How strongly do Nordic reservoir levels predict household price regimes?

Abstract

This working paper asks whether Nordic reservoir levels predict household-relevant electricity-price regimes. The preregistered target was a Pearson association involving an actual reservoir-level series. The public-safe frozen snapshot contains no such series. The primary result is therefore not a number: its value is null, its sample size is zero, its bootstrap interval is null, and its status is insufficient_evidence.

The snapshot does contain 462 available Nordic capture-rate rows, but capture rate is not reservoir level. Substituting it would change the physical quantity, estimand, and interpretation after the question had been registered. The analysis correctly refuses that substitution. Border flows, daily prices, outage events, generation mix, and zone load are also unable to stand in for stored hydrological reservoir levels without a separate proxy-validation protocol.

Accordingly, this paper does not estimate the strength, sign, timing, or predictive value of a reservoir-price relationship. It documents why the question is currently unanswered and what evidence would be required to answer it. Publishing an honest non-result protects the corpus from filler claims and makes absence of the necessary contract visible. This is not a null finding that reservoir levels have no relationship with prices; it is a finding that the relationship is not estimable from the frozen evidence. The paper is not peer reviewed and is not trading advice.

Plain-language answer

We do not know from this dataset. The required reservoir-level measurements are absent, so there are zero valid reservoir-price pairs and no correlation to report.

There are 462 Nordic capture-rate rows in the snapshot. Those rows measure a different concept. A capture rate generally concerns the relationship between a generation profile’s realized value and a market price benchmark; it does not state how full a reservoir is. Using those rows as though they were hydrological storage would produce a precise-looking answer to the wrong question.

The responsible household message is therefore simple: do not treat this paper as evidence that reservoir levels do or do not predict Nordic prices. A future answer needs a licensed, timestamped, consistently defined reservoir series aligned with bidding-zone prices on a prespecified decision clock. Until then, the status remains insufficient_evidence.

Research question

The registered question is: How strongly do Nordic reservoir levels predict household price regimes? Its central explanatory variable must be reservoir level, with a defined unit, geographic aggregation, release calendar, revision policy, and timestamp. Its outcome must likewise define “household price regime,” such as a prespecified price level, quantile, or state based on the local bidding-zone wholesale curve.

The frozen artifact labels the intended metric “Nordic reservoir-level association” with unit Pearson r. However, the value is null because no reservoir observations exist in the public-safe snapshot. A predictive claim would require more than contemporaneous correlation: it would need lagging reservoir information to a household decision time and evaluating out-of-sample price outcomes. No such pairs are available here.

This distinction prevents two common errors. First, missing evidence is not a zero effect. Second, a nearby variable is not automatically a valid proxy. Capture rate, hydropower generation, cross-border flow, precipitation, and price history may be related to hydrological conditions, but each measures something different and can be influenced by market decisions. None may be silently renamed “reservoir level.”

Data and provenance

The canonical public aggregate is /research-data/home-papers/how-strongly-do-nordic-reservoir-levels-predict-household-price-regimes.json. It records status insufficient_evidence, publication cutoff 2026-08-30T00:00:00Z, primary value null, sample size zero, and no bootstrap interval. It also reports one secondary inventory fact: 462 available Nordic capture-rate rows.

The corpus metadata lists daily prices from 2021-01-01 through 2026-08-29, detailed intervals from 2025-10-01, and long price history from 2015-01-01. None of these date windows creates reservoir data. Long price coverage cannot compensate for an absent explanatory series.

The registered family contracts are border_flows, day_ahead_prices, outage_events, generation_mix, and zone_load. Reservoir levels are not among the available public-safe tables. The implementation searches the existing Nordic capture records only to document what nearby data exist, then refuses to calculate a reservoir association from them.

Provenance binds the outcome to a SELECT-only transaction with a 180-second statement timeout. Snapshot SHA-256 is 7e97489fc8528c8cc8c38830e05b48d949ce1f67b98575dff26f5d7c321e4c67; analysis-code SHA-256 is 57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9; protocol SHA-256 is adb36bf6b447af9f96339249b8becaefc20422499cca1977242866347a97bd4b; paper-registry SHA-256 is 7bcb91d7476d0a69fe9fa75a5c7782f8117e0153f82f9112b7e1d307d3943717; and source-registry SHA-256 is 07949550ac443ff673fda5c0209b99f137544f3ffecf6775f109bb9d09663bd6.

The five source-registry references provide market and consumer context only. ENTSO-E describes SDAC; the European Commission documents 15-minute day-ahead trading; the IEA discusses grids and flexibility; ACER and CEER discuss consumer flexibility and contract choice; and Voltcast’s architecture describes the internal data plane. None is used to invent a reservoir time series or a numerical finding.

Method

The preregistered method family covers border-day panels, event studies, placebo windows, and association-only network models. For this paper, the required first step is a matched reservoir-price panel. Because the reservoir field is unavailable, the procedure stops before estimation.

The frozen implementation identifies Nordic capture rows whose zone codes begin with NO, SE, FI, or DK and whose capture rate is present. It counts 462 such rows as a secondary data-availability fact. It calculates no reservoir value from those rows and returns an outcome with null primary value, empty sample, and insufficient_evidence status.

That fail-closed behavior is methodologically important. A proxy choice made after discovering the required variable is missing would introduce researcher discretion and could answer a materially different question. For instance, generation output reflects dispatch and water use, while capture rate combines production timing and prices. Either may respond to the very price regime being predicted, producing reverse causality.

A future valid method would register the reservoir source before outcome analysis, document level or energy-content units, align publication timestamps to local market decisions, handle revisions and seasonal cycles, and define train and test periods. It would compare reservoir-based prediction with simple seasonal and price-persistence controls. Those are requirements for later work, not tests claimed in this artifact.

Results

The primary reservoir-level association has sample size zero, value null, and bootstrap interval null. There is no Pearson coefficient.

The only reported secondary quantity is 462 available Nordic capture-rate rows. They are explicitly not used as reservoir observations. Their existence does not move the paper from insufficient to measured evidence because they cannot identify reservoir fullness under the frozen contract.

The figure contains one label, “required reservoir series,” with a plotted value of 0. That zero visualizes the absent required series; it is not a zero-valued Pearson estimate. The canonical primary value remains null, the sample remains empty, and the status remains insufficient_evidence.

No sign, magnitude, statistical significance, forecast improvement, zone ranking, seasonality, household saving, or regime threshold is reported. The paper cannot say that reservoir levels strongly predict prices, weakly predict prices, or fail to predict prices. All three would be unsupported.

This outcome should not be described as a conventional null result. A null test would require valid observations and an estimate compatible with no association. Here there is no test. Insufficient_evidence is the complete result.

Robustness and placebo checks

No empirical robustness or placebo check can rescue an absent primary variable. Resampling an empty set, changing a correlation method, or adding price controls would not create reservoir observations.

The key robustness safeguard is refusal to substitute. The implementation preserves an empty primary sample even though 462 nearby Nordic capture records are available. It also avoids treating border flows, generation mix, load, or outage rows as hydrological storage. This protects construct validity.

For a future study, robustness should include alternative official reservoir aggregations registered in advance, seasonal demeaning, lag variations tied to publication clocks, zone exclusions, revision-vintage checks, and comparisons with naive seasonal and price-only models. Placebos could use future reservoir releases that were unavailable at the prediction time or non-hydrological zones, but only under a protocol that prevents lookahead. None of these has been run in the current artifact.

Multiplicity control is not applicable to a missing estimate. The family statement about Holm correction cannot turn a non-estimable question into evidence. The safest robustness conclusion remains that no unrelated proxy was substituted.

Limitations

The decisive limitation is absence of reservoir levels. The source contracts do not include the needed hydrological series, and no primary sample can be formed. This is not repaired by having a long price history.

“Nordic” itself would require a future aggregation rule. Norway, Sweden, Finland, and Denmark have different generation systems and bidding zones. Reservoir measures may be published at country, area, basin, or producer level, with differing units and calendars. A future panel must map those identities without pretending that one regional number represents every household zone.

“Household price regime” is also undefined in the current measured artifact. Wholesale daily mean prices are not final retail tariffs. Contracts may hedge or average prices and add taxes, supplier margins, and network charges. A predictive study should define whether it targets wholesale price states or actual tariff exposure.

Hydrological data can be seasonal, revised, and released after the period they describe. Using final historical values at an earlier forecast clock would create lookahead. Correlation would remain observational and could be confounded by weather, demand, generation availability, transmission, and market conditions.

Finally, the paper’s contextual references do not provide an authorized replacement series in the frozen snapshot. Public discussion of hydropower cannot be converted into row-level evidence. The result must remain incomplete.

Practical implication

Households, journalists, and product builders should not use this page as a reservoir-based price signal. The honest interface behavior is to show that reservoir evidence is unavailable, rather than fill the gap with capture rate or another convenient metric.

For future research, the priority is a data contract, not a more elaborate model. Secure and document a reservoir series with geographic identity, units, publication and revision clocks, licensing, and historical vintages. Then preregister the price-regime definition and benchmark. Until those steps are complete, no reservoir-derived automation or forecast claim is supported.

Reproducibility

An exact reproduction loads the frozen public-safe snapshot and checks for the required reservoir-level series. It finds none, creates no matched reservoir-price pairs, and returns status insufficient_evidence, sample size zero, null value, and null interval. Separately, it filters available capture rows to zone codes beginning NO, SE, FI, or DK with non-null capture rates and returns the inventory count 462.

Reproduction fails if it converts capture rate, generation, precipitation, price, or flow into an undeclared reservoir proxy. Adding a future reservoir dataset would be a new study with new provenance, not a rerun of this result. The public figure is a visual statement of the required missing series, not an empirical coefficient. Licensing is disclosed at /legal/data-licensing and the canonical licensing document.

Disclosure

Analysis and drafting were model-assisted. This working paper is not peer reviewed. It intentionally publishes insufficient evidence and does not replace the missing reservoir variable with another proxy. No causal, predictive, household-savings, or trading conclusion is made. Volt has no live traders or live capital. This is not trading advice, financial advice, or a recommendation for household energy scheduling.

References

• ENTSO-E — Single Day-ahead Coupling (SDAC)
• European Commission — EU electricity trading in the day-ahead markets becomes more dynamic
• International Energy Agency — Electricity 2026
• ACER and CEER — Rewarding flexibility: How retail contract choice can help unlock consumer flexibility
• Voltcast — Voltcast Architecture