Why They Matter and How They Are Formed

Speech

Notes

Inflation Expectations – Why They Matter and How They Are Formed Sarah Hunter [ * ] Assistant Governor (Economic) Citi Australia and New Zealand Investment Conference Sydney – 16 October 2024 Introduction I would first like to pay respect to the traditional and original owners of this land, the Gadigal people
of the Eora Nation, to pay respect to those who have passed before us and to acknowledge today’s
custodians of this land. I also extend that respect to any First Nations people joining us here today. A low and stable inflation rate is critical to preserving macroeconomic stability. Having a good idea of
what’s going to happen to prices allows businesses to plan for investment and expansion. It also
makes things like budgeting and financial planning easier for households. This is particularly true for
those on low incomes, who typically have smaller financial buffers than others and spend more of their
income on essentials. ¹ And with more stable household and business balance
sheets, the financial system is more stable. ² The experience of the last few years has clearly highlighted this. Everyone across the economy has felt
the increased cost of living. ³ This is very clear in the data we monitor, such as
household spending, but it’s perhaps more apparent in survey metrics such as consumer confidence,
which remains much lower than its pre-pandemic average (Graph 1). So there are a number of good
reasons to bring inflation down and keep it at a low and stable rate. Graph 1 In addition to the tangible impact of elevated inflation today, central bankers often note that they want
to make sure that inflation expectations remain anchored. But why is this the case? And what impact do
current inflation outcomes have on expectations? Why do inflation expectations matter? Macroeconomists generally think that a prerequisite for consistently achieving low and stable inflation
over time is well-anchored inflation expectations. That is, people across the economy believe inflation
will generally average a low rate (in Australia’s case, 2–3 per cent), and they make decisions based on this
underlying belief that becomes self-reinforcing. Indeed, this is a key lesson from economic history;
there are multiple episodes that demonstrate the damage de-anchored expectations can cause, and the
policy effort and welfare costs associated with re-anchoring them. Türkiye’s current experience is
just one example (Graph 2). ⁴ Graph 2 So why do expectations matter at all when it comes to economic outcomes? We think they matter because
people don’t just make decisions based on what is happening today, they also factor in what they
think will happen tomorrow. In other words, inflation expectations are at least partly self-fulfilling. For example, our decision over how much to save for retirement today is determined by how much income we
think we’ll need once we stop working, and this is partly influenced by what we think will happen to
prices between now and then. In addition to changing the behaviour of households, inflation expectations also directly feed into all of
the decisions firms make – for example, over capital investment, pricing and staffing. One way this
occurs is through the wage-setting process (Graph 3). ⁵ This could be workers, or their union
representatives, bargaining for higher wages if they think inflation will be higher. ⁶ Or it
could be firms’ expectations of higher future prices giving them the confidence to offer higher
wages today to attract workers. Graph 3 And given that this is an investment conference, I’d be remiss not to mention how important inflation
expectations are to the domestic and international portfolio allocation decisions made by financial
market participants. These expectations then feed into long-term interest rates, exchange rates, and the
prices of assets in our superannuation funds and all other investment portfolios. In short, inflation
expectations are a factor in pretty much every economic decision that’s made every day. The fact that expectations feed into actual inflation outcomes means de-anchored expectations typically
leads to greater inflation volatility (Graph 4). Volatility breeds uncertainty, and uncertainty
makes decisions harder for everyone. As a business, how do you decide when it’s right to invest if
you’re less sure of the financial returns? And to go back to the example of households deciding how
much to save for retirement or to buy a home, a bout of unexpectedly high inflation is very hard to plan
for. Both the effort required to make decisions with uncertainty, and that some otherwise good decisions
will not be made, makes us all worse off. Graph 4 Tracking inflation expectations Given the enormous damage that such de-anchoring can cause, and that policy can be enacted more flexibly
while expectations remain anchored, the RBA Board is constantly alert for signs that this risk might
emerge here in Australia. ⁷ It does that by tracking a range of inflation
expectations measures, including multiple financial market measures, and surveys of households, unions
and professional forecasters. ⁸ That analysis indicates that inflation expectations
have not become de-anchored through the current high-inflation experience (Graph 5). ⁹ Graph 5 So we’re not currently concerned that expectations could become de-anchored in the near term. But we
do think it’s important that we track how they’re evolving and that we understand how
expectations are formed, so we can monitor whether there are any signs of this risk materialising in the
future. As I’ve already alluded to, there are a number of different groups across the economy, and each plays
a part in determining aggregate macroeconomic outcomes. To understand what’s happening to
expectations, we therefore need to understand how different groups form their inflation expectations, as
they each play critical roles in determining how the economy evolves over time. For consumption/savings decisions, households’ own expectations matter the most. For wage bargaining
and competition for labour, unions’ and firms’ expectations likely matter most. And when it
comes to how inflation expectations feed into long-term interest rates, it’s the financial
markets’ expectations that matter. In short, given the importance of inflation expectations as a driving force of many economic decisions, we
need to understand how all of the different groups across the economy form their inflation expectations
so that we can do our best to keep them anchored. So today I’m going to discuss some of the latest research in this area, which we have conducted
ourselves and in partnership with our colleagues in academia. This includes a Research Discussion Paper
that has been released in parallel with this event, which explores some of the points below in more
detail – I encourage you all to have a look at my colleagues’ work. ¹⁰ The presentation I am giving today draws heavily on a presentation at one of the first ‘Policy Issues
Meetings’ with RBA Board members earlier this year. As previously highlighted by Governor Bullock,
these meetings: … assemble a group of staff with the right experience and expertise to give the members insights and
diversity of perspectives on the key issues relevant for policy. It will provide analysis of issues that
are relevant to a few upcoming [Board] meetings, not just the immediate one. ¹¹ These new meetings have been very well received by Board members. They have appreciated the opportunity to
explore policy-relevant topics in more depth and to meet with more of the staff that are engaged in the
work. In turn, staff have valued the additional engagement with their work, so it’s been a clear
win-win. For most of this speech, I’ll be focusing on household and union expectations, and mostly on
short-term expectations. In the past, how these groups form expectations has been less well-understood,
and this is why we’ve focused our latest research here. But before turning to unions and households, it is worth mentioning that we have a reasonable
understanding of how financial markets form expectations. Financial markets efficiently incorporate
signals about the likely future direction of inflation into market prices; by taking active positions
that are contingent on economic outcomes, it’s no surprise that market participants keep themselves
very well-informed about what’s happening. From these prices, we can discern whether their short-
and long-term expectations remain anchored to the RBA’s inflation target. ¹² To understand how households and unions form their expectations, we’ve collaborated with academic
colleagues to develop a very general model approach that we’ve then applied to different data
series. The model assumes that some people form their expectations by extrapolating from their previous
experience. That is, they assume that their experience of price increases in the past are a good guide
for what they’ll experience in the future. The model also assumes that some people build on this and
take account of forward-looking information as well. For example, they might expect to see a sharp
increase in grocery prices in the future if it’s reported that the harvest has been poor. The first iteration of the model was run through to around the middle of the pandemic. The graph shows the
fit of the model to actual data. In the grey lines are unions’ one- and two-year-ahead expectations,
and households’ one-year-ahead expectations (Graph 6). ¹³ And then the blue lines are
the model estimates of each of these. ¹⁴ Graph 6 We think the model did a reasonable job over the historical period. Especially for unions, where the model
pretty much captured every major wiggle in their expectations. We’ve learned a lot from this process, but there are three key insights that I want to highlight: We estimate that around three-quarters of households and unions form their expectations by
extrapolating from their lived experience. That is, they observe what inflation was yesterday and
compare it to what they expected. Every time inflation turns out higher than what these people
expected, they partially adjust their expectations up. This extrapolation process happens a lot slower for households than it does for unions. That is,
households only adjust their expectations a small amount each time they are surprised. As a result,
inflation has to be persistently higher or lower than previously expected for expectations to change
significantly. The remaining one-quarter of unions and households don’t just extrapolate, they incorporate a
lot more of the broader economic information available to them (beyond inflation outcomes themselves)
to make forward-looking judgements about where inflation is likely to go. In principle, this is
similar to the RBA’s forecasting process – we look at past outcomes and forward-looking
indicators to assess how we think inflation will evolve from today. Of the roughly 25 per cent who take on board additional information, this could come from a
number of different sources. To carry on my groceries example from earlier, in 2011 this group might have
expected that banana prices would shoot up in the months after Tropical Cyclone Yasi struck northern
Queensland, given the reporting of the damage to that year’s crop. ¹⁵ Or this group could be looking
at economic forecasts – including the RBA’s – to get a sense of where inflation may be
heading. With this better understanding of how people form their inflation expectations, we can now assess how they
have evolved recently, relative to what the models expected they would do. Less extrapolation recently could reflect greater attention to inflation or recognition that the recent
episode is temporary The orange line is the model’s prediction for how inflation expectations would evolve during the
recent high-inflation period (Graph 7). ¹⁶ While inflation was rising, expectations were
evolving in-line with the model’s output. But the model suggested that the turning point in
expectations would come later. So expectations are currently lower than our models thought would be the
case. ¹⁷ Graph 7 As best we can tell, the models missed the turning point because unions and households have been
extrapolating less from the recent high inflation outcomes. ¹⁸ The model attributes part of
this to an increase in the share of people who take on board forward-looking information, from around
one-quarter to over two-thirds for unions. This finding is consistent with a theory known as the ‘rational inattention’ hypothesis. ¹⁹ The
idea being that when inflation is low and stable, extrapolation from the past provides a reasonably
accurate expectation of the future, so it is not worth paying more ‘attention’. Conversely,
when inflation does not fit this pattern – for example, in the recent past when it was much higher
– extrapolation might provide a poor forecast. So it is ‘rational’ for people to put more
effort into thinking about where inflation will head next. Another finding from the model is that those who use previous inflation to form their expectations, that
is they use yesterday’s experience to guide today, have been adjusting their view more slowly in
recent years. A possible reason for this is that some people have seen the recent experience as atypical
and so don’t expect it to continue – given the nature of the shocks (the pandemic and then the
conflict in Ukraine), it’s easy to understand this. So while this group only use previous inflation
outcomes to form their expectations, they do appear to adjust how much weight they put on specific
outcomes to take account of broader economic conditions. Unfortunately, these are just plausible hypotheses at this point, we don’t have enough evidence to be
definitive. If once inflation sustainably returns to the target band expectation formation reverts to how
it was before the recent episode, that would provide further evidence in favour of these hypotheses. But
more importantly, it would give us comfort that in future inflationary episodes, expectation formation
might similarly change in a way that mutes the increase in expectations. Another possible explanation is that some more ‘salient’ prices have evolved differently to
average prices In everything I’ve shown so far, we assume that the price increases that matter most are the ones
that people spend most of their money on. Which is exactly how the Consumer Price Index, or CPI, is
constructed. But that might not be how people extrapolate from what they have previously observed to form their
expectations. Our lived experience is that we ‘see’ some prices much more frequently than
others, and that some price changes are more noticeable than others. Prices that change regularly or that people pay often may be particularly influential when people form
their expectations – they’re more visible, and they could be seen as a proxy for what’s
happening to all prices across the economy. These are known as salient prices. ²⁰ While there are some obvious candidates for prices that may be salient – such as fuel, groceries,
rent, and energy prices – determining how salient they are has unfortunately proven difficult. The strongest result we have obtained is with respect to petrol and diesel prices – that is, the
prices you see changing every day when you drive past a petrol station or fill your car up. For other
potentially salient prices, whether or not our models identify them as salient depends on the various
other modelling decisions that are made. But for fuel prices, it doesn’t seem to matter what you do
to the model, these prices almost always show up as salient. Having said all that, allowing for fuel to be a salient price in the model does not significantly change
the model’s estimate of inflation expectations most of the time. This occurs because fuel prices are
volatile and households learn slowly. So it actually takes an extended period of fuel prices evolving
differently to other prices before there would be a meaningful impact on expectations (according to the
model). But that’s exactly what we have seen in the past few years (Graph 8). From the beginning of 2021
until mid-2022, fuel price inflation was much higher than average price inflation, increasing
61 per cent over this period. ²¹ But for most of the period since then, fuel price
inflation has been around its historical average, while much of the broader consumption basket has
continued to experience above-target price inflation. Graph 8 So, for household’s expectations, accounting for the salience of fuel prices can at least partially
explain why the simpler inflation expectations model presented earlier predicted that short-term
inflation expectations would remain higher for longer. Conclusion To conclude, recent research has improved our understanding of how people form inflation expectations. As
a result, we have been able to better analyse how expectations have evolved during the recent
high-inflation period. And it’s a good news story with respect to expectations: Short-term expectations appear to be converging towards long-term expectations, and these have
remained anchored through the recent past. There’s no evidence of expectations being more persistent than normal. And there’s even some evidence of households and unions extrapolating less from recent
inflation, at least during the period of higher inflation. We need to be mindful of certain prices that may be particularly ‘salient’ for households.
But such prices work in both directions, and recently have been working to bring expectations down
faster. Endnotes I would like to thank Anthony Brassil and Callum
Ryan for their assistance in writing this speech. Helpful feedback was provided by Michelle
Bergmann, Samuel Evangelinos, Andrew Hauser, Brad Jones, Christopher Kent and Martin McCarthy. [*]
For further discussion, see Bullock (2024); Wood, Chan and Coates (2023); Beckers et al (2024). ¹
For a fuller discussion of the costs of high inflation, see Bullock (2024). ²
For surveys of US and European households, see Charm et al (2022); Bazzoni et al (2022). ³
Economic history is clear that hyperinflation, where expectations are completely
de-anchored and money supply growth is extremely rapid, produces very bad outcomes (e.g.
the experience of Germany in the early 1920s, Hungary in the 1940s, through to Zimbabwe
and Venezuela in the last 20 years). There is also plenty of evidence of poor outcomes
under high inflation experiences, such as the United States in the 1970s (see Blinder
(1982) for a short summary) and Türkiye today. On the other side, Japan’s experience
over the last 30 years highlights the difficulties of zero inflation/deflation.
⁴
On the pass-through of labour costs into prices, see Ampudia, Lombardi and Renault (2024); Chin and Lin (2023); Suthaharan and Bleakley (2022). ⁵ For an analysis of union bargaining power, see Bishop and Chan (2019). ⁶ For discussion on the policy flexibility afforded by anchored expectations, see Bernanke (2013); Borio et al (2023). ⁷ For detailed expositions, see Moore (2016); Beckers and Brassil (2022).
⁸ See RBA (2024) for a recent assessment
of inflation expectations. In Graph 5, the ‘adjusted bond market measure’ extracts the
expected inflation rate from bond market prices using the Hambur and Finlay (2018)
affine term-structure model, the parameters of which are disciplined by survey data. ⁹ See Brassil et al (2024). ¹⁰ Bullock (2023). ¹¹ The anchoring of these expectations is discussed further in RBA (2024). ¹² For the model estimates of unions’
two-years-ahead expectations, dots are used prior to 2004 due to model data
availability. For more information, see Beckers and Brassil (2022).
The household measure of expectations used in this research is the average of responses
excluding responses in exact multiples of five (i.e. 5, 10, 15 etc) as constructed by
Haidari and Nolan (2022). This measure is used because the average that includes these
rounded responses has a large upward bias relative to historical inflation outcomes.
While there are other methods of removing this bias, previous research has found that
changes in rounded responses tend to reflect changes in uncertainty: see, for example,
Binder (2017); Reiche and Meyler (2022). And while uncertainty affects decision-making,
the mechanism is separate to how inflation expectations affect decisions. The authors of
this research therefore conclude that removing rounded responses is a more transparent
and well-justified approach to removing this bias than other methods.
¹³ For details of the model, see Beckers and Brassil (2022); Brassil et al (2024); Brassil, Gibbs and Ryan (forthcoming). ¹⁴ Banana prices increased by over 400 per cent between June 2010 and June 2011. ¹⁵ The ‘prediction’ uses the original
model estimates and then projects the model forward using the model inputs available at each
point in time. See Brassil et al (2024) for model details. ¹⁶ Unions’ expectation data are available
until September 2023. ¹⁷ See Brassil et al (2024) ¹⁸ See, for example, Sims (2003); Coibion et al (2020); Afrouzi and Yang (2021); Maćkowiak, Matějka and Wiederholt (2023). ¹⁹ For further discussion and exploration of salient prices, see Binder (2018); D’Acunto et al (2019); De Fiore et al (2022); Kilian and Zhou (2022); Chua and Tsiaplias (2024); Anesti, Esady and Naylor (2024). ²⁰ Percentage increase in automotive fuel between December 2020 and June 2022 (from monthly CPI data). ²¹ References Afrouzi H and C Yang (2021), ‘ Dynamic Rational Inattention and the Phillips Curve ’,
CESifo Working Paper No 8840. Ampudia M, MJ Lombardi and T Renault (2024), ‘The Wage-price Pass-through Across Sectors:
Evidence from the Euro Area’, BIS Working Paper No 1192. Anesti N, V Esady and M Naylor (2024), ‘Food Prices Matter Most: Sensitive Household
Inflation Expectations’, CFM Discussion Paper Series CFM-DP2024-34. Bazzoni E, M Jacob, S Land, M Mijer, J Moulton and S Welchering (2022), ‘European Consumer
Pessimism Intensifies in the Face of Rising Prices’, McKinsey & Company, October. Beckers B and A Brassil (2022), ‘ Inflation Expectations in Australia ’, The
Australian Economic Review , 55. Beckers B, A Clarke, A Gao, M James and R Morgan (2024), ‘ Developments
in Income and Consumption Across Household Groups ’, RBA Bulletin ,
January. Bernanke B (2013), ‘ A Century of US Central Banking: Goals, Frameworks,
Accountability ’, Journal of Economic Perspectives , 27(4). Binder CC (2017), ‘ Measuring Uncertainty Based on Rounding: New Method and Application to
Inflation Expectations ’, Journal of Monetary Economics , 90. Binder CC (2018), ‘ Inflation Expectations and the Price at the Pump’, Journal of
Macroeconomics , 58. Blinder AS (1982), ‘The Anatomy of Double-Digit Inflation in the 1970s’, in Hall RE
(ed), Inflation: Causes and Effects , University of Chicago Press,
pp 261–282. Borio C, M Lombardi, J Yetman and E Zakrajšek (2023), ‘The Two-regime View of
Inflation’, BIS Papers No 113. Brassil A, C Gibbs and C Ryan (forthcoming), ‘Boundedly Rational Expectations and the
Optimality of Flexible Average Inflation Targeting’, RBA Research Discussion Paper. Brassil A, Y Haidari, J Hambur, G Nolan and C Ryan (2024), ‘ How Do Households Form Inflation
and Wage Expectations? ’, RBA Research Discussion Paper No 2024-07. Bullock M (2023), ‘ A Monetary Policy Fit for
the Future ’, Australian Business Economists Annual Dinner, Sydney, 22 November. Bullock M (2024), ‘ The Costs of High
Inflation ’, Keynote Address to the Anika Foundation Fundraising Lunch, Sydney,
5 September. Charm T, JR Saavedra, K Robinson and T Skiles (2022), ‘The Great Uncertainty: US Consumer
Confidence and Behavior during Inflationary Times’, McKinsey & Company, August. Chin M and L Lin (2023), ‘The Pass-through of Wages to Consumer Prices in the COVID-19 Pandemic: Evidence from Sectoral Data in the U.S.’, IMF
Working Paper No 2023/233 . Chua CL and S Tsiaplias (2024), ‘ The Influence of Supermarket Prices on Consumer Inflation
Expectations ’, Journal of Economic Behavior and Organization , 219. Coibion O, Y Gorodnichenko, S Kumar and M Pedemonte (2020), ‘ Inflation Expectations as a
Policy Tool? ’, Journal of International Economics , 124. D’Acunto F, U Malmendier, J Ospina and M Weber (2019), ‘ Salient Price Changes, Inflation
Expectations, and Household Behavior ’, June. De Fiore F, T Goel, D Igan and R Moessner (2022), ‘ Rising Household Inflation Expectations:
What are the Communication Challenges for Central Banks? ’, BIS Bulletin, No
55. Haidari Y and G Nolan (2022), ‘ Sentiment,
Uncertainty and Households’ Inflation Expectations ’, RBA Bulletin ,
September. Hambur J and R Finlay (2018), ‘ Affine
Endeavour: Estimating a Joint Model of the Nominal and Real Term Structures of Interest Rates
in Australia ’, RBA Research Discussion Paper No 2018-02. Kilian L and X Zhou (2022), ‘ Oil Prices, Gasoline Prices, and Inflation Expectations ’, Journal of Applied Econometrics , 37(5). Maćkowiak B, F Matějka and M Wiederholt (2023), ‘ Rational Inattention: A Review ’, Journal of Economic Literature , 61(1). Moore A (2016), ‘ Measures of Inflation
Expectations in Australia ’, RBA Bulletin , December. RBA (2024), ‘ Box A:
Are Inflation Expectations Anchored? ’, Statement on Monetary Policy ,
August. Reiche L and A Meyler (2022), ‘Making Sense of Consumer Inflation Expectations: The Role of
Uncertainty’, ECB Working Paper Series No 2642. Sims C (2003), ‘ Implications of Rational Inattention ’, Journal of Monetary
Economics , 50(3). Suthaharan N and J Bleakley (2022), ‘ Wage-price
Dynamics in a High-inflation Environment: The International Evidence ’, RBA Bulletin , September. Wood D, I Chan and B Coates (2023), ‘ Inflation and
Inequality: How High Inflation Is Affecting Different Australian Households ’,
Working paper prepared for the RBA Annual Conference, Sydney, 25–26 September.