Food Regime Index

Pillar 6 — Food. An 8-component ag/input pulse that nowcasts the IMF Food Price Index 3 months forward (in-sample corr +0.68, both halves stable). Updated .

Regime read ():

The 3-month forward food-CPI nowcast is the historical predictive use of the composite (in-sample corr +0.68, train-only OLS coefficients locked at end-2017; OOS residuals visible from 2018+).

Composite — 8-component pulse

Equal-weight z-score of YoY changes across 8 components — wheat / maize / soybeans / barley / palm oil / sunflower oil / rapeseed oil / fertilizer headline — using an expanding-window mean/std (min 24m). Phase bands at −0.5 / 0.5 / 1.5 mirror the Energy Regime convention. Higher z = more broad-based ag-input inflation.

Component breakdown — current z-scores

Each bar = one component's expanding-window z-score, signed so right-tail = inflation pressure. Sources: .

Forecast vs actual — train-only OLS

Cyan = actual food-index YoY at t+3. Gold dashed = train-only OLS forecast (coefficients fit on 2003 → 2017, applied throughout). Shaded band = 95% prediction interval (residual σ ≈ ). The forecast is informational, not a tradeable expression — food index is not a market.

Wheat — structural case study (locality)

The composite above operates on broad ag commodity prices. Underneath it, wheat is the deepest-data case study for the structural-locality channel — concentration, inventory, and Black Sea dynamics — because USDA PS&D coverage is most complete for wheat. The wheat-only locality composite is a separate annual indicator from the monthly broader nowcast above.

The wheat locality composite read at — note that locality NORMAL is consistent with the 2026 wheat tape moving on something other than concentration (weather, dollar denominator, restrictions cycle, fertilizer pass-through). Source: .

Sovereign stress — wheat-importer EM cohort

EM food-import cohort: largest wheat-net-importers with free-data reserves coverage. The reserve-adequacy floor of months follows IMF Assessing Reserve Adequacy guidance. Source: .

Active export restrictions

As of · active restriction(s) tracked · next review due

Operator-maintained snapshot reviewed against IFPRI Food and Fertilizer Trade Policy Tracker. Each entry cites a verifiable source URL and last-verified date. .

Fertilizer pass-through

Fertilizer cost is the second-order channel from energy and ammonia/potash supply concentration into food production. Nitrogen specifically (urea family) is the Russia/Belarus exposure; phosphate is the Belarus/Morocco/China exposure. Source: .

Animal-protein pass-through

Global feed-wheat share:

of world wheat consumption goes to animal feed. The peak-lag-correlation column above is the empirical wheat-to-meat transmission window for each protein — cattle 9m, poultry contemporaneous, hogs 6m.

Transmission chain — what's loaded, what hasn't passed through yet

Live: wheat $/MT (

% YoY) × broad dollar · fertilizer N PPI

% YoY · cattle PPI

% YoY. Source:

Methodology

The Food Regime Index has two layered indicators:

The pillar composite (this page's headline) is an 8-component equal-weight z-score of YoY changes in major ag commodity prices and fertilizer input cost. The composite is validated as a 3-month forward nowcast of the IMF Food Price Index, in-sample correlation +0.68 with both halves of the sample stable.
The wheat locality case study is a separate, narrower, annual composite (Russia/Ukraine/Kazakhstan share + exporter HHI + stocks-to-use) built on USDA PS&D wheat data. It captures the structural-locality channel where wheat data is deepest. Read in conjunction with the pillar composite, not in lieu of.

Composite construction

Component	Source
Wheat	FRED `PWHEAMTUSDM`
Maize	FRED `PMAIZMTUSDM`
Soybeans	FRED `PSOYBUSDM`
Barley	FRED `PBARLUSDM`
Palm oil	FRED `PPOILUSDM`
Sunflower oil	FRED `PSUNOUSDM`
Rapeseed oil	FRED `PROILUSDM`
Fertilizer (headline)	FRED `WPU0652`

Z-score convention. Each component's YoY % change is standardised with an expanding-window mean/std (minimum 24 months). Composite = simple average of the eight component z-scores. No IC-fitted weights (the composite_v7 line was killed precisely because in-sample IC weights had +0.42 sample bias — equal-weight is the deliberately-naive parameter-free combination).

Phase classification (matches Energy Regime convention so cross-pillar co-firing is comparable):

$z < -0.5$ → BENIGN
$-0.5 \le z < 0.5$ → NORMAL
$0.5 \le z < 1.5$ → ELEVATED
$z \ge 1.5$ → STRESS

3-month forecast. Train-only OLS regression of food-index YoY at $t+3$ on composite z at $t$, fit on 2003-01 → 2017-12 (n=). Coefficients $\alpha = $, $\beta = $. Residual $\sigma = $ — that's the source of the 95% CI band. Coefficients are locked at end-2017 and applied throughout (including OOS 2018+) so out-of-sample residuals are visible to the reader, not concealed by re-fitting.

What's in v1, what's in v2

v1 (this page):

Composite + 3m nowcast + component breakdown
Wheat locality case study (annual)
EM sovereign overlay, restrictions tracker, fertilizer pass-through, animal-protein pass-through

v2 candidates (deferred):

Trading sleeves on top of the composite — inflation_be_v0 pre-reg locked at docs/inflation_be_v0_pre_reg.md, not run; cattle_v0 documented in findings doc. Both gated on capital availability per launch sequencing.
CFTC COT positioning data on grains and cattle
Full Comtrade volume decomposition for fertilizer (Russia urea, Belarus potash export volumes)
Rice integration (PRICENPQUSDM only starts 2010 — adding it shortens the composite window; held until rice history matures or a deeper pre-2010 rice series surfaces)

Canon

Headey (2011, Food Policy) — wheat-to-meat pass-through; Trostle (2008, USDA ERS) — global commodity price transmission; Anderson & Nelgen (2012) — export-restriction cascade dynamics; Coibion & Gorodnichenko (2015) — household inflation expectations responsive to food more than energy; same expanding-window z-score discipline inherited from the BCI.

Headline inflation linkage

The food pulse was tested as a candidate input to the broader Inflation Nowcast composite (forward CPI YoY at 3-6m). The validation suite found pipeline-stage indicators (PPI YoY + Core PCE 6m annualized) dominated, and the food pulse was dropped from the final composite. The food pulse → food-CPI 3m relationship on this page (corr +0.68) remains the food-specific nowcast; the pipeline composite is the broader headline-CPI nowcast. They're complementary at different scopes.

inflation_be_v0 (pre-reg locked, not run) — cross-commodity-ratio signal targeting 10Y TIPS BE 12m forward; tradeable as TIP/IEF pair.

Inputs: FRED (8 IMF/BLS commodity series + fertilizer + protein PPI), USDA PS&D bulk download, World Bank reserves + import cover, operator-maintained IFPRI snapshot. Nightly rebuild. See methodology index for the full indicator manifest.