Econometric Robustness Requirements for Antitrust Claims

1. Introduction

Econometric robustness refers to the requirement that economic and statistical analysis relied upon in an antitrust case must be sufficiently reliable, transparent, reproducible, and capable of distinguishing the alleged anticompetitive effect from legitimate market forces.

Modern competition-law disputes increasingly depend on econometric evidence. Courts and competition authorities may be asked to determine whether:

prices were artificially increased;

output was restricted;

a cartel caused overcharges;

a dominant undertaking engaged in exclusionary conduct;

a merger substantially reduced competition;

predatory pricing occurred;

a platform's conduct caused foreclosure;

customers suffered measurable harm;

an alleged practice caused the claimed loss rather than unrelated economic factors.

Econometric evidence may include regression analysis, event studies, difference-in-differences models, price-concentration analysis, demand estimation, critical-loss analysis, panel-data analysis, synthetic controls, before-and-after comparisons, benchmark comparisons, and damages models.

The central principle is that an econometric model does not become legally persuasive merely because it produces a statistically significant result. The methodology must correspond to the legal theory of harm, use appropriate data, control for relevant variables, and withstand reasonable alternative explanations.

2. Meaning of Econometric Robustness in Antitrust

Econometric robustness has several dimensions.

A. Conceptual robustness

The economic model must actually test the theory of competition harm being alleged.

For example, if the allegation is that a dominant platform excluded competitors through discriminatory ranking, a regression showing that the defendant's prices increased is not necessarily relevant. The model must connect the conduct to foreclosure, reduced competition, increased prices, reduced quality, reduced innovation, or another legally cognizable harm.

B. Statistical robustness

The result should not disappear merely because a reasonable alternative specification is used.

An expert should ordinarily consider:

different functional forms;

alternative control variables;

different time periods;

alternative geographic markets;

alternative samples;

alternative estimation techniques;

treatment of outliers;

clustering of standard errors;

serial correlation;

heteroskedasticity;

endogeneity;

missing observations.

C. Data robustness

The underlying data must be:

sufficiently complete;

accurately measured;

representative;

relevant to the market being investigated;

appropriately cleaned;

temporally appropriate;

capable of supporting the inference being made.

D. Causal robustness

Correlation is not necessarily causation.

An increase in prices after a merger, for example, does not automatically prove that the merger caused the increase. Costs, demand, inflation, supply shocks, regulation, exchange rates, capacity constraints, or technological changes may have produced the same result.

E. Legal robustness

The economic analysis must correspond to the applicable legal test.

A statistically significant effect may nevertheless be insufficient if it does not establish an element required under the relevant competition statute.

3. Why Econometric Evidence Matters in Antitrust

Competition law frequently deals with counterfactual questions.

The court may effectively have to ask:

What would the market have looked like if the allegedly anticompetitive conduct had not occurred?

This is the counterfactual problem.

For example:

Observed world

Defendant engages in alleged exclusionary conduct → price = ₹120

Counterfactual world

Defendant does not engage in alleged exclusionary conduct → hypothetical competitive price = ₹100

The alleged antitrust overcharge would therefore be:

₹120 − ₹100 = ₹20

But determining the hypothetical ₹100 price is often extremely difficult.

Econometric methods are designed to estimate this counterfactual.

4. Major Econometric Requirements

I. Clear Identification of the Theory of Harm

The first requirement is identification of the precise economic mechanism.

An expert should specify:

What conduct occurred?

Why could that conduct harm competition?

Which market mechanism transmits the harm?

Which economic variable should change?

What counterfactual is being estimated?

What evidence would falsify the hypothesis?

For example:

Conduct: exclusive dealing

→ mechanism: foreclosure of rival distribution channels

→ effect: reduced rival access

→ competitive consequence: increased market power

→ ultimate effect: higher prices/reduced output/quality/innovation.

A regression that jumps directly from "exclusive contract" to "higher prices" without establishing the intermediate mechanism may be economically weak.

5. Appropriate Market Definition

Econometric robustness also depends upon defining the relevant market correctly.

Market definition may involve:

product substitutability;

geographic substitutability;

cross-price elasticities;

diversion ratios;

switching behaviour;

customer preferences;

supply substitution.

Econometric techniques can be used to estimate whether consumers substitute between products.

For example:

If Product A's price increases by 10% and demand for Product B increases substantially, this may indicate meaningful substitution between A and B.

But market definition cannot always be reduced to a single regression coefficient.

The economic evidence must be integrated with:

documentary evidence;

business records;

customer evidence;

industry characteristics;

internal strategy documents;

competitive constraints.

6. Control Variables

One of the most important requirements is controlling for variables that independently affect the outcome.

Suppose an antitrust claimant argues:

"Prices increased after the defendant's conduct, therefore the conduct caused the increase."

This is potentially defective.

Prices might simultaneously have increased because of:

input costs;

fuel prices;

wages;

inflation;

exchange rates;

taxation;

supply shortages;

transportation costs;

seasonal demand;

regulatory changes.

A simplified regression might be:

Pit=α+βAit+γXit+ϵitP_{it}=\alpha+\beta A_{it}+\gamma X_{it}+\epsilon_{it}

where:

PitP_{it} = price;

AitA_{it} = alleged anticompetitive conduct;

XitX_{it} = control variables;

β\beta = estimated effect of the conduct.

The credibility of β\beta depends heavily on whether important confounding variables have been properly addressed.

7. Statistical Significance Is Not Enough

A recurring mistake is treating statistical significance as equivalent to legal proof.

Suppose:

β=0.04\beta=0.04

with a p-value below 0.05.

That establishes evidence against a particular null hypothesis under the assumptions of the model.

It does not automatically establish:

substantial market power;

causation;

anticompetitive intent;

consumer harm;

damages;

exclusion;

monopoly maintenance.

Courts therefore need to distinguish:

Statistical significance

Whether an estimated effect is distinguishable from zero under the model.

Economic significance

Whether the magnitude of the effect is sufficiently large to matter competitively.

Legal significance

Whether the evidence satisfies the applicable legal standard.

These are three different questions.

8. Confidence Intervals and Precision

Experts should ordinarily provide confidence intervals rather than merely point estimates.

For example:

Estimated overcharge = 12%

but the 95% confidence interval is:

2%–22%.

That result is considerably less precise than:

Estimated overcharge = 12%, 95% confidence interval = 11%–13%.

Courts should therefore examine both:

magnitude; and

uncertainty.

A large estimate with enormous uncertainty may be less persuasive than a smaller but highly precise estimate.

9. Sensitivity Analysis

A robust econometric analysis should test whether the principal conclusion survives reasonable methodological changes.

For example:

The consistency of these estimates strengthens the inference.

By contrast:

would raise serious robustness concerns.

10. Counterfactual Construction

Antitrust damages frequently require estimating a counterfactual.

Common approaches include:

A. Before-and-after analysis

Compare prices before the conduct with prices after the conduct.

Problem:

Other market changes may coincide with the conduct.

B. Yardstick or comparator analysis

Compare the affected market with a similar unaffected market.

C. Difference-in-differences

Compare changes over time between:

treated units; and

control units.

A simplified model is:

Yit=α+β(Treatedi×Postt)+γi+δt+ϵitY_{it}=\alpha+\beta(Treated_i\times Post_t)+\gamma_i+\delta_t+\epsilon_{it}

where β\beta represents the estimated treatment effect.

D. Event study

Examine market outcomes around a particular event.

E. Structural demand estimation

Estimate consumer demand and simulate the competitive counterfactual.

F. Synthetic control

Construct a weighted combination of other markets or firms that approximates the affected market before the alleged conduct.

Each methodology has different assumptions and limitations.

11. Difference-in-Differences and Parallel Trends

Difference-in-differences is particularly useful but depends upon assumptions.

The important assumption is generally that, absent the alleged conduct, the treatment and control groups would have followed sufficiently comparable trends.

If the groups already had radically different trajectories before the conduct, the resulting estimate may be unreliable.

Therefore, experts should examine:

pre-treatment trends;

placebo interventions;

alternative control groups;

alternative treatment periods;

dynamic effects.

12. Endogeneity

Endogeneity is one of the most serious problems in antitrust econometrics.

Suppose a firm charges higher prices in areas where competition is weak.

A regression may show:

High concentration → high prices.

But the causal direction could be:

High expected demand → entry → concentration changes → prices change.

Or firms may deliberately enter markets where they anticipate higher prices.

Thus:

Correlation≠CausationCorrelation \neq Causation

Potential responses include:

instrumental variables;

fixed effects;

natural experiments;

lagged variables;

difference-in-differences;

structural models.

The chosen method must itself be justified.

13. Fixed Effects

Panel-data antitrust studies often use firm, geographic, product, or time fixed effects.

For example:

Pit=αi+δt+βAit+ϵitP_{it}=\alpha_i+\delta_t+\beta A_{it}+\epsilon_{it}

where:

αi\alpha_i controls for unit-specific characteristics;

δt\delta_t controls for common time shocks.

Fixed effects can remove certain forms of omitted-variable bias.

But they are not a universal solution.

An expert must explain:

what variation remains after fixed effects;

whether the alleged conduct itself varies sufficiently;

whether important variables are absorbed;

whether the fixed-effects structure matches the economic question.

14. Serial Correlation and Clustering

Antitrust datasets frequently involve observations across:

firms;

products;

geographic markets;

months;

years.

Observations may not be statistically independent.

Ignoring serial correlation or clustering can produce artificially small standard errors and exaggerated statistical significance.

Therefore, experts should consider appropriate:

clustered standard errors;

robust standard errors;

panel-data techniques;

autocorrelation structures.

15. Outliers

A small number of observations can sometimes determine the result of an antitrust regression.

Experts should therefore examine:

leverage points;

influential observations;

unusual transactions;

extraordinary market events;

data-entry errors.

However, removing inconvenient observations merely because they weaken the claimant's theory is itself problematic.

The exclusion criterion should be objective and economically justified.

16. Multiple Testing and Specification Searching

Another major concern is specification shopping.

If an expert runs hundreds of regressions, some may produce statistically significant results merely by chance.

This creates the risk of:

"Finding" an antitrust effect after searching through many specifications.

Robust analysis should therefore disclose:

alternative specifications;

model-selection decisions;

pre-existing hypotheses;

excluded variables;

robustness checks.

Transparency becomes especially important when the final model is substantially different from the initial analytical approach.

17. Econometric Robustness and Expert Evidence

Courts must also consider whether the methodology used by an expert satisfies applicable evidentiary requirements.

In the United States, Daubert and related cases have significantly influenced judicial scrutiny of expert economic evidence.

The court may consider matters such as:

testability;

peer review;

known error rates;

methodological standards;

general acceptance.

Antitrust litigation therefore creates an intersection between:

competition law + economics + statistical methodology + evidence law.

18. Case Law

1. Matsushita Electric Industrial Co. v. Zenith Radio Corp., 475 U.S. 574 (1986)

This is a foundational antitrust decision concerning economic plausibility.

The U.S. Supreme Court emphasized that courts should consider whether the alleged conspiracy makes economic sense and whether the evidence supports the inference of anticompetitive conduct.

Importance for econometric robustness

Economic evidence cannot be considered in isolation from the underlying economics of the market.

Where an alleged conspiracy would be economically irrational or implausible, weak statistical evidence may not be sufficient to sustain the claim.

Principle

Econometric results must fit a plausible economic theory of anticompetitive conduct.

19. Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209 (1993)

Brooke Group is central to predatory-pricing analysis.

The Supreme Court required consideration of:

pricing below an appropriate measure of cost; and

a reasonable prospect of recoupment.

Econometric significance

A claimant cannot establish predatory pricing merely by showing that prices were low.

The analysis must examine:

relevant costs;

pricing;

competitive conditions;

recoupment;

market structure.

Principle

The economic model must address every economically necessary component of the legal theory.

20. Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993)

Although Daubert was not itself an antitrust case, it is extremely important for antitrust economic experts.

The Supreme Court established a framework for assessing scientific expert testimony.

Relevant considerations include:

whether the theory can be tested;

whether it has been subjected to peer review;

potential error rates;

methodological standards;

acceptance in the relevant scientific community.

Antitrust significance

Econometric testimony cannot become reliable simply because an economist presents it in court.

The methodology itself must be sufficiently reliable.

21. Concord Boat Corp. v. Brunswick Corp., 207 F.3d 1039 (8th Cir. 2000)

This case is important concerning econometric evidence and exclusionary-contract theories.

The court scrutinized the economic evidence offered to establish competitive harm arising from allegedly restrictive arrangements.

Significance

The case illustrates that an expert must connect statistical or economic evidence to the actual mechanism through which competition allegedly has been harmed.

Evidence that merely establishes correlation or a difference in market outcomes may be insufficient.

Principle

An econometric model must explain the competitive mechanism rather than simply identify an association.

22. Comcast Corp. v. Behrend, 569 U.S. 27 (2013)

Comcast is one of the most important modern cases concerning antitrust damages models.

The Supreme Court rejected a damages methodology that did not correspond adequately with the theory of antitrust injury that remained in the case.

The Court emphasized the need for a damages model capable of measuring damages attributable to the specific theory of liability.

Econometric significance

This is a crucial robustness requirement:

The damages model must match the theory of antitrust injury.

An expert cannot construct a general price-impact model and assume that it measures damages caused by every alleged form of anticompetitive conduct.

23. In re High Fructose Corn Syrup Antitrust Litigation, 295 F.3d 651 (7th Cir. 2002)

This litigation involved alleged price fixing and extensive economic evidence.

The Seventh Circuit emphasized the importance of examining the total evidentiary record, including economic evidence, when determining whether coordinated conduct can reasonably be inferred.

Econometric significance

Parallel pricing alone does not necessarily establish a cartel.

Economic evidence becomes more persuasive when combined with:

communications;

suspicious conduct;

market structure;

opportunities for coordination;

conduct inconsistent with independent competition.

Principle

Econometric evidence should generally be integrated with documentary and circumstantial evidence rather than treated as an autonomous proof of conspiracy.

24. United States v. Microsoft Corp., 253 F.3d 34 (D.C. Cir. 2001)

The Microsoft litigation is particularly important for understanding economic analysis of exclusionary conduct in technology markets.

The court considered Microsoft's contractual and technological practices and their effect on competing technologies and distribution channels.

Econometric significance

The case demonstrates the importance of examining:

foreclosure;

network effects;

barriers to entry;

technological advantages;

distribution arrangements;

consumer effects.

In digital markets, conventional price-based econometric analysis may be inadequate because competitive harm can manifest through:

reduced innovation;

reduced access;

degraded interoperability;

foreclosure of emerging technologies.

25. FTC v. Indiana Federation of Dentists, 476 U.S. 447 (1986)

The Supreme Court considered a restraint involving dentists' refusal to provide insurers with dental X-rays.

The case is particularly significant for the treatment of quality and information effects.

Econometric significance

Competition is not limited to price.

An econometric analysis may therefore need to consider:

quality;

service;

information;

output;

consumer choice.

Principle

An antitrust model should measure the competitive variable actually affected by the challenged conduct.

26. Ohio v. American Express Co., 585 U.S. 529 (2018)

This case is highly relevant to modern platform economics.

The Supreme Court considered the two-sided nature of the credit-card platform and emphasized the importance of considering both sides of the platform in assessing competitive effects.

Econometric significance

Traditional single-market regression models may be misleading when a platform simultaneously serves:

consumers;

merchants;

advertisers;

developers;

service providers.

The economic model must account for cross-side network effects where relevant.

Principle

The econometric unit of analysis must correspond to the economic structure of the market.

27. FTC v. Actavis, Inc., 570 U.S. 136 (2013)

Actavis concerned reverse-payment arrangements in pharmaceutical markets.

The Court rejected an approach that would treat the underlying patent as automatically determining the competition question.

Econometric relevance

The economic analysis may need to examine:

expected litigation outcomes;

payment magnitude;

entry probability;

expected generic competition;

settlement incentives.

This illustrates the importance of counterfactual modelling.

The relevant question may be:

What would competition have looked like absent the challenged settlement?

rather than simply:

Was there a valid patent?

28. European Competition-Law Dimension

Econometric robustness also has substantial significance in EU competition law.

The European courts and Commission increasingly rely on quantitative economic evidence in:

cartel damages;

merger analysis;

abuse of dominance;

exclusionary conduct;

rebates;

predatory pricing;

market definition.

The underlying principles are similar:

identify the relevant counterfactual;

establish causation;

ensure methodological consistency;

distinguish correlation from causation;

test alternative explanations;

quantify uncertainty.

29. Cartel Damages and Econometric Robustness

Cartel cases create a particularly important econometric problem.

The claimant must frequently estimate:

Overcharge=Pcartel−Pcompetitive counterfactual\text{Overcharge} = P_{\text{cartel}} - P_{\text{competitive counterfactual}}

For example:

Cartel price = ₹150

Estimated competitive price = ₹120

Estimated overcharge:

₹150−₹120=₹30₹150-₹120=₹30

or:

150−120120×100=25%\frac{150-120}{120}\times100=25\%

But the ₹120 counterfactual is not directly observable.

This creates substantial methodological challenges.

30. Common Approaches to Cartel Overcharge Estimation

Before-and-after method

Compare prices:

before cartel → during cartel

During-and-after method

Compare:

cartel period → post-cartel period

Benchmark method

Compare affected transactions with unaffected markets.

Difference-in-differences

Compare:

affected market versus unaffected market

over:

cartel versus non-cartel periods.

Regression analysis

Control for:

input prices;

demand;

capacity;

geography;

product characteristics;

seasonality.

Structural modelling

Estimate demand and supply relationships and simulate the competitive counterfactual.

31. The "Counterfactual Stability" Requirement

A particularly important robustness principle is counterfactual stability.

If the estimated antitrust harm changes dramatically depending upon whether the expert uses:

2018 or 2019 as the baseline;

one competitor or another;

monthly or quarterly data;

one control market or another;

then the court should investigate why.

The more sensitive the result is to arbitrary modelling choices, the weaker the inference may become.

32. Regression Discontinuity and Natural Experiments

Where available, natural experiments can provide stronger identification.

Suppose a regulatory change affects one category of firms but not another.

The researcher can examine:

ΔYtreated−ΔYcontrol\Delta Y_{\text{treated}}-\Delta Y_{\text{control}}

This can sometimes provide a stronger causal inference than simple correlation.

However, the natural experiment must genuinely isolate the relevant competitive effect.

33. Demand Estimation

Demand estimation is increasingly important in:

merger cases;

pharmaceutical markets;

digital markets;

transportation;

consumer goods;

telecommunications.

Economists may estimate:

Q=f(P,Y,S,X)Q=f(P,Y,S,X)

where:

QQ = quantity demanded;

PP = price;

YY = consumer characteristics/income;

SS = substitutes;

XX = other relevant variables.

The resulting elasticities can inform:

market definition;

market power;

diversion;

merger effects;

unilateral effects.

But demand estimates can be highly sensitive to model specification.

34. Price Elasticity and Market Power

A basic elasticity measure is:

Ed=%ΔQ%ΔPE_d=\frac{\%\Delta Q}{\%\Delta P}

A low absolute elasticity may indicate that consumers are relatively insensitive to price changes.

But elasticity alone does not establish dominance.

Market power depends upon the broader competitive environment, including:

substitutes;

entry;

capacity;

buyer power;

innovation;

switching costs;

network effects.

35. Econometric Robustness in Digital Markets

Digital markets create special challenges.

Traditional price-based analysis can be inadequate because products may be:

free to consumers;

subsidized;

advertising-supported;

bundled;

monetized through data;

dependent on network effects.

Consequently, econometric models may examine:

user engagement;

search quality;

click-through rates;

advertising prices;

conversion rates;

ranking positions;

switching;

churn;

developer entry;

app availability;

interoperability.

For example:

UserQuality=f(Ranking,Relevance,Ads,Competition,Time)UserQuality = f(Ranking,Relevance,Ads,Competition,Time)

The expert must explain why changes in the measured variable constitute competitive harm.

36. AI and Algorithmic Markets

The robustness requirement becomes even more significant where algorithms determine:

prices;

rankings;

advertising allocation;

search results;

procurement;

product recommendations.

An econometric model may attempt to determine whether algorithmic pricing produces:

coordinated outcomes;

discriminatory prices;

reduced consumer surplus;

exclusion;

increased margins.

However, the expert must distinguish:

algorithmic correlation

from:

algorithmically caused anticompetitive coordination.

A simultaneous increase in algorithmically determined prices does not by itself prove an unlawful agreement.

37. Econometric Evidence in Merger Cases

Merger analysis frequently uses econometric evidence to estimate:

diversion ratios;

price elasticities;

unilateral effects;

coordinated effects;

efficiencies;

entry;

innovation effects.

A simplified merger simulation might estimate:

PPost=f(MarketShare,Elasticity,Diversion,Costs)P^{Post}=f(MarketShare,Elasticity,Diversion,Costs)

The model's robustness depends on the assumptions concerning:

consumer substitution;

firm conduct;

efficiencies;

entry;

capacity;

competitive responses.

38. Econometric Robustness and Efficiencies

A defendant may argue that a merger creates efficiencies.

For example:

Costpost<CostpreCost_{post} < Cost_{pre}

But courts and authorities must ask:

Are the efficiencies real?

Are they merger-specific?

Are they sufficiently verifiable?

Will consumers benefit?

Are they sufficiently likely?

Are they offset by anticompetitive effects?

Econometric evidence should therefore not model only the predicted price increase.

It may also need to model predicted cost reductions.

39. Robustness to Alternative Market Definitions

Suppose an expert finds a 10% price effect using Market A.

The court may ask:

What happens if Market B is used?

A robust analysis should ideally examine reasonable alternative market definitions.

This is particularly important in:

digital platforms;

rapidly evolving technology markets;

pharmaceutical markets;

differentiated products;

transportation networks.

40. Transparency and Reproducibility

A strong econometric submission should make it possible for opposing experts and the court to understand:

data sources;

variable construction;

sample selection;

model equations;

estimation method;

assumptions;

omitted observations;

robustness tests;

sensitivity analysis;

limitations.

The inability to reproduce the result can significantly undermine confidence in the analysis.

41. Common Weaknesses in Antitrust Econometric Evidence

1. Post-hoc model construction

The model is developed only after observing the results.

2. Cherry-picking

Only favourable specifications are disclosed.

3. Ignoring alternative explanations

The model attributes all changes to the defendant's conduct.

4. Weak control group

The comparison market is not genuinely comparable.

5. Endogeneity

The explanatory variable is correlated with unobserved determinants of the outcome.

6. Incorrect standard errors

Serial correlation or clustering is ignored.

7. Small sample

Too few observations are used to support a complicated model.

8. Measurement error

Important variables are poorly measured.

9. Specification sensitivity

The result disappears under reasonable alternative specifications.

10. Legal-economic mismatch

The model estimates an economic effect different from the legally alleged harm.

42. Relationship Between Econometric Evidence and Documentary Evidence

Courts should generally avoid treating econometrics as an independent substitute for all other evidence.

The strongest antitrust cases may combine:

Economic evidence

internal documents

communications

market structure

business conduct

consumer evidence

econometric evidence

For example, if internal documents show a deliberate plan to exclude a rival, competitors independently report foreclosure, and an econometric analysis identifies a statistically robust reduction in rival sales following implementation, the combined evidence may be considerably stronger than any one component.

43. Burden of Proof

Econometric robustness also depends on the procedural stage.

At different stages, the court may consider:

whether a claim is plausible;

whether sufficient evidence exists for trial;

admissibility of expert evidence;

proof of liability;

proof of causation;

proof of damages.

A methodology sufficient to survive an early procedural challenge may not ultimately establish liability or damages.

Therefore:

Admissibility, evidentiary sufficiency, causation, and quantification should not be treated as identical questions.

44. Antitrust Injury Versus Damages

An important distinction is:

Antitrust injury

The claimant must establish injury resulting from harm to competition.

Damages

The claimant must quantify the economic loss attributable to that injury.

An econometric model may successfully establish a price effect but fail to establish that the effect resulted from the specific antitrust violation alleged.

Conversely, a damages calculation may be mathematically sophisticated but legally irrelevant if the underlying injury theory is defective.

This distinction is particularly important under the reasoning in Comcast v. Behrend.

45. Robustness Checklist for Courts and Competition Authorities

A court or competition authority examining an econometric antitrust claim can ask:

Theory

What precise anticompetitive theory is being tested?

Does the model correspond to that theory?

Data

Where did the data come from?

Is the dataset complete and representative?

Are important observations missing?

Causation

What is the counterfactual?

How is causation distinguished from correlation?

Specification

Why were these variables included?

Why were others excluded?

Does the result survive alternative specifications?

Statistics

Are standard errors appropriate?

Is clustering required?

Are confidence intervals reported?

Sensitivity

What happens if outliers are removed?

What happens if the sample period changes?

What happens with alternative control groups?

Economic significance

How large is the estimated effect?

Is it competitively meaningful?

Legal relevance

Does the estimate prove an element of the relevant legal test?

Does the damages model correspond to the theory of antitrust injury?

46. Six Core Principles Emerging from the Case Law

The cases discussed above collectively support several important principles.

Principle 1 — Economic plausibility matters

Matsushita demonstrates that economic evidence must be assessed against the realities of the market.

Principle 2 — The model must fit the legal theory

Brooke Group illustrates that the economic analysis must correspond to the elements of the applicable antitrust test.

Principle 3 — Expert methodology must be reliable

Daubert establishes the broader evidentiary importance of methodological reliability.

Principle 4 — Statistical association is not automatically causation

Concord Boat illustrates the importance of connecting economic evidence to the alleged competitive mechanism.

Principle 5 — Damages must correspond to the theory of harm

Comcast provides perhaps the clearest modern illustration of this requirement.

Principle 6 — Economic evidence must be considered within the entire evidentiary record

High Fructose Corn Syrup demonstrates the importance of combining economic evidence with documentary and circumstantial evidence.

47. Practical Framework for an Econometrically Robust Antitrust Claim

A strong antitrust econometric analysis can therefore be structured as follows:

Step 1 — Define the legal theory

Identify the precise alleged infringement.

↓

Step 2 — Identify the economic mechanism

Explain how the conduct could harm competition.

↓

Step 3 — Identify the relevant outcome

Price, output, quality, innovation, entry, foreclosure, consumer welfare, etc.

↓

Step 4 — Construct the counterfactual

Determine what would have happened without the challenged conduct.

↓

Step 5 — Obtain appropriate data

Use sufficiently detailed and reliable observations.

↓

Step 6 — Select an appropriate methodology

Regression, difference-in-differences, event study, structural estimation, comparator analysis, etc.

↓

Step 7 — Address identification problems

Endogeneity, omitted variables, selection bias and reverse causality.

↓

Step 8 — Conduct robustness testing

Alternative specifications, samples, controls, time periods and estimation techniques.

↓

Step 9 — Quantify uncertainty

Confidence intervals and appropriate statistical inference.

↓

Step 10 — Connect economics to law

Explain exactly which legal element the empirical result establishes.

48. Conclusion

Econometric robustness is increasingly central to modern antitrust litigation. Courts and competition authorities cannot simply accept a regression coefficient, statistically significant result, or damages estimate at face value.

A persuasive econometric antitrust claim should demonstrate:

a clearly defined theory of harm;

an economically plausible mechanism;

appropriate and reliable data;

a credible counterfactual;

proper causal identification;

appropriate statistical techniques;

controls for relevant confounding factors;

sensitivity and robustness testing;

appropriate treatment of uncertainty;

economically meaningful effects; and

a direct connection between the empirical findings and the applicable legal standard.

The most important lesson from Matsushita, Brooke Group, Daubert, Concord Boat, High Fructose Corn Syrup, Comcast, Microsoft, Indiana Federation of Dentists, American Express, and Actavis is that econometric sophistication alone does not establish an antitrust violation.

The decisive question is whether the methodology reliably demonstrates the specific competitive harm alleged, under the legally relevant counterfactual, with sufficient empirical and economic robustness.

In modern digital, platform, AI, algorithmic-pricing and data-driven markets, this requirement is becoming even more important because competitive harm may involve quality, innovation, access, ranking, interoperability, data, network effects and foreclosure, rather than merely observable price increases.