What caused inflation after 2020?
The drivers and the measurement limits.

Section 1 · Why this distinction matters

Rate changes move borrowing, spending, asset valuations and expectations. They do not produce oil or repair a supply chain. Raising rates can still restrain demand after a supply shock, and the cost lands on employment and output with lags nobody can time.^[3]

Inflation, disinflation and deflation: in a same-basket example, a $100 basket rises 9% to $109 in year 1, then 3% to $112.27 in year 2. Disinflation means the inflation rate fell from 9% to 3%; prices still rose. If year 3 brings 2% deflation, the basket falls to $110.02 (rounded from $110.0246), still above the original $100. A lower rate does not undo price increases that already happened. This is price-level arithmetic, not a claim about what caused past inflation.^[1]

Section 2 · Type 1: Demand-pull inflation

Too much money chasing too few goods. Buyers want to spend more than producers can supply, and producers respond by raising prices.

The Fed raises rates to restrain demand. Transmission is hard to time and can weaken hiring; a rate increase does not guarantee a recession or instantly restore supply.^[3]

Section 3 · Type 2: Cost-push inflation

A supply shock drives up the cost of producing everything else. Energy, labor, raw materials, when the cost of inputs rises, the cost of outputs rises downstream.

A rate decision cannot create scarce physical inputs. It can still move demand and inflation expectations, and whether that restrains prices more than it weakens activity depends on the shock.^[3]

The 1970s stagflation era is the historical warning. The Fed's attempts to fight cost-push inflation with rate hikes produced multiple recessions without solving the underlying energy problem. It took Volcker's willingness to crush the economy with 20% rates to break the inflation expectations that had embedded, plus the eventual resolution of the energy supply shock itself.

Section 4 · Type 3: Monetary inflation

Expansion of the money supply causes all prices to rise over time. This is Milton Friedman's definition: "Inflation is always and everywhere a monetary phenomenon."

Money and spending changes transmit with variable lags. No formula converts M2 growth into household purchasing-power loss. CPI measures consumer-price changes; an asset-price index answers a different question.^[1][2]

QE changes central-bank assets and reserve liabilities, and a nonbank seller also receives a deposit. Bank lending creates deposits through a different transaction. Higher rates still reach new lending, repayment, spending and asset prices after money has already been created. Neither reserves nor M2 alone sets the path of CPI.^[2][3]

Section 5 · Type 4: Asset-price inflation

New money flows into assets (stocks, real estate, Bitcoin, art) before it reaches consumer goods. This is the Cantillon Effect in action: whoever receives the new money first can buy assets at pre-inflation prices and sell them at post-inflation prices.

The wage side of this divergence is reinforced by a labor-market feature called sticky wagesSticky wagesThe tendency of pay to adjust slowly. Employers rarely cut pay in dollar terms, and raises often trail rising prices, so real (inflation-adjusted) pay can shrink even as paychecks grow. Example: a 3% raise with 5% inflation is roughly a 2% real pay cut, one way inflation shifts buying power away from wage earners.: nominal pay adjusts downward slowly or not at all, even when conditions warrant it. When inflation runs hotter than wage growth, employers do not formally cut nominal salaries; the inflation gap quietly cuts real wages instead verify×DON'T TRUST, VERIFYClaim: Nominal wages exhibit downward rigidity ("sticky wages"); inflation reduces real wages without explicit nominal cuts.Verify at: Akerlof, Dickens, and Perry (1996), "The Macroeconomics of Low Inflation," Brookings Papers on Economic Activity ↗ Source 1 in the list belowAkerlof, Dickens, and Perry document downward nominal wage rigidity in US data. The mechanism is well-established in labor economics.. From the worker's perspective the effect is the same as a pay cut, but framing it as inflation rather than as a salary reduction is part of why the transfer is politically tolerable.

CPI does not measure investment asset prices. It measures consumer goods and services, including shelter services rather than the capital value of a home you buy. That scope is a design choice. It does not show where monetary expansion lands, and it does not prove CPI understates every household’s inflation.^[1]

See The Cantillon Effect: Who Gets New Money First for the 300-year-old framework that explains why new money moves through assets before goods.

How inflation is measured: CPI vs PPI vs PCE

There is no single inflation number. Three measures matter and they often disagree.

Consumer Price Index (CPI)

CPI measures average price changes for a representative basket of consumer goods and services. CPI-U covers urban consumers; CPI-W covers urban wage earners and clerical workers. BLS publishes them monthly and, since January 2023, updates their expenditure weights annually; the weights then remain fixed for 12 months. Most basic indexes use a geometric mean that allows substitution within item categories, not a permanently fixed shopping list.^[1][6]

Producer Price Index (PPI)

Price changes received by domestic producers for their output, before goods reach consumers. PPI changes typically flow through to CPI with a lag of a few months, which makes it a leading indicator verify×DON'T TRUST, VERIFYClaim: The Producer Price Index measures wholesale-stage prices and is published monthly by the BLS.Verify at: BLS Producer Price Indexes ↗ Source 2 in the list belowPPI covers final demand, intermediate demand, and crude materials; "PPI Final Demand" is the headline most often cited..

PCE deflator (Personal Consumption Expenditures)

PCE is published by BEA; the Fed’s longer-run 2% target is the annual change in the total PCE price index, not CPI. The substitution story is more specific than it sounds: CPI-U and CPI-W allow modest within-category substitution in most basic indexes, but not contemporaneous substitution between item categories. Chained CPI (C-CPI-U) is designed to capture that upper-level substitution using current spending patterns, and its initial values are revised as spending data arrive. PCE and CPI also differ in coverage and weights, so substitution alone does not explain the gap.^[1][6][8]

Core vs headline

Headline includes everything. Core excludes food and energy, which are volatile because of weather, supply shocks, and geopolitics. The Fed focuses on core PCE as its primary trend-inflation gauge, on the theory that food and energy prices add noise without changing the underlying signal.

What this means for your money

• The number the Fed targets (PCE) is typically lower than the number you see in headlines (CPI). The Fed may declare victory on inflation while groceries and gasoline still feel high.
• Social Security uses CPI-W, a wage-earner variant of CPI. Your actual basket may grow faster or slower than CPI-W, so the COLACost-of-Living Adjustment (COLA)An automatic raise to Social Security or pension payments each year, sized to match how much prices have gone up. may not protect your specific spending pattern.
• TIPS coupons adjust to CPI-U. If your spending tracks CPI closely, TIPS preserve real purchasing power. If it does not, the protection is partial. Detail at Cash Management and Social Security.

Section 6 · What CPI can and cannot tell you

CPI is a population average, not your personal inflation rate. Your housing, healthcare, location and spending weights can push your rate above or below the published number. That cuts both ways; these limits do not show systematic understatement for every household.^[1][6]

• Hedonic adjustments. If a TV's features improve, the BLS adjusts the price downward to reflect "quality-adjusted" cost, even if the sticker price went up. Defensible in theory, but it means you can see the sticker price rise while CPI shows it as flat or declining.
• Substitution: CPI-U and CPI-W account for some substitution within categories; C-CPI-U also handles substitution between categories. The beef-to-chicken story does not describe all three the same way. Chained CPI approximates a cost of living; it does not claim that switching to chicken leaves you as well off as before.^[6]
• Owners’ equivalent rent (OER): CPI measures the shelter service a homeowner consumes, not the investment price of the home. BLS estimates owners’ equivalent rent from observed rents on rental units; it does not poll homeowners about what their house is worth. What a recent buyer actually pays can look very different.^[1]
• Headline versus core: headline CPI includes food and energy; core CPI strips them out to show the underlying trend. You still pay for food and energy. Say which series you mean, and do not let core stand in for the whole official inflation rate.^[1]
• Shrinkflation and quality degradation. Package-size reductions and quality downgrades are supposed to be caught by CPI adjustments but the corrections lag and quality changes are modeled rather than measured. See Shrinkflation for the full breakdown of what CPI misses at the grocery shelf.

The full downstream effect of under-measured inflation, how it cascades into housing, family formation, government expansion, and political instability, is documented on Downstream Consequences.

Real vs nominal interest rates

Nominal rate is the rate printed on your statement. Real rate is the nominal rate minus inflation. The real rate tells you whether your purchasing power is growing.

A 5% savings account is good when inflation is 2% (3% real). The same 5% account is a real loss when inflation is 7% (negative 2% real). The bank still pays you 5%, but each dollar buys less than it did when you deposited it.

For decisions that span more than a year, always think in real terms. Mortgages, student loans, retirement projections, and bond yields all need the inflation deduction to be comparable across time.

What this means for your money

• "High-yield savings accounts" advertised at 4% to 5% can still lose real purchasing power when inflation runs hotter.
• Fixed-rate mortgages get cheaper in real terms when inflation rises faster than expected. Borrowers benefit from unexpected inflation; savers and lenders lose.
• Bond yields quoted in newspapers are nominal. The real yield on TIPS is what you keep after inflation.

The Phillips Curve

A.W. Phillips documented a historical pattern in UK data published in 1958: low unemployment correlated with high wage growth (a proxy for inflation), and high unemployment with low wage growth verify×DON'T TRUST, VERIFYClaim: A.W. Phillips's original paper was "The Relation Between Unemployment and the Rate of Change of Money Wage Rates in the United Kingdom, 1861-1957," Economica 25(100), 1958.Verify at: JSTOR ↗ Source 3 in the list belowThe original paper plotted wage growth against unemployment. Later economists translated wage growth to consumer-price inflation; Phillips himself did not..

The intuition was straightforward. When unemployment is low, workers have bargaining power; wages rise, and higher wages flow through to higher prices. When unemployment is high, workers compete for jobs and price pressure cools.

By the 1960s, this had become the basis for activist policy: governments could pick a point on the curve. Slightly higher inflation in exchange for lower unemployment, or higher unemployment in exchange for lower inflation. The trade-off was treated as stable.

The 1970s broke it

Stagflation: simultaneous high inflation and high unemployment. The textbook trade-off said this should not happen. Milton Friedman and Edmund Phelps had independently predicted it would, on the argument that workers and firms adapt their inflation expectations. If workers expect 5% inflation, they demand wage increases of 5% or more; firms agree, expecting to charge higher prices. Prices and wages both rise. Unemployment returns to its "natural rate" but with higher inflation baked in. To get the same temporary unemployment benefit next time, the central bank has to deliver even more inflation, creating a spiral verify×DON'T TRUST, VERIFYClaim: Friedman's 1968 American Economic Association presidential address argued that the Phillips trade-off would break down once inflation expectations adjusted.Verify at: Friedman, "The Role of Monetary Policy," American Economic Review 58(1), 1968 ↗ Source 9 in the list belowPhelps's parallel argument appeared in 1967-1968 papers. Both were vindicated by the 1970s experience..

The modern synthesis distinguishes a short-run Phillips Curve (the trade-off is real temporarily) from a long-run vertical Phillips Curve (no lasting trade-off; the economy returns to the natural rate of unemployment regardless of the inflation level).

What this means for your money

The Fed uses the Phillips Curve framework to set interest-rate policy. When unemployment falls below what the Fed considers the natural rate, the Fed raises rates preemptively to head off wage-price pressure. When unemployment rises, the Fed cuts rates to stimulate hiring. This is why your mortgage rate, your car-loan rate, and your savings yield all move on the unemployment report.

The 2021-2023 episode is the most recent stress test of the framework. The Fed held rates near zero while unemployment fell rapidly, partly because it had revised down its estimate of the inflation-triggering level. Inflation reached 9.1% (CPI-U headline) in June 2022. Whether this confirms or refutes the Phillips framework is actively debated; the empirical answer turns on whether the inflation was driven by demand (a Phillips story) or by supply shocks (a different story).

Section 7 · Why the Fed targets 2% inflation (and why that's not crazy)

The Federal Reserve's 2% inflation target is not arbitrary and not malicious. The strongest version of the pro-central-bank case deserves a hearing on its own terms before anyone rebuts it. There are real reasons mainstream economists settled on this number.

Even 2% a year, hit perfectly for 50 years, compounds: 1 − 1/(1.02^50) = 62.8% purchasing-power loss on cash that earns nothing. That is the success case. It is not the return on an interest-paying account or a claim about what caused past CPI changes.^[1]

Section 8 · What this means for Bitcoin

Bitcoin's fixed supply removes the question of whether the target gets hit. There is no target. There is no discretion. 21 million coins, issued on a schedule written into the protocol in 2009. Whether the Fed succeeds or fails at 2% does not affect what a satoshi represents.

A fixed-supply asset is neutral for which type of inflation is happening. All four types erode the purchasing power of unbacked fiat over long horizons. None of them can inflate Bitcoin's supply.

Whether inflation shows up in consumer prices (Type 1 or 2), in monetary aggregates (Type 3), or in asset prices (Type 4), the denominator is always the currency. Bitcoin is the only major asset whose supply cannot be expanded to paper over the problem. That is why Bitcoin holders care less about which type of inflation is running hot in any given quarter and more about the long-term trajectory of the currency itself.