Emi Nakamura on the methods and madness of inflation

Episode Summary

Executive Summary: The episode features economist Emi Nakamura explaining how U.S. inflation is measured, why CPI is generally trustworthy but imperfect, how CPI differs from PCE, and what historical evidence says about the costs of inflation and the Phillips curve. She argues inflation measurement is transparent yet noisy, that many quality improvements bias CPI upward, and that large 1970s disinflation was driven more by regime change and expectations than by unemployment alone.

Main Topics: How CPI is measured and why it is credible (Priority: 5/5): Nakamura explains that BLS collectors repeatedly observe the same items in stores, compare matched products month to month, and average roughly 100,000 monthly price quotes. She says this makes CPI transparent, replicable, and hard to manipulate. Biases and limitations in inflation measurement (Priority: 5/5): The discussion covers quality change, new goods, and substitution bias. Nakamura argues that CPI can miss improvements in product quality and new products, which tends to bias inflation upward rather than downward. Why the Fed prefers PCE in some contexts (Priority: 4/5): She contrasts CPI with PCE, noting PCE better accounts for substitution and uses production-based weights, while CPI relies on consumer surveys and is more useful for indexing contracts because it is simpler and timelier. The costs of inflation and the 1970s episode (Priority: 5/5): Using microdata from the 1970s, Nakamura and coauthors found little evidence that moderate inflation around 10% strongly distorted relative prices, though she stresses this does not mean inflation is harmless because other costs remain. Volcker, regime change, and inflation expectations (Priority: 5/5): Nakamura argues that breaking the Great Inflation required both high unemployment and a major change in policy credibility. She emphasizes Milton Friedman’s expectations-based view and Paul Volcker’s ability to establish a new anti-inflation regime. The Phillips curve today (Priority: 5/5): She says the Phillips curve appears fairly flat in modern data: inflation does respond to unemployment, but slowly and modestly. Large historical swings are better explained by shifts in expectations and policy regime than by a stable unemployment-inflation tradeoff. Measurement noise, market reactions, and wage inflation (Priority: 4/5): Nakamura notes that monthly inflation and wage figures are noisy and composition-sensitive. She warns against overreacting to single data prints, especially average hourly earnings, because they do not hold workers or jobs constant.

Key Arguments: CPI is trustworthy because it uses a simple matched-model approach with direct price collection in stores, not because it is perfect. Claims that inflation data are routinely 'rigged' are implausible; the methodology is transparent and difficult to manipulate. The main legitimate critique of CPI is not downward bias from manipulation but upward bias from missing quality change and new goods. Many quality-improving innovations, such as better coffee or smartphones, effectively make measured inflation look higher than consumers’ true cost of living growth. PCE is often preferred by economists and the Fed because it better accounts for substitution and uses production-account weights, which may be more accurate. The 1970s high-inflation period did not show strong evidence that moderate inflation severely distorted relative prices, at least along the channel studied. This does not mean inflation is harmless; inflation still creates contract complexity, redistributes between debtors and creditors, and threatens policy credibility. The biggest historical fall in inflation was likely due to a regime change in expectations and credibility, not unemployment alone. Modern evidence suggests the Phillips curve is relatively flat, so unemployment changes have only modest effects on inflation in the short run. Inflation data are noisy at high frequency, so monthly changes should not be overinterpreted, especially when market narratives are built on tiny moves. Wage inflation data are even harder to interpret than price inflation because average wages reflect worker and job composition changes, not just pure pay growth.

Data Points: Monthly CPI price quotes: about 100,000 - BLS collects roughly 100,000 price observations each month for CPI measurement. Typical CPI method: matched model index - Prices are compared for the same product at the same store over time. Inflation in the 1970s studied period: about 10% - Nakamura describes the Great Inflation as moderate-high inflation rather than hyperinflation. Period for comprehensive CPI methodology reforms: about every 10 years - She notes BLS methodology changes occur infrequently and slowly. Boskin Commission reform: substitution bias adjustment - Referenced as a major methodology update to CPI. Long-term inflation target debate: 2% vs 3% - Used as an example of the current policy debate over modest inflation differences. Unemployment during Volcker disinflation: high unemployment rates - The inflation collapse in the 1980s coincided with elevated unemployment. Great Recession unemployment: about 10% to 4% - Describes the large fall in unemployment after 2009 with little effect on inflation. Core inflation during Great Recession: small dip - She notes only a limited decline in core inflation during the recession. Average hourly earnings surge cited by markets: 2.9% annualized - Mentioned as a data point that triggered market excitement. Sampling uncertainty: substantial at monthly frequency - BLS standard errors indicate monthly CPI movements contain meaningful noise. TIPS data availability: late 1990s onward - Used in Nakamura’s FOMC/real interest rate paper.

Pivotal Quotes: "I think that at a basic level, you can be very comfortable that they're accurate in the sense of capturing accurately what the CPI says it's capturing." — Emi Nakamura: On whether BLS inflation statistics are trustworthy and how CPI is compiled. "I think it's essentially impossible." — Emi Nakamura: On the idea that CPI data could be rigged or manipulated to hide inflation. "We really didn't see any evidence that this higher inflation during this period ... led to important distortions in relative prices." — Emi Nakamura: On her research into the costs of moderate inflation in the 1970s.

Implications: Listeners should treat monthly inflation and wage releases cautiously: the underlying data are credible but noisy. For policymakers, credibility and expectations matter as much as unemployment, and small changes in reported inflation should not be mistaken for regime shifts.