PCE vs CPI (Rickyboy’s argument flies in his face)
The PCE vs the CPI
The confusion begins with the two headline figures. The CPI and the PCE are designed to measure the same thing, so one should expect they will generally agree. Until Q2 2021 – when our inflation outbreak really got going – they did agree.
But since then, they have diverged significantly. The long term “gap” between the two metrics had been small (the CPI was less than 2/10ths of a percent higher). But as inflation has accelerated, the gap has widened by a factor of 8.
This is troubling. If two bathroom scales give different answers, but the difference is small and consistent, we accept it as an ordinary variation in the manufacture or calibration of the two mechanisms. But if one scale starts producing answers that differ by a large and growing amount from the other, the conclusion must be that there is something wrong with one or both of them.
Part of the problem is clear: the CPI is badly broken. This has been known for a long time. Congressional hearings and formal studies of the problem dating back to the 1990’s identified a systematic over-estimation of inflation by the CPI. Economists have estimated that the measurement error has contributed trillions to the federal deficit. The CPI is the benchmark for cost of living adjustments for social security payments, military pensions, and many other entitlements. (These problems are detailed in a previous column, here.)
The Federal Reserve itself recognized the problems with the CPI over twenty years ago, and replaced it with the PCE for purposes of setting monetary policy.
But the recent divergence is related to a more serious problem. It calls into question whether our conceptual understanding of inflation, and our techniques for measuring it, remain valid in a post-industrial, service-dominated and increasingly digital economy.
This intellectual reckoning is overdue. Inflation was “invented” as an economic concept when the economy was based on principally mass production of commoditized products. It “works” for assessing the cost of gasoline, say, or eggs. It works (to a point) for labor costs involving farm labor, hourly wages for “metal bending” jobs in a factory, or piece-work in the garment industry. But economists today struggle to apply it to services, to the compensation for knowledge work (e.g., doctors, chip designers, educators), to housing costs, and to products that embody high-tech features enabled by software, realtime data, and network connectivity.