Ricardo A.M.R. Reis, Losing the Inflation Anchor (2022)
Summary
Ricardo Reis's paper Losing the Inflation Anchor, published in 2022 in the Brookings Papers on Economic Activity and held in LSE Research Online. The paper states that inflation has an anchor in expectations of its long-run value. It uses data on expectations from market prices, professional surveys and the cross-sectional distribution of household surveys to measure shifts in that anchor, with its main application to the US Great Inflation.
Summary drafted by a model from the document's text below and checked by script against that text before publication. It is a navigation aid, not a reading of what the document proves. Where AI is used
Cite this document
“Ricardo A.M.R. Reis, Losing the Inflation Anchor (2022).”
Original at researchonline.lse.ac.uk. Archived at Pandemic Darlings, https://pandemicdarlings.com/source-documents/ricardo-a-m-r-reis-losing-the-inflation-anchor-2022-97de9ed287/
, SHA-256 503026fd4afc66ad011792fe2f3136200f383db5f39890a54a5f2097e7830ca8.Full text
Reis, Ricardo (2022) Losing the inflation anchor. Brookings Papers on Economic
Activity, https://doi.org/10.1353/eca.2022.0004
https://researchonline.lse.ac.uk/id/eprint/112462/
Version: Published Version
LSE Research Online is the repository for research produced by the London School
of Economics and Political Science. For more information, please refer to our Policies
page or contact lseresearchonline@lse.ac.uk
Losing the inflation anchor∗
Ricardo Reis
LSE
October 2021
Abstract
Inflation has an anchor in people’s expectations of what its long-run value will be.
If expectations persistently change, then the anchor is adrift; if they differ from the
central bank’s target, the anchor is lost. This paper uses data on expectations from
market prices, from professional surveys, and from the cross-sectional distribution of
household surveys to measure shifts in this anchor. Its main application is to the US
Great Inflation. The data suggests that the anchor started drifting as early as 1967 and
that this could have been spotted well before policymakers did. Using this approach
on expectations data from Brazil, Turkey, South Africa, the US in the 1970s, and the
US in 2021, confirms their usefulness to measure the inflation anchor in real time.
∗ Contact: r.a.reis@lse.ac.uk. I am grateful to Alan Blinder, Janice Eberly, Yuriy Gorodnichenko, and
Jens Hilscher for comments. Discussions with Joao Ayres, Carlos Viana de Carvalho, Steve Davis, Erik
Hurst, and Valerie Ramey stimulated this project at its start. Gustavo Araujo, Viviane Bittencourt, Aloisio
Campelo Junior, Pierre Siklos, and Justin Wolfers helped to collect some of the data series. I am grateful
to Aleksandra Furmanek, Patrick Leitoff, Samantha Ong, and Annie Wang for their research assistance,
and especially to Adrien Couturier, Rui Sousa, and Borui Niklas Zhu. This paper was prepared for the
Fall 2021 Brookings Papers of Economic Activity conference on September 9, 2021 (first draft, August 2021). It
has received funding from the European Union’s Horizon 2020 research and innovation programme, INFL,
under grant number No. GA: 682288.
1 Introduction
Where is inflation heading? A natural way to answer this question is to measure where
people expect inflation to go. Even if each individual is likely surprised by transitory
shocks, there is a general principle that on average, across people and time, they will be
collectively right. Therefore, expectations data will reveal persistent changes.
When it comes to inflation, economic theory puts a particularly special role on expec-
tations as its driver. The classical dichotomy dictates that in the long run any inflation
rate is consistent with the same real outcomes and welfare. Pinning down inflation (in
the literature on determinacy) consists almost entirely of pinning down what people ex-
pect inflation will persistently be, and how this is influenced by rules, policy regimes, and
monetary standards (Castillo-Martinez and Reis, 2019). As for the short run, if expecta-
tions of inflation rise, households will buy more goods today, savers will shift away from
nominal assets, workers will demand higher wages, and firms will post higher prices, all
of which lead inflation to rise. The Phillips curve captures the core idea that if expected
inflation rises, then only a deep recession can keep inflation down (Mankiw and Reis,
2018). It is therefore no surprise that central bankers see it as a key part of their job to
keep expectations near the target.
Of course, in equilibrium, both inflation and expected inflation are determined to-
gether. Also because of this co-determination, in order to gauge future inflation, beyond
data on present inflation, a superior source of information ought to be data on expected
inflation. To answer the question of where inflation is heading, we are like a person sitting
on the beach, seeing a boat float with the sea current, trying to figure out if it is drifting
away or staying near shore before it is too late. A screen showing the feed from an un-
derwater camera with even a grainy picture of the boat’s anchor would be very valuable.
This paper tries to build this grainy picture of the inflation anchor.
In particular, it asks whether data on expectations in the past have helped to detect
persistent deviations of inflation from target. Looking back at the US Great Inflation of
the 1970s, could we see signs of what was coming in the expectations data? Do these data
help advance the debate on what caused it? Were the same data patterns present in two
other experiences of drifting inflation, Brazil and Turkey in the past decade? Or would
the data generate false positives, missing the stabilization of US inflation in the 1980s, or
the stability of inflation in South Africa in the last decade in spite of successive adverse
shocks? Finally, what do the expectations data reveal about inflation in 2021, after all the
price volatility brought about by the pandemic?
1
These are difficult questions to answer empirically. They require having, for the same
period in time, both an explosion in inflation and available data on inflation expectations.
Yet, while inflation scares were common until the mid 1990s, most reliable measures of in-
flation expectations start only later. Survey measures in most countries were promoted by
central banks under inflation targeting, and this policy regime has so far been successful
at keeping inflation stable. As for sovereign bonds whose returns are indexed to inflation,
or markets in which one can trade inflation swap contracts, they are also recent and grew
in counties where inflation was reasonably under control. In short, between 1970 and
1995, many anchors were lost, but there are almost no expectations data; between 1995
and 2020, there are data, but no lost anchors.
This paper starts by focusing on the US experience between 1966 and 1973. In the
previous seven years, annual inflation had not been above 2% for a single month. But,
between January of 1966 and September of 1973, just before the first oil price shock, infla-
tion averaged 4.3%, rising from 1.9% to 7.4%. It is well appreciated that in the following
seven years, from 1974 to 1981, US inflation fluctuated widely and was never below 5%.
But the drift in inflation happened before, gradually and steadily from 1966 onwards.
To measure the inflation expectations anchor requires introducing data that have rarely
been used. The standard surveys of expectations about CPI inflation in their current form,
the Survey of Professional Forecasters and the Michigan survey, start in 1981 and 1978,
respectively. Inflation-indexed bonds started only in 1997, and the inflation swap mar-
kets was only reasonably liquid starting in 2009. Some research has used the Greenbook
forecasts, and the mean of earlier vintages of the Survey of Professional Forecasters or the
Livingston Survey to assess expectations in the 1960s and 1970s. But their samples are
small, making it hard to estimate any cross-sectional moment beyond central tendency.
This paper brings in some less-used sources of data on inflation expectations. First,
between the second quarter of 1966 and the second quarter of 1977, the Michigan survey
included a qualitative question about whether inflation was expected to go up, down, or
remain unchanged and supplemented the up answers with a follow-up question on by
how much, with a choice of 5 bins. I use those answers to provide measures of cross-
sectional mean, variance and skewness of inflation expectations among households dur-
ing this key period in the history of US inflation. Second, I use data from the Zurich mar-
ket on gold forward contracts in USD, one of the few international markets where these
were freely traded. Third, I complement these continuous series with more scattered data
on expectations by firms and measures of media attention to inflation.
2
At first sight, it may appear that expectations lagged the drift away of US inflation.
The cross-sectional median (or mean) survey expectation for professionals or households
only rose sluggishly and with a lag relative to actual inflation during the 7 years. How-
ever, between 1968 and 1971, the disagreement captured by the cross-sectional standard
deviation and skewness combined with the rising forward prices reveal a clear drift of
the inflation anchor. The wage and price controls adopted in 1971 seem to have halted
this drift for a little under two years, but after the oil price shock of late 1973, the anchor
was clearly lost. Using the approach in Reis (2021) to combine the different expectations
data, I estimate a fundamental expected inflation. It shows the anchor drifting by 2.5
percentage points between 1967 and 1971.
The next section of the paper looks at other episodes, across countries and time, to
see whether the expectations data provides a reliable measure of the anchor. First, it
looks at two cases where inflation drifted away: in Brazil between 2011 and 2016 and in
Turkey starting in 2018. In each, looking back today with the benefit of hindsight to what
happened to actual inflation and what were the stances of monetary and fiscal policy, the
upwards drift of inflation is clear. The paper shows that looking at expectations data at
the time would have signaled that this was the case in real time. Second, it looks at a
possible false positive, South Africa between 2010 and 2016, which had inflation at the
very edge of its target corridor. At the time, it was unclear whether the anchor had been
lost, or whether an unlucky succession of supply shocks had temporarily raised prices.
Today, hindsight points to it being the latter, and again the real time data on inflation
expectations points in the right direction. Third, it looks at a case where inflation drifted
down, instead of up, the US experience of the early 1980s. The expectations anchor moved
accordingly, and again the data in real time reveals it. Finally, it looks at the data on
expected inflation in 2020 and 2021, where the jury is out on whether inflation is going to
drift away.
All combined, this paper concludes that, while market and survey data on expecta-
tions are noisy and imperfect, they provide valuable signals on where the expected infla-
tion anchor is.1
1 The closest work to this paper is Jalil and Rua (2016) who use news records and scattered business
forecasts to infer whether inflation expectations shifted in 1933, triggering the start of the recovery from the
Great Depression. See also Binder (2016) for other uses of expectations inflation data in economic history.
3
2 U.S. inflation 1966-73: Was the anchor adrift?
The top panel of Figure 1 plots inflation in the United States during the 1960s and 1970s
according to three commonly-used price indices. Before 1965, inflation was remarkably
stable, oscillating in a narrow band between 1% and 2%. The (incomplete and noisy) data
on inflation expectations from this time show an anchor that was firm in the seabed. After
the Treasury-Fed accord of 1951, monetary policy had been focussed on inflation control,
and an independent central bank was able to deliver it. Fast forward a decade, to 1975,
and there is little doubt that the expectations anchor was lost, with actual inflation being
high and volatile. Low and stable inflation only returned after a painful recession in the
early 1980s.
The focus of this section is on the decade between the start of 1965 and the end of
1974. Inflation gradually rose between 1965 and 1967, its pace accelerated between 1967
and 1971, and then slowed down between the middle of 1971 and the end of 1973. While
at that time there was a legitimate argument that the anchor was re-settling, as soon as a
price shock hit in the form of the rapid rise in oil prices at the end of 1973, it became clear
that inflation, whether going up or down, was definitely adrift. There are many discus-
sions of what was behind these movements, which I summarize below, before turning to
the focus of this paper: the role that inflation expectations played in policy and academic
debates of this time.2
2.1 Inflation 1965-68: All the signs were there
Inflation and monetary policy. Under the William McChesney Martin Jr. chairmanship
of the Board of Governors, the Fed settled into a policy regime which he termed “leaning
against the wind,” and that resulted in short-term interest rates moving countercyclically.
After a brief recession, the US economy went through an expansion between 1961 and
1965, with the unemployment rate falling by 4% but inflation only moderately increasing.
Things changed in the second half of 1965. Inflation rose above 2%, and the Fed stepped
on the brakes, raising the discount rate in December of 1965. It proceeded with other
measures to restrict credit in the first half of 1966 that were enhanced in its August 1966
meeting. This policy was a response to both the current state of the economy, as well as
to projections of growing military expenditures because of the Vietnam War and grow-
2 Some excellent discussions are in Blinder (1982), DeLong (1997), Romer and Romer (2002), Meltzer
(2010), Hetzel (2020).
4
Figure 1: The evolution of the US economy
(a) Actual inflation
Anchor In Seabed A Drifting Anchor First Oil Unanchored Inflation
Shock
12%
CPI Inflation
CPI Core Inflation
GDP Deflator Inflation
8%
USD Off
Gold
4%
0%
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981
(b) Other macroeconomic series
12
Federal Funds Rate
10−Year Treasury Yield
Unemployment Rate
Annual Federal Deficit (% GDP)
8
4
0
1965 1970 1975
5
ing welfare state spending under President Johnson’s Great Society legislation (Meltzer,
2005).
However, the monetary restraint was gone by the end of 1966. At the time, the Fed
was accused of causing a credit crunch. The United States was going through turbulent
political times, due to social divisions on race and the Vietnam War. Fiscal policy had
shifted from being a stabilizer of the business cycle in the Keynesian post-war tradition
to instead seeing its role as lowering unemployment as much as possible to raise the tide
on all groups in the labor market. Martin thought policy coordination with the adminis-
tration was important, and the Fed no longer tightened, especially as it subscribed to the
view that an unemployment rate above 4% justified loose monetary policy.
Not only did the Fed not tighten further in 1966, in spite of rising inflation (already
above 3%), but when unemployment slightly rose in the Spring of 1967, it reversed course
with cuts in interest rates to arrest any chance of a recession. The mild slowdown of 1967
gave way to a strong expansion in 1968, and inflation rose to the 4.5% region.
Debates in the literature. Most accounts of the sources of the Great Inflation find its seeds
during this period. A first seed was a view of a static Phillips curve such that any inflation
was deemed temporary and acceptable in order to keep unemployment low (Romer and
Romer, 2002). A second one were the estimates that the economy’s acceptable rate of
unemployment was 4% in spite of a significant deceleration of productivity (Orphanides,
2003). Third, there was a misperception of the link between monetary policy and inflation
that ignored the role of money supply and did not distinguish between nominal and real
interest rates (Nelson, 2005). Fourth, despite a permanent and growing external deficit, as
well as a large decline in the US stock of gold, the FOMC deferred the defense of the gold
peg to the Administration and its active use of capital controls, but the Treasury after 1965
subordinated the commitment to the Bretton Woods system to its goals for fiscal spending
and low unemployment (Bordo and Eichengreen, 2013).
Common to all four factors — the Phillips curve, the natural rate of unemployment,
the role of monetary policy, and the defense of the gold peg — was Martin’s rejection
of explicit economic models, and the general lack of a coherent framework (Meltzer,
2005). Martin gave many speeches opposing inflation and emphasizing its costs, and
in his meetings with the Johnson Administration he openly disagreed with their sole fo-
cus on employment. But, to control inflation, he paid attention to money market data and
participants, not economists. Discussions in the FOMC more often focussed on current
technical market details, especially with regards to the globalization of financial markets
6
and banks’ response to capital controls, rather than reflecting on the economic fundamen-
tals of inflation.
The role of inflation expectations. This short-run market-centered view created a spe-
cial role for inflation expectations. Martin placed a strong role on investor psychology in
financial markets as driving inflation. Early in his term, testifying to Congress about the
slight increase in inflation in the summer of 1959, he stated: “About this time inflationary
expectations began to spread. The abrupt upward shift of interest levels in central money
markets ... reflected investor demand for an interest premium to cover the risk of a depre-
ciating purchasing power of invested funds.... The experience in the government bond
market ... is a vivid example of the influence of inflationary expectations in financial mar-
kets. To the extent that such attitudes come to be reflected in decisions on wages, prices,
consumption, and investment, they help to bring about their own realization.” Later, to
explain the clear rise in inflation in 1967-68, FOMC member speeches frequently refer-
enced that ”inflationary expectations remained widespread,” and in December of 1969 to
”the prevailing inflationary psychology.” Part of the drift to loose monetary policy during
this time can be explained by the subdued values of long-term interest rates. They were
used as the measure of both expected inflation and inflation risk premia. Since they were
low, Martin inferred that the anchor was firmly in place.
In sum, monetary policy (in relation to inflation expectations):
• saw the inflation expectations anchor as important to keep inflation stable;
• inferred expectations from bond rates and market movements;
• because bond rates remained low, the Fed was growingly surprised by the drift in
inflation in 1965-68;
• so it increasingly relied on appeals to unobserved psychology to justify the changes.
2.2 Inflation 1968-71: The anchor comes loose
Inflation and monetary policy. In 1969 and 1970, the economy experienced a mild down-
turn. GDP fell by less than the fall in federal defense spending. Disposable income actu-
ally rose between the peak and the trough of the recession, as the tax surcharge of 1968
expired and the automatic stabilizers in the welfare state kicked in. Monetary aggre-
gates kept rising though, and inflation accelerated to 6%. Martin significantly tightened
7
monetary policy; December of 1968 is one of the Romer and Romer (1989) dates for a con-
tractionary monetary shock. Yet, this arrived late and did not last long. In February of
1970, Arthur F. Burns became the new chairman of the Board of Governors. The decision
at his first FOMC meeting was to ease monetary policy.
Debates in policy and academia. At the time, there was an active debate about the
Phillips curve, which Burns was keenly aware of. The economists in the recently-elected
Nixon administration accepted the analysis in Friedman (1968), and thought that any
reduction in unemployment achieved by increasing inflation was temporary (Meltzer,
2005). But many economists in policy still thought that an overheated economy would re-
sult in higher inflation, but not accelerating inflation.3 On the other side, some academic
economists had embraced the view of an expectations-augmented Phillips curve, where
the expectations term was proxied by lagged inflation, leading to the so-called accelera-
tionist Phillips curve.4 By the end of 1971, the latter group had won the debate in light of
the continued rise in inflation in spite of the recession.5
Even then, an accelerationist Phillips curve implied that reducing inflation even by
a few percentage points back to 2% would come with large sacrifices in terms of higher
unemployment. Andrew Brimmer, a Board member between 1966 and 1974, affirmed
that there was little doubt within the Fed: “Fighting inflation, checking inflation was the
second priority”. Even in April of 1969, after all the signs of lower productivity and a
rising natural rate of unemployment (Orphanides, 2003), Arthur Okun, chair of the CEA,
wrote that it was compatible to have a long-term “4 percent rate of unemployment and a
2 percent rate of annual price increase” (Okun, 1970).
The role of inflation expectations. In the Martin Fed throughout 1969, there were men-
tions that “inflationary expectations remained widespread” and of ”expectations of con-
tinuing inflation” (Romer and Romer, 1989). The tightening of monetary policy was jus-
tified by appealing directly to expectations: ”the Committee agreed that, in light of the
persistence of inflationary pressures and expectations, the existing degree of monetary
restraint should be continued at present” and that “The Committee took note of the signs
of some slowing in the economic expansion and of the indications of stringency in finan-
cial markets. In view of the persistence of strong inflationary pressures and expectations,
3 See, for instance page 95 in the Economic Report of the President of 1969.
4 However, there was an active debate on whether to interpret the lags on inflation as arising because of
adaptive expectations or because of institutional rigidities (Gordon, 1980, Tobin, 1980).
5 See Sargent (1988), Primiceri (2006) for formal models of this change in ideas.
8
however, the members agreed that a relaxation of the existing degree of monetary re-
straint would not be appropriate at this time” (1969 Annual Report, pages 121 and 145,
respectively). In July of 1969, Martin stated that “inflationary psychology remained the
main economic problem ... It would be a mistake ... to take any action that might rein-
force inflationary expectations just at the time when some weakening in those expecta-
tions might be developing” and in August that “it was important for the System not to
get into a position of validating the expectations of numerous skeptics who believed the
System would ease its policy as soon as it heard the words ‘recession’ or ‘overkill’” (cited
in Weise, 2012).
With respect to inflation expectations, during the period 1968-71:
• they became central in economic discussions as a shifting term in the Phillips curve;
• but their measurement was still inexistent, beyond using past actual inflation to
proxy for it;
• policymakers kept appealing to shifts in psychology with no hard data to back it.
2.3 From 1971 to 1974: The anchor adrift
Inflation and monetary policy. In 1972, the economy turned sharply to a strong expan-
sion. Real GDP grew by an astounding 6.9% by the end of the year (but the unemploy-
ment rate fell by only 0.8%). Policy was strongly procyclical, as the decline in government
purchases of the previous years was halted, and the Fed kept policy loose. From the per-
spective of inflation though, two milestone events in 1971 drove a fall in CPI inflation
during that year, and an acceleration in the next.
The first of these was the end of Bretton Woods, with the suspension of dollar gold
convertibility on the 15th of August 1971. The steady increase in inflation in the previous
6 years had caused a large outflow of gold from the United States, and such a large imbal-
ance in balance of payments was no longer sustainable. Without this constraint, inflation
was free to rise much more, and the dollar depreciated continuously between 1971 and
1973. Expectations of this depreciation could now drive inflation.6
The second key event was the imposition of a wage and price freeze by the Nixon Ad-
ministration on the 15th of August of 1971 as part of its New Economic Policy (NEP). This
6 By itself, a flexible exchange rate did not have to cause inflation. Free from the influence of US inflation,
West Germany went through a disinflation in 1971, almost a decade before the United States had its own.
9
brute-force approach to keep CPI inflation down was fully supported by Burns. Nominal
wage growth had risen by more than 8% per year in 1970, in spite of the recession and
stagnant productivity, and he blamed this on the monopoly power of labor unions and
the generosity of welfare programs. An automobile labor strike in 1971 cemented this
view, and the annual report of the Federal Reserve was filled with analyses of unit labor
costs. Burns dismissed the fast growth in monetary aggregates, and thought that the con-
trols and other guideposts would break the expectations of workers. Hetzel (1998) argues
that, even well before this episode, Burns had long regarded wage-price spirals driven
by the expectation of inflation as a key driver of high inflation: “One of the main factors
in the inflation that we have had since the end of World War II is that many consumers,
businessmen, and trade union leaders expected prices to rise and therefore acted in ways
that helped to bring about this result” (Burns, 1957, page 71). Government interference in
private price and wage setting was to be the way of breaking the spiral.
From then on, between 1972 and 1974, inflation accelerated. Besides the bounce-back
from the price controls and the continued effect of the depreciation of the dollar, the most
well known direct cause was the rapid increase in oil and food price shocks. Crop failures
in 1972, an El Nino event that made Peruvian anchovies disappear, and large sales of US
wheat to the Soviet Union led to a surge in world food prices in 1973-74. The OPEC oil
embargo after the 1973 Yom Kippur War led to oil prices increasing by a factor of 4 by
1974. By themselves, they can account for more than half of the increase in inflation and
its fall in the following years once oil prices stayed high but stopped rising (Blinder and
Newton, 1981, Blinder and Rudd, 2013).
Yet, the reversal down of inflation after the oil shocks was far from complete. A possi-
ble explanation is that the temporary inflation shock raised inflation expectations becom-
ing inertial. Policymakers at the time thought so, leading to the famous “whip inflation
now (win)” public relations campaign in 1974 to try to lower the psychology of infla-
tion expectations, one of the most infamous failures of policy communication that is not
backed by actual policy actions.
Theories and expectations. Beyond supply shocks, the literature has offered two other
factors that contributed to the acceleration of inflation in these years. Each involves ex-
pectations at their center.
The first of these was a confusion between transitory and permanent shocks at the
Federal Reserve. At the helm, Burns was confident of his forecasting ability guided by
reduced-form empirical analysis of data, shaped by his experience as president of the
10
National Bureau of Economic Research and its measurement without theory tradition
(Hetzel, 1998). In his words, testifying to Congress in 1974, “The current inflation began
in the middle 1960s when our government embarked on a highly expansive fiscal policy.”
In turn, he thought the inflation of 1973 was due to food and oil prices, and the further in-
crease in 1974 was due again to budget deficits (even though those had been small). There
was always a temporary shock to explain the persistent drift. Meltzer (2005) describes this
period as one when, among the Federal Reserve staff, “... they gave special explanations
— a relative price theory of the general price level — in effect claiming that the rise in
the price level resulted from one-time, transitory changes that they did not expect to re-
peat.” Likewise, by the end of 1972, an unemployment rate of 5% was seen as close to full
employment, higher than the previous 4% but still well below what our estimates today
suggest (Reis, 2003). This sluggish adjustment upwards of the natural rate in spite of the
higher unemployment and acceleration of inflation of previous years implicitly reveals a
view that this was due to temporary shocks to the natural rate of unemployment, coupled
with a belief that expected inflation had not changed.
Burns viewed monetary policy as affecting credit conditions and through it the con-
fidence and investment proclivity of businessmen, with money at best reflecting these
rather than causing them (Hetzel, 1998). Shifts in expectations were then always tenden-
tially transitory, with a multitude of policies being able to bring them back in line. For
instance, testifying to Congress in 1964 about deficit reduction, he stated: “In this new
psychological environment, our trade unions may not push quite so hard for a large in-
crease in wage rates, since they would no longer be anticipating a higher inflation rate.
And in this new psychological environment, our business people would not agree to large
wage increases quite so quickly”.
This last quote points to the second factor: the emergence of an inflation bias through
expectations. As articulated by Tobin (1980), this was a break with the Keynesian thought
of the previous two decades, since it implied that monetary and fiscal policy manage-
ment would be unable to cure stagflation. Arthur Burns described this in his lecture that
reflected on his experience (Burns, 1987): “In earlier times, when a central bank permit-
ted excessive creation of money and credit in times of prosperity, the price level would
indeed tend to rise. But the resulting inflation was confined to the expansion phase of
the business cycle; it did not persist or gather force beyond that phase. Therefore, people
generally took it for granted that the advance of prices would be followed by a decline
once a business recession got under way. That is no longer the case. Nowadays, business-
11
men, farmers, bankers, trade union leaders, factory workers, and housewives generally
proceed on the expectation that inflation will continue in the future, whether economic
activity is booming or receding.”
One of the reasons for this change was the political influence over monetary policy. Es-
pecially through 1972, the Nixon Administration, and the President personally, strongly
pressured the Federal Reserve to not lean against the economic expansion and jeopar-
dize the Fall reelection, and the minutes and transcripts of the Federal Reserve show a
concern about the political repercussions of its measures (Weise, 2012). However, they
also reflect a view that this would not compromise long-run price stability because the
election would be over soon. The inflation bias that yielding to political pressure would
have on expectations given recurrent elections was dismissed, and yet it was visible in
the data for the two decades that followed (Alesina, 1988). As Gordon (1980) writes: ”The
overheated expansion of 1972—73 is perhaps the leading postwar example of Nordhaus’
(1975) “political business cycle” in action.”
In short, while the oil price shocks were significant and unforeseen, the shoot up of
inflation in 1971-74 and its lasting effects through the rest of the decade worked in part
through their effect on expectations since:
• the gold peg was not replaced by another regime that would keep the expectations
anchor steady;
• hope was placed on price and wage controls affecting inflation expectations;
• communication, not backed by actual polices, failed to bring down inflation expec-
tations;
• internal forecasts that all shocks had temporary effects led to a belief that inflation
expectations remained anchored, even as there were growing suspicions of an infla-
tion bias in expectations emerging, but no data confirmed either.
2.4 Conclusion: Expectations as add-factors
The decade between 1965 and 1974 was filled with shocks, changes in regime, and con-
tributing causes to the rise in inflation. The literature that has inspected and debated
them has fallen into three competing hypothesis: a politics view, that focuses on the pres-
sure from expansionary fiscal policy and the lack of independence by Martin and Burns
relative to the Administrations; the ideas hypothesis, that focuses on incorrect views of
12
the Phillips curve, the value of the natural rate of unemployment, the power of monetary
policy, and the role of Bretton Woods; and the supply shocks hypothesis, that focuses
on declining productivity, the Nixon price controls and the OPEC oil shock. Across all of
them, the political context of the time played an important role, and debates will continue
on the relative weight of each view.
The discussion in this section highlights that expectations play an important role in
these discussions. Policymakers used changes in expectations to justify why policies had
failed in the past, but would not in the future. Academics appealed to shifts in expecta-
tions to explain why economic relationships changed, why the monetary policy regime
led to high inflation, or why inflation stayed elevated. Throughout, however, data on ex-
pectations was conspicuously missing. Like add-factors in forecasting models, accounts
of the Great Inflation use private inflation expectations as unobservables that help explain
some of the data.
At times, data on expectations were replaced by appeals to the psychology of markets
or businessmen. Other times, they were only indirectly inferred from changes in macroe-
conomic variables using particular economic theories. With the widespread use of ratio-
nal expectations from the 1980s onwards, expectations became tightly tied to observed
fundamentals, so the obstacles in measuring and observing them were circumvented al-
together. None of these options is very useful when trying to assess whether inflation
expectations are anchored in the present. Appeals to how people may be feeling are too
subjective. Macroeconomic variables can only be used with a long delay to infer what
were expectations. And, while expectations are undeniably forward looking, they are
also not strictly rational in the rational-expectations sense, leaving us with a wide set of
models of where they may be drifting. The next section turns instead to real-time data
from surveys and market prices to see whether they can provide timely signals of how
solid the anchor is.
3 Data on inflation expectations
This section canvasses sources of data on the inflation expectations anchor during the
period when inflation went adrift.
13
Figure 2: Expected inflation by professionals
Livingston median expected inflation
6% SPF median expected inflation
Greenbook inflation forecast
4%
2%
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
3.1 Inflation expectations: Professionals
The first source of data was surely looked at by policymakers: the forecasts made by
the professionals that work at the Federal Reserve System. These are published in the
Greenbook, and are nowadays made public by the FRB Philadelphia. Starting in 1967Q1,
the Fed staff produced forecasts for the GDP deflator over the next 3 months. Starting in
1969Q4, there are also forecasts over the next twelve months, and so for annual inflation
one-year ahead.
These forecasts are plotted in figure 2 where in the horizontal axis is the date at which
the forecast was made (so in quarter t it shows the report made at t for inflation between
t and t + 4). The data is consistent with the narrative analysis of Fed statements in the
previous section that it thought that the successive rises in inflation were temporary one-
off deviations. The forecasts were significantly below the realizations of inflation, and
only adjusted upwards sluggishly. The top panel of figure 3 confirms this view by plotting
both the 3 months ahead expectation, the 12 months ahead, and the 9-12 months ahead
expectation. Inflation was almost always expected to be higher in the nearer term than in
the farther away term.
At the same time as the Greenbook forecasts become available, the American Statisti-
cal Association and the NBER started surveying a group of professional forecasters work-
ing for major banks and large corporations. The typical survey at the time had 10 to 30
14
Figure 3: Expectations of inflation at different horizons
(a) Greenbook, Fed staff (b) Livingston, private-sector economists
8%
12 months ahead 12 months ahead median
9−12 months ahead 6 months ahead median
3 months ahead 6% 6−12 months ahead
6%
4%
4%
2%
2%
1967 1968 1969 1970 1971 1972 1973 1974 1975 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
respondents, and took place every quarter. The series in these data for 12-month ahead
CPI inflation forecasts is commonly used, but it only starts in 1981Q3. However, the sur-
vey started earlier, and from 1968Q4, it included questions on the expected growth of
nominal and real GDP; combining the two yields their forecasts for the GDP deflator. A
less used but similar survey of professionals was undertaken by journalist Joseph Liv-
ingston who asked between 12 and 48 economists in academia, business, and markets,
twice per year, starting in 1946, about CPI inflation over the next 6 months and over the
next 12 months.
The cross-sectional median in the two surveys of expected inflation over the next 12
months is plotted in figure 2. Because the size of the samples are small, one cannot esti-
mate reliably the cross-sectional distribution beyond a first moment. The median forecast
was, similar to the Fed’s forecasts, sluggish and slow to keep up with the rapid increase in
inflation at the time. There is a telling difference between the two revealed by comparing
the two panels of figure 3. The private professionals seemed less convinced that the infla-
tion rise was temporary, as they would raise their expectations at the different horizons
in lockstep.
3.2 Inflation expectations: Households
The main survey of household expectations of inflation is the Survey of Consumer Atti-
tudes and Behavior administered by the University of Michigan Survey Research Center
(SRC). Since January of 1978, it has surveyed around 500 to 700 households every month,
asking them: “By about what percent do you expect prices to go up, on the average, dur-
ing the next 12 months?” An early version of the quantitative inflation question was tried
15
Figure 4: Household qualitative expectations of inflation going up/down
75%
50%
Percentage up
Percentage down
25%
0%
1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
in 1976 and 1977, only every 6 months, still outside our period of interest.
Before 1976, and starting in 1946, the survey existed but it was quarterly and it only
asked a qualitative question about inflation. The precise question changed over time, but
between 1961Q2 and 1977Q2 it was: “Speaking of prices in general, I mean the prices of
the things you buy–do you think they will go up in the next year or go down?” Figure 4
shows the time series of the answers to that question between 1965 and 1974. It is hard
to see much of a change. At the same time, with the majority of respondents already
anticipating prices to rise from the start, a shift up in how much they think so would not
be detected.
Luckily, the SRC asked a supplementary quantitative question after this one between
1966Q2 and 1976Q4. Respondents who expected prices to rise were asked to state in
which of a few bins their mean expectation would be. The precise question was: “How
large a price increase do you expect? Of course, nobody can know for sure, but would
you say that a year from now prices will be about 1% or 2% higher, or 5% higher, or
closer to 10% higher than now or what?” While the answers to this question were used
in academic articles published in the early 1980s, the data seem to have been forgotten
(and it is not easy to access); I found no paper using them that was published in the last
twenty years. To convert the bins answers into numerical answers, I use the standard (but
16
Figure 5: Cross-sectional mean of expected inflation by households
12%
Michigan quantitative survey mean
Mankiw−Reis−Wolfers qualitative survey mean
10%
8%
5%
2%
0%
1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
imperfect) procedures in the literature (e.g., Pesaran and Weale, 2006).7
Figure 5 shows the resulting cross-sectional mean expected inflation. Also plotted in
the figure is a series built by Mankiw, Reis and Wolfers (2004) that used only the answers
to up/down and assumed a normal distribution to back out a mean and a standard de-
viation at every date. As with the first moment from the professional surveys, there is no
discernible trend in the central tendency of the household survey until the very end of the
decade. As the literature has found for other periods, also during this time, the average
household expectations appears very sticky (Coibion and Gorodnichenko, 2012).
3.3 Disagreement and the cross-sectional distribution of expectations
In light of what we know today from the literature on expectations, it is not so surprising
that measures of the first cross-sectional moment of households surveys are seemingly
unresponsive to shocks, and only very sluggishly adapt to new regimes (Coibion, Gorod-
nichenko and Kamdar, 2018). Most people are uninformed, inattentive, or just ignorant
about inflation. However, the literature that has focussed on the data since 1980s has
also found that the cross-sectional distribution in the expectations contains plenty more
information (Mankiw, Reis and Wolfers, 2004, Coibion and Gorodnichenko, 2012). Mod-
ern theories of inattention, learning, or mental biases highlight that people differ in how
7 My estimates are not too different from those in Juster et al. (1972) and Juster and Comment (1980).
17
Figure 6: Disagreement about expected inflation among households
(a) Cross-sectional moments (b) Three snapshots of the distribution
0.75
5% 1967
0.10 1970
1974
0.50
Standard deviation (right axis)
Skewness (left axis) 4%
0.25 0.05
0.00
3%
0.00
1967 1968 1969 1970 1971 1972 1973 1974 1975 1.5 3.5 5.0 7.5 12.0
fast they adjust to news in systematic ways (Mankiw and Reis, 2010, Angeletos, Huo and
Sastry, 2020). By comparing their expectations, we can infer where they are eventually
heading, and thus where the anchor is. The distribution captures the disagreement be-
tween people, some better informed than others. It can be quite informative of what the
anchor is.
The top panel of figure 6 plots the cross-section standard deviation and skewness of
household expectations using the Michigan data. Starting with the forecasts from 1968
about inflation in 1969, the disagreement in the series clearly starts rising for about two
years. For the following three years, disagreement is approximately unchanged. But, after
the oil shock at the end of 1973, disagreement about expected 1974 inflation jumps up, and
stays elevated after that. Skewness, which throughout is positive, also changes around
1968, clearly falling for the first two years. When disagreement settles, skewness starts
rising, and when at the end of 1973 disagreement jumps up, skewness jumps down.8
Combined, these two series tell the following story, which the bottom panel with the
cross-sectional distribution at three points in time confirms. While inflation may have
started drifting up in 1965, it is only in the forecasts made around 1967-68 that expec-
tations seem to show some drift in the anchor. The previous section discussed how the
1969-70 downturn marked a turning point in academic thought, with its clear rejection of
a static Phillips curve as inflation accelerated. In household expectations, there is also a
difference. The mass of agents expecting low inflation thinned out as more households
moved to the long right tail of the positively-skewed distribution. This raised dispersion
8 Estimates of kurtosis or higher-order moments show no appreciable trend. This is also the case in the
quantitative Michigan survey answers after 1978: three cross-sectional moments seems to suffice to capture
all the relevant time-series variation in the distribution.
18
Figure 7: Media mentions and public concerns related to inflation
15%
Central bank % mentions in NYT (left axis)
Inflation % mentions in NYT (left axis)
Concern about inflation − about unemployment Gallup (right axis)
60%
10%
40%
5%
20%
0% 0%
1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975
while lowering skewness. Then, between the 1971 and the middle of 1973, the incomes
policy seems to have worked in the way Burns hypothesized, by reversing the increase in
expected inflation. This is consistent with the robust empirical finding that people form
their expectations using a subset of goods, and that large changes in public policy trigger
attention to inflation.
Yet, the oil shock, which had only a small and delayed effect on the first cross-sectional
moment, had a large impact on the second and third moments in 1974. There was an
immediate large movement of mass from the left and middle of the distribution to its
right tail. If the wage and price controls had an effect on expected inflation, it was only a
temporary one.
3.4 Other, imperfect, proxies for peoples’ expectations
Figure 7 plots a proxy for attention to inflation news by households: the number of news
references to inflation or to the Federal Reserve, in articles published in the New York
Times. Some research suggests that the news can proxy for people updating their expec-
tations more often (Carroll, 2003). The uptick in news stories starts in 1968. Its relative
decline in 1972, and the surge upwards after the oil price shock, tracks the patterns of
disagreement in the household survey.
The figure also shows the difference between the share of households that report that
19
inflation is the nation’s biggest problem minus those that mention unemployment as a
concern instead in the Gallup poll. The uptick here is higher, and it happens already at the
end of 1966, with no decline in 1970-72. The rise after the oil shock is again considerable.
One concern with the survey data used so far is that its horizon was for the most part
one year. Yet, economic theory would suggest that it is longer-horizon inflation expecta-
tions that provide the anchor for the persistent component of inflation. Moreover, insofar
as inflation may be sometimes affected by shocks with a half-life of 2 or 3 years, then
1-year ahead expectations may be more volatile than their 5 or 10-year ahead counter-
parts giving a misleading impression of an anchor that is not firm. The Michigan survey
started asking about long-term inflation expectations after 1990. Focussing on the last ten
years, since 2010, the correlation between the standard deviation of the 1-year ahead and
the long-term inflation answers is 0.91. The correlation between the skewness measures
is lower, 0.4, but mostly because of the considerable month-to-month noise in estimating
it, well visible in figure 6. The low-frequency movements that we emphasized are quite
similar.
Finally, because firms choose prices, the expectations of their managers may be par-
ticularly important as an anchor. Most of them will not employ a professional economist,
but at the same time may have a greater financial incentive than households to get their
forecasts right. However, research has found that firms do not seem to make more ac-
curate forecasts than households, and too often confuse industry-specific relative price
changes for overall inflation (Candia, Coibion and Gorodnichenko, 2021, Andrade et al.,
2021).
The only series that I was able to find on firm expectations was asked by the Bureau
of Economic Analysis as part of the annual survey of business expenditures on plant and
equipment published in the Survey of Current Business from 1970 onwards. Approxi-
mately 3200 firms answered the question in year t: “What are your best estimates of av-
erage price changes from t − 1 to t and from t to t + 1? (Approximations are acceptable.)”
The question was asked separately for capital goods purchases, and for the goods and
services sold; Figure 8 reports both, as tabulated by de Leeuw and McKelvey (1981). The
pattern in the figure roughly confirms the fact that by 1971, expectations were elevated,
but over the next two years they seemed to fall, before shooting up in 1974 following the
oil price shock.
20
Figure 8: Firm expectations of inflation
12%
% change in prices of goods and services sold
% change in prices of capital goods purchased
10%
8%
5%
2%
1971 1972 1973 1974 1975
3.5 Inflation expectations: Markets
Most central banks today look at market prices to infer inflation expectations at high fre-
quency. These can be useful because of the strong financial incentive that traders have to
get this right. Moreover, the market price reflects the expectation of the marginal trader.
Comparing it with that of the average trader, or person, provides a signal of the distribu-
tion of inflation expectations. At the same time, market prices are often polluted by noise
arising from financial and liquidity frictions.
Market expected inflation is usually measured using either the price of inflation swap
contracts, or the difference between yields on inflation-indexed and nominal government
bonds. Neither existed in 1965-75. McCulloch (1980) argued that these contracts could
not even exist, as they were illegal. In 1933, Congress had approved a “Gold Clause”
prohibiting bonds indexed to the price of gold as being against public policy. Under
Bretton Woods, court rulings extended this prohibition to any domestic contract indexed
to the price level. The clause was only explicitly repealed in 1977, at which point futures
contracts for gold started being traded in the United States.
The US Treasury issued foreign currency denominated bonds, in West German marks,
Swiss francs, Italian lira, Belgian francs, and Austrian schillings. These “Roosa bonds”
were medium-term notes issued between 1962 and 1971 to help relieve the pressure on
US gold holdings during its attempt to defend the Bretton Woods system in spite of rising
21
US inflation. But they were non-marketable, held by foreign governments. Their interest
rate does not reflect market expectations.
Chairman Martin often referred to long-term bond yields as a measure of market ex-
pected inflation. If real interest rates and long-term interest-rate risk were constant, then
these would be market forecasts of the persistent component of inflation. However, dur-
ing this decade, the argument that the natural rate of unemployment and productivity
growth changed significantly would imply with equal force that the real rate would have
changed as well. Moreover, the literature on the term structure of interest rates almost al-
ways finds that risk premia moves around considerably (Ang, Bekaert and Wei, 2007). An
alternative used by Friedman and Schwartz (1963) is the equity premium, but again to-
day it is understod that it is more likely to reflect changes in risk premia than in expected
inflation (Campbell, Sunderam and Viceira, 2017).
A final dead end is the USD price of gold that was traded freely in the London Mercan-
tile Exchange. The Fed and other foreign central banks formed a “Gold Pool,” combining
their stocks of gold to intervene in that market to keep the price of an ounce of gold close
to its peg of $35. The pool failed in March of 1968, the London market temporarily closed,
and even after it re-opened, there were US capital controls and regulations of the London
market. Perhaps as a result, there was no trading of organized gold futures in the London
market until 1975.
Across the channel, in Zurich, there had been a just-as-active spot gold market since
the end of World War II. There were no capital controls in Switzerland, no legal restric-
tions on writing gold contracts, and no limits to holding, exporting or importing gold. In
the immediate post-war, Zurich was the major gold trading centre, only losing to London
later, as the market in London opened in 1954. The market boomed in activity during the
gold pool in the London market, and there was trading in multiple currencies. American
residents and corporations became large investors in the late 1960s. Figure 9 plots the spot
USD price of a gold ounce in the London and Zurich markets during these times. They are
very close confirming that the Zurich market price signals were arguably as informative
as those from the large London market that are commonly used by researchers.
More important, for the purposes of this paper, is that throughout the 1960s, 6-month
and 12-month gold future contracts were traded in Zurich payable in USD. They arose
following the London “gold bubble” of 1960. During the Bretton Woods system, the loss
of an inflation anchor had to be tied to leaving convertibility with gold. Therefore, the
USD prices of future contracts of gold delivery would provide an indicator of expected
22
Figure 9: Gold prices, spot, per ounce
Zurich price
London price
USD to gold ounce convertability
42
40
38
36
1966 1967 1968 1969 1970 1971
inflation. Picks (1971 edition) reports representative annual premia in these contracts
throughout the 1960s. The implicit expected devaluation of the USD relative to gold can
be taken as a proxy for market expected inflation. Between 1966 and 1971, this implicit
market expected inflation was: 1.4%, 1.8%, 3%, 4.4%, 3.1%, and 4.1%. That is, again there
was a large surge in 1968-69 that persisted until 1971.
3.6 Conclusion
Looking only at the expectations from the Federal Reserve staff and the cross-sectional
mean of professional forecasts or households, the inflation expectations anchor seemed to
be firm until 1971, with the ever-rising inflation being treated as a sequence of unfortunate
temporary deviations. Only after the end of Bretton Woods is there a sustained drift
upwards.
This section noted that looking at other, arguably richer, data paints a different picture.
Already between 1967 and 1970, the anchor went adrift as revealed by market prices for
gold futures, public concerns and media reports and, perhaps more convincingly, by the
shift in the cross-sectional distribution of household expectations.
After mid-1970, household expected inflation fell even as professionals and markets
were rising, perhaps because of the wage and price controls. This effect was temporary,
and by 1973 the inflation expectations anchor was definitely lost.
23
4 A measure of fundamental expected inflation
In almost every dynamic macroeconomic model, both actual inflation and inflation expec-
tations are endogenous objects that are closely related. Under the extreme assumption of
perfect foresight, they coincide, while with full-information rational expectations, the dif-
ference between the two is serially uncorrelated and has a mean of zero. Therefore, a
traditional approach to measure the expected is to measure the actual inflation. Then,
given a goal of foreseeing future inflation movements, since expected inflation is their
anchor, economists have focussed on current inflation as a predictor, supplemented with
its lags. But the Lucas critique dictates that the relation between expected inflation and
lags of inflation can change, and that is is unreliable especially when inflation is about to
drift away.
This paper has taken a different approach: it has measured inflation expectations di-
rectly, and investigated whether they provide signals about future inflation. It found that
it does, but only if one uses the whole cross-sectional distribution of survey expectations
from both households and professionals complemented with market-price expectations.
It would be useful to combine these into a single estimate of expected inflation. This re-
quires having a model of expectations, yet there is a myriad of them, each with its own
virtues and flaws. However, almost all of them share one feature: expectations ultimately
converge to the rational expectations fundamental. Whether agents use past data to learn,
have incomplete signals, limited attention or heuristic biases, their expectations will differ
from rational expectations but gradually converge to them.9 If an economic fundamental,
like a policy regime or a shock, makes the rational expectation change, then in almost all
modern models expected inflation will change in the same direction as well. The rational
expectation serves as a fundamental for expected inflation.
4.1 A model as a measurement tool
This section uses a parsimonious model following Reis (2021) to combine the different
expectations data and estimate this fundamental. The goal is not to test theories or distin-
guish between models of expectations, but to combine their key ingredients to extract as
much signal as possible from the data. The model has three equations for three measures
9 Most models of expectations eventually converge to the rational expectations equilibrium, much like
most business-cycle models eventually converge to the Ramsey-Solow steady state.
24
of expected inflation:
vth = πt∗ + cth + θt (eth + πte − πt∗ ) with cth ∼ E(λt ), eth |πte ∼ N (0, σt2 ) (1)
yt (πte ) gt ( Ft−1 (ωt )) f t ( Ft−1 (ωt ))dπte
R
qt = with: ωt ∼ B( β) (2)
gt ( Ft−1 (ω )) f t ( Ft−1 (ωt ))dπte
R
Etb = Et (πt |vmedian
t , qt ) (3)
The fundamental rational expectation of inflation is πte , while each equation reports how
the expectation of households vth , market prices qt , and professionals Etb relates to it.
The first equation implies that the expectations of inflation vh reported by a house-
hold h in a survey follows an exponentially-modified Gaussian (EMG) distribution Ft (.),
which is the sum of three parts: The first part is just a prior mean πt∗ for what fundamental
expected inflation πte might be. The second is a bias cth drawn from an exponential dis-
tribution. It is important to match the fact that survey expected inflation is persistently
above actual inflation, perhaps as a long scar from the high inflation of the 1970s and
80s. The third is an individual-specific signal of the fundamental, with noise eth , drawn
from a normal distribution. The relevant parameters are: the reaction of expectations to
signals θt , the informativeness of the signal affecting the dispersion of expectations σt2 ,
and the average bias 1/λt . Reis (2021) shows that this model captures the main proper-
ties from the large literature on imperfect information, over-reaction, experience learning,
and sticky information.
The second equation states that the market price expected inflation qt is equal to the
expected risk-adjusted inflation with risk-adjustment yt (πte ). That expectation is formed
by traders that start from a prior which is the EMG Ft (.) (traders are people after all), but
which is updated following Bayes rule after observing the market price itself, as written
out in the second equation. The price is contaminated by a noise ω drawn from a symmet-
ric Beta distribution with parameter β. The density gt (.) is the endogenous distribution
that results from markets having to clear given the noise and fundamental. Reis (2021)
shows that this model proxies for many models of imperfect information in financial mar-
kets that are in the spirit of Grossman and Stiglitz.
The third equation states that the professionals in the surveys are similar to traders of
inflation risk. They use Bayes rule to combine the information that households have with
market prices, but what they report in a survey is neither risk-adjusted nor identifies who
is the marginal player in the markets. Rather, from the small samples in these surveys we
25
can only see the objective beliefs of the median trader.
4.2 The information in different data, and the model to filter it
Individually, each different source of expectations data carries useful information. The
household expectations are sluggish, less informed and subject to biases, but we have
more cross-sectional data with which to estimate the three moments of their cross-sectional
distribution and to separate the fundamental from bias and noise. The market price is er-
ratic and polluted by financial noise, but it is fast to react to fundamentals. Surveys of
professionals or traders have small samples, so only a first cross-sectional moment can be
calculated. But, combined with market prices, they can separate the fundamental from
the market noise. While each series has virtues and flaws, it is by combining all of these
pieces of data that the model is able to uncover the fundamental expected inflation.
With observations of the first three moments of the cross-sectional distribution of sur-
vey inflation expectations, together with a market price for expected inflation, and a first
moment from a survey of professionals or traders, one has 5 numbers at each date. With
these, and the three equations above, one can back out the reaction, dispersion, and bias
coefficients θt , σt2 , λt , as well as financial market noise ωt and fundamental expected infla-
tion πte . The model gives a way to filter these five pieces of information and recover the
anchor for expectations that underlies all of them.
If the market prices’ expected inflation rises, but there is no change in survey expec-
tations, the model will attribute most of it to financial market noise and not change the
estimate of fundamental expected inflation. At the other extreme, if the cross-sectional
mean of survey expectations of households and professionals both moved in one direc-
tion, the model would update its πte in the same direction. In between, and more relevant,
if the cross-sectional household survey mean barely moves, but the cross-sectional survey
standard deviation and skewness of households moved, together with the market survey
and price-implied expectations in a consistent direction, then the model infers that the
fundamental has changed. Some people have started updating, and if they have done so
in the same direction as traders, then the fundamental must have changed.
To take the model to the 1966-71 annual sample I use the three moments from the
Michigan quantitative survey, the median expectation from the subset of professionals in
the Livingston survey that work in the finance industry, and the annual expected inflation
implied in Zurich gold futures. From 1972 onwards, there is no clear market price to use
to implement the model. Note that all inputs are series of expectations; nowhere does
26
actual inflation enter the calculations.10
For the prior inflation πt∗ , I use the previous year’s fundamental inflation estimate
πte−1 . This way, the econometrician’s prior is to find no change in the fundamenal. Finally,
the parameter β measures how noisy financial prices are. Following Reis (2021), I set it
equal to 2, which allows for a significant amount of noise in financial markets.
There are three important limitations in using this model with the 1967-73 data. First,
the data on market prices comes from the Zurich gold market, but the data on surveyed
market participants comes from professional economists in the United States. These are
not as close to each other as would be desirable. Second, there are no data with which to
calibrate changes in the inflation risk premium, so I assume it is constant (and since the
model only identifies changes I set it to zero: y(π e ) = π e ). Third, as discussed in Reis
(2021), this is a model of long-horizon expectations. For short-horizon expectations, the
different structural models of expectations in the literature are too diverse to be captured
in a parsimonious way. Yet, there are no long-horizon expectations data. So, I treat the
data on one-year expectations as proxies for longer-horizon expectations.11
4.3 The estimates of fundamental expected inflation
Figure 10 plots the estimates of the model. Panel (a) has the fundamental expected in-
flation. It rose from between 1967 and 1970 by 2.2 percentage points. During this time,
actual inflation rose by 2.7 percentage points.12 The bond yields or the Greenbook fore-
casts that policymakers looked at during that time would have given a false sense of
stability. Instead, the rising market prices combined with the large increase in standard
deviation and fall in skewness across households reveal that fundamental expected infla-
tion was changing significantly. The estimates also confirm the temporary respite from
this increase in 1971.
To understand what is going on, panel (b) plots the difference between the market
10 Eusepi et al. (2019) provide an alternative measure of the inflation expectations anchor using actual
inflation data.
11 In the last decade, there are data surveying the traders in financial markets, and asking households
about both 1-year and longer-horizon expectations. The correlation coefficient between the quarterly me-
dian answer to the FRBNY survey of dealers and the median in the SPF is 0.69; the correlation between the
cross-sectional standard deviation and skewness of the answer to long-horizon inflation expectations and
1-year ahead in the Michigan survey are 0.91 and 0.41, respectively.
12 The initial level of expected inflation is not identified. It is pinned down by the choice of prior π ∗ in
the initial period, so the appropriate comparison between actual and expected is their predicted changes.
In the figure I set initial expectation in 1967 to 2.0%, even though actual inflation was 3.2%.
27
Figure 10: Fundamental expected inflation
(a) Estimates (b) The market-people discrepancy
4.5 2
4
1.5
3.5
1
3
% % 0.5
2.5
0
2
-0.5
1.5
1 -1
1967 1968 1969 1970 1971 1968 1969 1970 1971
Year
(c) Standard deviation of signals σt (d) Responsiveness to signals θt
6
3.1
3
5
2.9
2.8 4
2.7
% 3
2.6
2.5 2
2.4
1
2.3
2.2
1968 1969 1970 1971 0
1968 1969 1970 1971
Year
Year
expectation and the household expectation, and decomposes it into the difference within
traders (marginal versus median) and across people (professionals versus households).
At first, market expected inflation rose, while surveys of professionals showed little change.
The model interprets this disagreement within as partly due to a financial shock. How-
ever, because of the increase in variance and fall in skewness, the model puts a significant
weight on this instead being a result of expected inflation rising. Therefore, the estimate of
expected inflation is updated upwards. The following year, this suspicion is confirmed, as
the median expected inflation rises across surveys converging to the market expectation.
In 1970, the rise in survey expectations combined with the movements in disagreement
confirm that the fundamental has risen, so the model updates its estimate further up, even
though market prices have fallen.
Panels (c) and (d) show the time series of two important model parameters, σt , which
28
measures the dispersion of the idiosyncratic signals, and θt ,which measures by how much
an individual will respond to his/her individual signal. If people were pure Bayesians,
then θt would be proportional to the inverse of σt , but the model allowed this to not be
the case following theories of over-confidence, over-reaction, and diagnostic expectations.
The estimates strongly supports this. Throughout the period, the estimated standard
deviation rises. Agents are confused and disagree on where inflation is heading. But
at first they are also more responsive to their individual signals, as if they were paying
more attention to the inflation path. Only after they return to their usual responsiveness.
Finally, 1971 continues to stand out as a time where there was a reversal in these drifts.
5 Measuring the expected inflation anchor in other episodes
Measures of expected inflation are sufficiently noisy that using them to forecast inflation
during tranquil times will likely bring little benefit. The previous two section showed
that, when there was a regime change in the US, and inflation persistently drifted, then the
expectations data were useful to diagnose this shift. This section investigates whether the
same is true in other episodes of regime change, when inflation moved significantly and
persistently. Each episode provides a different test of our ability to measure the expected
inflation anchor.
Searching through the history of inflation over the last 15 years, there were only a
handful of countries that had both good survey and market-price expectations of infla-
tion, as well as inflation above their central banks’ targets for a few years. Three episodes
stood out. Brazil between 2011 and 2016 provides the clearest counterpart to the US in-
flation of the 1970s. Turkey since 2018 has also had a drifting inflation, but because it is
more recent, the jury is still out in terms of how persistent this will be. Looking at the
expectations data today in 2021 gives an example of how to try to measure the anchor in
almost real time. Finally, South Africa between 2010 and 2016 provides a placebo, or a
possible example of whether the expectations data can deliver false positives. While in-
flation was high during those years, this turned out to be temporary, and the rate of price
changes was back on target afterwards.
After these three cases in other countries, this section returns to US history to look at
what the expectations data showed in the early 1980s, when inflation came down sharply.
Finally, it concludes with a real time look at expectation data after the 2020 pandemic.
29
5.1 Brazil, 2011-16
Introduction. After two decades of struggling against high, and sometimes hyper, infla-
tion, Brazil managed to reform its monetary regime in 1994 and to bring inflation down
(Ayres et al., 2019). In 1999, the Banco Central do Brasil (BCB) adopted an inflation tar-
geting regime with a target for the CPI and wide upper and lower bounds that are pe-
riodically updated. Between 2002 and 2003 the regime came under a trial by fire, as the
combination of a domestic drought, contagion from the financial crisis in Argentina, and
a contested presidential election cycle led to a sharp increase in sovereign spreads, a de-
preciation of the currency, and inflation spiking to 17% in 2003, well outside the target
range. Aggressively tight monetary policy followed, and inflation came down sharply
back to target.
Actual inflation. In August of 2010, the inflation rate (year-on-year changes in the IPCA
index) was 4.5%, exactly its target. The BCB had just gone through a tightening cycle,
but inflation was modestly rising, perhaps due to large public deficits and increases in
oil and commodity prices. A new president came into power in January 1st, 2011 and
in the August policy meeting, according to press accounts, the BCB unexpectedly and
significantly eased policy. The main policy interest rate (Selic) was reduced from 12.5%
to 7.25% continuously over the next 14 months, this in spite of GDP in 2010 having risen
by a remarkable 7.5%, and inflation in July of 2011 having been 6.9%, above the upper
bound of the target of 6.5%.
Panel (a) of Figure 11 shows the evolution of Brazilian inflation. Between 2011 and
the end of 2015, inflation was always close to the upper bound, peaking in June of 2015
at 11.4%. Public deficits were large and growing, and the government resorted to hidden
fiscal expansions by persuading the main oil producer to sell domestically below market
prices, and having the State-owned development bank (BNDES) and mortgage lender
(Caixa) run persistently large losses that amounted to a quasi-fiscal deficit. Monetary
policy was timid in its response to inflation rising, only starting to gradually raise the
interest rate after May of 2013. This led to concerns about fiscal dominance of inflation
and lack of independence of the BCB, especially as the central bank reports and policy
minutes are persistently optimistic about inflation, fiscal consolidation, and the public
credit subsidies. Moreover, the government kept the administered prices of gasoline and
diesel very low between 2011 and 2014, and cut energy tariffs and sales taxes for some
commodities, all directly lowering measured consumer prices. This created a large gap
30
between prices that were set in markets, and those that were set by the government, and
contributed to rising public deficits (Bonomo, 2018).
After the re-election of president Dilma Roussaef in 2014, administered prices were
adjusted back in line, which immediately led to a very large jump of overall inflation
by more than 3 percentage points. Accumulating the change in the price level between
August of 2010 and August of 2016, the annualized rate of inflation was 7.3%, well above
the 6.5% target. After a controversial impeachment process, Roussaef left office at the end
of August of 2016. The new government replaced the members of the monetary policy
committee, monetary and fiscal policy were tightened, and by March of 2017 inflation
was back at the 4.5% target. Some good luck followed in the form of strong agricultural
production lowering food prices, and inflation was always below 3% between mid 2017
and mid 2018.
Expected inflation. Panel (b) in figure 11 shows series for inflation expectations in Brazil
during this time. The professional forecasters series corresponds to the median answer
in the Focus survey ran by the BCB. It is an unusual survey because it includes as many
as 140 respondents drawn from financial market participants, and it provides incentives
by giving out a prize every year to the most accurate forecaster, which is then widely
reported in the financial media. The household expectations correspond to the survey
of consumer sentiment that includes approximately 2000 consumers and asks a precise
numerical question about inflation over the next 12 months. The series corresponds to
the mean over the sample that excludes the top and bottom quartile of answers, and
is reported monthly by FGV-IBRE. Finally, the market price expected inflation refers to
the breakeven rate—the difference between the rates in inflation-indexed and nominal
bonds—over the next year, corrected for seasonal factors and indexation lags.13
Both the survey of professionals and the market prices track actual inflation quite
well. This is perhaps not too surprising, given the strong incentives to do so. Insofar as
inflation was being kept artificially low through the administrative prices though, this
is less informative about the anchor. The first moment of the household survey instead
starts clearly drifting up from the start of 2013 onwards. It strongly suggests that the
anchor was drifting away.
Panel (c) shows the second and third cross-sectional moments from the household
13 For more on the Focus survey, see Alves, Areosa and Carvalho (2020), on the FGV-IBRE series see
Gaglianone, Issler and Matos (2017), and on the breakeven series see Val and Araujo (2019).
31
survey.14 This shows a clear increase in both series since the middle of 2011. The starting
positive skew in the Brazilian household expectations is much lower than its counterpart
in the United States. This is consistent with a smaller inflation bias and higher rate of
information updates by Brazilian citizens given the country’s history.15 But during this
period, the loss of the anchor appears, as more households move to the right of the distri-
bution. This both raises the variance and increases the positive skew.
Panel (d) confirms this by plotting the distributions at the start of the year in 2011, the
middle of 2014 and the start of the year in 2016. The increase in variance and skewness
clearly arises from a subset of households breaking off to the right of the median. They
expected higher and higher inflation as months went by.
Figure 11: Brazil’s drifting expected inflation anchor: 2011-16
(a) Actual inflation and its target (b) Markets and survey first-order moments
18.0% Consumer Price Inflation Market−price implied
Consumer price inflation − admin prices Survey of Professionals
Consumer price inflation − free prices 10.0% Survey of Households
Inflation Target Inflation Target
Upper Bound Upper Bound
14.0%
6.5%
10.0%
4.5%
6.5%
4.5% 2.5%
2.5%
0.0% 0.0%
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
(c) Cross-sectional disagreement of households (d) Cross-sectional distribution of households
8.0%
Cross−sectional survey skewness (lhs) 30
Cross−sectional survey standard deviation (rhs)
2011
2014
6 2016
6.0%
20
4
4.0%
10
2
2.0%
0 0.0% 0
2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 0% 5% 10% 15% 20% 25%
14 I calculate this by using the micro-data directly, and excluding only the top and bottom 1% of answers
to deal with outliers. Results cutting the top and bottom 5% are very similar. Also, the mean calculated this
way is not too different from the one calculated with the official procedure of excluding the bottom and top
25% (which would be too aggressive to calculate higher-order moments).
15 Fitting the EMG distribution from section 4, the λ parameter is much higher for Brazil than for the US.
32
Lessons. The Brazilian experience in 2011-16 adds to the lessons from the US experi-
ence of 1967-74. A similar combination of timid monetary policy, fiscal dominance, and
persistent belief that inflation is only temporarily high were the harbingers of rising infla-
tion. Moreover, the government’s direct interference on the CPI via administrative prices
lowered professional and market-price expectations, but only appeared to have delayed
the movements in household expectations. Like in the US, such policies seemed to be
effective only temporarily.
As in the US, measures of disagreement in the cross-sectional survey provide a good
image of the loss of the inflation anchor, but one should not expect them to be captured
by a naive time-series correlation between standard deviation and skewness. In Brazil
they moved together, while in the US, initially skewness and variance moved in opposite
directions. The reason is revealed by the cross-sectional distribution plots. The initial
drift of the anchor in the US came with a thinning out of the left tail of the distribution:
those that expected very low inflation joined the median. Instead, in Brazil, the median
moved to the right and the right tail thickened. A naive correlation of these two measures
of disagreement will not be particularly informative.
5.2 Turkey: 2018-...
Introduction. Turkey had a brush with very high inflation during the 1990s associated
with political instability and the monetary financing of deficits. After an IMF program
and a law giving independence to the Central Bank of the Republic of Turkey (CBRT)
in 2001, a sharp disinflation program lowered it from 49% to 8% in 4 years with little
output costs. In 2006, an inflation targeting regime was adopted, with a target around
5% although with some annual changes. Actual inflation was always above target, but
steadily so, averaging 8% between 2006 and 2017.
Actual inflation. In the three years between the start of 2018 and the end of 2020, inflation
shot up to an average of 15%, three times the inflation target (Kara, 2021). The precipitat-
ing event behind the rise in inflation seems to have been the re-election of Recep Erdogan
in June of 2018. This was a period of political instability, following a failed coup in July
of 2016, a constitutional referendum in April of 2017, and the premature election that
should have taken place only in November of 2019. As the president consolidated his
power, but the economy was faltering, he started commenting on inflation and interfer-
ing with the CBRT’s independence. In May of 2018, in a campaign speech in London he
33
expressed desire to take greater control of the economy, and defended lower interest rates
in order to control inflation. In June, he announced he would investigate Moody’s after
a downgrade of the country’s debt, and the lira depreciated strongly against the USD. In
September, Erdogan repeated his theory that lower interest rates are desirable and that
inflation was not a monetary phenomenom, sharply criticizing the CBRT.
On the fiscal side, defense spending increased significantly since 2015 associated with
the wars in Syria and Iraq. The early elections of 2018 came with several new spending
measures in 2017-18: 750,000 contract workers in the public sector became permanent em-
ployees, a minimum wage subsidy estimated to cost 0.2% of GDP was extended together
with a monthly subsidy for new hires, and a bonus was offered to all pensioners at two
religious holidays costing 0.7% of GDP (IMF, 2018, OECD, 2018). As for monetary policy,
right after the election in July of 2018, the CBRT surprisingly did not raise interest rates, as
had been widely expected since inflation was at a 14-year high. However, in September,
it sharply increased them from 17.75% to 24%, right after the criticisms from Erdogan.
Expected inflation. Panel (a) of Figure 12 also shows monthly inflation expectations
from t to one year ahead. The market-price series is a measure of breakeven inflation
rates made available by the CBRT. Throughout, they have been very significantly under-
predicting inflation.
The second series is the median from a monthly survey of professionals that is dom-
inated by financial-market participants who are asked to put their expected inflation in
bins. This is also collected by the CBRT, starting in 2001, and has 60-70 respondents.
The figure also shows the median response from the Business Tendency Survey, also con-
ducted by the CBRT since 1987 and that every month surveys approximately 500 firms on
business confidence, including a question about PPI inflation.
All three measures of expected inflation move closely together. All of them missed the
2018 burst in inflation back in 2017; all of them expected some of that inflation to revert
back; and all of them expected some of it to persist into the future
The micro data from the survey of firms is not available. An imperfect proxy is the
survey of professionals. This is imperfect for three reasons. First, because these are pro-
fessionals, and at least for the US, their expectations tends to be closer to a normal and
less informative beyond the cross-sectional mean. Second, because the number of respon-
dents is not large enough to accurately calculate a third-order cross-sectional moment.
And third, because the survey’s top bin is 20% or more, which given the rise in inflation
will truncate the right tail.
34
Figure 12: Turkey’s drifting expected inflation anchor: 2018-...
(a) Actual inflation, markets and survey first-
(b) Cross-sectional survey distribution
order moments
30%
2017
2019
Actual inflation 20% 2020
Cross−sectional median of professionals... survey
25% Market−price implied expectation
Weighted mean of firms... survey
Target
15%
20%
15%
10%
10%
5%
5%
0%
0% 2% 4% 6% 7% 9% 10%12%13%15%16%18%19% 2% 4% 6% 7% 9% 10%12%13%15%16%18%19% 2% 4% 6% 7% 9% 10%12%13%15%16%18%19%
2015 2016 2017 2018 2019 2020 2021
Subject to all these caveats, already by the end of 2017, the standard deviation almost
quadrupled, while the skewness went from being negative at -1% to positive at 0.25%.
Panel (b) of Figure 12 shows the distributions in December of 2017, January of 2019 and
June of 2021. In 2017, the uncertainty is evident, with a bimodal distribution and more
than half of the respondents expecting inflation to exceed 17%. The events of 2018 re-
moved some of the disagreement by consolidating a view that inflation would be well
above the target. By 2021, more mass has moved rightwards, and the inflation anchor
seems definitely lost.
Lessons. The Turkish experience of a lost anchor leads to two additional lessons. First,
even close to real time, and when inflation is bouncing up and down, like it did in Turkey
in 2018 and 2019, the expectations data can paint a clear picture of a lost inflation anchor.
Second, in countries where arguably the anchor was not firm in the seabed to start with,
the shifts in the cross-sectional distribution can be large and fast. The loss of the inflation
anchor can come fast and need not be gradually building up like it did in the US in the
late 1960s.
5.3 South Africa: 2010-2016
Introduction. The South African Reserve Bank (SARB) adopted inflation targeting in
2000, with a target range of CPI inflation between 3% an 6% and no stated midpoint.
The first few years of the new regime were rocky, with oscillations in the exchange rate
and reversals of policy, but after 2005, transparency increased, so that after one decade
of inflation targeting, outcomes were on average solidly within the range. The global
35
financial crisis of 2008-10 brought a recession, affecting the country mainly through its
international trade linkages to Europe.
Actual inflation. Between 2010 and 2016, the SARB had to face an unusual, difficult
sequence of supply shocks. Inflation averaged only slightly less than the 6% upper bound
of its target as a result, as shown in the top panel of Figure 13. First, the rise in global oil
and commodity prices in 2010-12 pushed inflation towards the upper bound of the range.
At the same time, the recession in Europe due to its sovereign debt crisis implied that the
South African economy was slow to recover. The SARB kept interest rates steady so as to
not prevent that recovery.
Second, as inflation was falling in the first half of 2012, with the economy recovering
and oil prices stabilizing, inflation was hit by a different shock throughout 2013 and 2014.
The producer of electricity ran perennial deficits and needed large investments to solve
the persistent outages. The increase in costs led the regulators to approve a large increase
in prices, well above the inflation targeting range. Headline inflation continued to exceed
core inflation, but monetary policy held steady for a while until finally raising the repo
interest rate in 2014, for the first time in 6 years. By 2015 these effects were over and infla-
tion fell sharply to the middle of the target range, leading the SARB to loosen monetary
policy.
A third shock arrived soon though. In 2016, an El Nino storm led to a large decline
in soil productivity together with a drought, which raised food prices that are a large
component of the consumption baskets. Again headline inflation moved to the upper
range of the target, even as core inflation was well within it, and again the SARB held
monetary policy steady. Finally, by 2017, no more shocks arrived, and since then inflation
has stayed near the midpoint of the range.16
Expected inflation The top panel of figure 13 shows the median expected inflation from
three separate samples: financial market analysts, non-financial firms, and trade unions.
The survey is conducted by the Bureau of Economic Research for the SARB on a South
African panel quarterly since 2000 (for Economic Research, 2021). All three behaved sim-
ilarly during this period, persistently forecasting inflation to be around 6%. On the one
hand, these forecasts by professionals proved to be reasonably accurate ex post and they
are remarkably stable in light of the volatility in actual inflation. On the other hand, they
16 For a description of the SARB’s reaction to the successive shocks, see Kabundi, Schaling and Some
(2015) or Mminele (2019).
36
Figure 13: South Africa’s unlucky run: 2010-16
(a) Actual inflation, markets and survey first-
(b) Cross-sectional survey distributions
order moments
2014
2015
2016
0.2
6.0%
4.5% Inflation (CPI Headline) 0.1
Inflation (CPI Core)
Analysts' forecast
Businesses' forecast
Trade unions' forecasts
Upper bound
Lower bound
3.0%
0.0
−1% 0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15%
2012 2013 2014 2015 2016 2017 2018 2019
missed the fall in inflation in 2015 and were slow to catch up to the lower inflation from
2017 onwards.
The bottom panel instead reproduces the cross-sectional distribution among house-
holds at three successive months of October between 2014 and 2016, calculated by Du
Plessis, Reid and Siklos (2021). These come from a remarkable survey of between 2000
and 2500 individuals in urban and rural environments at a quarterly frequency, con-
ducted by AC Nielsen under contract with the SARB and the BER. As inflation moves
up and down, the distributions shift right and left. However, note that disagreement,
measured by either second or third moments, does not change much.
Lessons. The survey data throughout the 2011-16 period seemed consistent with a stable
anchor. Shocks hit the economy, the central bank justifiably kept monetary policy steady
letting inflation rise, and both outcomes and expectations reflected this with higher ex-
pected inflation. Yet, there was no permanent rise of either actual or expected inflation,
unlike we saw in the US, Brazil, or Turkey. Disagreement did not increase during this
period. Contrary to the US in the 1970s, inflation did not drift up as the expected inflation
anchor remained steady. In the South Africa case, the price controls worked in the op-
posite direction of what they did in the US and Brazil, yet their effects were qualitatively
similar: significant but temporary.
5.4 The United States in the 1980s: dropping the anchor
Between 1979 and 1983, under chair Paul Volcker, the Fed adopted highly restrictive mon-
etary policy and inflation fell significantly. If the main episode studied in this paper is the
37
Figure 14: Dropping the anchor: the US 1980s
(a) Actual and survey first-order moments (b) Survey disagreement
15.0% 4.0 20%
Standard deviation (right axis)
Skewness (left axis)
3.5
10.0% 3.0 15%
2.5
SPF median 2.0 10%
5.0%
Michigan median
Inflation (CPI)
Inflation (Core CPI)
1.5
1978 1979 1980 1981 1982 1983 1984 1985 1979 1980 1981 1982 1983 1984 1985
loss of an inflation anchor, this episode is the dropping a new anchor, which persists in
place until today. It adds a reversal situation and again tries to measure the anchor.
Figure 14 shows that the survey of households, which now corresponds to the stan-
dard Michigan quantitative series, was quick to catch on. The decline was swift, keep-
ing up with inflation. Professional forecasters were slower (or perhaps more skeptical).
Digging deeper into the household cross-sectional distribution, the figure shows that as
some households started expecting lower inflation, this increased the standard deviation,
while it lowered the positive skew. As the remaining households gradually caught up,
the median fell, the standard deviation started declining after reaching a peak, and the
skew started rising. Altogether, this behavior is consistent with the model described in
section 4, where people that are inattentive, overconfident, and use sticky information in
updating their biases would react in this gradual way to a change in policy regime. As
in section 3, it shows that looking at the distribution of expectations, from first to third
moment, can provide some signals of where the expected inflation anchor is and where it
is going.
5.5 United States 2020-21: Where is it heading?
The final application of the ideas in this paper is more tentative. The pandemic recession
of 2020 and the swift recovery in 2020-21 interrupted three decades where the expected
inflation anchor was steadily at 2% and actual inflation had only small transitory move-
ments near its anchor. As panel (a) in Figure 15 shows, inflation fell sharply with the
lockdown in the first half of the year, and rebounded very strongly in the first half of 2021
reaching levels that had not been seen for decades. Many ask today the question that this
38
Figure 15: The expected inflation anchor through the pandemic
(a) Actual inflation (b) Markets and survey first-order moments
6% CPI Urban Core
CPI Urban All Goods
Trimmed Mean PCE 5%
5% Inflation Target
People: Michigan 1 Year
4% People: Michigan 5 Years
4% Traders: SPF 1 Year
Market: 10 year Breakeven Inflation Rate
3%
2% 3%
1%
2%
0%
1%
2018 2019 2020 2021 2018 2019 2020 2021
(c) Cross-sectional disagreement of households (d) Cross-sectional distribution of households
0.15
5
Standard deviation (right axis)
Skewness (left axis)
2020 January
1.0 4 2020 September
0.10 2021 June
3
0.05
0.5 2
1
0.00
2018 2019 2020 2021 0.0 1.5 3.5 5.0 7.5 12.0
paper asked at the start: Where is inflation heading?
The economic disruptions associated with the pandemic have many peculiar features
that make this a hard question. Still, we can look at the expectations data to see if the same
patterns are present as in the other episodes of a lost anchor. Panel (b) in Figure 15 shows
that professional forecasters have raised their expectation only slightly. If they were the
single measure of the anchor, one would confidently conclude that inflation will soon
quickly revert back to this anchor. Market-price expected inflation instead has increased
significantly in 2021, albeit only after having fallen significantly in 2019 and 2020. These
sharp movements in market expectations might shake that confidence, but they could be
dismissed as another illustration of the market noise ω in the model.
The household survey is more worrying. Both the long-horizon expectations and the
one-year ahead have jumped up in just 6 months faster than almost ever before in the
post-1980s sample. Given its usual sluggish slow adjustments, this unusual quick jump is
hard to interpret. Candia, Coibion and Gorodnichenko (2021) report a very high discrep-
39
ancy between the expectations of households and professionals across different surveys
and for other countries as well.
Again, looking at higher order moments and, more generally, at the whole distribution
provides some guidance. Between the start of 2020 and the second half of the year, mass
moved away from the peak of the bell-shaped distribution and towards the right tail.
That is, some households started believing that inflation might be very high. As a result,
both the standard deviation and the skewness increased rapidly. In the first half of 2021,
instead it is the left tail that has been hollowed out. Few US households believe today
that inflation will be below target in the near future, and just as many expect that an
inflation disaster could happen. As Knotek et al. (2020) document, households have been
particularly attentive and responsive to news about inflation during this time period.
Perhaps, like in the early 1980s, households are ahead of the curve in detecting a
change in regime. Or, perhaps through the overconfidence discussed in the model, a
few price changes in key goods have had an over-influence on overall inflation expecta-
tions, much like the wage and price controls in the US 1970s, the administered prices in
Brazil 2010s, and the food prices and electricity prices in South Africa in 2012. If so, then
those past experiences suggest that the rise in expectations will be transitory and brief.
One lesson from the 1967-73 experience is that policymakers should keep a close eye on
measures of the expected inflation anchor, and not give in to the temptation to dismiss
them as temporary noise or as vague psychological factors.
6 Conclusion
Every respectable central banker talks endlessly about credibility, steering expectations,
and how well anchored inflation is. Yet, the experience of the US Great Inflation illus-
trates that expectations are sometimes treated as mystical objects that the policymaker
infers from his/her wisdom and personal experiences. Economists as well often refer to
inflation expectations as an add-on factor to complement their model of fundamentals for
why inflation rose or fell. Theorists assume that expectations are strictly rational, and so
need not be measured, while empiricists despair at how first-order moments for surveys
have persistent bias and move so sluggishly.
This paper argued that we can measure inflation expectations in multiple ways. Look-
ing deeper at households surveys, we can examine their distributions, and especially the
disagreement in them as measured by second and third order moments. Combining sur-
40
veys of professionals with market prices provides valuable information to separate noise
from signal. The wisdom of the crowd gives hints of what lies ahead. A parsimonious
model can be used as a measurement tool to combine these hints into estimates.
During the US Great Inflation, the expectations data shows a drifting anchor already
between 1967 and 1970, well before the end of Bretton Woods or the oil price shocks.
This paper adds a new element to the bad measurement class of explanations of the Great
Inflation: the bad measurement of inflation expectations. Not only did policymakers
had incorrect estimates of productivity growth, the natural rate of unemployment, or the
natural rate of interest, but they also failed to accurately measure expectations of inflation.
These mistakes reinforced each other in leading them to miss the drift of the expected
inflation and hence not preventing its loss. To the bad theories explanations, it adds the
observation that policymakers had no theory of expectations, appealing instead to vague
animal spirits that led them to over-rely on transitory explanations for persistently rising
inflation, and to believing that wage and price controls sufficed. Finally, to the bad luck
stories, it adds a new channel to the role of oil price shocks. The rise in the salient price of
gas made usually-sluggish inflation expectations jump quickly, and by a lot.
Confirming the usefulness of using inflation expectations data, the paper showed that:
(i) they also detected the loss of the inflation anchor in Brazil in 2011-16; (ii) even if Turkey
has not had a very firm anchor, and events are still too recent to be sure, the expectations
data points to a significant change in 2018, (iii) the expectations data do not give a false
positive in South Africa 2010-16, instead showing a steady expectations anchor in spite of
several adverse shocks to actual inflation, and (iv) they also move consistently when an
anchor is being dropped on the seabed as in the US 1979-84. Finally, looking at current
events, the data suggests that the anchor may have drifted in 2021, but it is still early
enough that good policy or good luck in the near future can keep it in place.
Looking forward, there is much work to do in order to provide better measurements
of the expected inflation anchor. The 1960s provide a warning that, even if a central bank
cares about inflation expectations, trying to steer them with communications without a
clear model or measurement of these expectations is a fruitless task. Ultimately, inflation
expectations are not an exogenous force, but a mirror of the monetary and fiscal policy
regimes being followed. Those policies are the fundamentals that anchor the expectations
anchor itself.
41
References
Alesina, Alberto. 1988. “Macroeconomics and Politics.” In NBER Macroeconomics Annual 1988,
Volume 3. 13–62. National Bureau of Economic Research, Inc.
Alves, Sergio, Waldyr Areosa, and Carlos Carvalho. 2020. “Beating the “pros” ith a semi-
structural model of their own inflation forecasts.” PUC-Rio manuscript.
Andrade, Philippe, Olivier Coibion, Erwan Gautier, and Yuriy Gorodnichenko. 2021. “No firm
is an island? How industry conditions shape firms’ expectations.” Journal of Monetary Economics.
Ang, Andrew, Geert Bekaert, and Min Wei. 2007. “Do macro variables, asset markets, or surveys
forecast inflation better?” Journal of Monetary Economics, 54(4): 1163–1212.
Angeletos, George-Marios, Zhen Huo, and Karthik Sastry. 2020. “Imperfect Macroeconomic Ex-
pectations: Evidence and Theory.” In NBER Macroeconomics Annual 2020. Vol. 35. NBER.
Ayres, Joao, Marcio Garcia, Diogo A Guillén, and Patrick J Kehoe. 2019. “The Monetary and Fis-
cal History of Brazil, 1960-2016.” National Bureau of Economic Research Working Paper 25421.
Binder, Carola Conces. 2016. “Estimation of historical inflation expectations.” Explorations in Eco-
nomic History, 61: 1–31.
Blinder, Alan S. 1982. “The Anatomy of Double-Digit Inflation in the 1970s.” Inflation: Causes and
Effects, , ed. Robert E. Hall, 261–282. University of Chicago Press.
Blinder, Alan S., and Jeremy B. Rudd. 2013. “The Supply-Shock Explanation of the Great Stagfla-
tion Revisited.” The Great Inflation: The Rebirth of Modern Central Banking, , ed. Michael D. Bordo
and Athanasios Orphanides, 119–175. University of Chicago Press.
Blinder, Alan S., and William J. Newton. 1981. “The 1971-1974 controls program and the price
level : An econometric post-mortem.” Journal of Monetary Economics, 8(1): 1–23.
Bonomo, Marco. 2018. “The Brazilian Economy in 2018: Diagnosis of a Crisis.” Lemann Lecture,
University of Illinois Urbana Champaign.
Bordo, Michael, and Barry Eichengreen. 2013. “Bretton Woods and the Great Inflation.” The Great
Inflation: The Rebirth of Modern Central Banking, 449–489. University of Chicago Press.
Burns, Arthur F. 1957. Prosperity Without Inflation. New York:Fordham University Press.
Burns, Arthur F. 1987. “The Anguish of Central Banking.” In Money and the Economy: Central
Bankers’ Views. Palgrave Macmillan Books, , ed. Pierluigi Ciocca, Chapter 7, 147–166. Palgrave
Macmillan.
Campbell, John Y., Adi Sunderam, and Luis M. Viceira. 2017. “Inflation Bets or Deflation
Hedges? The Changing Risks of Nominal Bonds.” Critical Finance Review, 6(2): 263–301.
42
Candia, Bernardo, Olivier Coibion, and Yuriy Gorodnichenko. 2021. “The Inflation Expectations
of U.S. Firms: Evidence from a new survey.” National Bureau of Economic Research Working
Paper 28836.
Carroll, Christopher D. 2003. “Macroeconomic Expectations of Households and Professional
Forecasters.” Quarterly Journal of Economics, 118(1): 269–298.
Castillo-Martinez, Laura, and Ricardo Reis. 2019. “How do central banks control inflation? A
guide to the perplexed.” LSE manuscript.
Coibion, Olivier, and Yuriy Gorodnichenko. 2012. “What Can Survey Forecasts Tell Us about
Information Rigidities?” Journal of Political Economy, 120(1): 116–159.
Coibion, Olivier, Yuriy Gorodnichenko, and Rupal Kamdar. 2018. “The Formation of Expecta-
tions, Inflation, and the Phillips Curve.” Journal of Economic Literature, 56(4): 1447–1491.
de Leeuw, Frank, and Michael J. McKelvey. 1981. “Price Expectations of Business Firms.” Brook-
ings Papers on Economic Activity, 1981(1): 299–314.
DeLong, J. Bradford. 1997. “America’s Peacetime Inflation: The 1970s.” Reducing Inflation: Moti-
vation and Strategy, 247–280. University of Chicago Press.
Du Plessis, Stan, Monique Reid, and Pierre Siklos. 2021. “What Drives Household Inflation Ex-
pectations in South Africa? Demographics and Anchoring Under Inflation Targeting.” Economic
Systems, 45.
Eusepi, Stefano, Emanuel Moench, Bruce Preston, and Carlos Viana de Carvalho. 2019. “An-
chored Inflation Expectations.” C.E.P.R. Discussion Papers CEPR Discussion Papers 13900.
for Economic Research, Bureau. 2021. “The BER’s inflation expectations survey: know your data.”
BER Research Note 3.
Friedman, Milton. 1968. “The Role of Monetary Policy.” The American Economic Review, 58(1): 1–17.
Friedman, Milton, and Anna J. Schwartz. 1963. A Monetary History of the United States, 1867–1960.
Princwton University Perss.
Gaglianone, Wagner Piazza, João Victor Issler, and Silvia Maria Matos. 2017. “Applying
a microfounded-forecasting approach to predict Brazilian inflation.” Empirical Economics,
53(1): 137–163.
Gordon, Robert J. 1980. “Postwar Macroeconomics: The Evolution of Events and Ideas.” The
American Economy in Transition, , ed. Martin Feldstein, 101–182. University of Chicago Press.
Hetzel, Robert L. 1998. “Arthur Burns and inflation.” Economic Quarterly, , (Win): 21–44.
Hetzel, Robert L. 2020. “Handbook of the History of Money and Currency.” , ed. Yago K. Bat-
tilossi S., Cassis Y., Chapter The Evolution of US Monetary Policy. Springer, Singapure.
IMF. 2018. “Staff Report for the 2018 article IV consultation.” IMF Country Report No. 18/110.
43
Jalil, Andrew J., and Gisela Rua. 2016. “Inflation expectations and recovery in spring 1933.” Ex-
plorations in Economic History, 62: 26–50.
Juster, F. Thomas, Paul Wachtel, Saul Hymans, and James Duesenberry. 1972. “Inflation and the
Consumer.” Brookings Papers on Economic Activity, 1972(1): 71–121.
Juster, Thomas, and Robert Comment. 1980. “A note on the Measurement of Price Expectations.”
SRC Michigan manuscript.
Kabundi, Alain, Eric Schaling, and Modeste Some. 2015. “Monetary policy and heterogeneous
inflation expectations in South Africa.” Economic Modelling, 45: 109–117.
Kara, Hakan. 2021. “Turkey’s Ceaseless Battle with Price Stability.” LSE lecture delivered on
February of 2021.
Knotek, Edward S., Raphael Schoenle, Alexander Dietrich, Keith Kuester, Gernot Müller, Kris-
tian Myrseth, and Michael Weber. 2020. “Consumers and Covid-19: A Real-Time Survey.”
Federal Reserve Bank of Cleveland Economic Commentary 2020-08.
Mankiw, N. Gregory, and Ricardo Reis. 2010. “Imperfect Information and Aggregate Supply.”
In Handbook of Monetary Economics. Vol. 3A, , ed. Benjamin Friedman and Michael Woodford,
Chapter 5, 183–230. Elsevier: North Holland.
Mankiw, N. Gregory, and Ricardo Reis. 2018. “Friedman’s Presidential Address in the Evolution
of Macroeconomic Thought.” Journal of Economic Perspectives, 32(1): 81–96.
Mankiw, N. Gregory, Ricardo Reis, and Justin Wolfers. 2004. “Disagreement about Inflation Ex-
pectations.” , ed. Mark Gertler and Kenneth Rogoff Vol. 18, 209–248. NBER Macroeconomics
Annual 2003.
McCulloch, J Huston. 1980. “The Ban on Indexed Bonds, 1933-77.” American Economic Review,
70(5): 1018–1021.
Meltzer, Allan H. 2005. “Origins of the Great Inflation.” Review, 87(Mar): 145–176.
Meltzer, Allan H. 2010. A History of the Federal Reserve, Volume 2, Book 1, 1951-1969. Univesity of
Chicago Press.
Mminele, Daniel. 2019. “Understanding the core drivers of inflation in South Africa.” Keynote
address at SP Dow jones Indices South African seminar.
Nelson, Edward. 2005. “The Great Inflation of the Seventies: What Really Happened?” The B.E.
Journal of Macroeconomics, 5(1): 1–50.
OECD. 2018. OECD Economic Surveys: Turkey 2018. Paris:OECD Publishing.
Okun, Arthur. 1970. The Political Economy of Prosperity. Washington, DC:Brookings Institution
Press.
44
Orphanides, Athanasios. 2003. “The quest for prosperity without inflation.” Journal of Monetary
Economics, 50(3): 633–663.
Pesaran, M. Hashem, and Martin Weale. 2006. “Survey Expectations.” In Handbook of Economic
Forecasting. Vol. 1, , ed. G. Elliott, C.W.J. Granger and A. Timmermann, Chapter 14, 715–776.
Elsevier: North Holland.
Picks. 1971 edition. Pick’s Currency Yearbook. Pick Publishing Co.
Primiceri, Giorgio E. 2006. “Why Inflation Rose and Fell: Policy-Makers’ Beliefs and U. S. Postwar
Stabilization Policy*.” The Quarterly Journal of Economics, 121(3): 867–901.
Reis, Ricardo. 2003. “Where is the Natural Rate? Rational Policy Mistakes and Persistent Devia-
tions of Inflation from Target.” Advances in Macroeconomics, 3(1): 1–40.
Reis, Ricardo. 2021. “The People versus the Markets: A Parsimonious Model of Inflation Expec-
tations.” CEPR discussion paper 15624.
Romer, Christina D., and David H. Romer. 1989. “Does Monetary Policy Matter? A New Test in
the Spirit of Friedman and Schwartz.” NBER Macroeconomics Annual, 4: 121–170.
Romer, Christina D., and David Romer. 2002. “The evolution of economic understanding and
postwar stabilization policy.” Proceedings - Economic Policy Symposium - Jackson Hole, 11–78.
Sargent, Thomas J. 1988. The Conquest of American Inflation. New Jersey:Princeton University Press.
Tobin, James. 1980. “Stabilization Policy Ten Years After.” Brookings Papers on Economic Activity,
11(1, Tenth Anniversary Issue): 19–90.
Val, Flávio, and Gustavo Araujo. 2019. “Breakeven Inflation Rate Estimation: an alternative ap-
proach considering indexation lag and seasonality.” Central Bank of Brazil, Research Depart-
ment Working Papers Series 493.
Weise, Charles L. 2012. “Political Pressures on Monetary Policy during the US Great Inflation.”
American Economic Journal: Macroeconomics, 4(2): 33–64.
45
File and source
- File
- reis-2022-losing-the-inflation-anchor.pdf
- Size
- 502,222 bytes
- SHA-256
- 503026fd4afc66ad011792fe2f3136200f383db5f39890a54a5f2097e7830ca8
- Original
- researchonline.lse.ac.uk