RESEARCH DOSSIER
Empirical Legal Research Methods
Abstract. This monograph investigates empirical legal research as a source-critical inquiry into the observable operation of law. It reconstructs the design of descriptive, qualitative, survey-based, administrative, observational, experimental, quasi-experimental and mixed-method studies, with attention to source authenticity, causal identification, legal relevance and comparative institutional interpretation. It explains the difference between the authority of legal norms and evidence concerning their operation; identifies recurrent threats from sampling, nonresponse, incomplete court records, measurement and policy timing; and examines standards of transparency, privacy and human-participant ethics. Three extended hypothetical research designs illustrate the consequences of these methodological choices for legal assistance, court digitalization and unmet legal needs. This scholarly synthesis reviews original literature, peer-reviewed methodological scholarship and selected official sources through November 21, 2026. No human-subject research, original survey, experiment or new statistical estimation was undertaken.
Keywords: empirical legal research; sociolegal methods; legal evidence; research design; sampling; legal needs; courts; administrative data; qualitative interviewing; legal ethnography; causal inference; difference-in-differences; comparative law; research ethics; reproducibility
Introduction
Empirical legal research studies the observable organization and operation of law. It examines institutions, professionals, populations, practices, disputes, decisions, enforcement processes and experiences under explicit rules for producing and evaluating evidence. Its distinctive research problem arises whenever a conclusion about law depends on observations of the world: how many people encounter a justiciable problem, why certain claims reach courts, what judges and officials do, whether procedural assistance changes outcomes, or how litigants experience a formally guaranteed right. Answering such questions requires defensible definitions, reliable evidence and an inferential relationship between them.
A court decision can have legal force and at the same time constitute one observation in a dataset. Those functions must be established separately. A published judgment may state an authoritative proposition while reflecting only disputes that survived filing, settlement, admissibility and appeal. A regression across judgments may describe the composition of that selected group, and a legal analysis must independently determine jurisdiction, hierarchy and temporal applicability. Epstein and King's (2002) rules of empirical inference supply the discipline of specifying what research can learn from quantitative and qualitative evidence. Theil's (2025) account of doctrinal and empirical contributions clarifies how observation informs source-based reasoning while remaining subject to its own epistemic conditions.
The monograph proceeds from questions and sampling to field methods, datasets, descriptive estimates, causal designs, comparison, ethics and reproducibility. Its governing analytical task is correspondence: each legal claim must be supported by a source and method appropriate to that claim, with the strength and limitations of the inference made explicit. The international perspective is methodological. Examples concern United States federal and state courts, European judicial evaluations, the European Court of Human Rights and legal-needs research across jurisdictions. These examples illustrate the distinct production processes of legal information, rather than make a universal claim about national legal systems.
The work offers independent interpretation and original worked examples of research design. It evaluates established methods and develops a comparative synthesis without creating a separately named theory or claiming any new empirical finding. The cutoff for factual and legal-source review is November 21, 2026. Every reference to a dataset is tied to its identified producer and reporting window. The examples labelled hypothetical describe proposed studies, not experiments, interviews or datasets undertaken by the author.
Historical Foundations of Empirical Legal Research
Empirical legal scholarship has roots in legal realism, sociology, institutional analysis, political science and anthropology. Its central historical development was an expanding understanding of the places where law could be observed. Legal texts and decisions remained important, while transactions, negotiation, claim-making, administration, noncompliance and informal resolution became legitimate subjects of systematic inquiry. This development changed the unit of research from an exclusively reported legal decision to a range of actions and experiences through which normative institutions operate.
Stewart Macaulay's (1963) study of commercial relationships is a landmark because it directs attention to how business participants organize transactions beyond the formal enforcement possibilities of their contracts. The historical contribution is a research orientation: asking when commercial actors invoke the law, how they understand an obligation and how relational arrangements influence conduct. It does not establish that formal contracts are irrelevant, nor that all industries share one informal practice. Those larger statements require new evidence from specified populations and periods.
Marc Galanter's (1974) analysis of repeat players similarly expanded inquiry from isolated litigation outcomes to structural characteristics of participation in the legal system. Repeated experience, resources, coordination and long-term interests can give institutional actors different opportunities to shape litigation strategy. The original work identifies questions and conjectures about such advantages. An empirical study would have to translate repeat participation and resource asymmetry into observable measures, compare relevant populations and examine counterexamples before reporting a general result.
The development of empirical legal studies within legal academia heightened concerns about design and inference. Epstein and King (2002) showed that law reviews had developed substantial empirical work without always applying the standards of uncertainty, observational comparability and inferential transparency practiced in other social sciences. Their critique should be read as a methodological demand rather than an argument to replace legal reasoning with numerical analysis. Legal questions about authority, validity and interpretation still call for legal materials, and questions about practice call for observations.
The contemporary interdisciplinary landscape therefore includes at least four tasks: reconstructing operative law, describing how institutions behave, explaining why behavior varies, and evaluating legal practices against explicitly identified standards. Their relationships are productive when scholars articulate which answer derives from which source. The emerging availability of large digitized corpora widens opportunities, yet intensifies the problems of completeness, classification, selection and the interpretation of legal language.
Legal Authority and Empirical Inference
An empirical proposition and a doctrinal proposition can concern the same event while requiring different grounds. “A statute was in force on a given date” requires an authenticated legal text and rules governing enactment, commencement and amendments. “Officials applied that statute to a stated proportion of applicants” requires an observed population, recorded decisions and a valid measurement rule. “The application was lawful” requires an additional argument about the operative norm, the relevant facts and interpretive authority. “The application improved welfare” requires yet another evaluation of outcome and criterion.
Theil (2025), in his methodological study of doctrinal scholarship, identifies the distinct role of legal sources in changing or establishing the content of law and specifies several ways empirical inquiry can contribute to understanding, assessment and the discovery of relevant sources. This distinction is practically consequential for study design. Interviews can establish that many participants misunderstood official guidance, but they cannot themselves settle the legally controlling interpretation. Conversely, a constitutionally guaranteed procedural entitlement does not establish the proportion of beneficiaries who can successfully exercise it.
The researcher must identify the level at which a claim operates. Descriptive claims concern distributions or records. Interpretive claims concern meanings reconstructed from texts or action. Explanatory claims compare mechanisms compatible with observed patterns. Causal claims ask what would have happened under different interventions. Normative claims apply legal or explicitly specified evaluative standards. Inferences move between these levels only with additional premises and evidence. A tendency to infer legal validity from statistical frequency, or policy effectiveness from the existence of formal rules, corrupts both modes of reasoning.
Law also structures the empirical field. Jurisdiction determines which institution can decide a dispute; procedure determines who can file, what will be registered and when an appeal can occur. Confidentiality and publication rules determine what the researcher can observe. A database must therefore be studied as a legally and administratively produced object. The interpretation of an empirical finding may itself change when the researcher discovers that the legal classification changed halfway through the period studied.
A strong legal research design links its substantive question to the competent legal authorities and the relevant observation process. The resulting report assigns each conclusion its correct status, including the precise institutional scope and time. The goal is a coherent interdisciplinary explanation in which empirical observation improves legal understanding and legal source criticism improves the validity of empirical analysis.
Research Questions and Unit Definition
Research questions define which evidence will be collected and what sort of conclusion can follow. A descriptive question asks how frequent an event or condition is in a stated population. A process question asks how a legal problem moves between individuals, advisers, officials and tribunals. A comparison asks how specified distributions differ across groups. A causal question asks what would have happened to eligible units under a different intervention. An interpretive question examines how actors understand and reproduce legal categories in situated action.
Legal research demands exceptional precision in defining a unit. A person may have several disputes, each dispute multiple claims, each claim multiple hearings, and each proceeding several dispositions or appeals. A judicial panel may issue a single judgment concerning several parties, whereas an administrative report may count individual defendants rather than files. The unit of assignment to an intervention can differ from the unit of observation and the unit used for analysis. Studies that move unannounced between cases, people, claims and court events create misleading denominators.
A causal estimand specifies the population and contrast. In a legal aid evaluation, assignment to an offer of representation and actual representation are different treatments. Greiner and Pattanayak's (2012) randomized legal-assistance investigation illustrates why this distinction changes interpretation. Actual use may depend on participant motivation, availability or case complexity after an offer. Comparing represented and unrepresented persons without accounting for those processes does not automatically estimate the effect of offering services.
The question should include jurisdiction and time. A legislative enactment date, legal commencement date and date of implementation can differ, and case outcomes may lag all three. The observer must decide whether a newly filed case is included by filing date, disposition date, exposure period or a person's first encounter with the institution. These decisions alter the population and can change the apparent result.
A useful protocol records the research question, scope, units, dependent and explanatory variables where relevant, legal definitions, observation window, recruitment or extraction process, anticipated mechanisms and contrary evidence. It states the effect size or degree of descriptive accuracy that would be substantively meaningful. Formal significance levels alone do not determine whether a reported difference matters to legal rights, costs or institutional legitimacy.
Research Design and Measurement
Research design constructs a relationship between a question and observations capable of answering it. A probability survey can estimate a population distribution under documented sampling and response assumptions. A purposive interview study can identify meanings and mechanisms through comparison of strategically chosen experiences. A court-level panel may describe an administrative trend; its causal interpretation requires a separately defended identification strategy. A close case study may expose a hidden institutional process while leaving its prevalence unknown. The appropriate method follows the inferential task.
Measurement turns a concept into a rule for classifying observations. “Access to justice” can involve awareness of rights, affordability, availability of advice, filing, voice in proceedings, timely resolution, substantive remedy, enforcement or participant experience. A count of claims filed measures one institutional behavior, not the entire construct. Similarly, “judicial efficiency” may be operationalized by disposition time, backlog or cost, each with different distributions and normative implications. Researchers should state the conceptual and operational definitions together and explain why the indicator is substantively valid.
Source provenance is part of measurement. A docket entry may reflect a contemporaneous event, a delayed clerical update, an automated migration or a later recoding. Administrative completeness can change across years. An interview transcript records what was said in a specific encounter, while a questionnaire records answers shaped by wording, mode and recall. Distinct source-generating mechanisms yield distinct opportunities for error.
Preanalysis planning is especially valuable when quantitative work presents many reasonable specifications. It can distinguish confirmatory hypotheses from exploratory findings, define missing-data treatment and set primary outcomes. Qualitative work also benefits from recorded research questions and analytic choices while retaining an appropriate role for discovery, reflexivity and emerging categories. A design can incorporate iterative refinement while still documenting what changed and why.
Triangulation should address different sources of error rather than merely accumulate citations. Two datasets copied from the same court management system are not independent corroboration. Agreement between public survey evidence, administrative records and authenticated legal rules may be more compelling because their sources and limitations differ. Disagreement can reveal different units, samples or concepts rather than a failure of one source.
Sampling and Selection into Law
Legal institutions receive a selected subset of possible disputes. Individuals may not recognize a problem as legally relevant, may lack information or resources, may negotiate informally, or may abandon a claim. A case can settle before filing, be dismissed early, never reach an appeal or remain unpublished because of legal rules. Court-based research therefore must document the selection processes that lie between the underlying social event and the observed record.
Sampling frames should be specified independently of sample size. A survey of people visiting a legal aid office describes eligible or attending office users under the recruitment conditions; it does not establish the unmet legal needs of every resident. An online panel can exclude those without digital access. Interviewing only litigants excludes people deterred by expected cost or risk. A probability sample supports design-based estimates only when inclusion probabilities and response adjustments have an adequate empirical foundation. A purposive sample supports context-rich analytical inferences, with claims tied to the sample's selection logic.
The OECD/Open Society Foundations (2019) legal-needs survey framework addresses the broader population of justiciable problems and the variety of ways people respond. This is a methodological expansion beyond court caseloads. AAPOR's survey research standards describe the consequences of sampling coverage, respondent recruitment, nonresponse and questionnaire wording. Those considerations remain relevant even when the subject is legal consciousness rather than public opinion.
Purposeful qualitative selection may include first-time claimants, repeat organizational litigants, advisers, administrators, people who settled and people who avoided formal action. Variation can help identify different pathways through the legal system. Researchers should record who was approached, who declined, which groups remained unreachable and whether intermediaries influenced participation. The integrity of a qualitative argument depends on the adequacy of the sources for the interpretive claim, not the mechanical conversion of interviews into an unqualified population percentage.
A particularly serious selection problem arises when both the putative cause and the outcome influence the probability that an observation enters the dataset. Conditioning on filing, appeal or publication may create an artificial association between resources and case results. More controls cannot automatically repair such sample construction. The empirical object must be rebuilt from the earliest observable relevant population wherever the research question requires it.
Observational and Ethnographic Research
Observation allows inquiry into the routines through which legal institutions function. A researcher in a courtroom, administrative office or legal aid clinic can study sequence, interaction, discretion, waiting, document handling, informal explanation and participant understanding. Docket codes may record that a matter is pending, while observation shows whether the delay concerns service, an interpreter, a missing file, an adjournment request or a congested calendar. Those mechanisms have different policy and legal significance.
A sound observational protocol states the sites and sessions observed, entry conditions, time windows, selection of cases and observer role. Field notes should distinguish contemporaneous description from later reconstruction and interpretation. Observers should record institutional language, exclusions, interruptions and events that contradict emerging hypotheses. Access to a public court does not necessarily authorize photographing documents, identifying vulnerable parties or violating orders that protect witnesses or confidential records.
The tradition represented by Macaulay's study of informal commercial practice shows why an empirical account should include the relational setting in which formal legal categories are invoked. This insight can guide observation of consumer complaints, employment grievances, family disputes and housing administration. Still, a handful of vividly described events cannot establish their frequency in an unobserved population.
Researcher positionality matters because a scholar's language, occupational identity, institutional affiliation and visible presence can influence conduct and access. Reflexive notes should record possible observer effects, dependence on gatekeepers and interpretive assumptions. Institutional documents and interviews can then be compared with observed action. The final report should make a path from the observations to the interpretation sufficiently clear for critical evaluation.
Ethics and methodology are connected. If researchers cannot observe a confidential process directly, they may use authorized aggregate descriptions, voluntary participant accounts or public documents, explaining the different evidentiary status of each. The correct response to inaccessible material is to narrow the inference rather than presume what happened behind closed doors.
Qualitative Legal Interviews
Interviews are evidence of participants' experience, understanding, recall and explanations. The central methodological issue is the relation between what a person reports and what the researcher concludes. An account of an unsuccessful hearing may establish a litigant's perceived unfairness, but a legally authoritative account of procedural compliance requires the governing rules and authenticated records. A lawyer's explanation of strategy is evidence of professed professional reasoning, which may differ from behavior recorded in the file. These differences are research subjects rather than reasons to discard one kind of evidence.
Semi-structured interviews balance a common set of topics with follow-up questions about concrete events. An effective interview guide asks respondents to reconstruct chronology, participants, decisions, alternatives, reasons and consequences. Questions that presume unlawful treatment, institutional bias or negligence distort the evidence before it is collected. The interviewer must explain the research purpose, participation conditions and confidentiality with language that respondents can understand. Recruitment through a court, employer or agency can create an impression that the interview is mandatory, so the voluntary character of participation requires careful protection.
Purposeful sampling can compare claimant and official perspectives, persons with and without legal representation, or cases resolved through different pathways. Recording the recruitment procedure is necessary to interpret silence and absence. Transcription and translation can change the available evidence: a technical legal term may have no straightforward local equivalent, while hesitation, irony or deference may disappear in summary notes. Where translation is required, store version provenance and permit methodological review consistent with confidentiality.
Braun and Clarke's (2006) thematic analysis account establishes the usefulness of a deliberate approach to patterns in qualitative material. A researcher should identify an analytic orientation, preserve memos, explain how codes became themes, consider disconfirming examples and distinguish participant language from the author's interpretation. A purely mechanical agreement score cannot substitute for a coherent interpretive account of legal meaning. Where coding reliability is relevant to classification, it should be assessed against an explicit category definition.
The COREQ interview reporting checklist of Tong, Sainsbury and Craig (2007) prompts disclosure of research team, participant selection, context, data collection, analysis and reporting. O'Brien and colleagues' qualitative reporting recommendations offer a complementary standard. Reporting checklists support transparency but do not establish that ethical safeguards were followed or that the interpretations are compelling. Quoted material requires careful consideration of indirect identification: a distinctive legal event can expose a participant even when the name is absent.
Survey Methods and Legal Needs
A survey measures answers to a specified instrument administered through a documented sampling and fieldwork process. It is therefore a combination of concepts, questions, response options, recall periods, languages, modes, recruitment procedures and data processing. A survey asking whether respondents “had a legal problem” can miss housing, work or administrative conflicts whose legal character is invisible to the participant. A question about concrete events may identify a wider set of justiciable problems without requiring respondents to possess specialized legal knowledge.
The Legal Needs Surveys and Access to Justice report by the OECD and Open Society Foundations (2019) develops a framework around problem experience, seriousness, response, assistance, costs, outcomes and impacts. Its central evidentiary advantage is the ability to study problems irrespective of whether they reach formal institutions. This makes legal-needs surveys an essential complement to filing statistics, particularly when the research object is access rather than court workload.
The design should define the target population, sampling frame, eligibility, language coverage, mode, contact procedure, expected exclusions and intended measures. Probability estimates require suitable inclusion probabilities and a transparent account of weighting and nonresponse. A large number of completed responses does not repair systematic exclusion of people without connectivity or stable housing. AAPOR's best practices for surveys emphasize question wording, mode effects, interviewer quality, nonresponse and confidentiality. Data quality should be monitored during collection and not inspected only after analysis.
Questionnaires should avoid compound questions such as “Were the court and your lawyer fair?” which combine institutions, experiences and standards. Response categories must be mutually intelligible, and the instrument should permit uncertainty or refusal when appropriate. Recall periods require consideration of forgetting and telescoping; a respondent may report a recent event that took place before the intended window, or omit a less salient unresolved problem. Translation must retain conceptual rather than literal equivalence when legal vocabulary differs.
Reported percentages must specify whether their denominator is respondents, disputes, claims or help-seeking episodes. A respondent can experience several problems and pursue several forms of assistance, so multiple-answer proportions may appropriately exceed 100 percent. Weighted estimates should include uncertainty that reflects the sample design. A comparison over time requires stable concepts and instruments or a justified method of adjustment. Surveys can reveal an important social distribution of legal problems, but a statistically observed barrier becomes a legally actionable violation only through an additional jurisdiction-specific legal analysis.
Official Court Statistics and Administrative Data
Courts and justice agencies collect information to administer their work. Researchers benefit from the volume and temporal continuity of these records, while inheriting the institutional definitions through which events were entered. A filing, defendant, claim, order, disposed matter, pending case or appeal can be counted differently across agencies and historical reporting versions. A case may be transferred, consolidated, reopened or terminated in the information system without corresponding to the intuitive image of a separate dispute.
The Administrative Office of the U.S. Courts' 2025 federal caseload report covers the twelve-month period ending March 31, 2025 and explains changes in the workload of federal appellate, district and bankruptcy courts. Its discussion of multidistrict litigation illustrates why large changes in aggregate civil filing or termination counts can be driven by concentrated procedural categories. One cannot infer a broad decline in demand for justice from such a series without analyzing case mix and institutional recording rules.
The official federal judicial caseload tables include a 2026 reporting vintage. An analyst must distinguish that vintage's March 31 reporting window from fiscal-year tables, earlier reports and case-level adjudication times. The Court Statistics Project, maintained by the National Center for State Courts with the Conference of State Court Administrators, provides a different institutional scope: state-court caseloads, not an extension of the federal series. Adding state and federal counts requires explicit definitions, duplicate checks, category mapping and a properly described population.
Before analysis, construct a data dictionary describing the official publisher, document title, extraction date, reference period, unit, inclusion and exclusion rules, event classification, available identifiers, corrections, suppression and unknowns. Audit duplicates, inconsistencies between opening and closing dates, impossible transition sequences and changes in fields after migrations. Identify which administrative transformations were performed by the source agency and which were introduced by the researcher. Record software and transformation versions to permit a later audit.
Missing data have several meanings. A blank field in an otherwise comprehensive extract can result from omission; an unpublished decision reflects publication rules; a sealed proceeding reflects legal restrictions; and a dispute settled before filing never entered the source population. Treating all four as random missing observations misconstrues the institution that produced the data. The value of an administrative dataset depends as much on the validity of its event construction as on the size of its rows.
Judicial Decisions and Corpus Construction
Judicial opinions are texts, institutional acts and possible legal authorities. Their analytical function depends on the research question. Counting references to a statute studies textual practice. Measuring appeals that succeed studies a defined procedural population and a specified outcome. Determining the binding effect of a judgment requires the legal rules governing jurisdiction, hierarchy, reasons and precedent. Large text corpora make the first two tasks scalable but do not automatically answer the third.
The European Court of Human Rights maintains the HUDOC case-law database covering several categories of judgments, decisions, communicated cases and related material. Its official description records limits of publication, including the absence of single-judge decisions from its published collection. A dataset extracted from HUDOC cannot therefore be treated as a census of all applications. An analyst of successful merits claims must also account for admissibility, case communication, settlement, referral, publication selection and procedural changes over time.
A corpus protocol should list the repositories searched, exact query strings, retrieval date, language restrictions, document types, temporal coverage, pagination, deduplication and version controls. For each item, retain the official citation or identifier, provenance and original text. A plain-language search result can contain translated excerpts, editorial summaries, quoted pleadings, dissenting opinions or superseded provisions. These should be separated from the court's actual holding when a legal proposition is coded.
Automated coding introduces further measurement problems. OCR may alter section numbers and negation; a classifier may confuse an allegation with a finding; training and test splits may leak duplicated judgments or shared case histories. Evaluation should include representative manual checks, review of rare categories, temporal holdouts where relevant and error analysis by language, court and procedural posture. Accuracy on an aggregate benchmark can conceal severe error on a legally important minority category.
A verified legal-source return path is required whenever computational output supports doctrine. Researchers should be able to move from a statistical item or embedding to the authenticated judgment version that generated it. This preservation of provenance turns computation into a research instrument serving interpretation and source criticism rather than a substitute for either.
Statistical Description and Uncertainty
Quantitative description begins with the population and unit, then a measure. Frequencies, rates, medians, quantiles, survival distributions, cross-tabulations and transition probabilities summarize different aspects of legal processes. Median time to disposition can decline while the number of exceptionally delayed cases increases. A clearance ratio compares terminations with the incoming workload during a reporting interval; it is not the probability that a particular claimant obtains a remedy. A simple average can be distorted by mass litigation, bulk processing or the mixing of radically different case types.
The inferential status of a statistic depends on the selection and measurement process. Administrative records can constitute a near-census of registered events within a defined system while remaining a selected view of underlying disputes. Sample surveys may represent a broader population under explicit design assumptions. A regression across courts may summarize an association even when its predictors include variables produced by earlier legal processes.
Uncertainty has statistical, measurement and institutional dimensions. Sampling error arises from observing a sample; model uncertainty concerns assumptions about functional form and residual dependence; selection uncertainty concerns which people or events entered the source; classification uncertainty concerns whether the variable captures the concept. Reporting a narrow confidence interval addresses only some of these sources. A statement such as “the reform improved access by 12 percent” is methodologically incomplete until the outcome, denominator, comparison, uncertainty, exposure and causal design are supplied.
Repeated observations are commonly clustered. Cases may share judges, courts, law firms and regional institutions; time observations within one court share unmeasured factors. Treating each case-month as independent can understate uncertainty. Clustering or multilevel methods should reflect the substantive source of dependence, with attention to the number of independent clusters. Researchers should also address missing observations, selective follow-up and rare outcome instability.
Rosenbaum and Rubin's propensity-score study provides a formal strategy for balancing observed covariates when comparing exposed and unexposed groups. Its methodological value depends on measured variables and overlap; a balanced propensity score does not eliminate omitted or unobserved confounders. If a policy applied only to one kind of court or claimant, extrapolating beyond the common support remains a substantive assumption. Model output should be evaluated in terms of what comparable legal populations the evidence actually contains.
Causal Inference and Legal Interventions
Causal research asks what would have happened under an alternative intervention for a defined population. A policy effect is a comparison of potential outcomes under specified treatments, not a synonym for change after an enactment. Legal interventions can include advice provision, procedural deadlines, electronic filing, changes in eligibility or enforcement resources. Each has its own assignment process, implementation date and possible spillovers.
Randomization can identify an effect of assignment within a defined setting when assignment is valid and follow-up, compliance, attrition and interference are understood. Greiner and Pattanayak's randomized evaluation of representation distinguishes an offer from actual uptake, illustrating how the estimand can be altered by participant choice. The results of one type of unemployment-benefit appeal should not be extrapolated to all civil legal aid, all claimant groups or the moral desirability of legal representation.
Nonexperimental studies require a strategy for understanding why some units received treatment and others did not. Legislators respond to social conditions; courts with severe backlogs may adopt digital systems earlier; legal aid recipients may differ systematically from nonrecipients. A credible design anticipates such selection and distinguishes variables preceding treatment from those affected by treatment. Post-treatment adjustment can conceal part of the effect or create new bias.
Hernán and Robins' Causal Inference: What If explains the importance of defining an intervention, time zero, eligibility, follow-up and causal contrast. Legal researchers can use the logic of a target trial as a design diagnostic even where actual random assignment is neither possible nor ethical. A process of referral, filing or appeal that occurs after exposure should not automatically become a baseline restriction on who is studied.
A legal evaluation must also interpret substantive importance. An intervention may change time to hearing without changing final disposition; improved measured throughput could come at the expense of participation. The causal estimate should report which outcome it describes and how the relevant legal standards evaluate that outcome. Sensitivity analyses, placebo periods, negative controls and rival mechanisms improve appraisal but cannot transform unsupported identifying assumptions into observed facts.
Difference-in-Differences and Policy Evaluation
Difference-in-differences compares observed changes across treated and comparison groups to estimate a treatment effect under assumptions about the counterfactual trend. Its use in legal reform studies requires identification of the operative intervention, the eligible institutions or people, comparable periods and the selection process governing adoption. A reform may be announced, enacted, commenced and implemented at different dates; each creates a different potential treatment marker.
The common parallel-trends condition concerns untreated potential outcomes. Similar observed levels before policy adoption are not sufficient; even similar short-run trends cannot prove what would have happened absent intervention. Anticipation, policy spillovers, contemporaneous economic changes and administrative reclassification can threaten the comparison. The researcher must defend a plausible comparison population and document relevant institutional shocks.
Bertrand, Duflo and Mullainathan's difference-in-differences assessment demonstrates why serial correlation can create misleading conventional standard errors in policy panel studies. Court-by-month observations may produce many rows while only a small number of courts independently adopt a reform. The inferential procedure must respect that actual assignment structure. A large database does not guarantee a large number of independent policy interventions.
When jurisdictions adopt at different dates, the conventional two-way-fixed-effects estimator may mix contrasts between newly treated, never treated and already treated groups. Goodman-Bacon's staggered treatment timing analysis makes the weighting problem explicit, particularly when treatment effects vary across cohorts and time. A substantive legal paper should identify the cohort-time effects it aims to estimate and the comparison groups that can support them. An unexplained single regression coefficient can conceal the estimand.
A responsible event-study report includes the legally operative adoption timeline, sources of implementation information, outcome coding, sample composition, effect heterogeneity, uncertainty assumptions and sensitivity to alternative preperiods. Where a new filing platform also changes how cases are registered, the researcher should consider an administrative measurement discontinuity before attributing observed changes to court behavior. Qualitative observation and technical records can identify whether such a discontinuity occurred.
Comparative and Transnational Data
Comparative empirical legal research is conceptually demanding because legal categories depend on institutional competences. A “court,” “case,” “administrative appeal,” “legal aid recipient” or “judicial clearance” may represent different entities across legal systems. Even when two countries use the same English label, the underlying process may have different jurisdiction, deadlines, classification and remedies. The first comparative task is therefore to establish functional equivalence rather than rush to aggregate counts.
The Council of Europe's CEPEJ evaluation of judicial systems gathers information through national correspondents and publishes data with explanations of national arrangements and collection methodology. These explanatory materials are part of the evidence needed to interpret cross-country tables. Differences between countries can reflect institutional design, financial resources, reporting definitions or legal culture as well as performance. A cross-national median is a descriptive statistic for a set of reported values, not an automatic criterion of justice.
A comparative protocol defines the legal function being observed, lists the competent institutions, maps the relevant procedure, identifies a shared temporal period and states which measures can be harmonized. If one jurisdiction resolves a dispute through an administrative tribunal while another uses ordinary courts, comparing court filing counts alone can make the latter appear to have more legal problems. The research must follow the social and legal function across different institutional arrangements.
Comparisons also require attention to common shocks and interdependence. Neighboring countries may adopt similar reforms, share European legal obligations or experience correlated economic changes. Treating each nation as an independent observation may be misleading, especially with a small set of countries and many covariates. Detailed process-based comparison may provide better mechanism evidence than a broad regression with poorly comparable variables.
A sound conclusion acknowledges where commensurability ends. It can report the kinds of procedures covered, the observed differences and the interpretation supported by national source material without converting them into a universal league table. The purpose of international evidence is to make institutions mutually intelligible while preserving their legal specificity.
Mixed Methods and Triangulation
Mixed-method research joins evidence-generating practices with different strengths. The relevant question is what each source can contribute to a common inquiry. A survey can estimate how legal problems are distributed across a population; administrative data can measure recorded institutional workload; interviews can reconstruct how persons understood available remedies; observation can reveal the procedural encounters through which choices were made; and doctrinal analysis can establish which rules, competences and obligations governed the events. The scholarly achievement lies in identifying the relation among these claims and the limits of their convergence.
An exploratory sequential study might begin with qualitative interviews to discover local understandings of a housing dispute, then build and test a questionnaire that captures those experiences in a broader sample. An explanatory sequential study might begin with a court-statistics anomaly and then examine institutional workflow and administrative changes to determine why it occurred. A concurrent study can analyze qualitative and quantitative data separately before bringing them into a common explanatory discussion. All approaches require a written account of selection, timing, analytic choices and integration.
Triangulation strengthens a proposition when observations come from sufficiently independent production processes and when the same construct is being compared. Two administrative exports produced from an identical docket do not provide independent confirmation, and a survey of court users cannot by itself validate findings about persons excluded from courts. The research must ask whether apparently corroborating measures share the same hidden errors. Differences across sources are equally important findings when they expose different units, populations, event definitions or institutional timing.
Consider a court reporting lower median case durations while users report worsening delay. Both observations may be true. The median may describe completed cases, while interviews concern still-pending complex disputes; case composition might have changed, and duration may begin at different events in the two systems. Reconstructing denominators and time origins resolves the apparent contradiction before any theory of institutional performance is asserted.
Integration should occur at identifiable points in the argument. A field observation might justify a new variable, an interview theme may explain a quantitative interaction, or a legal source may establish that the observed institutional category was changed by an amendment. The final manuscript should mark which conclusion is supported by each type of source. A mixed-method label does not itself justify generalization or causal identification: each constituent inference must remain independently defensible.
Research Ethics and Confidentiality
Research about law often involves people whose legal, occupational, family or immigration circumstances make disclosure consequential. A study of asylum procedures, criminal defence, debt enforcement, family violence or workplace complaints may expose participants to identification, intimidation, strategic disadvantage, distress or state scrutiny. Ethical design begins with the actual people, power relations, possible harms and intended uses of collected information, then specifies safeguards in recruitment, analysis, storage and publication.
The National Commission's (1979) Belmont Report articulates respect for persons, beneficence and justice for human-subject research. These ethical principles provide a foundation for voluntary participation, meaningful information about the study, fair participant selection, risk minimization and attention to burdens and benefits. Consent is a process rather than a signature alone. Researchers should consider whether a participant fears that refusing will endanger access to a lawyer, public service, visa process or employer.
Applicable legal duties require jurisdiction-specific analysis. The U.S. HHS 45 CFR 46 regulations govern covered research under their applicability conditions and provide particular rules for consent, institutional review and protected populations. A scholar should not assume that every legal interview falls under the federal Common Rule or that an institutional review exemption eliminates all professional and confidentiality obligations. In the European Union, the General Data Protection Regulation governs covered processing of personal data, including legal bases under Article 6, special-category conditions under Article 9 and research-related safeguards under Article 89. Scientific research is subject to the actual conditions of applicable law.
Anonymity and confidentiality should be distinguished. Data collected without identifiers can still become identifying when combined with unique dates, rare case types, locations and public judgments. A quotation from a small administrative tribunal might identify a lawyer or applicant through procedural details even after the name is changed. Researchers should conduct a disclosure-risk review of qualitative excerpts and microdata, including the risk that software searches will link records with publicly available information.
A defensible protocol specifies recruitment independence, informed consent, withdrawal conditions, data minimization, secure storage, access control, role separation, retention limits, translation handling and procedures for compelled disclosure or breach. Public availability of a docket does not automatically justify republishing sensitive information in a new searchable database. Ethical review and legal compliance interact with inferential quality: people who fear identification may decline participation or alter their accounts. This dossier conducted no participant recruitment, field interviews, original survey or access to confidential case material and claims no ethics-board determination for such work.
Research Reproducibility and Reporting
Reproducibility is the ability to examine how an inference was generated, from the original question and source selection through the transformations that produced its reported result. In legal research, that record must include the status of each document as an authoritative instrument, administrative record or participant account. Legal-data processing often embeds interpretive decisions: whether a case is considered disposed, whether transfer is a new filing, how an amended provision is dated and whether a statement belongs to the court's reasoning or a party's allegation.
A reproducible administrative study preserves a machine-readable source inventory, retrieval dates, queries, reporting windows, field definitions, coding rules, transformations, exclusions, software versions and analysis code. Version controls should make it possible to distinguish original sources from modified extracts. If privacy or legal restrictions prohibit public microdata, the researcher can still publish a detailed codebook, redacted examples, synthetic test data, aggregate statistics and an access procedure where authorized. The ethical limits of openness should be explained with the same precision as statistical limits.
The STROBE observational reporting guidance helps structure disclosures of study design, setting, eligibility, variables, bias, sample size, methods, missing data, results and limitations. It was developed for observational health studies and is useful as a reporting reference, not a legal-research certification or a substitute for defensible causal assumptions. Interview studies may use the COREQ reporting checklist, while the standards for qualitative reporting of O'Brien and colleagues (2014) provide another benchmark for making interpretive work inspectable. A researcher should adopt relevant items and explain method-specific exceptions rather than present borrowed instruments as a guarantee of validity.
Transparency includes reporting failed analyses, unexpected categories, decisions to combine groups and the consequences of unavailable data. Exploratory discovery is an important form of scholarship; its inferential status should be stated. A post hoc finding should not be rewritten as if it were a precommitted hypothesis. Sensitivity to different case definitions and source versions can be as important as changing a regression model.
The citation architecture of a scholarly synthesis is itself reproducible. Every external research link should point to the work or official document supporting the adjacent proposition, and its full bibliographic identity should appear in References and a claim-level evidence ledger. A DOI helps identify a work but does not prove that every assertion about it is accurate. Publisher abstracts and repository descriptions support the propositions they actually report; more detailed substantive claims need the relevant full text or corroborating evidence. This work records such source-review depths and does not claim that restricted full texts were completely examined.
Worked Design: Legal Assistance
Consider a hypothetical administrative appeals program that has limited capacity to offer additional lawyer consultations. The study asks whether assignment to a timely offer changes a specified procedural outcome for eligible appellants. This example is a methodological exercise; the study was not implemented, and no results or effect sizes are asserted.
The population might consist of individuals filing a defined type of administrative appeal during a fixed registration window, excluding persons whose entitlement to assistance makes random allocation impermissible. The unit of assignment would be the eligible person or case, explicitly chosen in advance where one person can file multiple appeals. The proposed primary outcome might be successful completion of required procedural steps within ninety days after an objectively defined eligibility time. Secondary outcomes might include delay, participation, actual receipt of advice, perceptions of procedural fairness and eventual disposition. Each outcome has a distinct operational definition and legal relevance.
The design distinguishes assignment to an offer from attendance and from sustained representation. Greiner and Pattanayak's (2012) randomized legal-assistance study illustrates the importance of these differences in a particular unemployment-benefits appeal context. Assignment-based intention-to-treat analysis asks whether being offered assistance changed an outcome under randomized assignment. Comparing those who ultimately used services with those who did not would generally reintroduce self-selection. A causal claim about actual use would require additional assumptions and an identification strategy adequate to the treatment.
Implementation would require checking the relevant statutory right to assistance, rules for service allocation and consent, the agency's capacity, independent ethics or institutional review when required, and the security of case data. A study must not withhold legally required services or create coercion through the appearance that research participation controls an ordinary entitlement. A lawful design could study the allocation of additional optional capacity while preserving universally required assistance, but that determination can only be made against applicable law at the actual institution.
The analytic plan should specify assignment integrity, randomization balance, missing outcomes, crossover, attrition, interference, preregistered primary contrasts where appropriate and the handling of multiple outcomes. Interpretation should state the exact legal and administrative setting. An observed effect on time to complete paperwork would not by itself prove improved substantive justice, and a null result with wide uncertainty would not establish an absence of meaningful benefit. This worked example shows how legal constraints are part of identifying the causal question rather than a downstream adjustment to its results.
Worked Design: Court Digitalization
Suppose regional trial courts adopt an electronic filing platform over several years. The question is whether deployment reduces time to disposition for a defined class of newly filed civil claims. This is a proposed research design, not a report of any observed deployment. Its central challenge is that digitalization changes both institutional behavior and the recording system used to observe behavior.
The first step is legal and administrative source reconstruction. The researcher must identify which courts and claim categories were included, the enactment and commencement of relevant procedural rules, actual dates when filing became available, transitional exemptions, staff training, outages and the migration of existing cases. It is essential to distinguish a policy announcement from actual exposure. Changes in fee schedules, jurisdictional thresholds or case-management standards introduced during the same period may become competing explanations.
The outcome must specify a clock. Time from initial lodging to first hearing differs from time from valid filing to final disposition and from the administrative interval between database entry and closure. A platform may improve timestamp precision without accelerating adjudication. The official federal caseload report illustrates how aggregate case counts can be distorted by procedural concentrations such as multidistrict litigation. A court digitalization study should therefore document case mix, bulk migrations, transferred records and duplicated or reopened proceedings.
A staggered difference-in-differences analysis could compare adopting courts with appropriate not-yet-treated or never-treated courts, provided the identifying assumptions are defensible. Courts with the greatest backlog might adopt first, making treatment selection endogenous to outcome trends. The serial-correlation problem identified by Bertrand, Duflo and Mullainathan matters when monthly observations are nested in a limited number of courts. Goodman-Bacon's treatment-timing decomposition further shows why a simple two-way-fixed-effects coefficient can be misleading if effects differ across cohorts and time. The number of court-month rows does not create independent jurisdictional reforms.
The research design would record adoption cohorts, implementation audits, actual data-field changes, pretrends, exclusion rules, case characteristics, missingness and sensitivity to alternative outcomes. Staff interviews and workflow observation could test whether registration practices changed at the same time as filing speed. A carefully bounded finding might show a change in recorded delay associated with deployment; a causal claim would additionally require an adequately supported counterfactual. The original manuscript claims neither an observed reduction nor a computed estimate.
Worked Design: Unmet Legal Needs
A third hypothetical study examines an apparently paradoxical pattern: official housing-court filings decline while survey respondents increasingly describe unresolved housing problems. A superficial interpretation could identify deteriorating access to justice, but two series constructed from different populations and units cannot yet establish either agreement or contradiction. The study must reconstruct each data-generating process.
The survey frame should cover residents at risk of housing problems, including people who have never approached a lawyer or court. The OECD legal-needs survey model supplies a structure for recording specific problems, impacts, help-seeking, responses and outcomes. Question wording, selection and nonresponse should follow AAPOR survey quality guidance. A fixed recall period and a clear definition of problem episodes are required because one individual may report several disputes and a single dispute may persist across survey waves.
Court records would cover a precisely defined housing jurisdiction and reference period, with rules for transferred, consolidated, withdrawn and terminated proceedings. Researchers should identify whether a reform changed filing eligibility, online registration, fee arrangements or early mediation. Qualitative interviews with residents, advisers and officials can investigate whether the decline in filings reflects more settlement, greater deterrence, jurisdictional relocation, improved informal resolution or a reporting artifact. Each mechanism predicts different kinds of observable traces.
The legal component maps the controlling landlord-tenant statutes and procedural entitlements at the relevant dates. Even convincing empirical evidence that people encounter barriers does not independently determine whether those barriers violate an enforceable legal right. A claim of statutory breach needs separate elements, legal thresholds and authority. Conversely, the presence of a statutory remedy does not demonstrate that people experience meaningful access in practice.
An integrated analysis might identify more unresolved problems in particular demographic or geographic groups and fewer filed proceedings in the same period, but causal explanation would remain conditional on evidence about complaint behavior, problem incidence and procedural changes. Interview evidence could narrow rival mechanisms but should not be presented as a population estimate. Data protection measures are particularly important where housing records can be linked to income, health, migration or family circumstances. This is a design illustration with no fieldwork, administrative extraction or numerical finding.
Methodological Synthesis and Limits
The preceding analysis identifies a recurring sequence of obligations in empirical legal research. The investigator must formulate a proposition whose inferential status is explicit, identify the relevant legal authorities and dates, define the target population and observational unit, examine how the institution or survey produced the record, select an analytic approach capable of answering the question and explain what alternative interpretations remain possible. These are practical research decisions, each grounded in established methodological literature and official source documentation.
The analysis also exposes three common transitions that require explicit justification. First, the transition from law as written to law as practiced depends on observation and cannot be inferred from normative text alone. Second, the transition from a recorded institutional population to the broader community of people with legal problems depends on sampling and a credible account of selection. Third, the transition from observed change to causal effect depends on a defensible comparison of potential outcomes. The legal significance of any result then requires a further account of competent authority, applicable normative standards and institutional context.
Judicial datasets are often exceptionally large yet limited to selected recorded events. The U.S. federal judicial caseload reporting, the state-focused Court Statistics Project, the Council of Europe's CEPEJ evaluation and the ECtHR's HUDOC database illustrate distinct institutional constructions of judicial evidence. The differences among them are part of the substantive explanation, not inconvenient noise to remove before comparison.
The qualitative and quantitative traditions address complementary aspects of legal life. Observation and interviews can reveal mechanisms, meanings and hidden institutional work; appropriately sampled surveys can describe wider populations; administrative datasets track recorded activities; and experimental or carefully justified quasi-experimental strategies can estimate specified interventions. Each method retains its own forms of bias and uncertainty. Their combination produces deeper explanations only when source dependencies, units and interpretive assumptions are stated.
The manuscript's direct evidentiary limits follow from its genre. It is a scholarly methodological synthesis based on selected official and academic sources at documented inspection depths. It does not establish a new dataset of legal outcomes or attempt a universal quantitative comparison of countries. Some academic works were checked through publisher or repository records, indexed abstracts and bibliographic metadata, and the argument confines their use to the supported methodological propositions. Legal and ethics references are jurisdiction-limited and time-sensitive. The three examples are hypothetical protocols rather than empirical studies or legal advice.
The authorial contribution is an integrated and original reading of how legal validity, institutional data production, sampling, identification and ethics constrain one another. It is intended to help researchers select adequate methods and recognize unsupported inferential transitions. It proposes no new named doctrine, conceptual system or theory, and makes no unsupported claim to first discovery of a research method.
Conclusion
Empirical legal research advances legal knowledge by examining how law is experienced, practiced, administered and enforced through observable evidence. Its essential methodological requirement is alignment between the research question, the population, the unit of observation, the source-production process and the inference claimed. A judicial disposition can establish an event in a legal record; a sample of dispositions cannot automatically establish how legal problems are distributed across society. A documented statute can establish a legal rule; it cannot alone establish how consistently institutions implement it.
Established sociolegal, qualitative, survey and causal methods supply a repertoire of complementary research practices. Interviews reconstruct situated experiences, observation reveals institutional workflows, legal-needs surveys reach beyond formal claims, administrative statistics describe recorded case processing, and experimental or justified observational comparisons estimate narrowly specified intervention effects. Causal estimands, meaningful denominators, coherent sampling frames and legally accurate timing determine the force of resulting conclusions.
The source comparisons in this monograph show that official evidence is always produced under institutional rules. Federal and state court statistics, European judicial-system evaluations and published human-rights judgments have different jurisdictions, coverage and reporting conventions. The scholar must keep original source identity, legal authority, text versions, procedural stage and publication selection visible through data extraction and analysis. These obligations are central to the meaning of the observations.
Research ethics and transparency reinforce methodological validity. Participants need appropriate safeguards, sensitive legal information requires lawful and secure handling, and a scholarly argument should expose its selection, measurement and interpretive decisions to scrutiny. A well-documented methodological limitation narrows a conclusion rather than nullifying all evidence; an unsupported legal or causal inference requires substantive correction.
The synthesis establishes a practical discipline for the study of law in operation: define the proposition, authenticate the law, reconstruct the observational population, choose a method suitable for that proposition, inspect competing mechanisms and report only the inference warranted by the evidence. The monograph's original contribution lies in the detailed articulation of those connections and the three worked hypothetical designs. Its conclusions are methodological, source-bounded and independent of any claim to original fieldwork, a novel theoretical system or unperformed external peer review.
References
- Epstein, Lee, and Gary King. 2002. “The Rules of Inference.” University of Chicago Law Review 69 (1): 1–133. Source
- Theil, Stefan. 2025. “Carefully Tailored: Doctrinal Methods and Empirical Contributions.” Oxford Journal of Legal Studies 45 (4): 1047–1075. https://doi.org/10.1093/ojls/gqaf029. Source
- Macaulay, Stewart. 1963. “Non-Contractual Relations in Business: A Preliminary Study.” American Sociological Review 28 (1): 55–67. Source
- Galanter, Marc. 1974. “Why the ‘Haves’ Come Out Ahead: Speculations on the Limits of Legal Change.” Law & Society Review 9 (1): 95–160. Source
- Greiner, D. James, and Cassandra Wolos Pattanayak. 2012. “Randomized Evaluation in Legal Assistance: What Difference Does Representation (Offer and Actual Use) Make?” Yale Law Journal 121 (8): 2118–2214. Source
- OECD and Open Society Foundations. 2019. Legal Needs Surveys and Access to Justice. Paris: OECD Publishing. https://doi.org/10.1787/g2g9a36c-en. Source
- American Association for Public Opinion Research (AAPOR). n.d. “Best Practices for Survey Research.” Accessed November 21, 2026. Source
- Braun, Virginia, and Victoria Clarke. 2006. “Using Thematic Analysis in Psychology.” Qualitative Research in Psychology 3 (2): 77–101. https://doi.org/10.1191/1478088706qp063oa. Source
- Tong, Allison, Peter Sainsbury, and Jonathan Craig. 2007. “Consolidated Criteria for Reporting Qualitative Research (COREQ): A 32-Item Checklist for Interviews and Focus Groups.” International Journal for Quality in Health Care 19 (6): 349–357. https://doi.org/10.1093/intqhc/mzm042. Source
- O’Brien, Bridget C., Ilene B. Harris, Thomas J. Beckman, Darcy A. Reed, and David A. Cook. 2014. “Standards for Reporting Qualitative Research: A Synthesis of Recommendations.” Academic Medicine 89 (9): 1245–1251. https://doi.org/10.1097/ACM.0000000000000388. Source
- Administrative Office of the U.S. Courts. 2025. Federal Judicial Caseload Statistics 2025. Washington, DC: United States Courts. Twelve months ending March 31, 2025. Source
- Administrative Office of the U.S. Courts. 2026. “Federal Judicial Caseload Statistics: Data Tables.” Washington, DC: United States Courts. Includes March 31, 2026 reporting series. Accessed November 21, 2026. Source
- National Center for State Courts and Conference of State Court Administrators. n.d. “Court Statistics Project.” Accessed November 21, 2026. Source
- European Court of Human Rights. n.d. “HUDOC Database.” Council of Europe. Accessed November 21, 2026. Source
- Rosenbaum, Paul R., and Donald B. Rubin. 1983. “The Central Role of the Propensity Score in Observational Studies for Causal Effects.” Biometrika 70 (1): 41–55. https://doi.org/10.1093/biomet/70.1.41. Source
- Hernán, Miguel A., and James M. Robins. 2020. Causal Inference: What If. Boca Raton, FL: Chapman & Hall/CRC. Author's online book page with current freely downloadable edition, accessed November 21, 2026. Source
- Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2004. “How Much Should We Trust Differences-in-Differences Estimates?” Quarterly Journal of Economics 119 (1): 249–275. https://doi.org/10.1162/003355304772839588. Source
- Goodman-Bacon, Andrew. 2021. “Difference-in-Differences with Variation in Treatment Timing.” Journal of Econometrics 225 (2): 254–277. https://doi.org/10.1016/j.jeconom.2021.03.014. Source
- European Commission for the Efficiency of Justice (CEPEJ). 2026. “Evaluation of Judicial Systems: 2026 Cycle.” Council of Europe. Evaluation methodology and cycle documentation, accessed November 21, 2026. Source
- National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research. 1979. The Belmont Report: Ethical Principles and Guidelines for the Protection of Human Subjects of Research. Washington, DC: U.S. Government. Source
- U.S. Department of Health and Human Services, Office for Human Research Protections. n.d. “45 CFR 46: Protection of Human Subjects.” Federal regulations and official guidance, accessed November 21, 2026. Source
- European Parliament and Council. 2016. Regulation (EU) 2016/679 of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data (General Data Protection Regulation). Official Journal of the European Union L 119: 1–88. Source
- von Elm, Erik, Douglas G. Altman, Matthias Egger, Stuart J. Pocock, Peter C. Gøtzsche, and Jan P. Vandenbroucke. 2007. “The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies.” PLoS Medicine 4 (10): e296. EQUATOR Network guideline. Source