The 100% threshold

The textbook link between money supply and inflation has held up badly in recent decades, and the usual response is to argue about whether the relationship exists at all. The question I asked instead was whether it is conditional: whether the fiscal position of the government determines if money growth reaches consumer prices.

The candidate condition is federal debt at 100% of GDP, a level the United States crossed around 2013 and now sits well above, at roughly 122%. That gives two subsamples of real data rather than a hypothetical, which is what makes the question testable.

1,891 weekly observations from the Federal Reserve Economic Database, January 1990 through March 2026: headline CPI, M2, federal debt, energy prices, unemployment, and inflation expectations from both consumer surveys and TIPS breakevens.

The sample is split at 100% debt to GDP and the same specification is estimated separately on each side, with M2 lagged 78 weeks to allow for transmission delay. Alongside the inflation model I estimated a Federal Reserve reaction function, to see whether the Fed's own behavior changes across the same boundary, and ran Monte Carlo simulations with correlated input draws to price a monetary shock under each regime.

A single split invites the objection that the result is an artifact of where the line was drawn, so the analysis carries 17 robustness checks across alternative specifications in a separate notebook.

The M2 coefficient does not weaken across the threshold, it changes sign. In the low-debt regime from 1990 to 2013 it is −0.129, and in the high-debt regime from 2013 to 2026 it is +0.104, both significant at p < 0.001.

Inflation expectations pass through to prices 4.6 times more strongly once debt is above the threshold, 1.623 against 0.354. Expectations are the channel that changes most, which is the part with the clearest implication for policy: credibility does more work under high debt than it does under low debt.

The reaction function tells a matching story from the Fed's side. A Taylor Rule explains 4.5% of the variation in its decisions in the low-debt period and 51.1% in the high-debt period, with the Taylor gap coefficient five times larger and the mandate-tension coefficient roughly doubled.

Simulating a 20% surge in M2 produces 0.54% inflation under the low-debt regime and 4.26% under the high-debt regime: a gap of 3.72 percentage points from the same shock, which is the practical size of the effect.

Two Jupyter notebooks, one for the passthrough models and one for the robustness checks, over a small Python layer that handles the data: pandas, NumPy, statsmodels, scipy, and matplotlib, with the FRED fetching and scoring factored out into modules so the notebooks stay about the economics.

It reuses my own tooling rather than restating it. The series come in through the same loader work described in FRED and Census loaders, and otter is vendored in for the simulation machinery.

Read the full write-up on Substack · Code and notebooks on GitHub · Back to all work