the formation of expectations is still quite
limited, and there is still substantial room for improvement in the

Author contact: Stephen M. Ross School of Business and Department of Eco-
nomics, University of Michigan, 701 Tappan Street, Ann Arbor, MI 48109; e-mail:
[email protected].

207
208 International Journal of Central Banking September 2016

measurement of expectations. The two papers in this session advance

research on these two important fronts.
The paper by Jonathan Wright titled Options-Implied Proba-
bility Density Functions for Real Interest Rates (this issue) looks
at market-based measures of the (risk-neutral) distribution of real
interest rates extracted from prices of traded options. Market-based
measures oer the promise of a high-frequency assessment of the
entire term structure of distributions of future outcomes, although
their interpretation is not entirely straightforward due to conation
with risk premia, illiquidity eects, and other confounding factors.
In my commentary, I will focus on long-term ination
expectationsarguably the most important place to look for threats
to price stability. And I want to focus on one aspect of the long-
run outlook for which market-based measures like those in Wrights
paper are particularly useful: subjective uncertainty. Surveys often
focus on point forecasts without eliciting the subjective uncer-
tainty of the respondent (the U.S. Survey of Professional Forecast-
ers (SPF) is a notable exception). Option price data are therefore
a useful source of information about the uncertainty perceived by
investors.
Subjective uncertainty measures can help shed light on a num-
ber of important policy questions. One example that I will discuss
is anchoring of ination expectations. The anchoring hypothesis not
only makes predictions about the (lack of) reaction of point forecasts
to news but also about the (lack of) subjective uncertainty about
long-run ination rates. Data on long-run ination uncertainty
market-based or otherwiseshould therefore be informative about
the degree to which monetary authorities succeeded in anchoring
expectations.
The issue of subjective uncertainty is also closely related to the
issue of expectations heterogeneity. If there is substantial uncertainty
about outlook for ination, people may also disagree about this out-
look. For this reason, the article by Boneva et al. titled The Eect
of Unconventional Monetary Policy on Ination Expectations: Evi-
dence from Firms in the United Kingdom in this issue performs
an important task by providing insight into the expectations forma-
tion of rms. The expectations of decision makers in rms are not
necessarily identical to those of commonly surveyed households or
professional forecasters. The literature on rms expectations is still
Vol. 12 No. 3 Long-Run Ination Uncertainty 209

quite small, even though rms, as price setters, play a crucial role
in determining the rate of ination.
I will argue below that expectations heterogeneity is likely to
aect the risk-neutral distribution of outcomes and hence market-
based measures of uncertainty. As far as I am aware, this is not an
issue that has received a lot of attention in the literature. On the one
hand, the fact that market-based measures of uncertainty could be
informative about disagreement is another reason to be interested in
studying these measures. On the other hand, the potential presence
of disagreement eects complicates the interpretation of these meas-
ures. The eect of belief dispersion on market-based measures of
uncertainty could be particularly relevant in recent years, as uncon-
ventional monetary policies have, at least anecdotally, led to a sub-
stantial disagreement about the ination outlook among nancial
market participants.

2. Market-Based Measures of Long-Run Ination

Uncertainty

For some years now, ination caps and oors have been traded in
the United States. The owner of an ination cap (oor) receives a
payment if the average CPI ination rate exceeds (is lower than) the
strike of the cap (oor). Prices of these derivative instruments can
be used to extract market-based measures of investors beliefs about
ination. Unlike ination-indexed bond prices or ination swaps,
these option-like derivatives allow the extraction of a whole prob-
ability distribution at various horizons. The extracted probability
distribution reects beliefs under the risk-neutral measure, i.e., the
probabilities are adjusted for risk premia. For now, I will interpret
the probability distribution extracted from long-maturity ination
caps and oors as reecting investors perception of long-run ina-
tion uncertainty. I will discuss the eect of risk premia further below.
Figure 1 shows the percentiles of the risk-neutral density of ve-
year average (annualized) CPI ination based on data from the Fed-
eral Reserve Bank of Minneapolis. The gure presents the median,
the 10th percentile, and the 90th percentile of the risk-neutral den-
sity. As the gure shows, investors seem to perceive a substantial
degree of uncertainty about long-run ination. In October 2015, the
210 International Journal of Central Banking September 2016

Figure 1. Percentiles of the Risk-Neutral Density of

Five-Year Ination

Note: The gure shows the 10th percentile, median, and 90th percentile of
the risk-neutral distribution of ination over a ve-year horizon extracted from
ination options by the Federal Reserve Bank of Minneapolis.

spread between the 90th and the 10th percentile is about 3 per-
centage points. This spread has come down a little bit since 2011
when it maxed out at around 4 percentage points, but it is still
substantial.
The high degree of dispersion in the risk-neutral distribution
of long-run ination is striking and dicult to reconcile with the
view that ination expectations are well anchored. In the United
States, the persistence of ination has decreased substantially in the
past few decades (Williams 2006). Ination expectations seem to be
less sensitive to macroeconomic news in recent years than in ear-
lier decades (Davis 2012). Evidence of this kind has given increasing
support to the idea that the Federal Reserve has gained credibility
in its intention and ability to keep ination close to a stable long-run
target. Bernanke (2007) expresses this view, for example, albeit with
the caveat that the anchoring of expectations is not perfect. Indeed,
Gurkaynak et al. (2007) nd that prices of nominal and real bonds
in the United States still imply some sensitivity of long-run ination
expectations to macroeconomic news.
Vol. 12 No. 3 Long-Run Ination Uncertainty 211

There are two somewhat separate notions of anchored expecta-

tions: (i) insensitivity of ination expectations to macroeconomic
news, including recent surprise ination; and (ii) a high degree of
condence in the intention and ability of the monetary authority to
keep ination close to a target. It is possible that ination expecta-
tions are anchored in the sense of (i) but that there is nevertheless
substantial uncertainty about the long-run target. For illustration,
consider the following model of perceived ination dynamics:

t+1 = t + t+1 , (1)

t = t1

+ ut , (2)

where t and ut are uncorrelated standard normal shocks. In this

model,  shocks have no eect on ination expectations. In this sense,
expectations are well anchored. However, if > 0, there can be
considerable uncertainty about the future path of the target and
hence long-run ination rates. The signals u that move the perceived
target need not be news associated with regular macroeconomic
announcements. In this sense, substantial target uncertainty could
very well be consistent with a low sensitivity of ination expectations
to macroeconomic news.
The term structure of uncertainty about ination should be infor-
mative about the relative importance of  and u shocks. Figure 2
shows the forecast uncertainty in terms of the spread between the
90th and 10th percentile of the distribution over various horizons
from one to ten years ahead (with = 0.01). The solid line shows
that in the absence of target rate uncertainty ( = 0), there is strong
decay of the uncertainty over longer horizons as the  shocks average
out. If there is target rate uncertainty ( = 0.005), the random-walk
nature of the perceived target process induces higher uncertainty for
ination rates averaged over longer horizons, which can oset the
decrease in the transitory shock uncertainty. Similar implications
about long-run ination uncertainty follow from a more sophisti-
cated unobserved-components stochastic volatility model (Stock and
Watson 2010) as estimated in Kitsul and Wright (2013).
Figure 3 plots the 90th10th percentile spreads from the
extracted risk-neutral distribution of ination for one-year, two-
year, and ve-year horizons. As the gure shows, there is about
as much uncertainty about the ve-year ination rate as there is
212 International Journal of Central Banking September 2016

Figure 2. Long-Run Ination Uncertainty with and

without Uncertainty about Target

Note: The gure shows the spread between the 90th and 10th percentile of the
distribution of ination rates.

about ination rates over one-year and two-year horizons. This pat-
tern would be dicult to explain in a model in which the public has
little uncertainty about the ination target. It looks very similar to
the target rate uncertainty case in gure 2.

3. Interpreting Dispersion in the Risk-Neutral

Distribution

3.1 Lack of Anchoring?

The relatively high degree of dispersion in the risk-neutral distribu-
tion of long-run ination evident in gures 1 and 3 raises questions
about the success of the monetary authority in anchoring long-run
ination expectations. Apparently, a substantial degree of uncer-
tainty about long-run ination still exists.
This evidence is also reason to be somewhat skeptical of expla-
nations that attribute the relative stability of ination expectations
Vol. 12 No. 3 Long-Run Ination Uncertainty 213

Figure 3. Interpercentile Range (90th10th Percentile) of

the Risk-Neutral Density of Ination over Various
Horizons

Note: Smoothed with local linear regression and tricube kernel with bandwidth
0.10 (of sample size).

(missing disination) in the recent Great Recession to rm anchor-

ing of ination expectations. If ination expectations really were so
rmly anchored, why would there be so much uncertainty priced into
long-maturity ination options? Clearly, more research is needed on
this point, but the evidence suggests that it may be worth looking for
alternative explanations. For example, Coibion and Gorodnichenko
(2015) show that much of the missing disination can be explained
by the fact that households ination expectations were very sen-
sitive to the rise in oil prices in the years leading up to 2013. This
explanation, based on lack of anchoring, would be consistent with the
high degree of uncertainty about long-run ination rates reected in
market-based probabilities. It may also help explain why the median
of the risk-neutral distribution of long-run ination in gure 1 has
declined substantially in 2015 coincident with a fall in oil prices.

3.2 Risk Premia?

So far I have treated the percentile spreads of the risk-neutral dis-
tribution of ination as a representation of subjective uncertainty of
214 International Journal of Central Banking September 2016

nancial market participants. Of course, since the risk-neutral prob-

abilities reect the product of actual probabilities and the stochastic
discount factor, the risk-neutral probabilities are aected by risk pre-
mia. For example, if nancial market participants dislike states of
very low and very high ination and are willing to pay a premium
to insure against these states, the risk-neutral probabilities of these
tail states will be higher than the actual probabilities. However,
while such risk premia can make the risk-neutral distribution more
dispersed compared with the actual distribution, it is not obvious
that there is a simple risk premium explanation for the fact that
the dispersion does not decay with longer horizons. It seems di-
cult to explain this feature of the data without substantial actual
uncertainty about long-run ination rates.
It is also worth keeping in mind that one could argue, along
the lines of Feldman et al. (2015), that policymakers should base
their decisions on the risk-neutral distribution. A social-welfare-
maximizing policy should take into account not only the subjec-
tive probabilities that the public attaches to future states of the
world but also the price that the public is willing to pay to insure
against these states of the world. Risk-neutral probabilities capture
both of these aspects. With regard to long-run ination, a dispersed
risk-neutral distribution indicates some combination of high uncer-
tainty about long-run ination and a high willingness to pay for
high-ination and deation insurance. This indicates that the pub-
lic would value policiese.g., credible commitment to an ination
targetthat reduce the likelihood of reaching these states of the
world.
However, in imperfect markets subject to segmentation fric-
tions, illiquidity, and limited participation, the interpretation of risk-
neutral probabilities is not straightforward. I illustrate this next by
pointing out how investor disagreement could potentially inuence
the dispersion of the risk-neutral distribution.

3.3 Expectations Heterogeneity?

Discussions of the interpretation of risk-neutral probabilities typi-
cally focus on how risk premia, and, occasionally, illiquidity distor-
tions in option prices drive a wedge between actual and risk-neutral
probabilities. Interpretation of risk premia is typically based on a
Vol. 12 No. 3 Long-Run Ination Uncertainty 215

representative-agent framework, where the stochastic discount fac-

tor is interpreted as revealing the marginal valuation that households
place on resources in dierent states of the world. However, the risk
premia embedded in the risk-neutral distribution, especially in its
tails, can also reect the disagreement between agents about the like-
lihood of future events rather than their subjective uncertainty. The
literature on market-based probabilities has not paid much attention
to this possibility.
To see how expectations heterogeneity can aect the risk-neutral
distribution, it is easiest to consider a model in which the public can
buy options, but only specialized, risk averse, imperfectly hedged
intermediaries can sell options. This setup is broadly consistent with
the evidence in Garleanu, Pedersen, and Poteshman (2009) that mar-
ket makers are net sellers of options in the stock index option market.
Applied to the ination options market, this would mean that prices
of ination caps reect the beliefs of investors expecting high ina-
tion (while those expecting deation cannot sell these options) and
the price of oors reects the beliefs of investors expecting deation.
Intermediaries can smooth out some of these eects, but if they are
risk averse and cannot perfectly hedge, they will only partly do so. As
a consequence, the tail probabilities of the risk-neutral distribution
are elevated relative to actual probabilities. In this interpretation,
risk premia embedded in risk-neutral probabilities are the manifes-
tation of dierences in beliefs. Buraschi and Jiltsov (2006) provide
a more sophisticated asset pricing model with disagreement that
implies similar eects on the risk-neutral distribution.
For these reasons, it is possible that the relatively high disper-
sion of the risk-neutral distribution of long-run ination in gures
1 and 3 reects disagreement between market participants rather
than a high level of subjective uncertainty of market participants
in their own assessment of future ination. Anecdotally, judging by
the diversity of opinions that various investors, pundits, and others
expressed in reaction to the Federal Reserves unconventional mon-
etary policy measures, there was plenty of disagreement in recent
years. Moreover, research has identied various dimensions of sys-
tematic disagreement between individuals. For example, Coibion
and Gorodnichenko (2015) highlight dierences between professional
forecasters and households. Malmendier and Nagel (2016) docu-
ment disagreement between individuals of dierent age with dierent
216 International Journal of Central Banking September 2016

accumulated ination experiences. The paper by Boneva et al. in

this issue shows that there is disagreement between dierent types
of rms in the United Kingdom.
The view that considerable disagreement about ination per-
sisted in the wake of the Great Recession in the United States
would also be consistent with the evidence in Cecchetti and Hakkio
(2009) that adoption of an explicit ination-targeting regime does
not reduce the dispersion of private-sector ination expectations by
much. Since the Federal Reserve has not adopted an explicit ination
target, there is even more reason to expect disagreement about the
long-run ination outlook to persist. Disagreement could potentially
oer a common explanation of the facts that ination rates and the
mean or median of professional forecasts were not sensitive to real-
ized ination rates and macroeconomic news in recent years, but at
the same time substantial dispersion in survey forecasts and in the
risk-neutral distribution extracted from option prices persisted.

4. Concluding Remarks

To sum up, option price data oer useful insights into the macro-
economic uncertainty perceived by investors. In particular, data on
ination options in the United States indicate that there is substan-
tial dispersion in the risk-neutral distribution of long-run ination.
Even though realized ination was quite stable in recent years and
measures of the central tendency of the perceived ination distribu-
tion did not move much, market participants still seem to perceive
uncertainty about the ability and willingness of the Federal Reserve
to keep long-run ination close to a stable target.
However, challenges remain in the interpretation of market-
based measures of uncertainty. One issue that seems to deserve
more attention of researchers is the role of disagreement in gen-
erating the risk premia that are embedded in the risk-neutral dis-
tribution of macroeconomic outcomes. Conventional interpretations
take a representative-agent approach. But there is reason to believe
that the tails of the risk-neutral distribution extracted from option
prices could be inuenced by disagreement among investors. If so,
a high degree of dispersion in the risk-neutral distribution could
indicate dierences in opinion among investors (who may be quite
condent in their own forecasts) rather than high levels of subjective
uncertainty.
Vol. 12 No. 3 Long-Run Ination Uncertainty 217

References

Bernanke, B. S. 2007. Ination Expectations and Ination Forecast-

ing. Speech at NBER Summer Institute Monetary Economics
Workshop, Cambridge, MA, July 10. Available at http://www.
federalreserve.gov/newsevents/speech/Bernanke20070710a.htm.
Buraschi, A., and A. Jiltsov. 2006. Model Uncertainty and Option
Markets with Heterogeneous Beliefs. Journal of Finance 61 (6):
284197.
Cecchetti, S. G., and C. Hakkio. 2009. Ination Targeting and
Private Sector Forecasts. NBER Working Paper No. 15424.
Coibion, O., and Y. Gorodnichenko. 2015. Is the Phillips Curve
Alive and Well after All? Ination Expectations and the Miss-
ing Disination. American Economic Journal: Macroeconomics
7 (1): 197232.
Davis, J. S. 2012. Ination Expectations Have Become More
Anchored Over Time. Economic Letter (Federal Reserve Bank
of Dallas) 7 (13).
Feldman, R., K. Heinecke, N. Kocherlakota, S. Schulhofer-Wohl, and
T. Tallarini. 2015. Market-Based Probabilities: A Tool for Poli-
cymakers. Working Paper, Federal Reserve Bank of Minneapo-
lis.
Garleanu, N., L. H. Pedersen, and A. M. Poteshman. 2009.
Demand-Based Option Pricing. Review of Financial Studies
22 (10): 425999.
Gurkaynak, R. S., A. T. Levin, A. N. Marder, and E. T. Swanson.
2007. Ination Targeting and the Anchoring of Ination Expec-
tations in the Western Hemisphere. Economic Review (Federal
Reserve Bank of San Francisco) 2007: 2547.
Kitsul, Y., and J. H. Wright. 2013. The Economics of Options-
Implied Ination Probability Density Functions. Journal of
Financial Economics 110 (3): 696711.
Malmendier, U., and S. Nagel. 2016. Learning from Ination Expe-
riences. Quarterly Journal of Economics 131 (1): 5387.
Stock, J. H., and M. W. Watson. 2010. Modeling Ination After
the Crisis. NBER Working Paper No. 16488.
Williams, J. C. 2006. Ination Persistence in an Era of Well-
Anchored Ination Expectations. Economic Letter No. 2006-27,
Federal Reserve Bank of San Francisco.