The Inflation Expectations Fiction

Central bankers spend a lot of their public communication talking about "inflation expectations." When expectations are "well anchored," the story goes, inflation behaves. When expectations "become unanchored," inflation runs hot even after the original shock has passed, because households and firms bake the new pessimism into wage negotiations and price lists. The policy tool that acts on this is the interest rate, wielded as an instrument of credibility.

I want to take this seriously for a moment and then explain why, the more you look at it, the more it starts to resemble a theory with no empirical referent.

Start with a simple exercise. Ask anyone you know — an electrician, a teacher, a shop owner, a web developer — whether they successfully negotiated a wage increase ahead of inflation because they expected next year's CPI to run hot. Almost no one can tell you a story like that. Wages are sticky and backward-looking. You get a raise in December because you suffered through inflation all year, not because you or your union priced in a forecast. The wage side of the wage-price spiral is, in most economies, reactive.

Now ask a merchant why they raised a price. They will tell you their supplier raised theirs. They bought diesel at a higher wholesale price this month, eggs cost more at the farm gate, the shipping agent sent a new tariff sheet. So they passed it through — not aggressively, but enough to protect their margin and be able to restock next month at replacement cost. This is real, and it is preemptive in a narrow sense: the baker raises bread today so she can afford flour tomorrow. But she is not reading the nation's psychology. She is reading her supplier's invoice.

Neither of those behaviors — the backward-looking wage, the replacement-cost markup — is "expectations" in the sense the central bank's models require. Both are mechanical, observable, and tied to realized costs, not forecasts. So what exactly are central banks talking about?

Because the model demands it

Modern central banks run some version of a New Keynesian DSGE model. Inside those models, inflation persistence is generated by a term that looks like $E_t(\pi_)$: the representative agent's expected inflation for next period. Without that term, the math cannot explain why realized inflation stays high after a shock. Take it out and the model collapses back to target on its own.

The trouble is that this term was inserted for mathematical tractability, not because anyone ever pinned down its counterpart in the real world. Survey measures of expectations — the ones that get cited as evidence that "expectations have deanchored" — are overwhelmingly adaptive. The Coibion and Gorodnichenko literature established this years ago: when you ask households and firms what they expect inflation to be, they tell you what it has been recently. Survey expectations are a lagged mirror of realized CPI dressed up in the grammar of forecasting.

So the story central banks tell goes: we estimate a model with an expectations term, the expectations term is large, therefore expectations are the cause. But the expectations term is whatever the model needs it to be. It soaks up all the persistence the model cannot otherwise explain — supply-chain unwinding, sectoral bottlenecks, contractual indexation, corporate markup behavior — and relabels that residual "psychology."

It is the econometric version of writing dragons on the blank parts of a medieval map. The map is not wrong exactly; the dragons are a placeholder for "we did not measure what is actually here."

The MMT critique, and the inconvenient Fed paper

MMT economists are noisy about interest-rate policy for political reasons — they think hiking rates disciplines workers while rewarding bondholders, and they prefer credit controls and fiscal tools. That is the loud part. Underneath it is a sharper methodological claim: there is no empirical evidence that consumer inflation expectations cause inflation.

The canonical reference here is not a fringe heterodox paper. It is a 2021 Federal Reserve working paper by Jeremy Rudd, pointedly titled "Why Do We Think That Inflation Expectations Matter for Inflation? (And Should We?)". Rudd, a mainstream Fed economist, walked through the empirical literature and concluded that the belief is theoretically shaky and empirically weak. The paper caused a small scandal inside the Fed because it said out loud what many economists already suspected: the edifice of "expectations anchoring" rests on correlation, convention, and modeling convenience rather than identified causal evidence.

The 2021–2024 disinflation cycle did not help the expectations theory. Real wages fell across most of the developed world. Workers lost purchasing power. The runaway wage-price spiral central bankers were warning about never materialized. Inflation came down roughly in step with supply chains repairing, energy shocks rolling off, and shipping capacity returning — exactly the variables MMT economists and supply-side analysts had been pointing at from the start. Central banks took credit for the disinflation anyway. The post-hoc narrative was: we hiked, expectations re-anchored, inflation fell. The actual causal chain is less flattering. The central bank was riding the denominator.

Why the story persists

If the expectations story is this weak, why does it dominate? Three reasons, none of them about science.

The models need it. DSGE models cannot generate realistic inflation persistence without an expectations term. Dropping it would require rebuilding the core toolkit used for forecasting and policy simulation. Institutional momentum inside central banks runs hard against that.

It preserves the illusion of control. If inflation is caused by global oil prices, avian flu, shipping backlogs, or — in Iceland's case — a structural housing shortage amplified by contractual CPI indexation, the central bank is largely a spectator. It cannot print houses, lay eggs, or drill for oil. But if inflation is caused by expectations, the central bank has a direct tool. The framing converts the central bank from a passive observer of real-economy shocks into the hero of the story.

It shifts blame. Treating inflation as an expectations problem quietly encodes the idea that the public is at fault — consumers accepting price hikes too easily, workers demanding raises too aggressively — and that the cure is to discipline them through higher unemployment. The alternative diagnoses — corporate margin expansion, fiscal choices, institutional design flaws like Iceland's verðtrygging — all point at politically costlier targets.

What the honest answer looks like

The Iceland case is particularly instructive because the CBI has been running the expectations argument as its explicit diagnosis while the actual persistence mechanism is mechanical and contractual. When Icelandic CPI rises, indexed mortgages, rental contracts, and service agreements adjust automatically. No expectation formation is required. A Bayesian trend-cycle decomposition will pick up this automatic contractual pass-through as a slow-moving "trend" component and call it deanchored expectations. The label is a modeling choice, not an empirical finding. The same trend is observationally equivalent to mechanical cost pass-through in a contractually-indexed economy with a real-resource bottleneck.

An honest central-bank diagnosis of persistent inflation would have to say: we cannot distinguish, from our models alone, between behavioral-expectations persistence and mechanical cost-pass-through persistence; the two are observationally equivalent within our toolset; the microdata suggests workers are not negotiating ahead of inflation and merchants are reacting to realized upstream costs rather than forecasts; and the canonical citations for expectations-as-cause are methodologically thin.

No central bank will say this, because it leads directly to the awkward question: then what are you for? If the policy rate is not actually acting on expectations — and the expectations term in the model is a placeholder for mechanical persistence the central bank cannot directly touch — then the rate is a blunt instrument being applied to the wrong variable.

That is the question MMT is really asking, stripped of the culture-war packaging. Not "are interest rates good or bad," but: what exactly do you think you are acting on, and where is the evidence?

The answer from Washington, Frankfurt, and Arnarhóll is consistent: we are acting on expectations, we have models that say so, and the models have the coefficients the models need. It is a circular defense, and the public is starting to notice.