Climate, Grids, Balance Sheets

Days with high GFS-ECMWF ensemble disagreement over the Baltic coast show significantly fatter intraday-minus-day-ahead spread tails in the LT zone. Forecast dispersion, not the forecast itself, is the tradable state variable.

Test · Eighteen months of ENTSO-E LT day-ahead and imbalance prices against archived Open-Meteo multi-model wind forecasts; quantile regression of spread tails on dispersion; fix the spike threshold before looking.

The February 2025 synchronization to Continental Europe produced a structural break in LT balancing costs and in LT price coupling with SE4 and Poland — detectable, datable, and attributable.

Test · Regime-switching model and event study across the desync date on coupling coefficients and balancing-energy volumes; falsified if no break survives controls for fuel prices and interconnector outages.

The Baltic offshore buildout will raise, not lower, price volatility for years — capacity is arriving faster than the transmission and storage that would absorb it, so cannibalization and negative-price hours lead the smoothing.

Test · Cross-sectional check against zones further along the same curve (Denmark, northern Germany): volatility and negative-price frequency vs wind penetration, controlling for interconnection ratio.

In hours when gas sets the uniform clearing price, wind owners collect inframarginal rents financed through the bills of households in some of the EU’s most energy-poor member states — and the post-BRELL security premium is likewise socialized onto ratepayers by convention. The rent transfer is calculable, and it lands on the inequality frame directly.

Test · Reconstruct hourly inframarginal rents for LT wind from ENTSO-E prices and a merit-order proxy; join spike frequency to EU-SILC energy-poverty indicators across the Baltics; headline number is euros transferred per household per year under the clearing convention.

ENSO indices lead PLD regime shifts by one to two quarters through the inflow channel; a parsimonious reservoir-state model beats naive persistence in forecasting PLD regime, and the lead is long enough to act on.

Test · NOAA ONI vs ONS reservoir levels vs monthly PLD, 2001-2026; Markov regime model with ENSO as exogenous driver; falsified if ENSO adds nothing beyond current reservoir level.

Listed Brazilian generators and utilities underreact to reservoir-state changes — the physical state variable leads earnings revisions and returns, because the hydrological balance sheet updates faster than the analyst one.

Test · Event study on ADRs (Eletrobras EBR, Cemig CIG, Copel ELP, Sabesp SBS) around large reservoir-state moves; long-short conditioned on hydrology vs sector benchmark; pre-registered horizon.

PLD deviates from a physically grounded shadow price in systematic, directional ways attributable to the model’s own conventions — and each convention change (the 2013 CVaR introduction, the 2021 move to hourly PLD) measurably redistributed money between market segments.

Test · Reconstruct a simple physical benchmark price from reservoir and load data; regress the PLD-benchmark gap on convention-change dates; the redistribution estimate is the paper’s headline number.

The 2021 drought’s costs were distributed by convention, not by weather: tariff flags — including the special water-scarcity flag — moved scarcity costs onto household bills, regressively, since electricity is a far larger budget share for poor households, while the CVaR convention cushioned generator margins. The incidence of ENSO is an institutional choice.

Test · ANEEL tariff-flag history joined to IBGE POF electricity budget shares by income decile; compute the 2021-22 scarcity burden as a share of income per decile against a counterfactual where generators bear the hydrological risk.

CAISO solar and net-load forecast error is conditionally predictable from marine-layer indicators, and spike-day probability with it — the grid inherits the forecast skill ceiling of coastal stratus.

Test · CAISO forecast-vs-actual archives against coastal inversion and cloud-cover data; does a marine-layer index improve spike-day classification over temperature alone?

Scarcity events cluster in multi-day heat-dome regimes; capacity and RA constructs underprice that clustering, so the realized distribution of consecutive-day scarcity beats the independence assumption by a measurable factor.

Test · Historical CAISO / NYISO scarcity hours fitted as a regime process vs independent draws; the likelihood ratio is the underpricing estimate.

NYISO Zone J price separation is predictable on sea-breeze-failure days — when the marine cooling that the load forecast implicitly assumes does not arrive.

Test · Zone J vs rest-of-state spreads conditioned on sea-breeze onset from coastal station data; effect must survive controlling for raw temperature.

A portfolio long merchant generators, short regulated utilities, conditioned on regional scarcity state, outperforms — the market prices the sector, not the clock-speed difference within it.

Test · Backtest on US names with CAISO / ERCOT scarcity states; then out-of-sample structurally in Brazil with the hydro state; pre-registered rules, walk-forward only.

Insurers and utilities with concentrated exposure to correlated climate tails — wildfire, drought, heat-dome outage clusters — systematically underprice them until an event forces repricing. There is a screenable "next MBIA" in the intersection of Lane II’s drought states and Lane III’s correlation error.

Test · Build the exposure screen from regulatory filings and physical state data; the falsifiable form is that screened names show asymmetric drawdown behavior around climate events versus unscreened peers.