Government Reliance on Private Forecasting Models and Dependency Risks

Introduction

Governments increasingly rely on privately developed forecasting models for decisions involving economic growth, inflation, energy demand, electricity generation, climate risk, traffic, public-health demand, procurement volumes, disaster planning, credit risk, migration, defence logistics, and infrastructure investment.

The competition-law concern arises when a private forecasting provider moves from being an ordinary supplier of analytical services to becoming a critical informational intermediary. If government departments, regulators, public undertakings, or procurement authorities become dependent on one firm's proprietary forecasts, the provider may acquire significant information-based market power.

This creates a distinctive competition problem:

A government may formally retain decision-making authority while becoming practically dependent upon a privately controlled forecasting infrastructure.

The relevant risks include foreclosure of competing forecasting providers, discriminatory access to data, interoperability restrictions, exclusive contracts, self-preferencing, algorithmic lock-in, excessive switching costs, and the possibility that the model itself becomes an unavoidable input for downstream public-sector markets.

Importantly, the leading cases discussed below do not all concern government forecasting models specifically. They establish broader competition-law principles concerning indispensable inputs, data, interoperability, information advantages, exclusionary conduct, algorithmic coordination, and market access that can be applied by analogy.

1. What Is a Private Forecasting Model?

A forecasting model may combine:

• historical government data;
• proprietary datasets;
• satellite or sensor information;
• financial information;
• real-time market data;
• machine-learning models;
• proprietary assumptions;
• scenario-generation systems;
• probabilistic predictions;
• natural-language processing;
• digital twins;
• economic models; and
• continuously updated algorithms.

Examples include models predicting:

Economic

• GDP;
• inflation;
• unemployment;
• tax revenues;
• government expenditure.

Infrastructure

• electricity demand;
• transport congestion;
• water consumption;
• housing demand;
• telecommunications demand.

Public services

• hospital demand;
• school enrollment;
• emergency response requirements;
• social-security expenditure.

Climate and environmental

• flooding;
• wildfire;
• drought;
• emissions;
• renewable-generation output.

The competition issue becomes particularly serious when government institutions use one privately controlled model as a central reference point for procurement, regulation, investment or allocation decisions.

2. Why Government Dependency Can Create Competition Concerns

Government dependence can operate through several mechanisms.

A. Technical dependency

The government may build internal systems around:

• one API;
• one data format;
• one forecasting architecture;
• one model;
• one proprietary software environment.

Once integrated, replacing the provider becomes expensive.

B. Data dependency

A provider may possess datasets unavailable to competitors because of:

• exclusive licences;
• historical accumulation;
• proprietary collection systems;
• privileged commercial relationships;
• government contracts.

The forecasting advantage can therefore become self-reinforcing.

C. Institutional dependency

Once government departments have repeatedly relied upon a particular model, officials may develop:

• workflows;
• procurement templates;
• regulatory methodologies;
• dashboards;
• performance indicators;
• budget assumptions

around that model.

Switching then involves more than changing software. It may require changing the institution's decision-making architecture.

D. Reputation dependency

A model used by government can acquire a powerful reputational advantage:

government adoption → credibility → private-sector adoption → more data → improved model → stronger government dependence.

This can create a feedback loop favouring the incumbent.

3. Government Procurement Can Entrench a Private Forecasting Provider

Government procurement is particularly important because governments can represent enormous demand.

Suppose a government awards a long-term exclusive forecasting contract to Provider A.

Provider A subsequently becomes the reference provider for:

• energy forecasting;
• transport planning;
• infrastructure investment;
• insurance risk;
• public procurement.

Competing providers may then lose access to government-generated feedback, benchmarking opportunities and commercially important datasets.

The procurement decision can therefore have market-structuring effects beyond the immediate contract.

Competition law should distinguish between:

1. legitimate procurement based on quality and reliability; and
2. procurement arrangements that unnecessarily exclude competing suppliers.

4. Relevant Competition-Law Theories

Several doctrines can potentially apply.

4.1 Abuse of dominance

If the forecasting provider is dominant in a relevant market, conduct such as:

• exclusive dealing;
• discriminatory access;
• tying;
• refusal to supply;
• interoperability restrictions;
• predatory pricing;
• self-preferencing;
• leveraging into adjacent markets

may raise Article 102 TFEU or equivalent domestic competition-law concerns.

4.2 Essential-facility-type concerns

A forecasting model is not automatically an essential facility merely because government relies upon it.

The stronger case arises where:

• the provider is dominant;
• the model or underlying data is practically indispensable;
• duplication is technically or economically difficult;
• government procurement has made the model a de facto industry standard;
• access is objectively necessary for competing downstream businesses.

The classic essential-facilities jurisprudence therefore becomes relevant.

5. Case Law

Case 1 — Commercial Solvents v Commission

Cases: Commercial Solvents Corp and Istituto Chemioterapico Italiano v Commission, Joined Cases 6/73 and 7/73.

The Court recognised that a dominant undertaking controlling an upstream input could not use that position to eliminate competition in a downstream market.

Relevance

The forecasting analogy is:

proprietary data/model → forecasting service → downstream government/public markets.

If a forecasting provider controls an indispensable upstream informational input and then uses that position to favour its own downstream services, competition concerns can arise.

For example, a provider might supply forecasts to government but simultaneously compete for:

• infrastructure contracts;
• energy-management services;
• transport optimisation;
• insurance;
• public-sector analytics.

The concern is vertical leveraging.

6. Case 2 — United Brands v Commission

Case: United Brands Company and United Brands Continentaal BV v Commission, Case 27/76.

The case is a foundational authority on dominance and exclusionary conduct.

The Court considered factors including market power, barriers to entry and the ability of an undertaking to behave independently of competitors and customers.

Relevance

A forecasting company could acquire substantial market power where it possesses:

• unique datasets;
• superior historical information;
• significant computational infrastructure;
• network effects;
• reputation;
• government certification;
• switching-cost advantages.

Government dependence can consequently become evidence of the provider's commercial importance, although government use alone does not establish dominance.

7. Case 3 — Bronner v Mediaprint

Case: Oscar Bronner GmbH & Co. KG v Mediaprint Zeitungs und Zeitschriftenverlag GmbH & Co. KG, Case C-7/97.

The Court established a demanding test for imposing compulsory access to an infrastructure controlled by a dominant undertaking.

Important considerations included whether the facility was genuinely indispensable and whether duplication was realistically possible.

Relevance

This is highly useful for private forecasting infrastructure.

A government or competitor cannot simply argue:

"The government uses this forecasting model, therefore everyone must receive access."

Instead, the legal inquiry would ask whether the model or its underlying infrastructure is genuinely indispensable.

Relevant factors could include:

• availability of substitute forecasting models;
• cost of reproducing the model;
• availability of comparable datasets;
• technical interoperability;
• switching costs;
• time required to develop an alternative.

Thus, dependency must be demonstrated rather than assumed.

8. Case 4 — IMS Health v Commission

Case: IMS Health GmbH & Co. OHG v NDC Health GmbH & Co. KG, Case C-418/01.

The Court dealt with access to a proprietary information structure and the circumstances in which refusal to license intellectual property could constitute abusive conduct.

The case emphasised the exceptional circumstances required before competition law compels access to protected infrastructure or intellectual property.

Relevance to forecasting models

A private forecasting model may be protected through:

• copyright;
• trade secrets;
• database rights;
• contractual restrictions;
• proprietary software;
• confidential training datasets.

Government reliance does not automatically eliminate those rights.

However, if a particular proprietary forecasting infrastructure becomes indispensable to effective competition in a downstream market, IMS Health provides a framework for analysing whether exceptional circumstances exist.

9. Case 5 — Microsoft v Commission

Case: Microsoft Corp v Commission, Case T-201/04.

The General Court upheld findings concerning Microsoft's refusal to provide interoperability information and the use of technological control to restrict competition.

Relevance

This is particularly significant for modern AI forecasting systems.

Suppose a government procurement environment becomes dependent upon:

• Provider A's forecasting API;
• Provider A's proprietary data schema;
• Provider A's model outputs;
• Provider A's authentication system.

If competing forecasting providers cannot effectively interoperate with the government infrastructure, technical compatibility can become a competitive bottleneck.

Potential concerns include:

• closed APIs;
• proprietary formats;
• data-export restrictions;
• incompatible interfaces;
• contractual prohibitions on portability.

The broader lesson is that interoperability can be a competition variable, not merely a technical matter.

10. Case 6 — Slovak Telekom v Commission

Case: Slovak Telekom a.s. v European Commission, Joined Cases C-165/19 P and C-166/19 P.

The Court considered exclusionary conduct involving access to infrastructure controlled by a dominant undertaking.

Relevance

The case illustrates how an infrastructure owner can potentially use control over an important input to disadvantage downstream competitors.

Applied to forecasting:

forecasting infrastructure → downstream analytical services

could produce concerns where the provider:

• supplies its own downstream services;
• controls access to the underlying data;
• offers competitors inferior access;
• imposes discriminatory technical conditions;
• uses contractual restrictions to prevent multi-sourcing.

The more government procurement makes the forecasting infrastructure unavoidable, the more important access conditions become.

11. Case 7 — Google Shopping

Case: Google and Alphabet v Commission, Case T-612/17.

The EU courts considered Google's conduct concerning comparison-shopping services and the treatment of its own service within its broader search infrastructure.

Relevance

The broader principle is particularly useful for forecasting ecosystems:

A company controlling an important intermediary infrastructure may potentially favour its own downstream service.

Imagine a private forecasting provider supplying government infrastructure while also offering:

• infrastructure investment advice;
• energy trading;
• insurance analytics;
• procurement optimisation.

If the provider's forecasting system systematically gives preferential treatment to its own downstream businesses, the situation could raise self-preferencing or leveraging concerns, depending on the applicable legal framework and evidence.

12. Case 8 — Eturas

Case: Eturas UAB and Others, Case C-74/14.

The Court considered the competition-law implications of a common electronic platform through which competing businesses received information capable of coordinating their commercial behaviour.

Relevance

This case is important because forecasting platforms can become information infrastructures.

A forecasting platform used by multiple firms could expose:

• expected demand;
• expected prices;
• capacity;
• market forecasts;
• future commercial strategies.

If competitors receive commercially sensitive predictive information through the same platform, the system could facilitate coordination.

Thus, the competition problem is not limited to government dependency.

There can also be:

government forecasting infrastructure → private-sector information dissemination → reduced competitive uncertainty.

13. Case 9 — T-Mobile Netherlands

Case: T-Mobile Netherlands BV and Others v Raad van bestuur van de Nederlandse Mededingingsautoriteit, Case C-8/08.

The Court examined exchanges of information between competitors and the circumstances in which such exchanges could restrict competition by reducing strategic uncertainty.

Relevance

Forecasts are inherently forward-looking.

A forecasting system that distributes information concerning expected:

• prices;
• capacity;
• demand;
• production;
• investment;

can have a particularly strong effect on competitive uncertainty.

Government-sponsored or government-procured forecasting platforms therefore require careful governance where competing private firms contribute information to the system.

14. Case 10 — Hoffmann-La Roche

Case: Hoffmann-La Roche & Co AG v Commission, Case 85/76.

This is a foundational dominance case concerning exclusionary conduct and loyalty-inducing arrangements.

Relevance

A dominant forecasting provider could potentially use:

• exclusivity clauses;
• loyalty discounts;
• bundled forecasting/data services;
• long-term contracts;
• minimum-purchase obligations

to prevent government bodies or major institutional customers from using competing forecasting providers.

The competition question is whether contractual arrangements foreclose rivals beyond what is justified by legitimate efficiency considerations.

15. The Special Problem of Government as a "Demand Anchor"

Government is different from an ordinary customer.

A government contract can have effects throughout an industry.

Example

Assume Government Department X announces:

"All infrastructure planning will use Provider A's forecast."

Provider A now obtains:

1. government revenue;
2. reputational legitimacy;
3. industry-standard status;
4. additional private customers;
5. additional data;
6. additional model-training opportunities.

Competitors may consequently lose:

• customers;
• data;
• benchmarking opportunities;
• credibility;
• scale.

This can create a demand-anchor effect.

Government purchasing power therefore has the potential to influence market structure even without the government intending to exclude competitors.

16. The "Forecasting Lock-In" Cycle

A particularly important modern risk is a feedback loop:

Government adopts Model A
↓
Model A becomes institutional standard
↓
More organisations adopt Model A
↓
Provider obtains more data
↓
Model A becomes more accurate or commercially valuable
↓
Competitors face reduced scale and data
↓
Switching becomes more difficult
↓
Government becomes even more dependent

This can create a self-reinforcing informational ecosystem.

17. Data Advantages

Forecasting competition is often really a competition over data.

A government contract may give the incumbent access to:

• procurement data;
• infrastructure usage;
• transport flows;
• energy consumption;
• demographic information;
• environmental measurements;
• public-service utilisation.

If those data are subsequently combined with proprietary datasets, the provider may acquire an informational advantage that competitors cannot readily reproduce.

Competition authorities should therefore distinguish between:

Legitimate advantage

Better forecasting because of superior innovation and investment.

Potentially exclusionary advantage

A contractual or technical arrangement prevents competitors from accessing information necessary to compete effectively.

18. Model Opacity and Competition

Machine-learning forecasting models introduce another problem: explainability.

A government may become unable to independently reproduce why a particular forecast was generated.

This creates a form of institutional dependency:

Decision authority remains with government, but epistemic capacity moves to the private provider.

Competition concerns may arise where the provider controls:

• model architecture;
• training data;
• validation procedures;
• error measurements;
• model updates;
• confidence intervals;
• historical revisions.

A government that cannot independently audit or reproduce forecasts may have greater difficulty switching suppliers.

19. Model Updates as a Lock-In Mechanism

A particularly subtle risk occurs when the provider continuously changes the model.

For example:

Year 1: Version A
Year 2: Version B
Year 3: Version C
Year 4: AI-enhanced Version D.

If historical government decisions were based on previous versions, switching providers may make longitudinal comparisons difficult.

The government may therefore become dependent upon the provider's:

• historical datasets;
• version history;
• calibration methodology;
• archived predictions;
• error metrics.

This produces historical lock-in in addition to technical lock-in.

20. Forecasting Models as De Facto Standards

A forecasting model can become a de facto standard even without formal governmental designation.

For example, banks, insurers, infrastructure investors and government agencies may all use the same provider.

Eventually:

"What does the leading model predict?"

may become more important than competing independent forecasts.

This can produce standardisation effects.

The competition question is whether standardisation results from:

• superior performance;
• legitimate network effects; or
• exclusionary conduct preventing alternative models from competing.

21. Risks of Exclusive Government Contracts

Long-term exclusive contracts can create several problems.

Foreclosure

Rivals cannot obtain sufficient scale.

Entry barriers

New firms cannot develop a viable customer base.

Innovation reduction

Alternative forecasting methodologies may receive less investment.

Switching costs

Government systems become dependent on one vendor.

Vendor leverage

Renegotiation becomes difficult once migration costs become substantial.

Information asymmetry

The incumbent knows that the customer cannot easily switch.

Therefore, procurement design itself can become relevant to competition analysis.

22. Competition Issues in Government Procurement

A competition-sensitive procurement framework should consider:

None of these arrangements is automatically unlawful. Their legality depends upon market power, effects, justification and applicable procurement/competition rules.

23. Public-Interest and Efficiency Defences

Government reliance may be entirely legitimate where the provider offers:

• substantially greater accuracy;
• lower cost;
• superior reliability;
• better cybersecurity;
• specialised expertise;
• unique technology;
• demonstrably better disaster forecasting.

Competition law should therefore not require governments to use inferior forecasts merely to preserve competitors.

The appropriate question is whether the government's procurement decision or the provider's conduct produces unnecessary exclusionary effects.

24. Remedies

Potential safeguards include:

1. Multi-vendor procurement

Government should avoid unnecessary dependence upon one forecasting provider.

2. Data portability

Government-owned or government-generated data should be exportable in interoperable formats.

3. API interoperability

Contracts can require reasonable technical interoperability.

4. Exit provisions

Procurement contracts can provide:

• migration assistance;
• historical-data transfer;
• model documentation;
• archival access.

5. Independent validation

A second forecasting provider can periodically benchmark the incumbent.

6. Model audit rights

Government should be able to test:

• accuracy;
• bias;
• model drift;
• methodology;
• version changes.

7. Forecast diversity

For high-impact decisions, governments can maintain multiple independent forecasts rather than treating one private model as authoritative.

25. Competition-Law Analytical Framework

A regulator examining government reliance on a private forecasting model could proceed through the following questions:

Step 1 — Define the market

Is the relevant market:

• general forecasting services;
• specialised energy forecasting;
• government forecasting;
• climate-risk modelling;
• AI forecasting infrastructure;
• forecasting data?

Step 2 — Assess market power

Examine:

• market share;
• entry barriers;
• data advantages;
• switching costs;
• government adoption;
• network effects;
• technological superiority.

Step 3 — Identify the dependency

What exactly does government depend upon?

• model;
• data;
• API;
• software;
• methodology;
• historical forecasts;
• technical infrastructure?

Step 4 — Examine exclusion

Has the provider:

• excluded rivals;
• imposed exclusivity;
• denied interoperability;
• restricted data access;
• tied products;
• discriminated between users?

Step 5 — Examine downstream effects

Does the conduct affect:

• infrastructure markets;
• energy;
• transport;
• finance;
• insurance;
• public procurement;
• other forecasting services?

Step 6 — Consider objective justification

Are restrictions genuinely necessary for:

• security;
• reliability;
• confidentiality;
• intellectual-property protection;
• system integrity?

Step 7 — Evaluate remedies

Can competition be preserved without compromising legitimate governmental objectives?

26. Key Legal Principle

The most important distinction is between dependence created by superior performance and dependence created by exclusion.

A government becoming dependent on a private model because the model is demonstrably better is not, by itself, a competition-law violation.

The competition concern becomes stronger where the provider uses:

control over forecasting → control over information → control over government procurement → foreclosure of competing forecasting or downstream markets.

The cases such as Bronner, IMS Health, Microsoft, Slovak Telekom, Commercial Solvents and Google Shopping provide useful doctrinal analogies for analysing these circumstances.

27. Overall Conclusion

Government reliance on private forecasting models represents a developing form of informational and infrastructural dependency.

The central competition-law risk is not simply that government buys a private forecast. It is that repeated procurement, technical integration, data accumulation, institutional standardisation and reputational effects may transform a private forecasting provider into a de facto critical information infrastructure.

The most significant risks are:

1. vendor lock-in;
2. exclusive government procurement;
3. data accumulation;
4. interoperability restrictions;
5. foreclosure of rival forecasting providers;
6. vertical leveraging into downstream markets;
7. self-preferencing;
8. algorithmic information concentration;
9. reduced competitive diversity of forecasts; and
10. institutional dependence on a privately controlled epistemic infrastructure.

The emerging legal question can therefore be expressed as:

When does government procurement of a private forecasting service cease to be merely a purchasing decision and begin to shape the competitive structure of an entire information market?