ICEQC-R-2005-06 — Defining Minimum Evidence for Educational Quality in Expanding Systems

Expansion is commonly observed through more institutions, places, enrolments or expenditure. Each indicates scale but not necessarily the service received. An additional enrolment can represent a child attending a functioning school, a duplicate registration, a learner placed in an overcrowded class, or a name retained after withdrawal.[REF-12]

The appropriate starting question is therefore not only how many places were added, but what educational service became available to which population. This requires evidence at several transitions: population to entry, entry to participation, participation to progression, provision to teaching and teaching to learning.[REF-13]

Expansion changes the composition of learners, institutions and staff. New entrants may come from poorer, more remote or linguistically diverse populations. New schools may be smaller, temporary or difficult to staff. Newly recruited teachers may have different preparation. A stable average can therefore conceal offsetting improvement and deterioration.[REF-14]

Time series also face structural breaks. The school census may add previously unrecorded institutions; age data may improve; definitions may change; a new assessment may cover a different grade. The break is part of the evidence and is not smoothed away merely to preserve an apparent trend.[REF-01]

Evidentiary substitution occurs when one stage is reported as if it proved another. An appropriation is treated as expenditure, expenditure as resource receipt, receipt as service, service as participation or participation as learning.[REF-02]

Each substitution can produce a favourable statement unsupported by the observed source. The minimum-evidence framework interrupts the chain by requiring an explicit claim class and a source that observes that class.[REF-03]

Source: ICEQC minimum-evidence claim classification.

System finance and policy become education through institutions, classes, teachers and scheduled time. Expansion can expose bottlenecks at any of those points. National teacher growth may be sufficient in total but poorly distributed; classrooms may be built without recurrent staff; materials may reach districts but not pupils.[REF-04]

The 2004 *World Development Report* demonstrates why inputs and provider records cannot be treated as service by themselves. Its service-delivery evidence supports a general measurement lesson: administrative status requires confirmation at the point where the public service is received. It does not justify automatic attribution of individual fault.[REF-08] [REF-05]

The statement that expansion “dilutes quality” is too general. It can refer to fewer resources per learner, larger classes, less prepared teachers, reduced time, weak curriculum coverage or lower measured learning. Each is tested separately.[REF-06]

Nor is dilution inevitable. Expansion may bring new finance, better organisation and reduced exclusion. The evidence compares defined conditions across cohorts, institutions or periods and examines composition. It does not assume that growth causes decline.[REF-07]

As participation rises, those still excluded may face more complex barriers: remoteness, disability, conflict, poverty, language, discrimination or household responsibilities. Aggregate growth can make this smaller population statistically less visible while its rights claim remains material.[REF-08]

Minimum evidence therefore includes an exclusion account. It identifies what is known about children outside the recorded system, the limitations of population and household sources, and the actions needed to improve visibility. It does not calculate an exact residual by subtracting two uncertain totals without reporting their errors.

The 2005 development debate places substantial emphasis on scaling finance and essential services. Additional resources are necessary in many systems, but their educational value depends on planning, procurement, staffing, distribution and recurrent capacity.[REF-06] [REF-07]

Minimum evidence follows the resource to operation. It also records implementation time. Funds released near the end of a fiscal year may not produce teachers, facilities or materials within the same school year. Comparing annual expenditure with an annual outcome without alignment can misstate both performance and delay.

Expansion produces strong public claims and legitimate expectations. Credibility is protected when reports distinguish preliminary counts, estimates, administrative completion and verified service. Revision of an early figure is not institutional weakness where the method and reason are clear.

The Fundamental Principles of Official Statistics provide a relevant public-evidence model: professional methods, source accountability, correction and confidentiality. Education-system evidence is wider than official statistics, but the same disciplines apply to public quantitative claims.[REF-14]

An evidence threshold states what support is necessary for a proposition. A service threshold states the educational condition required. A policy target states the result and time to which an authority has committed. These must not be merged.

A system can possess strong evidence that a service threshold is not met. That is high-quality evidence of an adverse condition. Conversely, it can report that a target has been met on weak or incomplete evidence. The evidentiary judgement and the service judgement remain separate.

The word minimum imposes discipline on what cannot be omitted. It does not encourage the least possible data collection regardless of decision. The minimum increases with the breadth, consequence and precision of the claim.

A local repair decision may require a verified facility condition and users affected. A national assertion of equitable quality requires representative and disaggregated evidence across several domains. The same instrument cannot satisfy both merely because it contains many questions.

Every material claim includes proposition, population, period, unit, source and evidence-quality account. The last covers coverage, missingness, definition, timing and known error. A citation or data source name is not a substitute for these elements.

Where a value is calculated, the numerator, denominator and exclusions are available. Where a judgement is qualitative, the criterion and basis for classification are stated.

Direct evidence observes the asserted condition. A payroll is direct evidence of payment; it is indirect evidence of teacher availability. A timetable is direct evidence of planned instruction; it is indirect evidence of delivery. Directness is always defined in relation to the claim.

Indirect evidence can support screening or corroboration. It cannot carry a stronger conclusion merely because it is official or comprehensive.

Coverage identifies institutions, learners, staff, events or periods represented. It distinguishes eligible, observed and missing units. A national label does not prove national coverage.

Coverage also has a distributional dimension. If remote institutions, non-government provision or particular learner groups are systematically absent, the aggregate can be precise for the wrong population.

The reference period is the period observed, not the publication year. International tables can bring together values from different years because data availability differs. Every value retains its reference year.

For changing systems, a value several years old may be unsuitable for current allocation while still valid as historical evidence. Its use follows the decision and pace of change.

Consistency across sources supports a claim only where the sources are independently derived and definitions align. Identical figures copied from one administrative return do not constitute corroboration.

Differences are reconciled rather than averaged. A teacher total can differ because one source counts persons and another posts or assignments. The discrepancy contains information about the system and cannot be corrected until the unit is known.

Uncertainty comes from sampling, population estimates, missing records, measurement, classification and model assumptions. The report names the material sources and quantifies them where the method supports doing so.

False numerical precision is not preferred to an honest range. A range does not weaken a decision where every plausible value leads to the same conclusion. Where plausible values cross a decision boundary, more evidence or a precautionary rule is required.

Routine institutional records are indispensable and need not be rejected because the provider created them. Higher-consequence claims require verification where incentives, error or control weakness could materially affect the result.

Verification can use source reconciliation, sample visits, independent observation, audit or repeated measurement. It is matched to the claim and risk. A financial audit cannot by itself verify learning, and a classroom visit cannot verify national expenditure.

The evidence burden is proportionate to decision consequence, uncertainty, variation and cost. Universal observation is not necessary for every system claim; a probability sample may provide stronger national evidence at lower burden. Every individual safety case, however, may require direct action despite the absence of an aggregate estimate.

Proportionality prevents both excessive collection and evidentiary neglect. It does not permit a high-stakes national claim to rest on convenient examples.

Some evidence gaps cannot be offset by strength elsewhere. Learning results do not establish safe access. High enrolment does not resolve discrimination. Adequate national finance does not prove school operation. These dimensions are connected but not interchangeable.

A summary judgement preserves critical exceptions. If an overall index is produced for analytical screening, its components and non-compensable conditions remain visible and govern action.

Source: ICEQC minimum-evidence sufficiency test.

The test produces four outcomes: sufficient for the defined claim; sufficient for a qualified or provisional claim; insufficient but remediable through specified evidence; or incapable of supporting the proposed inference. The outcome is recorded before the result is used for ranking, funding or public assurance.

Insufficient evidence does not prove poor service, and it does not prove acceptable service. It defines what remains unknown and the decision that can safely be made in that uncertainty.

An access measure begins with the population to whom the educational commitment applies. School-age population estimates can derive from a census and projections whose reference date, age reporting and migration assumptions differ from education records. The source and projection year are therefore part of the measure.

Where civil registration or a recent census is weak, the denominator can carry substantial uncertainty. The system reports that uncertainty and supplements ratios with administrative counts and local identification. It does not describe a residual number of out-of-school children as exact when both population and enrolment totals are uncertain.

The institution register identifies operating schools or other authorised service points, location, level, provider, status and opening period. Planned, registered, constructed, temporarily closed and operating institutions are distinct states.

Coverage is reconciled with district lists and finance or staffing records. Expansion can bring community, private or temporary provision into operation before national registers are updated. Omitting it distorts access; including unverified sites can overstate service.

Enrolment counts distinguish unique persons from registrations. Learners attending more than one programme, transferring during the year or repeating a grade can appear more than once. Where stable identifiers are unavailable, reconciliation uses name, age, sex, location and prior school under appropriate protection, while acknowledging residual duplication.

The purpose is a credible aggregate, not creation of unnecessary personal records. Public evidence remains de-identified, and access to individual linkage is controlled.

The gross enrolment ratio compares enrolment at a level, regardless of age, with the population of the official age group for that level. It captures system participation and the effect of early or late entry and repetition, but can exceed 100 per cent.

The net enrolment ratio restricts the numerator to learners of official age enrolled at the level. It more closely addresses age-specific access but depends on accurate ages and excludes enrolled learners outside the official range. Neither ratio establishes attendance, progression or learning. International comparison retains the level definition and reference year.[REF-09] [REF-10]

Entry evidence distinguishes new entrants from repeaters and transfers. The gross intake ratio can exceed 100 per cent because new entrants may be above or below the official entry age. A net intake measure narrows the age basis.

Minimum evidence for an expansion claim includes the number of new entrants, age distribution, sex and material geographic distribution, with the relevant population estimate. Where first-grade enrolment is used as a proxy for entry, repetition is disclosed.

Enrolment is an administrative status. Attendance records whether a learner was present under a defined school calendar and counting rule. Active participation may require more, such as engagement in distance, non-formal or irregular provision where daily attendance is not the organising form.

The denominator distinguishes possible attendance days for the learner from school-open days. Absence caused by school closure is not attributed to the learner. Partial days, late entry and transfer follow declared rules.

An enrolled child cannot attend a school that is closed or lacks the scheduled service. Minimum evidence therefore includes operating days, unscheduled closure and the affected population. The official calendar is a criterion, not evidence of delivery.

Closure reasons are classified: weather, insecurity, facility failure, public event, teacher availability, health condition or administrative action as applicable. The system does not infer fault from the count alone.

Administrative records do not observe children who never enrol or have left without transfer. Household surveys, population censuses, civil registration and community identification can provide additional evidence, each with coverage and timing limits.

The out-of-school estimate states age range, education level, attendance or enrolment concept, survey period and treatment of non-formal provision. It examines those who have never attended separately from those who left where evidence permits, because policy responses differ.

Formal abolition of a fee does not establish cost-free access. Minimum evidence examines authorised charges, actual payments and material indirect costs such as uniforms, materials or transport where they constrain participation.

Household evidence is required for costs not recorded by institutions. It is interpreted with income, location and household composition where feasible. A national average cost can conceal a prohibitive burden for the poorest group.

Distance, transport, disability access, language, safety and discrimination affect whether an available place can be used. Minimum evidence follows the risk. It need not create a uniform global variable for every barrier, but it must not report access as universal without examining known material exclusions.

Verified individual cases require remedy even where prevalence cannot be estimated. Aggregate measurement and case responsibility proceed together.

Source: ICEQC access and participation interpretation table.

A claim of expanded access requires evidence of additional unique participation within a defined population and period, together with material distribution and operation. Enrolment growth can support the claim only when duplication, level, age and coverage are understood.

A claim of universal access requires a stronger record: credible population denominator, observation of those outside administrative systems, relevant accessibility evidence and unresolved cases. Where these are absent, the correct statement is progress in recorded participation, not universality.

Enrolment at a grade is a stock measured at a date or during a period. Progression, repetition, transfer and dropout are flows between states. Comparing stocks in adjacent grades can suggest loss but cannot identify each flow without further information.

Expansion increases intake and can make upper-grade enrolment rise even when progression rates do not improve. Cohort and reconstructed-flow measures are therefore necessary for claims about continuity.

Promotion evidence identifies learners enrolled in grade g during year t who enter grade g+1 in year t+1 under a declared rule. It distinguishes transfer to another institution, repetition and unknown status.

Where only aggregate grade totals exist, promotion can be estimated through reconstructed flows with assumptions. The result is described as an estimate and retains the method. It is not represented as linked individual progression.

Repeaters are learners enrolled in the same grade for a second or subsequent year. Definitions differ where learners repeat only examinations, repeat selected subjects or return after interruption.

Minimum evidence states the rule and level. A high repetition rate can reflect learning difficulty, assessment policy, curriculum, attendance or deliberate retention. The measure identifies a condition requiring enquiry; it does not assign cause.

Dropout is not simply absence from the originating institution. The learner may transfer, migrate, defer or enter another recognised form of education. A system without reliable transfer linkage reports unknown destination separately.

Household follow-up or local tracing can improve classification, subject to privacy and protection. The estimate states the follow-up period because temporary interruption and permanent exit cannot be distinguished immediately.

Survival rates estimate the proportion of a cohort expected to reach a specified grade, often through reconstructed cohort methods where individual data are unavailable. They depend on assumptions about repetition and flow stability.

The method, entry cohort or synthetic basis and terminal grade are reported. A survival estimate is not a learning measure and does not show whether the learner completed the grade.

Completion requires a defined programme endpoint and evidence that requirements were fulfilled. Administrative completion counts, graduates and final-grade enrolment can differ.

A gross primary completion ratio may use non-repeating enrolment in the final grade divided by the population at the theoretical graduation age. It provides a proxy where direct completion data are incomplete but can be affected by age and repetition. It is not the proportion of an actual entry cohort known to complete.

On-time completion and eventual completion answer different questions. Expansion can increase late entry and repetition, making a fixed-age measure lower even where more learners ultimately finish.

The report states the observation window. Cohorts still capable of completion are not coded as non-completers merely because the reporting date arrives.

Transition measures movement from one completed level to the next. The numerator and denominator require consistent completion and entry definitions. Capacity constraints at the receiving level can limit transition despite completion at the prior level.

The measure is disaggregated where material inequalities are plausible. A higher aggregate transition rate can coexist with reduced access for a particular region if capacity grows elsewhere.

Retention is not always favourable. Learners can remain enrolled without regular attendance or meaningful instruction. Conversely, a transfer or movement to appropriate alternative provision is not necessarily failure.

Minimum evidence therefore connects continuation with service operation and educational opportunity. It avoids using retention alone as evidence of quality.

Entry cohorts can differ in age, prior preparation, geography and social conditions as expansion reaches new populations. Outcome comparison reports these changes and avoids attributing the whole difference to institutional performance.

Equity requires that lower outcomes among newly included groups do not become a reason to restrict access. The evidence identifies additional support and system conditions needed for successful participation.

Source: ICEQC progression and completion interpretation table.

An expansion report supports a continuity claim only when it observes or validly estimates movement through the programme, distinguishes relevant statuses and identifies the cohort or model. Cross-sectional enrolment is insufficient by itself.

Where linkage is weak, the minimum responsible conclusion reports grade distribution, repeaters, known completions and unresolved flow. Improving learner-flow evidence becomes a system action, not a reason to present the available proxy without qualification.

Teacher evidence distinguishes authorised posts, funded posts, filled posts, unique persons, full-time equivalents and assignments. The appropriate count depends on whether the claim concerns finance, employment, available service or school deployment.

Duplicate persons and multiple assignments are reconciled. Temporary, volunteer, contract and private-sector teachers are included or excluded according to the published scope.

The proportion of trained teachers uses the national minimum organised teacher-training requirement. It does not establish a common content, duration or level across systems. Unknown status remains separate.

The report identifies education level, provider sector, teacher category, definition and reference year. It does not treat recognised status as direct evidence of current classroom practice.

National teacher sufficiency can coexist with local vacancy or subject shortage. Minimum evidence includes distribution by school or district, level and relevant subject or language where those conditions affect service.

Remote and small schools require careful interpretation. A low ratio can reflect the indivisibility of a minimum staff, while a favourable district average can conceal a school with no assigned teacher.

The pupil–teacher ratio divides pupils by teachers at a stated level and sector, preferably using consistent full-time-equivalent concepts where available. It is not average class size. Class formation, shifts, contact hours and non-teaching duties alter the relationship.

A target ratio used for planning is identified as a parameter, not a universal quality standard. The system reports the distribution because the national average can be produced by severe surplus and shortage together.

Employment and assignment do not establish presence. Observation or reliable attendance evidence is necessary for a presence claim. The reference day, schedule, reason and repeated coverage matter.

Authorised leave, official duty, training and unexplained absence are distinguished. Presence evidence is used first to assess service availability and system arrangements; individual employment conclusions require due process and a suitable evidentiary basis.

An available teacher must be assigned to the required class, grade or subject. Minimum timetable evidence identifies scheduled periods, assigned staff, uncovered periods, substitutions and conflicts.

The planned timetable is compared with delivery for a sample or full period appropriate to the claim. An issued timetable cannot establish instructional time without an operating record.

Instructional time is progressively reduced from intended calendar time through school closure, late start, teacher unavailability, timetable loss and interruption. Each stage answers a different responsibility question.

Minimum evidence states the unit—days, hours or periods—the scheduled denominator, the observed period and affected learners. Estimated duration is identified as such.

Class size counts pupils in an instructional group at a stated time. It differs from enrolment, average attendance and the pupil–teacher ratio. Multi-grade classes, shifts and subject groups require additional definition.

No universal class-size threshold is derived here. Minimum evidence supports local interpretation through distribution, room capacity, teacher organisation and educational purpose.

Materials evidence distinguishes procurement, school receipt, usable relevant stock, learner access and classroom use. Language, curriculum alignment, condition, sharing and storage can materially alter the service.

An aggregate textbook–pupil ratio can conceal wrong titles or concentration. Minimum evidence follows the intended grade and subject and identifies usable stock.

Facilities are counted as usable capacity only when safe, functional and accessible for the intended period and population. Construction completion and inventory status are earlier stages.

Water and sanitation evidence includes functionality, access, privacy, maintenance and seasonal reliability where material. The FRESH framework demonstrates why facilities, policy, skills-based education and service links should be considered together rather than as isolated counts.[REF-13]

Source: ICEQC teaching-service evidence chain.

A claim that expansion maintained teaching conditions requires evidence that appropriate staff, time, organisation, materials and facilities remained available to the added population. It does not require one universal numerical profile.

Where evidence is incomplete, the system reports the domains verified and the schools or populations not observed. A favourable national resource ratio cannot be used to close a known local service failure.

The authorised curriculum establishes what learners are intended to encounter. The minimum evidence record identifies the applicable version, level, subject, language and effective period.

A national curriculum does not prove that institutions possessed the time, staff, materials or preparation to deliver it. Expansion analysis therefore connects entitlement to opportunity.

Institutional plans, schemes and timetables show how the authorised curriculum was scheduled. They can reveal omission before delivery and permit analysis of time allocation.

Plans are compared with the authorised requirement. Local adaptation is identified and assessed under the applicable authority; it is not assumed invalid or sufficient merely because it is documented.

Minimum evidence of taught content can draw on teacher records, observation, learner work and assessment tasks. Each has limitations. Teacher records can be comprehensive but self-reported; observation is direct but narrow; learner work demonstrates opportunity for selected tasks but not every learner or topic.

A representative or risk-based method is chosen according to the claim. The system does not require exhaustive surveillance of teaching, and it does not infer full coverage from one exemplary class.

Opportunity concerns exposure to the required content under workable conditions. It includes time, appropriate teacher, language, materials, accessibility and participation.

An assessment difference can reflect unequal opportunity as well as learning. Minimum evidence interprets outcomes alongside these conditions rather than treating scores as a complete institutional explanation.

An assessment can support classroom diagnosis, certification, selection, system monitoring or research. The design and evidence threshold differ. A classroom test can inform immediate teaching without supporting national comparison.

The public claim identifies purpose. Results are not repurposed for high-stakes comparison when administration, content and population were not designed for it.

The assessment states what knowledge, skill or competence it measures and how tasks represent that construct. Alignment with curriculum is necessary for curriculum attainment claims, while broader assessments require a separate interpretive basis.

Coverage of one domain does not establish overall learning. A literacy measure cannot stand for mathematics, civic purpose, practical skill or the full aims of education.

Minimum evidence includes administration conditions, language, time, accommodations, scoring rules, marker preparation and quality checks. Differences in these conditions can create apparent performance differences.

For constructed responses or performance tasks, a sample of scripts is double-marked or otherwise reviewed where judgement variation is material. Scoring correction is documented.

The assessed population, eligible population and actual participants are reported. Exclusion of absent, disabled, low-performing, remote or non-dominant-language learners can raise the observed result.

Participation is disaggregated where exclusion risk is material. Results for participants are not described as results for all eligible learners without qualification.

Trend claims require sufficiently stable construct, scale, administration and population, or a valid linking method. Reusing the same title for a changed test does not preserve comparability.

Expansion changes learner composition. The system can report overall performance and subgroup distribution while examining whether like-for-like change is supportable. Adjustment methods disclose assumptions and do not erase the achievement of newly included learners.

A pass rate depends on threshold, candidate entry, assessment, marking and rules for absence or resit. Minimum evidence includes those elements and the candidate population.

Pass-rate change can identify an enquiry but cannot establish improved teaching or learning without stable assessment and supporting evidence.

An average score can rise while the lowest-performing group falls or remains without foundational learning. Minimum evidence reports a distribution suited to the assessment: performance levels, percentiles or shares reaching defined competencies, with counts and uncertainty.

Differences by relevant population are interpreted with opportunity, participation and sample size. The report does not label a group as deficient without examining service and contextual conditions.

Learning evidence informs curriculum, teaching support, resource allocation and equity analysis. It does not authorise an intervention merely by identifying a low score. Cause and institutional responsibility require additional evidence.

Publication protects learners and avoids school rankings unsupported by design. Sampling and measurement uncertainty are retained near the result.

Source: ICEQC minimum learning-evidence record.

A system can claim learning at a stated level only for the construct, population and period supported by the assessment. It reports participation, distribution and uncertainty. It does not convert attendance, completion or teacher qualification into learning evidence.

Where no suitable learning assessment exists, the system reports that limitation and strengthens opportunity-to-learn evidence while developing an appropriate measure. The absence of assessment is not filled with an input proxy presented as outcome.

An equity claim concerns how access, service, participation and learning are distributed. A national average provides no evidence of distribution unless the underlying variation is examined.

The minimum evidence identifies groups and places for which exclusion or unequal service is plausible in the system concerned. It does not require collection of every characteristic without purpose. It does require that known structural disadvantage is not omitted because it is difficult to measure.

Sex-disaggregated enrolment, attendance, progression, completion and learning evidence are basic to the international commitments as at 2005. Ratios and parity indices retain the underlying female and male values because a single parity value can conceal low participation for both.

Parity in enrolment does not establish parity in attendance, completion, treatment or learning. Evidence follows the stage relevant to the claim and examines programmes or regions where the national result hides difference.

Institutional records rarely contain reliable household income. Household survey, consumption, asset or geographic measures may provide a basis, each with limitations. The system states the measure and does not present one proxy as exact poverty status.

Cost, distance, work and household responsibility can affect entry and continuity. Equity evidence relates the barrier to the educational stage and avoids attributing low participation to household preference without examining the service and cost environment.

National and regional averages can conceal travel distance, sparse settlement, weather and weak transport. Minimum evidence includes location and a meaningful access or service classification. Administrative urban–rural categories are used with their national definition.

Remote schools can have higher unit costs and lower pupil–teacher ratios because minimum service is indivisible. Efficiency comparison accounts for geography and does not infer over-resourcing from the ratio alone.

Evidence concerning learners with disabilities is often incomplete because identification rules, stigma, inaccessible assessment and non-enrolment differ. A recorded count is not assumed to represent prevalence.

Minimum evidence distinguishes the purpose of identification, the educational barrier, accommodation or support and participation. It protects personal information. Verified inaccessibility requires action even where the system cannot estimate the national population reliably.

Language affects enrolment, understanding, teaching and assessment. The evidence identifies the language of instruction, learner language where lawfully and feasibly recorded, availability of materials and assessment language.

A group result in a non-primary language cannot be interpreted solely as a learning difference without examining language opportunity. Minority status is collected and reported with protection against identification or discriminatory use.

Population denominators, school lists and learner records can change rapidly under conflict or displacement. Exact coverage may be impossible. The system reports dated counts, source, mobility and unobserved areas rather than presenting a stable rate.

Protection limits detail. Location or personal information is not published where it could create harm. Temporary provision is assessed for safe operation, continuity, language and recognition without being treated as an acceptable permanent minimum.

Equity evidence includes distribution among schools, not only learner groups. Expansion can create a new tier of poorly staffed or temporary institutions while established institutions retain stronger service.

The system compares institutions on defined service conditions and context. It does not use raw outcome ranking as an equity measure. Higher results can reflect selection, prior opportunity and population difference.

Disadvantage can combine across sex, poverty, disability, language and location. Analysis follows combinations where policy relevance and sample size permit. It does not multiply categories until results become unstable or identifying.

Small populations are reported through counts, pooled periods or qualitative and case evidence as appropriate. Statistical imprecision does not remove the underlying right or need.

Schools and households least connected to administration may be missing more often. A national response rate can therefore conceal systematic absence. The evidence report compares response across known geography, provider type and other frame characteristics.

Follow-up and alternative sources focus on under-observed populations. Weighting can adjust for known selection differences but cannot recover a condition that was never measured.

The minimum equity threshold is not numerical parity across every measure. It is sufficient evidence to determine whether material groups receive the defined service and whether expansion reduces, preserves or widens the relevant gap.

Where no reliable group denominator exists, the system can use verified service maps and case evidence while developing stronger population information. It cannot declare equality from the absence of a measured difference.

Source: ICEQC minimum equity-evidence matrix.

A claim of equitable expansion requires distributional evidence across the stages relevant to access and quality. Parity at one stage cannot stand for equality across the chain.

The conclusion identifies remaining gaps, measurement limitations and responsibility. It avoids describing disadvantaged learners as the cause of lower system performance and directs analysis to opportunity, service and support.

Public finance passes through appropriation, release, commitment, expenditure, receipt and operation. Each state has a distinct record. A budget is not expenditure; expenditure is not receipt; receipt is not educational service.

Minimum evidence for a finance-to-quality claim follows the material resource at least to the operating condition it was intended to support. The length of the chain depends on the claim and timing.

Education expenditure is classified by level, function, economic type, government level and provider where feasible. The report states whether it includes capital, recurrent, external, local or household expenditure.

Changes in classification or decentralisation create breaks. A transfer between levels of government is not new education expenditure merely because it appears in both accounts.

Nominal expenditure can rise while purchasing power falls. Trend evidence identifies price adjustment, base year and currency conversion where used. Exchange-rate conversion serves a different purpose from domestic price adjustment.

Purchasing-power comparisons can improve international interpretation but depend on methods and do not directly price the particular education inputs required in each system. The method follows the claim.

Per-learner expenditure divides a defined expenditure total by a defined learner count. Full-time-equivalent and headcount learners can produce different results. The period and level must align.

During expansion, denominator growth can reduce the average even where total resources rise. The measure does not show distribution or efficiency without additional evidence.

Unit cost varies with school size, geography, salary structure, curriculum and capital stage. A remote minimum-service school may have a high unit cost because one teacher or facility cannot be divided.

Cost comparison specifies the service and population. A lower cost is not evidence of efficiency if instructional time, safety or curriculum is reduced.

Teacher recruitment creates a recurrent salary and support obligation. Minimum evidence for a staffing expansion includes funded posts, appointment, deployment, payroll entry, availability and projected recurrent cost.

Short-term finance can support urgent recruitment but does not establish sustainability. The system identifies the period financed and the consequence when temporary support ends.

New classrooms require teachers, materials, water, maintenance and operating funds. A capital completion claim is valid at the construction stage; a service claim requires the full operating package.

The expansion plan records commissioning date and recurrent readiness. Unused completed facilities remain visible rather than being counted as active capacity.

Procurement evidence includes specification, quantity, price, supplier, delivery and acceptance. Educational evidence adds relevance, usable condition, distribution and access.

Late delivery can make a resource ineffective for the intended school year even if expenditure is lawful. The report aligns financial and educational periods.

Public expenditure can increase while households continue to bear fees, materials, transport or labour. Minimum equity evidence examines material household costs where they affect access.

An intervention is not low-cost merely because teacher, family or community time is unpriced. Material non-cash contributions and maintenance responsibilities are disclosed.

Allocation formulas express intended distribution. Minimum evidence compares authorised allocation with actual release and receipt, then with the relevant service. Unspent balances are interpreted with timing and procurement, not automatically as efficiency or failure.

School-level receipt evidence is sampled or complete according to risk and system capability. Difference between central dispatch and local receipt identifies leakage, delay or classification error requiring investigation.

Efficiency relates resources to a defined output or result. A lower cost per enrolment is not necessarily efficient when attendance, completion or learning is weak. A higher cost can reflect difficult geography or inclusion.

The claim states the resource boundary, service or outcome, period and contextual difference. Causal attribution to management requires more than cross-sectional association.

Expansion plans test sensitivity to enrolment, salaries, attrition, construction cost, inflation and external finance. A point estimate can conceal a large unfunded obligation.

Minimum evidence identifies assumptions and the variables that could make the service unsustainable. Contingency does not replace the obligation to maintain essential provision.

Source: ICEQC finance-to-service evidence chain.

A claim that additional finance protected quality requires evidence beyond allocation and spending. It identifies which service expanded or remained functional, who received it and which recurrent obligation follows.

Where the service result is not yet observable because implementation is incomplete, the report states the financial stage and expected verification date. It does not anticipate the educational result.

System capacity includes authority, personnel, competence, information, logistics, finance and decision routes. Staff establishment or equipment alone does not establish capacity.

Minimum evidence examines whether the responsible level can complete the critical process within the required period and correct failure. This makes capacity observable through case flow and service reliability.

A current school register is foundational for allocation, statistics and inspection. It identifies operating status, location, level, provider and key service characteristics. New, closed, merged and temporary schools are dated.

Coverage is tested against local lists, finance, examination or population sources as appropriate. Unknown institutions remain an explicit frame limitation.

Learner records support entry, attendance, progression and completion. Minimum capability requires consistent status definitions and a method to handle transfer and duplication.

A national unique identifier can strengthen linkage but is not the only valid approach and requires privacy controls. Where unavailable, aggregate flows and local follow-up are reported with limitations.

Teacher evidence requires a unique person basis, employment status, qualification, assignment and dates. Payroll, establishment and school records are reconciled because each observes a different stage.

Unresolved discrepancies are reported by type and value where appropriate. They are not concealed in one adjusted total without a correction record.

Expansion depends on delivery of materials, construction inputs and support to new sites. Minimum evidence follows orders through dispatch, receipt, usable condition and timing.

Reliability is measured by complete and timely service, not only average delivery. Remote and difficult routes are reported separately where they experience systematic delay.

Intermediate authorities often convert national policy into school staffing, finance, supervision and maintenance. Their workload, delegated authority, travel and technical referral routes affect service.

Minimum evidence identifies case volume, stage, elapsed time, urgent backlog and resolution with operating verification. A count of visits or files processed is insufficient for a service claim.

Data capacity includes clear definitions, collection, validation, correction, analysis and publication. Faster collection is not sufficient if definitions and coverage weaken.

The system identifies critical indicators that can be produced reliably and a plan for improvement. It avoids expanding the form beyond the ability of schools and offices to maintain accuracy.

Verification can be risk-based. High-consequence claims, unusual values, new institutions and under-observed areas receive priority. Universal verification may be infeasible and can overwhelm the system.

The verifier has access, competence and sufficient independence. Findings produce correction and service action, not only an audit note.

Systems maintain a controlled route for schools and offices to correct errors. The original value, reason, authorisation and effect are retained. Publication revisions are dated and explained where material.

High correction rates can indicate active validation or weak original processes. Interpretation examines the type, source and recurrence rather than rewarding a low reported number of corrections.

In conflict, disaster or other disruption, the minimum set becomes more focused: operating sites, learner presence, responsible adults, safe space, essential materials, protection and service continuity. The evidence is dated frequently.

Reduced scope is reported openly. Temporary estimates and local reports can support urgent decisions but do not become stable annual statistics without reconciliation.

Repeated local failures are grouped to identify upstream causes. Similar teacher vacancies can reveal deployment rules; repeated missing materials can reveal procurement or route failure; inconsistent completion can reveal definitions.

The system keeps individual remedies open while addressing the common cause. Aggregation is a tool for institutional action, not a reason to close school cases.

Source: ICEQC minimum system-capacity evidence table.

A system can claim implementation capacity where it can identify eligible institutions and populations, allocate and trace critical resources, observe service operation, correct material error and act on unresolved conditions.

Capability can be partial and improving. The report identifies which claims the current system can support and which remain dependent on estimates or local verification.

Public reporting separates headline results from definitions and technical detail. The headline remains bounded by the evidence. A technical note does not repair an overbroad public statement.

The first layer states the claim, population, period, value and principal limitation. The second gives distribution and trend. The technical layer provides source, calculation, coverage, uncertainty and revision.

Every statistic shows the year or period observed. Publication year appears separately. Regional tables often combine different national reference years; each value retains its year and the aggregate method is disclosed.

A change in publication date does not create a new observation. Repeating an old value in a new report does not evidence current status.

Preliminary data support timely decisions when their coverage and expected revision are clear. The report labels them beside the value and later publishes material changes.

Final status does not mean error-free. It means the defined production and validation stage has closed. Subsequent correction remains possible and traceable.

An estimate infers an unobserved current or historical quantity; a projection describes a future quantity under assumptions. They are not presented as observations.

Methods, key assumptions and uncertainty are stated. A planning scenario is not described as a forecast merely because it contains a future date.

Targets express intended results and time. They are reported beside, not in place of, observed values. A target line in a chart does not imply that the target is an evidence-based universal threshold.

Progress is calculated on a defined basis and acknowledges breaks. The remaining gap is not extrapolated automatically to a conclusion about feasibility.

Missing remains missing unless an estimate is produced by a declared method. Regional aggregates state coverage and do not imply representation of all countries.

Symbols and notes distinguish not available, not applicable, zero and negligible. Blank cells are not open to favourable interpretation.

Material revisions state cause and affected series. A break arising from definition, source, coverage or method is marked at the point of change.

Where a bridge calculation is possible, both old and new bases are shown for an overlap period. The bridge does not erase the institutional reason for change.

Minimum evidence sufficient for national monitoring may be insufficient for ranking schools or countries. Comparison requires common constructs, definitions, periods and uncertainty.

The public report uses comparison to identify variation and enquiry. It does not assign responsibility or quality from a raw position where context and measurement differ materially.

Individual learner and teacher information is protected. Small cells, rare characteristics and narrative detail are reviewed for identification risk.

Confidentiality limits public detail but does not prevent competent authorities from acting on individual cases. Aggregate reporting preserves the material institutional condition.

The public has access to a correction route. Material errors are corrected with date, reason and effect on conclusion. Silent replacement undermines comparability and confidence.

The system distinguishes a corrected error from a later update. Both are legitimate but describe different events.

Source: ICEQC minimum public evidence statement.

A public quality claim is acceptable only when a reader can identify what occurred, to whom, when and on what evidence. Limitations appear with the result and alter the wording where material.

Transparency is not the publication of every record. It is sufficient openness to understand the claim and sufficient protection to prevent harm.

Education systems differ in administrative levels, provider mix, geography, population records and statistical capacity. The minimum-evidence principles are common; the collection architecture is not.

Application begins with the claims the system must support and the sources already capable of observing them. It does not begin with transfer of a large external indicator list.

Where institution, learner and teacher registers are stable, the priority is linkage, service verification, distribution and learning interpretation. High administrative coverage can still conceal definition error and provider incentives.

Sample verification and independent sources test whether records correspond to operation. Greater technical capacity supports more precise claims but also creates responsibility to report breaks and uncertainty accurately.

An annual census can provide broad coverage of enrolment, teachers, facilities and materials. Its minimum controls are an updated frame, clear reference date, response, validation and correction.

Annual frequency may be insufficient for rapidly changing service or emergency response. Targeted term or sample evidence supplements rather than duplicates the census.

Where census or age data are weak, enrolment ratios carry substantial uncertainty. The system reports administrative counts, age distribution where reliable, survey estimates and geographic evidence together.

It avoids exact universal-access claims and invests in denominator quality. Allocation decisions can still use verified school and local population evidence with appropriate contingency.

Remote systems require higher data-collection cost and can experience delayed returns. Minimum evidence protects their inclusion through planned routes, appropriate communication and longer but explicit verification periods.

Service interpretation accounts for small-school indivisibility, multi-grade teaching and seasonal access. Urban resource ratios are not applied as universal efficiency criteria.

Decentralisation can improve local decision-making but create definitional and system-coverage differences. A controlled core of units, classifications and reporting periods supports aggregation.

Local systems retain additional evidence relevant to their decisions. National aggregation tests comparability and reports units that do not meet the common basis.

Public, private, community and other provision may serve significant populations. National access and quality claims identify the provider scope and avoid treating unobserved provision as equivalent to no provision.

Minimum evidence follows the legitimate public interest—learner access, safety, required education and reliable statistics—under the applicable authority. It does not imply uniform institutional control where governance differs.

Routine annual measures may fail under population movement and inaccessible areas. The minimum shifts toward frequent operational mapping, safe access, responsible teaching adults, essential materials, protection and continuity.

Estimates are dated and revised. Public detail is reduced where it creates security risk. The system states which regions and populations remain unobserved.

Urban averages can conceal informal settlements, overcrowding, multiple shifts and migrant learners absent from population registers. Minimum evidence uses local service maps, actual class and shift conditions, and flexible denominator analysis.

Temporary or unregistered arrangements are not ignored. Their educational condition and route to safe, sustained provision remain visible.

Regional comparison uses common definitions where possible and publishes national variation. It does not force false comparability by removing country notes or substituting a regional model for missing national evidence without disclosure.

Aggregates use a stated weighting basis. Coverage accompanies every regional result. A high-coverage aggregate can still omit small countries or fragile systems whose policy importance is not proportional to population weight.

Source: ICEQC system-condition adaptation table.

The framework is applied through functional equivalence: different sources can support the same claim if they observe the relevant population and condition with disclosed limitations. Uniform forms are not required.

The common minimum is integrity of inference. No region is expected to manufacture precision it cannot support, and no region is treated as if resource constraint makes undocumented quality acceptable.

The minimum evidence floor consists of linked, claim-specific records for access, continuity, teaching service, curriculum opportunity, learning, equity, finance and implementation. Each record identifies population, period, source, coverage and limitation.

The floor is non-compensable in critical areas. A favourable learning average does not establish safe access; finance does not establish instruction; aggregate growth does not establish non-discrimination.

Systems can build from a small reliable core. The first priority is a current institution frame, unique and defined participation, teacher and timetable availability, essential functional conditions, basic cohort flow, selected learning evidence, and finance-to-service reconciliation.

Later stages improve linkage, timeliness, disaggregation, assessment, unit-cost analysis and independent verification. Staging order follows the decisions and risks of the system rather than a universal technology sequence.

Evidence has public value when it changes a legitimate decision or reveals an unresolved obligation. Indicator production without an owner, response route or review date can impose burden without improving education.

Every material minimum-evidence gap therefore has a capacity or verification action. Every material service gap has a responsible authority and interim protection where necessary.

Scarce resources require selection, but the selected measures must remain coherent. Removing the outcome or distribution stage can make a low-cost system incapable of detecting failure.

Sampling, rotating modules and risk-based verification can reduce cost. They are preferable to universal low-quality collection when the intended inference is preserved.

Trust follows from accurate scope, visible uncertainty, timely correction and consistency between claims and sources. It does not require that every result be favourable.

A system that reports an unknown condition honestly can improve its evidence and service. A system that converts unknown into compliance removes the signal required for action.

Authorities should define the claims required for planning and public accountability before expanding data forms. They should protect a common core of definitions, fund verification and support schools or districts with weak record capacity.

Targets should retain baseline, denominator and distribution. Finance plans should include recurrent service. Assessment programmes should protect participation and comparability. Public reports should show operating service and outcome rather than activity alone.

Schools and institutions record operation and learner service. Intermediate authorities reconcile, support and resolve cases. National authorities define, finance, aggregate and correct the system record. Statistical and technical bodies protect method and publication integrity.

Responsibility follows authority. A local school is not held responsible for a national population denominator, but it remains responsible for accurate local records and timely correction.

An evidence floor can be misused as a growing checklist. Excessive fields reduce completion, divert teaching time and encourage perfunctory data. Every item identifies the decision and frequency it serves.

Data already produced are reused where valid. Sensitive data are collected only with a legitimate purpose and protection. Fields without an active decision use are removed or sampled.

An overly narrow core can privilege what is easy to count: enrolments, teachers, buildings and expenditure. It then misses instructional time, functionality, curriculum opportunity, learning and exclusion.

The framework prevents this by requiring at least one direct service and one outcome or justified opportunity measure, alongside distribution and capacity. The exact indicator can vary; the evidentiary function cannot disappear.

Minimum evidence is the least body of information capable of supporting a defined quality claim without concealing material uncertainty or harm. It is established through the relationship between proposition, population, source and public consequence.

Expansion raises the value of this discipline. New enrolment changes who is served, where education operates and what capacity is required. A system demonstrates responsible progress when it can show not only that participation grew, but that learners received sustained, safe and equitable educational opportunity and that learning evidence is interpreted within its true limits.

An authority applying this framework first classifies every headline claim as commitment, resource, service, participation or learning. It then tests population, period, unit, source, coverage, uncertainty and distribution.

Claims that pass are published with their limits. Claims that do not pass are narrowed or withheld pending specified evidence. Critical service gaps discovered through the test are acted upon without waiting for a perfect statistical system.

The same datum can have different evidentiary sufficiency depending on its use. A teacher vacancy reported by a headteacher can support immediate district verification and temporary coverage. It may not support a national vacancy-rate estimate unless the school list, response and definition are adequate.

The system therefore records the decision attached to every minimum threshold. Evidence is not labelled “good” or “poor” in the abstract. It is adequate, provisional or inadequate for a stated use.

Screening identifies conditions that may require direct examination. It favours sensitivity where missing a serious case is the greater harm. School returns, complaints, unusual ratios and local reports can all support screening.

A screen is not a final finding. Cases that meet the screen receive verification under a defined rule. The public report does not publish the screened count as the prevalence of confirmed failure.

An individual service decision concerns a named learner, teacher, class or institution. It requires direct and current evidence of the condition, applicable criterion, competent authority and protection of personal information.

Statistical representativeness is not required because the decision does not generalise to a wider population. Fairness and verification are required where the consequence is material.

Allocation evidence identifies eligible units, need, available resource, formula or priority rule and exceptions. The minimum depends on whether allocation is universal, formula-based or discretionary.

Where schools have unequal reporting capacity, documentation quality is not allowed to become an unintended proxy for need. The authority assists verification and records unassessed cases.

Management decisions often require timely evidence before final statistical validation. Preliminary enrolment, attendance, stock and case-flow information can be sufficient when source, coverage and expected revision are known.

The decision remains reversible where uncertainty is material. A later revision is compared with the preliminary value so that systematic early error can be corrected.

Public monitoring requires consistent definitions, coverage, reference dates and transparent revision. It can use administrative data, survey estimates or assessments, but the public wording follows the method.

A measure sufficient to identify internal cases may be unsuitable for public comparison. Confidentiality and the risk of misinterpretation also affect the reporting level.

A policy evaluation asks whether an action contributed to a result, not merely whether both occurred. Minimum evidence includes the intervention, implementation, exposure, outcome, comparison or counterfactual reasoning, and relevant concurrent change.

Where a credible causal design is unavailable, the authority can still report implementation and descriptive outcome change. It does not attach causal language to administrative sequence.

Forecasts require a base population, model, assumptions and sensitivity. They are decisions about plausible future demand, not observed evidence. Enrolment projections include entry, progression, repetition and population assumptions where material.

Teacher and finance projections identify replacement, salary, training yield, deployment and recurrent cost. A single scenario is insufficient where small assumption changes materially alter the requirement.

International comparison adds requirements for common definitions, classification, population and reference periods. National values remain valid for their local purpose even when not directly comparable.

The publication can present non-comparable values with clear notes for descriptive context. It does not rank or calculate a regional average where differences defeat the intended inference.

Decisions affecting safety, individual rights, employment, award validity or substantial public resources require stronger verification and procedure. Aggregate probability is not substituted for the individual facts where an individual determination is made.

Evidence collected for system monitoring is not automatically repurposed for individual sanction. The source, opportunity to respond and applicable due process are examined separately.

Source: ICEQC decision-specific evidence thresholds.

Evidence can move from screening to individual verification, then to system estimation and policy evaluation. Each step requires additional design; it is not achieved by repeating the initial source.

The escalation path is recorded so that a local alert can produce both remedy and wider learning without using the case as if it were representative.

In emergency or severe capacity constraint, the evidence set can be reduced to the elements necessary for immediate safe service. The claim is reduced with it. A rapid count can support urgent supply; it cannot support a final national quality assessment.

The temporary threshold has an expiry and reconciliation plan. Emergency practice does not silently become the permanent statistical standard.

The scenarios are hypothetical and illustrate interpretation. They do not describe a country or establish universal planning values. Each begins with a plausible expansion claim and identifies the minimum evidence required to accept, qualify or reject it.

A system reports that first-grade enrolment rose from 480,000 to 560,000 in one year. The school census response increased from 88 to 96 per cent after previously unregistered community schools entered the frame. The reported increase of 80,000 therefore combines real participation, expanded coverage and possible duplication.

The minimum responsible statement reports the two totals with coverage and frame change. A comparable-school calculation can show growth among institutions observed in both years, while a separate count shows newly covered schools. Neither replaces the current total. A claim that 80,000 additional children gained access requires unique-learner and entry evidence beyond the gross difference.

A capital programme completes 1,200 rooms. Enrolment grows by 90,000 learners, 300 rooms are not operating because teachers or furniture are unavailable, and 150 replace unsafe rooms rather than add capacity.

The valid construction claim is 1,200 completed rooms under the programme's certification rule. Added operating capacity concerns 750 rooms, subject to safe use and scheduling. Replacement improves service but should not be counted as net capacity. The quality interpretation examines class distribution, shifts and instructional time, not the building total alone.

Ten thousand teachers are appointed nationally. Payroll records contain 9,700 new persons, school assignment records contain 9,250, and the first operating check confirms 8,900 at assigned schools. Some difference reflects processing time and authorised preparation.

The report presents each transition. It does not select the appointment figure for a claim about classroom service. Investigation distinguishes delay, duplicate or cancelled appointment, non-reporting, training and unexplained discrepancy. Regional distribution determines whether added service reached expansion areas.

The national pupil–teacher ratio remains 38:1 during expansion. District ratios range from 22:1 to 71:1, and a group of remote schools has one teacher covering several grades. The national stability can coexist with severe local pressure.

Minimum evidence for maintained teacher conditions requires the distribution, vacancy, class organisation and instructional availability. The 38:1 ratio remains a useful aggregate resource measure; it is not evidence that every learner encountered an adequately staffed class.

Reported completion rises from 62 to 70 per cent. In the later year, the terminal-grade rule excludes repeaters from the denominator and records learners completing through an alternative programme that was not included previously.

The system marks a break and reconstructs an overlap where records permit. If the earlier rule applied to the later population produces 65 per cent, that comparison can be reported with limitations. The 70 per cent remains valid under the new current definition but is not presented as an eight-point like-for-like improvement.

Mean assessment performance rises by 12 points while participation falls from 86 to 68 per cent. Absence is concentrated among remote schools and lower-attaining learners according to prior evidence.

The higher participant mean cannot support a system learning-improvement claim. The minimum public account reports participation and distribution beside the score and investigates access, administration and exclusion. A sensitivity analysis can illustrate the range under assumptions, but it cannot recover the missing results as observed facts.

The ministry dispatches 2 million books and reports one book for every two pupils. School receipt confirms 1.72 million, of which 120,000 are for the wrong grade or language and 80,000 are damaged. Usable relevant receipt is therefore 1.52 million before school-level distribution and access.

The dispatch claim, receipt claim and usable-stock claim are all reported accurately. Pupil access requires the intended-grade denominator and evidence that stock was distributed or available under the chosen sharing arrangement. Expenditure and logistics are examined separately from classroom use.

A policy abolishes an authorised primary fee and enrolment rises. Household evidence indicates continuing payments for materials and school maintenance in some districts. The policy commitment is established, but cost-free access is not.

Minimum evidence distinguishes official charge, actual payment, enforcement, household burden and distribution. The action may require financial replacement for schools as well as control of unauthorised charges. Enrolment change alone does not identify the policy effect.

Fifty new schools are registered in remote areas. Forty-two operate throughout the term, five open late, two operate intermittently and one cannot be verified. A headline of 50 new schools is valid as a register addition but not as sustained service.

The operating account shows each status, learners served, teacher coverage and essential conditions. The unverified school remains unresolved. Late and intermittent operation is measured through lost service rather than converted to simple open/closed classification.

An expansion programme finances construction and teacher salaries for three years. National projections assume continuation from domestic revenue, but the salary scenario omits annual increments and replacement recruitment.

The current finance claim is limited to the funded period. Sustainability evidence adds realistic salary, attrition, recruitment, maintenance and enrolment assumptions. A range is published; continuation beyond the funding agreement is not described as secured.

Source: ICEQC hypothetical expansion scenarios.

The examples share no single corrective indicator. They require preservation of states, populations and periods. Most misleading claims arise not from arithmetic error but from substituting one state for another.

Minimum evidence makes those transitions visible. It supports a more precise statement and identifies the next operational decision without dismissing genuine progress.

Before an indicator enters a national quality account, the responsible authority should record its purpose, population, unit, source, reference period, coverage, disaggregation and decision consequence. The record should also state which observations are unknown, which are estimated and which institutional office can correct the underlying condition. This discipline prevents a convenient measure from acquiring a meaning that its data cannot sustain. It is particularly important during expansion, when administrative coverage may grow at a different rate from service capacity and when an unchanged national average can conceal large shifts between districts, schools or learner groups.

Minimum evidence is therefore cumulative rather than singular. An enrolment record establishes participation in a defined administrative sense; it does not establish regular attendance, sufficient instructional time, trained staff or demonstrated learning. A school census establishes reported conditions on a stated date; it does not by itself establish that those conditions persisted throughout the year. Assessment evidence describes performance within its tested domain and population; it cannot repair missing information about exclusion. Each source should retain its proper claim, while reconciliation among sources identifies losses between entitlement, provision and educational experience.

The immediate planning obligation is to publish these boundaries with the result. Authorities should not postpone all reporting until every record is complete, but neither should they convert incomplete coverage into apparent compliance. A provisional figure can support action when its limitations, affected population and correction timetable are explicit. The public standard is a traceable account that permits correction and progressively narrows the areas in which educational quality remains unknown.

The specifications identify the minimum metadata required to interpret common expansion indicators. They do not prescribe that every system publish every indicator or that the listed measure is sufficient for a quality judgement by itself. Selection follows the decision and system context.

**Definition:** institutions delivering the defined educational service during the reference period. **Unit:** institution or service site. **Minimum fields:** level, provider, location, opening and closure status, reference dates and temporary operation. **Exclusions:** planned, constructed but unopened and permanently closed sites. **Limitation:** institution count does not measure capacity, attendance or quality.

**Definition:** unique learners meeting the system's enrolment rule at the reference date or during the stated period. **Unit:** person. **Minimum fields:** level, grade, age or date of birth where available, sex, provider and status. **Controls:** duplicate registration, transfer, late entry and withdrawal. **Limitation:** enrolment does not establish attendance.

**Numerator:** enrolment at the level regardless of age. **Denominator:** population of official age for the level. **Expression:** percentage. **Minimum metadata:** official age range, reference year for numerator and denominator, provider coverage and source of population estimate. **Limitation:** can exceed 100 per cent and does not measure the share of the official-age population enrolled.

**Numerator:** official-age learners enrolled at the level. **Denominator:** population of that official age. **Expression:** percentage. **Minimum metadata:** age basis, reference year, treatment of unknown age and enrolment in other levels. **Limitation:** sensitive to age reporting and population estimates; does not establish attendance.

**Numerator:** all new entrants to the first grade, regardless of age. **Denominator:** population at official entry age. **Controls:** repeaters are not new entrants; transfers within the grade are not new entrants to the system. **Limitation:** over- and under-age entry can raise the ratio above 100 per cent.

**Numerator:** new entrants of official entry age. **Denominator:** population of official entry age. **Minimum metadata:** entry rule, age source and unknown-age treatment. **Limitation:** does not represent late entrants who nevertheless gain access.

**Numerator:** learner-days or periods present. **Denominator:** possible learner attendance days or periods while the school service operated. **Minimum metadata:** counting unit, period, treatment of partial attendance, late entry and transfer. **Limitation:** attendance recording varies and does not establish learning.

**Numerator:** days or periods in which the defined school service operated. **Denominator:** scheduled days or periods. **Minimum metadata:** calendar, closures, affected grades and cause. **Limitation:** operation does not establish learner or teacher presence.

**Numerator:** learners from grade g promoted to g+1. **Denominator:** learners enrolled in grade g under the declared flow method. **Minimum metadata:** linked or reconstructed method, transfers, repetition and unknown status. **Limitation:** aggregate reconstruction depends on assumptions.

**Numerator:** repeaters in a grade in year t+1. **Denominator:** enrolment in that grade in year t under the adopted method. **Minimum metadata:** full or partial repetition and rule changes. **Limitation:** policy and recording can change the rate without a change in learning.

**Definition:** learners leaving the defined educational programme without known transfer or completion during the observation window. **Minimum metadata:** follow-up period, transfer linkage, deferment and unknown destination. **Limitation:** weak linkage can overstate dropout.

**Definition:** proportion of an entry cohort reaching a specified grade. **Minimum metadata:** actual or synthetic cohort, repetition assumptions, target grade and period. **Limitation:** a modelled survival rate is not observed individual progression.

**Definition:** completion of the programme under a stated rule, related to an entry cohort, relevant age population or another declared denominator. **Minimum metadata:** terminal requirement, time window, late completion, repetition and population base. **Limitation:** final-grade enrolment is a proxy and not confirmed cohort completion.

**Numerator:** completers of one level entering the next. **Denominator:** eligible completers under a consistent rule. **Minimum metadata:** interval, deferred entry and alternative provision. **Limitation:** receiving-level capacity and data linkage affect the result.

**Definition:** unique persons employed as teachers within the stated scope. **Minimum metadata:** level, sector, status, reference date and treatment of multiple assignments. **Limitation:** does not measure full-time service, presence or instruction.

**Definition:** contracted or assigned service converted to full-time units under a stated basis. **Minimum metadata:** full-time load, part-time treatment and reference period. **Limitation:** contractual time does not establish delivered instructional time.

**Numerator:** teachers meeting the national minimum organised training requirement. **Denominator:** teachers within the same scope with known status, or all teachers with unknown separately reported. **Minimum metadata:** national definition, level, sector and reference year. **Limitation:** not internationally equivalent preparation and not classroom-practice evidence.

**Numerator:** pupils at the stated level and sector. **Denominator:** teachers or teacher full-time equivalents within the matching scope. **Minimum metadata:** headcount or full-time basis, reference year and inclusion rules. **Limitation:** not average class size, distribution, presence or instructional time.

**Definition:** learners assigned to or present in an instructional group at a specified time. **Minimum metadata:** assigned or present, subject or grade, shifts, multi-grade organisation and distribution. **Limitation:** a mean can conceal very large classes and does not measure teacher availability across time.

**Numerator:** scheduled teachers observed present or validly recorded as available. **Denominator:** teachers scheduled at the observed time. **Minimum metadata:** observation design, visit timing, repeated coverage and reason categories. **Limitation:** one visit does not establish annual behaviour or teaching quality.

**Numerator:** instructional days, hours or periods delivered. **Denominator:** scheduled instructional time for the same population. **Minimum metadata:** unit, school operation, timetable, observation or recording method and cause of loss. **Limitation:** time does not establish instructional quality or learning.

**Definition:** intended learners with access to correct, serviceable and relevant material under the specified lesson or study arrangement. **Minimum metadata:** title, subject, grade, language, condition, sharing, distribution and period. **Limitation:** access does not establish pedagogical use or learning.

**Definition:** safe and functional instructional spaces available for the stated schedule and population. **Minimum metadata:** safety status, capacity rule, shifts, accessibility and unavailable rooms. **Limitation:** room capacity does not establish staffing or delivered instruction.

**Definition:** relevant facilities operating, accessible and usable during the reference period. **Minimum metadata:** source, reliability, sanitation function, privacy, accessibility, maintenance and seasonal condition. **Limitation:** facility service does not establish health outcome by itself.

**Definition:** learners had scheduled and delivered opportunity to engage with the required content. **Minimum metadata:** authorised curriculum, period, grade or course, time, teacher, material, observed coverage and exclusions. **Limitation:** opportunity does not establish attainment.

**Numerator:** eligible learners producing a valid assessment result. **Denominator:** all learners eligible under the assessment design. **Minimum metadata:** exclusions, absence, invalid responses, substitutions and relevant groups. **Limitation:** high participation does not establish measurement validity.

**Definition:** performance on a stated assessment scale among the defined participant population. **Minimum metadata:** construct, instrument, administration, scoring, sampling, participation, uncertainty and performance distribution. **Limitation:** comparison requires scale and population stability.

**Numerator:** defined public education expenditure for the level and period. **Denominator:** matching learner headcount or full-time equivalent. **Minimum metadata:** recurrent and capital scope, government levels, price basis and timing. **Limitation:** not a measure of efficiency, distribution or quality.

**Numerator:** intended units confirming timely and valid receipt. **Denominator:** units scheduled to receive the resource. **Minimum metadata:** dispatch, expected date, actual receipt, quantity, condition and non-response. **Limitation:** receipt does not establish operation or learner access.

**Numerator:** valid cases closed with the defined operating service verified. **Denominator:** cases due for resolution or all valid cases under a stated cohort. **Minimum metadata:** opening stock, new cases, severity, stage, age and reasoned removals. **Limitation:** closure count alone can conceal backlog and premature closure.

Source: ICEQC core indicator specifications.

The definition, source, reference period, coverage and limitation are not supplementary administrative detail. They determine what the value means. A number separated from them does not retain the same evidentiary status.

Systems may publish a concise headline, but the metadata remains accessible and is revised with the value. Common definitions are maintained over time; changes are introduced through a documented break rather than silent amendment.

The hierarchy orders evidence by its relationship to a claim, not by the prestige of the producing body. It prevents administrative documents from being treated as direct educational outcomes and prevents small observations from being generalised beyond their design.

Levels do not imply that higher evidence is always required. Screening and urgent provisional decisions can use lower levels with a defined follow-up. Public and high-consequence claims require evidence appropriate to their breadth and effect.

Level one includes laws, policies, authorised curricula, budgets, plans and standards. It establishes obligation, purpose or intended provision. Version, status and effective date are critical.

It does not establish implementation. A law requiring free education cannot by itself prove that families face no charge; a curriculum cannot prove instruction.

Level two includes appointments, payroll, procurement, transfers, registrations, construction certificates and recorded decisions. It establishes that an authorised transaction or administrative state occurred.

Its limitation is the gap between transaction and operation. It can be comprehensive and still observe the wrong stage for a service claim.

Level three observes whether the resource or service was functional at the point and time required. Examples include an operating school, assigned and present teacher, safe usable room, delivered period or accessible material.

This level is direct for service claims but not for longer outcomes. Observation timing and coverage remain necessary.

Level four records learner attendance, active use, curriculum exposure, service receipt or participation in the intended activity. It establishes contact with the educational service.

Participation can be recorded through registers, observation, work or user evidence. It does not establish learning or causation.

Level five observes the first result linked to the service: restored instructional time, completed required activity, usable feedback, progression or demonstrated task performance.

The measure is close enough to the action to guide management. Attribution remains bounded by the design.

Level six includes sustained participation, completion, learning, later progression and wider social or economic outcomes. It requires longer periods and stronger control of population and competing explanations.

The further the outcome lies from the action, the less one institutional source can support exclusive causation.

Distribution is not a separate final level. It applies at every level. A policy can exclude by design; a resource can be distributed unequally; service, participation and outcomes can differ among groups and places.

Minimum evidence tests distribution at the stage claimed. Equity is not postponed until outcome data become available.

Capability evidence shows whether records, decisions, logistics and verification function. It explains the reliability of each level and whether gaps can be corrected.

A weak information system does not prove weak education, but it limits claims and can delay remedy. Capability improvement is therefore part of the quality programme.

Source: ICEQC evidence hierarchy.

Triangulation combines sources that observe different aspects or have different errors. A payroll, school assignment and visit can trace a teacher from transaction to availability. A curriculum, timetable and work sample can trace intention to opportunity.

Triangulation does not permit skipping a level by accumulating indirect sources. Three procurement documents do not establish learner use. The source closest to the claimed state remains necessary where feasible.

A credible report of immediate harm can justify precaution before full verification. The initial evidence supports the protective decision, not a final prevalence estimate or attribution.

The hierarchy therefore separates action threshold from publication threshold. Waiting for level-six evidence before restoring a level-three essential service would be unreasonable.

An evidence package for a material claim contains the authoritative criterion where applicable, the operating source, participation or result source, distribution, and quality note. It can be concise. Its sufficiency depends on the links, not volume.

For routine low-risk claims, existing administrative and sampled operating evidence may be enough. For safety, award validity or national learning, the package requires stronger technical competence and verification.

The hierarchy protects verbs. It requires the words used in public reporting to correspond to the state actually observed. This linguistic discipline is an evidentiary control.

When a higher state is not observed, the lower state remains worth reporting. Accurate evidence of policy, expenditure or delivery supports accountability and identifies the next verification step without being inflated into an outcome.

The matrices specify the evidence without which a claim should be narrowed. They are interpretive aids rather than universal reporting forms. A system can use another source where it performs the same evidentiary function and its limits are stated.

The claim requires current and baseline counts of unique learners or valid enrolments, a stable level and provider scope, a population or other comparison basis where a rate is asserted, and evidence concerning frame or response change. It also requires a distributional test for groups or locations materially affected by expansion.

Minimum direct evidence is enrolment or participation under a dated rule. School capacity, funding or places offered cannot replace it. Where duplication cannot be fully removed, the system estimates or bounds the effect and narrows “additional learners” to “additional recorded enrolments.”

This claim requires evidence of prior non-participation and current entry for the relevant population, or a population-based design capable of estimating the change. Institutional enrolment alone cannot identify which entrants were previously excluded.

The evidence distinguishes first-time entry, transfer and return after interruption. It examines whether the learner can attend and receive the service; registration on one date is insufficient for sustained access.

Universal access requires a credible eligible-population denominator, coverage beyond institution records, operating provision, accessibility and unresolved-case evidence. The system examines never-enrolled and exited children as well as those registered.

Where population or household evidence is weak, the claim cannot pass merely because net enrolment approaches 100 per cent. Population estimates and enrolment can each contain errors larger than the apparent residual.

The claim requires comparable learner-days or periods present and possible attendance, with school closures, late entry, transfers and recording changes treated consistently. Coverage and missing registers are reported.

A higher attendance rate can result from removal of frequently absent learners from the register. The system therefore reconciles enrolment status and reports the number of eligible learners represented.

Minimum evidence identifies the completion rule, cohort or denominator, observation window, terminal requirement, late completers, repetition, transfer and data breaks. A proxy based on final-grade enrolment is identified explicitly.

Comparison requires a common rule and population or an overlap calculation. Current completion under a new rule can be reported without claiming a like-for-like trend.

The claim requires a stable exit definition, transfer linkage or qualified unknown status, follow-up period and comparable cohorts. School departure is not automatically system dropout.

Where transfer data improve, apparent dropout can fall because classification improves. The report separates change in known destinations from change in total unexplained exit.

The claim requires pupil and teacher measures with aligned levels, sectors, dates and headcount or full-time-equivalent basis. It includes replacement as well as net growth and shows distribution.

A stable national pupil–teacher ratio supports an aggregate resource statement. It does not establish subject coverage, school assignment, presence or delivered instructional time.

Minimum evidence identifies the recognised preparation rule, teacher population, known and unknown status, level, subject or grade where relevant, and distribution. It does not infer current knowledge or practice from qualification status.

If expansion relies on temporary or untrained staff, the system reports their assignments, support and progression plan without treating them as absent teachers or fully equivalent prepared staff.

The evidence includes actual instructional-group size or distribution, room use, shifts, timetable and teacher availability. A pupil–teacher ratio alone is insufficient.

The criterion is drawn from applicable policy, safety, pedagogy and context. This report does not supply a universal numerical threshold.

Minimum evidence identifies scheduled days, hours or periods and delivered time for the relevant classes, with losses by cause. The observation period covers material variation.

An official calendar, open-school count or teacher presence supplies part of the chain but not the full claim. The system identifies which learners lost time and whether restoration displaced another class or subject.

The claim requires correct and usable stock, the intended learner population, distribution, sharing rule, access period and material use appropriate to the claim. Procurement and dispatch are insufficient.

A numerical ratio is interpreted by title, subject, grade and language. Surplus of one material does not compensate for absence of another required material unless the curriculum permits substitution.

Minimum evidence identifies safe, functional, accessible rooms or spaces, scheduling and intended learners. Completed construction is separated from operating capacity and replacement.

The system reports unavailable rooms and multiple shifts. A gross floor-area or classroom count cannot establish actual class conditions without operation.

The claim identifies the service—safe water, sanitation, policy, skills-based education or service link—and observes its operation, access and reliability. A facility or programme inventory is insufficient for outcome language.

Health or attendance outcomes require suitable measures and consideration of concurrent conditions. The connected nature of the FRESH framework is retained.[REF-13]

Minimum evidence links authorised curriculum, planned time, taught content and learner opportunity across a defined sample or population. It includes schools and groups added during expansion.

Teacher declarations can support broad coverage but require work, observation or assessment evidence where the claim is high-consequence. A national curriculum document cannot establish local delivery.

The claim requires a stable or validly linked assessment construct and scale, comparable eligible population, participation, administration, scoring, uncertainty and outcome distribution. It also requires a reference period and a statement about what causal inference is intended.

A higher mean among a more selected participant group is not system improvement. The system reports participation and examines opportunity to learn.

Minimum evidence identifies the stage of the educational chain, relevant groups, values, counts, denominators, coverage and service conditions. A parity index alone is insufficient when both groups have low access or when later outcomes differ.

The claim specifies whether the absolute condition improved, the gap narrowed, or both. A narrowing gap caused by deterioration in the better-served group is not described simply as progress.

The evidence defines public expenditure, education level, government units, recurrent and capital coverage, fiscal year, price basis and any external finance. It reconciles transfers to avoid double counting.

The claim does not imply improved service. Per-learner or share measures require matched denominators and macroeconomic data.

This stronger claim requires the finance-to-service chain, implementation, population reached and a suitable service or outcome measure. Causal wording requires a design that examines other changes.

An association between higher spending and higher outcome can identify a question but does not establish the effect of spending or efficiency.

Minimum evidence includes enrolment and staffing projections, recurrent and maintenance costs, revenue or finance assumptions, implementation capacity and sensitivity. It identifies the period for which resources are secured.

Sustainability is not inferred from a balanced first-year budget. Future commitments and replacement demand are part of the claim.

The evidence defines stages and planned dates, then records actual status. Approval, procurement, completion and operation are separate. Delays are shown by stage and responsible authority.

An aggregate percentage complete is insufficient where critical tasks differ in importance. The operating service and urgent exceptions remain visible.

Minimum evidence concerns frame coverage, response, definition adherence, validation, revision, reconciliation and timeliness. Lower missingness alone can coincide with greater measurement error.

The system compares the same quality dimensions before and after. Increased corrections can reflect stronger detection rather than worse original quality and are interpreted by type.

The evidence includes timely publication, accessible definitions, material limitations, correction, distribution and evidence that responsible authorities responded to service gaps. More reports or indicators are not sufficient.

Participation in evidence and access to remedy are included where relevant. Publicity without protection can harm individuals and does not constitute stronger accountability.

Source: ICEQC claim disposition framework.

The matrices show that “minimum” varies by verb. Adopted, funded, delivered, attended, completed and learned describe different states and carry different sources. The public statement passes only when its verb matches the state observed.

This approach permits progress to be reported at every stage without inflating it. It also identifies exactly which transition requires evidence or action next.

Every calculation identifies source values, units, reference periods, inclusion rules, formula, rounding and responsible analyst. Where figures are revised, the calculation can be reproduced from the retained source version.

Numerators and denominators are aligned in level, sector, geography and time. A percentage is not calculated where the denominator is unknown or conceptually mismatched.

Let E0 be recorded enrolment in the baseline period and E1 the later period under the same definition and coverage. Absolute change is E1 − E0. Relative change is (E1 − E0) / E0.

If coverage differs, the calculation is decomposed. A common-institution change uses institutions observed in both periods. A coverage component reports institutions newly included or lost. The current total remains the total for the later frame; the decomposition prevents it from being misread as pure participation growth.

GER = enrolment at the level, regardless of age, divided by the population of official age for the level, multiplied by 100. Numerator and denominator reference years are stated.

If the official-age population is 1,000,000 and enrolment is 1,080,000, the GER is 108 per cent. This does not mean that every official-age child is enrolled; over-age, under-age and repetition can raise the numerator.

NER = official-age learners enrolled at the level divided by the official-age population, multiplied by 100. Unknown ages are not silently assigned to the official range.

If 820,000 official-age learners are recorded and the population estimate is 1,000,000, the NER is 82 per cent, subject to coverage and denominator error. The residual 18 per cent is not automatically an exact count of out-of-school children because some may be enrolled at another level and both sources have error.

For learner i, possible attendance equals the days or periods on which the relevant service operated while the learner was enrolled and expected to attend. Aggregate attendance rate equals total present units divided by total possible units.

If 500 learners are expected for 20 operating days, possible learner-days are 10,000. If presence totals 8,600, attendance is 86 per cent. Two whole-school closure days are excluded from learner possible days and reported as service loss, not learner absence.

An opening cohort can be reconciled as continuing, completed, transferred, repeated, deferred, known exit and unknown. Categories are mutually exclusive for the reference point and sum to the cohort after recorded corrections.

If 1,000 entrants yield 720 on-time completers, 90 repeaters still enrolled, 60 known transfers, 40 deferred, 50 known exits and 40 unknown, on-time completion is 72 per cent. It is not 72/ (1,000 − 40 unknown); unknown remains part of the cohort unless the definition justifies otherwise.

Where individual linkage is unavailable, promotion, repetition and dropout rates by grade can be used to reconstruct progression under assumptions. The report states whether flow rates are held constant and how entrants or repeaters are identified.

During rapid expansion, the stability assumption can be weak. Sensitivity to alternative flow rates is shown, and the output is labelled modelled survival or completion.

PTR = pupils in the stated scope divided by teachers in the matching scope. If pupils are headcounts and teachers are full-time equivalents, that basis is explicit.

For two districts, ratios cannot be averaged arithmetically to obtain a national ratio. Total pupils are divided by total teachers. A district with 900 pupils and 30 teachers has 30:1; another with 4,000 pupils and 80 teachers has 50:1. The combined ratio is 4,900 / 110 = 44.5:1, not 40:1.

Average class size equals pupils assigned to classes divided by the number of classes under a consistent rule. It differs from PTR because teachers may have non-teaching duties, classes may be taught by several teachers and shifts or subject groups alter organisation.

Distribution is reported. Five classes of 20 and five classes of 60 have an average of 40, but no class contains 40. The average alone conceals the two service conditions.

Delivery rate equals delivered instructional units divided by scheduled instructional units for the same population. Loss is classified by school closure, teacher availability, timetable conflict, interruption or other cause.

If 120 periods were scheduled and 96 delivered, the rate is 80 per cent. Restoring 12 periods raises it to 90 per cent, an increase of 10 percentage points and 12.5 per cent relative to the original delivered total.

Usable relevant stock equals received stock less damaged, incorrect, obsolete or otherwise unavailable items, subject to overlap among categories. The categories must not be subtracted twice.

Simultaneous demand follows the service design. In a two-shift school where books can be transferred safely, the larger shift rather than total enrolment may determine one-copy-per-user demand. The rule and transfer feasibility are stated.

A gross completion proxy can divide non-repeating enrolment in the last primary grade by the population at theoretical graduation age. It is affected by early or late entry, repetition and the use of enrolment rather than verified completion.

The value is reported as a proxy. A figure above 100 per cent is possible and does not establish universal completion.

A parity index commonly divides the female value by the male value. A value of 1 indicates numerical parity for that measure, not a high absolute level.

If female net enrolment is 60 per cent and male net enrolment 60 per cent, parity is 1 while 40 per cent of each official-age population remains outside the measured level. Both underlying values accompany the index.

An absolute gap subtracts group values; a relative gap divides them. Results can move differently as levels change. The report names the measure and direction.

If group A rises from 40 to 60 per cent and group B from 60 to 75 per cent, the absolute gap narrows from 20 to 15 percentage points while both improve. This differs from parity achieved through decline in the better-served group.

Per-learner expenditure equals defined expenditure divided by the matching learner measure. If recurrent primary spending is 120 million and primary enrolment is 600,000, the value is 200 currency units per enrolled learner.

If 30 million of the total is capital spending, it is excluded from a recurrent measure but can be reported separately. Inflation and currency basis are required for trend or international comparison.

Execution rate equals actual expenditure divided by the authorised or released amount, according to the stated denominator. The two versions answer different questions.

If 100 million is appropriated, 80 million released and 76 million spent, expenditure is 76 per cent of appropriation and 95 per cent of release. Neither rate establishes value, receipt or service.

Closing unresolved cases equal opening unresolved cases plus new valid cases, less verified closures and reasoned removals. Transfers to another authority remain open unless responsibility and receipt are confirmed under the defined system.

If opening stock is 150, new cases 70, verified closures 60 and duplicate removals 5, closing stock is 155. Reporting 60 closures without the stock and entries conceals that unresolved cases increased.

For a binary condition with O observed favourable cases, U unknown cases and N total eligible cases, the minimum all-case rate is O/N and maximum is (O+U)/N. These bounds require no assumption about unknown status.

If 72 of 90 observed cases are favourable and 10 are unknown among 100 eligible cases, the all-case rate lies between 72 and 82 per cent. A threshold of 80 per cent cannot be resolved without further evidence or an assumption.

A regional or national aggregate identifies the weight. Population-weighted enrolment rates use the relevant population; institution-weighted service rates give equal weight to institutions. These answer different questions.

Missing units are not assigned the observed average without a stated method. Aggregate coverage is the sum of represented denominator divided by the total eligible denominator where known.

Real expenditure converts nominal values using an appropriate price index and base year. General and education-specific price changes can differ; the selected index and purpose are stated.

An exchange-rate conversion does not remove domestic inflation. International price comparisons and domestic time trends require different treatment.

Source: ICEQC calculation controls.

Calculations use unrounded source values and round only the published result. Components may not sum to exactly 100 because of rounding; they are not altered to force agreement.

The number of decimal places follows data quality and decision use. Additional decimals do not compensate for uncertain definitions or coverage.

Most indicator disputes can be located in population, unit, period or state. The formula is often the simplest part. Reconciliation therefore precedes calculation.

The calculation record makes assumptions and exclusions visible. It supports correction and prevents a derived result from acquiring more certainty than its sources.

A census is appropriate when every unit requires an individual decision, the population is manageable or omission itself creates material risk. A sample is appropriate when the intended result is an aggregate estimate and complete collection would reduce timeliness or quality.

The word census describes intended coverage, not achieved coverage. A school census with 85 per cent response is an incomplete census and requires a non-response account.

The frame lists eligible institutions, learners, teachers, classes or events. It records source date, inclusion, duplicate and known omission. New and temporary institutions are especially important during expansion.

A sample cannot represent units absent from the frame. Comparison with district, finance, examination and provider records can identify frame gaps. The report states which populations remain outside sampling coverage.

Probability selection gives each frame unit a known chance of inclusion. Simple random, systematic, stratified and cluster designs can all be valid. The design follows the inference and field conditions.

Equal selection of schools estimates a school-level condition when weighted appropriately. It does not give learners equal selection chance where school sizes differ. Selection proportional to size can support learner-level inference but depends on the quality of the size measure.

Strata ensure coverage of regions, provider types, urban and rural settings, school levels or other conditions associated with the measure. They can improve precision and permit separate reporting.

Small or high-risk strata can be sampled at a higher rate. Combined results use design weights; the additional cases are not allowed to dominate the national estimate merely because they were intentionally oversampled.

Selecting schools and then classes or learners reduces field cost. Observations within the same school share conditions and are not statistically equivalent to observations spread across independent schools.

Sample-size and uncertainty calculations account for clustering. The report states numbers at every selection stage so that a large learner count does not conceal a small number of schools.

Risk-based verification deliberately selects unusual, high-consequence, new, remote or previously weak units. It is well suited to detecting control failure and confirming individual service. It does not provide a prevalence estimate without a probability component.

Reports separate purposive findings from representative estimates. Both can support action, but they answer different questions.

A rotating design observes part of the population each period while retaining some units for change measurement. It can reduce burden and broaden coverage over time.

The report distinguishes change among continuing sampled units from difference between independent samples. Rotation is not used to postpone examination of known critical cases.

Sample size reflects expected prevalence or variation, desired precision, confidence, design effect, subgroup analysis and non-response. It is not chosen through one fixed proportion of the population.

Where capacity is limited, the system may accept wider uncertainty and publish it. It does not retain narrow claims by ignoring clustering or non-response.

Follow-up effort is planned and documented. Schools not reached because of distance, insecurity or weak communication are not replaced by convenient schools without preserving their status.

Response is reported by stratum and relevant frame characteristic. A high overall rate can conceal low coverage of remote or private provision.

Teacher presence, instructional time, water availability and material use vary across days and seasons. Visits cover the periods relevant to the claim. A single observation supports a point-in-time finding and can screen for wider enquiry.

Announced and unannounced visits serve different purposes and carry different ethical and operational considerations. The report states the arrangement and does not infer misconduct from presence evidence alone.

Observers receive operational definitions and practise boundary cases. Supervisors compare independent classifications on a subset. Material disagreement leads to clarification, retraining or qualification.

Agreement is reported for the variables on which judgement depends. A general statement that observers were trained is not evidence of consistent classification.

Verification traces a value to its source and, where necessary, to the underlying event. It checks identity, date, unit, calculation and authorised correction. Recalculation from the same erroneous source is not independent verification.

Cross-source verification selects sources with different purposes: payroll and school assignment, dispatch and receipt, timetable and learner work. Discrepancy is investigated before one source is preferred.

Facilities and materials are verified for existence, condition, function, access and intended population. Photographs can support a dated condition but do not establish continued operation, scale outside the image or safety beyond the observer's competence.

Technical matters use competent inspection. A statistical fieldworker does not certify structural or water safety without relevant authority and method.

Users can confirm access, cost, teaching opportunity and service experience. Selection and power affect candour. The design provides safe participation and avoids asking individuals to determine technical or legal compliance.

User evidence is particularly valuable where administrative records end at delivery. It remains one source and is corroborated where high-consequence claims require it.

Intensity increases with consequence, uncertainty, novelty, prior error and incentive to report favourably. Low-risk stable variables can be checked through small samples. Safety, award integrity or large expenditure receives stronger direct and independent evidence.

The verification plan also considers cost and delay. It prioritises the link in the claim chain most likely to fail rather than checking every earlier document equally.

Source: ICEQC sampling and verification choices.

The sampling record contains population, frame, design, selection probabilities, strata, stages, replacements, response, weights and field dates. The verification record contains claim, source, selected cases, method, discrepancies, correction and action.

This information allows reproduction and appropriate reuse. It prevents a carefully designed sample from losing its inferential basis when only the final percentage is copied.

Sampling is not a concession to weak evidence. A well-designed sample can provide stronger national evidence than a nominal census with incomplete and unverified returns. Its uncertainty is explicit.

Individual harm and rare critical conditions remain subject to direct action. Representative estimation and case protection are complementary, not competing, functions.

Missing evidence is classified at frame, unit, item and linkage stages. A school can be absent from the frame, fail to return the form, omit one item or lack a link between years. These forms have different causes and remedies.

The record also distinguishes not applicable, true zero, below measurement limit and unknown. Blank is not a category.

Evidence can be missing independently of the condition, related to observed characteristics or related to the unobserved value itself. The last is particularly serious: schools with weak service may be least able or willing to report.

The system does not need to assign a formal mechanism where evidence is insufficient. It identifies plausible relationships and tests sensitivity.

Analysis limited to complete cases is transparent but unbiased only under restrictive conditions. It reports the number and characteristics lost.

A complete-case rate is labelled among respondents or observed units. It is not applied to all eligible units without justification.

Response weights can adjust estimates using observed classes such as region, school size or provider type. The method assumes that respondents and non-respondents are sufficiently similar within those classes for the variable concerned.

The classes, response rates and weight limits are reported. Weighting cannot correct omission from the original frame or unmeasured differences within classes.

Imputation supplies values under an explicit statistical method. It can support an aggregate estimate where relationships and uncertainty are adequate. Imputed values are flagged and are not used as if they were observed individual facts.

Simple replacement by the observed mean understates variation and can conceal distribution. The report compares results with and without imputation where the conclusion is sensitive.

Models can combine census, survey, administrative and contextual data to estimate unobserved values. The model, covariates, assumptions and validation are disclosed.

A model can improve planning but does not convert areas without evidence into directly observed compliance. Decision-makers know which values are estimated and the uncertainty attached.

Education denominators often rely on projections from an earlier census. Assumptions about fertility, mortality and migration affect age-specific populations. Rapid movement or weak age reporting increases uncertainty.

The report states projection source and base year. Sensitivity to plausible denominator variation is shown where an access conclusion is close to a threshold.

Measurement error can arise from definitions, age, attendance entry, observer classification, assessment scoring and respondent recall. It is distinct from sampling error.

Formal sampling intervals do not include all measurement error. The public note avoids presenting them as a complete boundary of truth.

Bounds provide a transparent range under simple extreme assumptions. They are especially useful when the unknown share is small enough for both extremes to lead to the same decision.

Where the range is wide, the correct conclusion is uncertainty. The system does not select the favourable end as an estimate.

Sensitivity analysis changes material assumptions—population, attrition, response, salary, assessment participation or classification—and observes the decision consequence. It focuses on plausible values rather than arbitrary extremes unless bounds are intended.

The result identifies which assumption requires better evidence. A robust decision remains the same across plausible values; a fragile decision requires caution or more data.

Preliminary values are revised when response, correction or improved denominators become available. The system records size and cause of revisions and examines whether early estimates show systematic bias.

Revision practice informs future provisional use. Frequent large favourable early errors may require stronger validation before publication.

Unknown status can itself identify a service or control problem: transfer not tracked, teacher qualification not verified, facility not inspected or assessment participation unexplained. The system assigns responsibility for resolving material unknowns.

Unknown is not automatically failure for statistical classification. For safety or rights decisions, precaution can operate while status is established.

Source: ICEQC missing-evidence treatment table.

Uncertainty appears in wording as well as technical notes. “Recorded enrolment increased among responding schools” is more accurate than “access increased” where frame and uniqueness are unresolved.

Ranges, status labels and limitations are placed beside the result. A public reader should not need to find a distant appendix to discover that the headline value is estimated or incomplete.

Uncertainty is not a reason to avoid decisions. It is information about the risk of error. The authority chooses verification, precaution, reversible action or qualified publication according to consequence.

The minimum standard is that uncertainty is neither hidden nor converted into favourable status. Better evidence is directed to the uncertainty most capable of changing an important decision.

An indicator profile keeps the meaning of a public value attached to it through collection, analysis and publication. The profile is updated when definition, source or coverage changes. It prevents the same label from carrying different quantities without a marked break.

The identity field contains title, short name, version, responsible statistical or administrative unit and first applicable period. The title names the actual state measured, not a broader policy objective.

“Primary learner attendance rate” is preferable to “primary education quality” for an attendance measure. The latter exceeds the evidence before any number is calculated.

The profile identifies the decisions and reports for which the indicator was designed. Uses can include planning, allocation, public monitoring, screening or international reporting.

Uses requiring stronger evidence are listed as excluded. A school-level administrative indicator can be valid for case management and unsuitable for ranking.

The concept explains the educational state. The operational definition gives numerator, denominator, population, unit and period. Terms imported from an international classification are mapped to national practice.

Where national law or policy uses a different term, the relationship is stated. The profile does not create the appearance that a statistical definition determines the legal position.

The source field identifies the institution, record or survey, collection mode, reference date, frequency and frame. Derived values identify every source component.

Administrative-source purpose is stated because it explains likely strengths and gaps. A record collected for payroll can be comprehensive for payment and incomplete for school presence.

Coverage lists education levels, grades, providers, regions and populations included and excluded. Response and item completeness are reported for each release.

Coverage changes create a break or adjustment. The profile prevents “national” from being used where a material provider or area is absent.

The formula, sequence of adjustments, weights, imputation, rounding and aggregation are provided. Complex models include a separate technical method while the profile gives the principal assumptions.

The calculation records treatment of zero, unknown and not applicable. It also identifies whether percentages can exceed 100 and why.

The profile gives principal sampling and non-sampling errors, validation checks, independent sources, revision practice and known bias. It does not use a generic quality statement unrelated to the indicator.

Verification frequency and selection are stated. Material discrepancies and their resolution are included in release notes.

The interpretation field states what a higher or lower value can mean and what it cannot establish. It identifies relevant companion indicators and contextual variation.

For the pupil–teacher ratio, the profile states that the value is a resource ratio, not average class size or delivered time. For completion, it states whether the measure is a cohort result or age-based proxy.

Available disaggregation and its evidence quality are listed. Categories follow lawful, meaningful definitions and protect confidentiality.

Where a national value conceals a material range, the standard public presentation includes the distribution or a link to it.

The profile states whether comparison over time, among regions or internationally is supported. It marks changes in level structure, provider coverage, age range, assessment or source.

Comparability is not binary. Some comparisons can be valid for counts and invalid for rates; some can be valid within a period and not across a break.

The release calendar, preliminary status, expected finalisation and correction route are public. Revised values retain date and reason.

Major revisions prompt examination of the dependent headline, target and policy conclusion. A corrected denominator can alter both the current rate and earlier trend.

Source: ICEQC public indicator profile.

The title is “share of scheduled teacher assignments observed available during sampled school visits.” The population is scheduled assignments in sampled schools during specified visit periods. The numerator is assignments for which the teacher was present and available under the protocol; the denominator is all scheduled assignments observed. Authorised absence and official duty are reported by reason.

The measure supports a system availability estimate under the sample design. It does not establish annual individual attendance, misconduct, teaching practice or learning. The profile reports schools, visits, timing, response, clustering and uncertainty.

The title identifies the numerator and denominator rather than calling the measure simply “completion.” It states that non-repeating final-grade enrolment is divided by the population of theoretical graduation age.

The interpretation describes it as an upper-bound proxy where final-grade loss is not separately observed, subject to age and repetition. It is not presented as an actual entry-cohort rate.

Changes require approval by a competent methodological role and a public version note. Operational offices can propose corrections, but the definition is not changed locally without record.

Profiles are reviewed when policy, level structure, record systems or assessment changes. The review removes indicators whose intended use no longer justifies their burden.

Evidence governance assigns responsibility for definition, collection, custody, analysis, publication, correction and service response. It does not centralise every task. It ensures that a material discrepancy has an owner and a route to resolution.

One role maintains the controlled definition and metadata. Changes are consulted with data producers and users and receive an effective date. Local clarifications are collected to identify ambiguity.

A technical definition does not override the competent legal or policy interpretation. Where they differ, the profile states which question each answers.

The source owner maintains the operational record and documents changes. It confirms coverage, reference period and known limitations. Ownership does not give authority to broaden the public claim.

For linked indicators, each source owner remains responsible for its component while an analytical owner controls linkage and calculation.

Forms and instructions define units, reference dates, categories and valid ranges. They avoid collecting totals that can be derived more reliably from components unless the total provides a reconciliation check.

Training uses boundary cases. Support channels record recurring questions so that definitions can be clarified system-wide rather than differently in each district.

Receipt is checked against the current frame. Late, duplicate and missing returns are identified. A return is not marked complete solely because a file arrived; critical items and internal consistency are tested.

Response follow-up gives priority to material gaps and under-observed populations. The release states the achieved coverage.

Logical rules test relationships that should hold: component totals, age and grade plausibility, progression states, staffing scope, operating status and chronology. A flag identifies a case for review; it does not automatically replace the value.

Rules are documented and updated when legitimate exceptions emerge. Excessive automatic correction can erase unusual but real conditions.

Large changes are compared with prior periods, but rapid expansion makes change plausible. The system asks for explanation and supporting evidence rather than forcing values toward historical trends.

Structural breaks are separated from errors. A newly included provider can cause a valid discontinuity that should remain visible.

School counts are compared with finance, staffing, examination and district sources; teacher data with payroll and assignment; enrolment with assessment and learner-flow records; deliveries with receipts.

Differences are classified by timing, unit, coverage, duplicate, missing record or error. Reconciliation decisions identify which source is authoritative for which state.

Field verification tests material values and source operation. It samples ordinary, high-risk, unusual and under-observed units according to purpose. Findings produce source correction and, where necessary, service action.

Verification reports do not disclose personal detail beyond authority. They distinguish statistical discrepancy from individual conduct.

Schools and offices can submit a correction with source evidence. The process records original value, proposed value, reason, approving role and downstream effect. Deadlines allow timely publication while preserving later material correction.

The route protects good-faith correction. A system that penalises every correction can encourage concealment of error.

Approval confirms that definition, coverage, calculation, limitations and reference dates support the proposed wording. It does not certify that the educational condition is favourable.

Headline statements receive the same evidentiary scrutiny as technical tables. A cautious footnote cannot cure a claim that exceeds the result.

Material errors are corrected promptly, with date and effect. The revised release does not silently overwrite the historical record where users need to understand the change.

The authority assesses whether targets, allocations or public decisions based on the error require revision. Correction therefore extends beyond the table.

Access to person-level and sensitive institution records is role-based. Linkage uses the least identifying information necessary. Public tables apply suppression or aggregation where disclosure risk is material.

Confidentiality review also covers maps and narrative. Geographic precision can identify a small group even when names are absent.

Sources, transformations, calculations and revisions are retained for a period suited to accountability and lawful requirements. Working material without continuing value is disposed of securely.

Retention supports trend reconstruction and correction. It does not justify indefinite retention of unnecessary personal data.

A statistical office can identify a material service gap but may not control the remedy. The governance record transfers the case to the competent authority and tracks acknowledgement without exposing personal data publicly.

The statistical value is corrected through its own method. The service case is not closed merely because the aggregate is published.

Source: ICEQC evidence governance responsibilities.

The evidence system periodically reviews response, revisions, discrepancies, user needs, reporting burden and unresolved data gaps. It examines whether indicators support decisions and whether important groups remain invisible.

Success is not the production of more indicators. It is a smaller or better-focused body of evidence that can support accurate public claims and timely educational action.

Minimum evidence depends on institutional arrangements that preserve meaning from source to publication. Definitions, corrections and responsibilities are therefore part of quality, not clerical detail.

Governance is credible when it allows an unfavourable value or error to remain visible, directs it to the correct authority and records what happened next.

A regional result can describe the experience of people, the position of countries, the prevalence among institutions or the distribution of systems. Each requires a different unit and weighting rule. The title states which is intended.

A population-weighted rate gives greater influence to larger populations and can describe the regional person-level condition. An unweighted country mean describes the average of available country rates but does not represent the region's people.

Before aggregation, definitions are compared for level, age, provider, teacher, completion, assessment and financial scope. National values can be mapped to a common concept where differences are documented and not material to the intended use.

Where no valid mapping exists, values remain in separate groups or are presented with country notes. Exclusion solely to obtain a cleaner table can systematically remove systems with different structures.

Regional sources often contain the latest available year for each country. The aggregate records the permissible reference-year range and the value used for each system. A wide range limits interpretation under rapid change.

Trend comparison uses sufficiently common periods or a method that accounts for changing country composition. An apparent regional trend can arise because different countries report in each year.

Coverage is reported by number of systems and share of the relevant regional denominator. A population coverage of 90 per cent can omit many small systems; a country coverage of 90 per cent can omit one large population. Both are informative.

Coverage for the numerator and denominator is consistent. A regional pupil–teacher ratio is not calculated from pupils covering more countries than the teacher total.

For a ratio, regional numerators are summed and divided by summed denominators within the same covered set. National ratios are not averaged unless the intended statistic is explicitly the mean country ratio.

The aggregation method can produce a regional result outside the simple range expected by readers when weights differ. The report explains the weight and retains national distribution.

Totals add comparable national values, adjusted for double counting where cross-border or multi-country programmes matter. Estimated national values are identified and uncertainty is reflected in the regional result.

Missing countries are not assigned zero. Modelled totals are labelled estimates and publish the contribution of imputation where material.

Regional learning comparison requires a common or validly linked construct, scale, population and administration. National examination results are rarely directly aggregable merely because they report percentages passing.

Where common assessment evidence exists, sampling weights and participation are respected. Systems with high exclusion or low response are not allowed to appear directly comparable without qualification.

Financial comparison states currency conversion, price year, purchasing-power treatment, public and private scope, level and capital coverage. Expenditure as a share of national income answers a different question from expenditure per learner.

Higher cost is not interpreted as higher quality without service, price and context. Regional finance evidence is used to examine resource capacity and distribution rather than to infer efficiency from one measure.

A regional overview is accompanied, where feasible, by country or system profiles that retain definition, trend, distribution and policy context. The profile prevents the aggregate from implying uniformity.

Profiles use a common core but permit national evidence that better observes a local condition. Difference in indicator availability is reported as a capacity issue, not automatically as educational failure.

Rankings discard the magnitude and uncertainty of difference. Adjacent positions can be statistically indistinguishable, while a small definitional difference can change order.

This framework favours value distributions, ranges and profiles. If rank is published for a legitimate use, ties, uncertainty, coverage and construct limits are shown.

Monitoring a common target requires a baseline, target definition, country applicability and aggregation method. Regional progress can be driven by large systems while smaller or fragile systems fall behind.

The report therefore shows both population-weighted progress and distribution among countries where relevant. A regional target is not treated as proof of a common national legal obligation.

Association between system features and outcomes can suggest hypotheses. It does not establish that transferring the feature will reproduce the outcome. Policy, culture, finance, population and measurement differ.

Comparative evidence identifies variation, consistent patterns and cases for closer study. Recommendations remain bounded by mechanism and implementation capacity.

Source: ICEQC regional aggregation controls.

The minimum for regional comparison is not identical administration. It is sufficient conceptual and statistical equivalence for the stated inference, with difference retained where it matters.

Regional evidence is strongest when it combines a transparent aggregate, the distribution among systems and concise profiles. This supports common learning without converting diversity into error.

A jurisdiction publishes a net primary enrolment rate of 99.2 per cent. The numerator covers public and registered private schools. The population denominator is a projection from a census conducted twelve years earlier. Recent migration into two urban areas is believed to be substantial, and a household survey reports a wider confidence interval.

The value is a valid administrative ratio under the stated sources. It does not support an exact universal-access claim. The evidence audit requires the denominator base, projection assumptions, private and unregistered coverage, enrolment at other levels, household estimate and unresolved local populations. The public wording becomes: “Recorded net primary enrolment is estimated at 99.2 per cent under the current population projection; denominator uncertainty and unregistered urban populations prevent an exact universal-access conclusion.”

The decision is not to discard the ratio. It is to improve urban population and service mapping and examine children outside school records.

A system adopts a planning parameter of 40 pupils per teacher and reports a national ratio of 39.6. School-level data show a median of 36, an upper tenth above 61 and 230 schools with no teacher for at least one scheduled grade. Some teachers serve administrative roles but remain in the denominator.

The aggregate target is met under the published counting basis, but the service claim is not. The audit recalculates the ratio using matching full-time-equivalent assignments, publishes the distribution and separates uncovered grades. It also states the planning status of the 40:1 parameter rather than calling it a quality threshold.

Allocation addresses the uncovered and high-pressure schools. The national total is not used to deny verified local need.

A report lists 800 completed classrooms. Completion certificates exist for 796; four remain under dispute. Field returns indicate 690 in use, 52 awaiting furniture, 24 awaiting teachers, 18 replacing condemned rooms and 12 without a return.

Several categories overlap: some replacement rooms are in use. The audit does not subtract categories without case-level reconciliation. It reports 796 certified completions, four unresolved construction cases, and 690 confirmed operating rooms. It then separately classifies the service effect of operating rooms as added or replacement capacity.

The 12 missing returns remain unknown. They are not coded as non-operating or operating. The public claim distinguishes infrastructure output from educational capacity.

Central records report 900,000 books delivered for 1.8 million pupils. District receipts total 850,000; school records cover 92 per cent of intended schools and confirm 760,000 copies. Of the confirmed stock, 65,000 are damaged and 40,000 are wrong for the intended grade or language, with an overlap of 5,000 copies that are both damaged and wrong.

Confirmed usable relevant stock is 760,000 − 65,000 − 40,000 + 5,000 = 660,000. It is not 655,000 because overlap would be subtracted twice. This figure covers responding schools, not all pupils. Learner access further depends on school distribution and sharing.

The audit retains four states: central dispatch, district receipt, confirmed school stock and usable relevant stock. No single ratio is labelled textbook availability without its state and population.

Attendance rises from 82 to 88 per cent. In the later year, 6 per cent of enrolled learners have missing daily registers, compared with 1 per cent previously. Schools with missing registers are concentrated in remote districts with historically lower attendance.

The respondent result improved, but system change is uncertain. The audit reports observed attendance, register coverage and bounds or adjusted estimates under stated assumptions. It does not classify missing days as absence automatically because some records may be lost despite presence.

The system strengthens register support and remote reporting. Allocation or sanction is not based on the unadjusted comparison.

A gross primary completion proxy is 104 per cent. The numerator contains non-repeating final-grade enrolment, including many over-age learners, and the denominator is the population at theoretical graduation age.

The value is possible under the proxy and is not capped at 100. It signals age and flow effects. The audit reports age distribution, repetition and actual completion evidence where available. It does not state that 104 per cent of children completed primary education.

A national sample assessment reports a mean of 510 compared with 498 in the prior cycle. School participation is 94 per cent in both cycles, but learner participation falls from 91 to 82 per cent. Excluded learners in the later cycle include more pupils requiring accommodations that were unavailable.

The mean difference among valid participants can be reported subject to scale comparability. A system-learning claim requires analysis of the participation change, exclusion and plausible performance of missing learners. The exclusion is also a service and equity finding independent of its statistical effect.

The immediate response secures appropriate accommodations and reviews administration. The system does not wait for a revised mean before addressing exclusion.

Nominal education expenditure rises by 18 per cent while the relevant price index rises by 12 per cent and enrolment by 9 per cent. Using an index base of 100 in the first year, real expenditure grows approximately by 1.18/1.12 − 1, or 5.4 per cent. Real expenditure per enrolled learner changes by approximately 1.18/(1.12 × 1.09) − 1, or −3.3 per cent.

The precise result depends on timing and index construction. The audit shows nominal, real and per-learner measures and avoids stating that resources per learner rose because total nominal spending increased.

A programme reports 85 per cent implementation because 17 of 20 planned activities are complete. The three incomplete activities are teacher deployment, water connection and distribution of learning materials; several completed activities are planning meetings.

The unweighted activity proportion misrepresents operating readiness. The audit groups critical service conditions and reports them individually. Completion of administrative tasks is retained as progress, but the school-opening decision follows teachers, safe water, materials and facility operation.

The gender parity index for lower-secondary enrolment improves from 0.82 to 0.94. Female enrolment rises substantially, while male enrolment falls slightly in conflict-affected districts. National parity improved, but the full distribution includes both positive female access and male loss in particular locations.

The audit publishes female and male rates, counts and district patterns beside the index. It does not treat every movement toward 1 as unqualified progress. Policy responds to both the persistent female gap and the new male participation risk.

Source: ICEQC hypothetical evidence audits.

Each audit preserves valid progress while removing the unsupported extension. This matters for public confidence. Correcting a claim does not require denying that classrooms were built, teachers appointed or enrolment recorded.

The corrected evidence also produces a clearer action. It identifies denominator, distribution, operation, exclusion or sustainability as the next decision point.

Most corrections in the worked cases do not depend on advanced statistics. They depend on states, populations, periods and transparent limitations. These are the core of minimum evidence.

Technical sophistication becomes necessary when sampling, assessment, modelling or price comparison supports a broader claim. It remains subordinate to the same interpretive discipline.

The least coherent evidence capable of supporting a defined claim for a stated population, period and decision without concealing material uncertainty or harm. It is not a minimum educational entitlement.

The linked claim-specific evidence that cannot be omitted from a responsible quality judgement. Critical dimensions are non-compensable.

The state asserted: commitment, resource, service, participation or learning. Classification determines the direct evidence required.

A statement that an authorised body adopted a duty, policy or target. It does not establish implementation.

A statement concerning authorised, financed, procured, delivered or held resources. The precise transaction state is named.

A statement that an educational service was available, functional or delivered for a defined population and period.

A statement concerning entry, enrolment, attendance, active participation, progression or completion under stated definitions.

A statement about demonstrated knowledge, skill or competence based on an assessment suitable for the construct and population.

The persons, institutions, events or services to which the claim and intended inference apply.

The entity counted or judged, such as person, registration, teacher post, assignment, class, institution, period or currency amount.

The time observed by the evidence. It is distinct from the date of collection or publication.

The eligible population represented by the source, including exclusions and missing units.

Evidence that observes the state asserted. Directness is relative to the claim.

A measure used to represent a related condition when the intended condition is not directly observed. Its relationship and limits are stated.

Support from a source or method with sufficiently different errors. Repetition of one source is not corroboration.

A controlled examination of a claim, source or operating condition through trace, comparison, observation or competent inspection.

A discontinuity caused by changed definition, source, coverage, assessment or system structure that limits comparison.

A value released before the defined validation cycle is complete and expected to be revised.

A value inferred for an unobserved quantity through a stated method. It is distinguished from an observation.

A conditional future value based on assumptions about population, flows, resources or other variables.

The range or risk of error arising from sampling, population, missingness, measurement, classification or modelling.

Enrolment at a level regardless of age divided by the population of official age for that level.

Official-age learners enrolled at a level divided by the population of that official age.

A measure representing completion through available final-grade or age-based evidence rather than direct observation of an entry cohort fulfilling requirements.

Persons or activities converted to the equivalent of full-time service under a stated load.

Pupils divided by teachers in a matching level and sector. It is not average class size.

Time during which scheduled teaching is delivered. Intended, scheduled, available and delivered time are separate.

A resource or arrangement operating and accessible for the intended population and period. Inventory existence is insufficient.

The learner's access to required content through time, teaching, materials, language and participation conditions.

A critical evidence or service condition whose absence cannot be offset by favourable performance in another dimension.

The ratio of one group's value to another's for the same measure. It does not show the absolute level of either group.

Unresolved valid cases carried at a point in time, reconciled through new entries, verified closures and reasoned removals.

The state in which the intended educational service functions. It follows administrative or physical completion.

The authority, competence, records, logistics and correction arrangements needed to produce and use reliable evidence.

A claim narrowed by a stated limitation in population, period, source, certainty or use.

A concise account of claim, population, period, source, coverage, distribution, uncertainty, status and revision.

The wording of a quality statement does not exceed the highest evidentiary state directly observed and validly represented. Where evidence remains incomplete, the system narrows the claim, strengthens verification and acts on critical risk without converting unknown into compliance.

This framework interprets the evidence necessary for claims during educational expansion. It does not determine the complete content of quality, replace national law or curriculum, or establish one international threshold for staffing, expenditure, class size or attainment.

The listed domains are linked because access without service and service without learning are incomplete accounts. They are not an exhaustive theory of education. Culture, purpose, relationships and learner development can require forms of evidence beyond the minimum system record.

A system with strong records can still provide poor education; a system with weak records can contain effective schools and teachers. Evidence capacity and educational condition are separate judgements.

Weak capacity limits what can be claimed and can delay identification of harm. It is therefore a quality-system concern without being a direct measure of every institution or classroom.

International definitions improve comparison but cannot remove all national difference. Education levels, official ages, teacher preparation, provider coverage and completion rules remain embedded in systems.

Mapping should preserve material difference. A value that cannot be compared can still be valid for national management; it should not be excluded from action merely because it lacks international equivalence.

Every indicator reduces a complex condition. Pupil–teacher ratios omit class formation and time; attendance omits learning; assessment omits parts of educational purpose; expenditure omits service and distribution.

Companion evidence is selected to address the most material omitted link. This does not require an unlimited indicator set. It requires that the public conclusion remains within the combined observation.

Evidence can be strong in coverage and weak in definition, timely and weakly verified, or precise for a narrow population. A single quality label conceals these dimensions.

The framework uses a reasoned sufficiency judgement for the claim. It does not calculate one data-quality score that permits strength in one dimension to offset a fatal weakness in another.

Most system-monitoring indicators are descriptive. They show conditions and changes but do not identify the effect of a policy without additional design. Expansion occurs alongside demographic, economic, institutional and social change.

The framework supports causal enquiry by separating intervention, implementation, service and outcome. It does not imply that this sequence alone establishes causation.

Learning assessments observe selected constructs under designed conditions. They do not measure every aim of education and can be sensitive to language, participation and opportunity.

The absence of a tested difference does not prove equal educational experience. A detected difference does not identify its cause or justify narrowing access.

Household surveys can observe children outside school records but may omit mobile, institutionalised, homeless or insecure populations. Population projections can be weak between censuses. Administrative data can be timely but exclude non-participants.

No single source is treated as complete by title. The system describes their complementary coverage and unresolved populations.

National evidence can guide policy while remaining insufficient for individual schools. A national sample estimate cannot determine the condition of an unsampled school.

Conversely, a verified school case cannot estimate national prevalence. Case response and system estimation have separate evidence routes.

Timely provisional evidence can be more useful for allocation than a final value released too late. Early evidence carries higher revision risk.

The framework does not prescribe one balance. It requires that status, expected revision and decision reversibility are explicit.

Public disclosure constraints can prevent detailed reporting for small or vulnerable groups. This may reduce external reproducibility.

Competent authorities retain protected access for action, and public reports state that suppression occurred. Confidentiality is not interpreted as absence of evidence or absence of need.

Collecting and verifying evidence requires finance and staff time. Expansion can overwhelm schools with reporting demands. A theoretically complete set can produce lower actual quality if it diverts teaching and encourages superficial entry.

The framework therefore prioritises claim-critical fields, sampling and reuse. It cannot eliminate the need for investment in registers, assessment and professional statistical capacity.

When an indicator controls money, recognition or sanction, providers can change behaviour around the measure. They may improve the intended service, alter recording, select participants or neglect unmeasured activity.

Balancing evidence and verification reduce this risk but do not remove it. Indicators and consequences are reviewed when behaviour changes.

The framework can be misread as requiring identical fields in every country. Its actual requirement is functional: every claim needs population, period, source, coverage, uncertainty and public-interest interpretation.

The source and indicator can differ. A local register, survey or sample can fulfil the function where it observes the claim credibly.

Missing evidence means the condition is unknown under the proposed method. It is not automatically adverse for statistical reporting.

Protective decisions can nevertheless act under uncertainty where the consequence of inaction is serious. The classification of evidence and the action threshold are separate.

Strong evidence can demonstrate that a service is inadequate. Passing the evidence test does not pass the educational criterion.

Reports display both judgements. Otherwise investment in honest measurement could appear to worsen quality while systems that do not measure weakness appear favourable.

An aggregate cannot establish that every group or institution meets a condition. Even a high national rate can coexist with complete exclusion of a small population.

Distribution follows known risk and non-compensable duties. The framework does not require every possible cross-tabulation, but it prohibits a universal or equitable claim without relevant examination.

Spending can be lawful and necessary without producing the intended service within the period. Procurement, receipt, function and outcome remain separate.

The report can acknowledge financial effort while keeping the educational result open. This is not a judgement that spending was wasted; it is a statement about what has not yet been observed.

Additional observations improve evidence only if they represent the intended population and measure the relevant condition consistently. Large biased datasets can provide precise but incorrect estimates.

Quality is strengthened by better definition, selection, verification and correction, not volume alone.

Source: ICEQC interpretive boundaries.

The framework's principal limitation is deliberate: it does not answer a broad question with one measure. It requires the claimant to state the educational proposition and accept the boundary of the source.

This restraint supports stronger policy. It makes uncertainty and unequal service visible and reduces the risk that expansion activity will be reported as quality before learners receive the intended education.

An expanding system may be unable to establish the full evidence floor immediately. Staging orders work by public consequence and dependency. It does not grant permanent exemption from observing critical service or learning.

The stages below are functional. A system can already have advanced assessment and weak institution registers, or strong finance and weak learner flow. Development occurs by domain rather than one overall maturity label.

The first stage identifies operating education institutions, location, level, provider and status. It assigns a stable institution key and records openings, closures and changes.

Without the frame, national coverage and sample selection cannot be established. Local lists are reconciled, and unverified institutions remain visible.

The system establishes enrolment unit, reference date, grade or programme, age where available, sex, entry, attendance and exit states. It develops proportionate methods to reduce duplicate registration and distinguish transfer.

Population denominators and household evidence are linked where feasible. Unknown ages and destinations are not filled without method.

Teacher person, post, payroll, qualification and assignment records are reconciled. Timetable and school-operation evidence identify whether instruction can be delivered.

The stage begins with high-need and expansion areas if national reconciliation is not yet feasible. It protects professional due process and separates individual records from public aggregates.

Schools record safe and functional spaces, water and sanitation, essential materials and accessibility under common definitions. Risk-based verification prioritises critical, new and under-observed institutions.

Inventory is linked to operation. The system identifies the authority and service response for unsafe or unavailable conditions.

Promotion, repetition, transfer, exit and completion states are defined and reconciled. Where individual linkage is unavailable, transparent reconstructed-cohort methods are used.

The system improves follow-up of unknown destination and keeps proxy measures clearly labelled. Cohort evidence is aligned with programme and assessment change.

Authorised curriculum, planned time and delivered opportunity are linked through school records, work samples, observation or selected modules. The method includes new and remote provision.

Collection is targeted to decision and burden. The purpose is to detect systematic omission and unequal opportunity, not to record every lesson centrally.

A suitable assessment programme defines construct, population, sampling, administration, scoring, participation and reporting. It can begin with representative samples rather than universal tests.

Results are connected with opportunity and distribution. The system protects against exclusion and does not attach high-stakes uses unsupported by the initial design.

Budget, release, expenditure, receipt and operation are linked for major expansion inputs. Per-learner and unit-cost measures align finance, period and population.

Sensitivity analysis identifies recurrent and fiscal risk. Household burden is added where it materially affects access.

The system develops risk-based field checks, cross-source reconciliation, correction routes and versioned public releases. Verification findings inform both statistical improvement and service remedy.

The stage is not postponed until all data systems are sophisticated. Basic correction and high-risk verification are necessary from the first public claims.

Once core systems operate, duplicate collections are removed, definitions aligned and valid records reused. Rotating modules and samples address less frequent questions.

Integration does not create one unlimited personal record. Linkage follows purpose, access and confidentiality. The system retains decentralised evidence where central collection adds no decision value.

Priority first protects life, rights, essential service and validity of educational decisions. It next secures the denominators and frames on which allocation depends. It then strengthens outcome and efficiency evidence.

A politically visible indicator does not automatically displace a less visible but decision-critical register. The authority states which claims remain unsupported during development.

Schools require clear definitions, manageable records and feedback. Intermediate authorities require reconciliation, support, logistics and case-resolution capacity. National authorities require methodology, aggregation, publication and finance. Technical bodies require professional independence and correction authority.

Training follows role. A school record keeper is not expected to conduct national sampling analysis; a national analyst must understand how local data are produced.

Evidence investments are assessed by the decisions improved, errors reduced, burden, recurrent cost and protection. Expensive technology is not presumed necessary or sufficient.

Paper or simple electronic systems can support credible evidence if identifiers, definitions, custody and correction operate. Technology choices reflect communications, maintenance and staff capacity available in 2005.

Milestones are operating states: percentage of institutions reconciled to the frame; proportion of learner records with resolved status; school assignments linked to unique teachers; critical facilities verified; cohorts reconciled; assessment participation established; and material corrections published.

Completion of software acquisition, training or form design is an implementation step, not the evidence-capacity result.

Source: ICEQC staged evidence-capacity sequence.

Progress is reviewed through operation, error, coverage and use. A new system that increases reported missingness can represent improvement because it reveals previously hidden uncertainty.

The authority examines whether evidence changed allocation, corrected service and improved public claims. A system producing timely indicators without a response route remains incomplete.

Stages can fail because forms are too complex, identifiers are unstable, staff lack time, communications are weak or data carry punitive consequences. The response addresses the binding condition rather than adding parallel collection.

Pilot evidence is examined in ordinary and difficult settings. Expansion of an evidence process follows demonstrated usability and decision value, not successful operation in the best-equipped sites alone.

Evidence capacity is built around public decisions and educational service. The sequence begins with knowing which institutions and learners exist and whether essential instruction operates; it advances toward stronger flow, learning, finance and comparative evidence.

At every stage, the public claim remains within the current capacity. This protects credibility while allowing systems to improve evidence progressively and direct scarce analytical resources to the conditions that matter most for learners.

The decision review is completed before a material quality claim is used for public reporting, allocation, target monitoring or evaluation. It examines whether the evidence supports the proposed wording and use. It does not decide whether the observed educational condition is favourable; that is a separate substantive judgement.

The reviewer records the exact sentence, indicator or conclusion proposed. Broad terms such as quality, access, efficiency or success are replaced by the particular state intended unless the full breadth is genuinely supported.

The verb receives special attention. Adopted, financed, supplied, available, delivered, attended, completed and learned require different evidence. A change to the verb can be the necessary correction even when the number remains valid.

The eligible population, observed population, exclusions and missing cases are stated. The reference period is distinguished from collection and publication date.

Where a claim says “national,” “all” or “universal,” the reviewer tests frame and population evidence beyond responding institutions. If material populations are unobserved, the wording is narrowed.

The reviewer identifies which state the source observes directly. Policy and budget sources establish intention or authorisation; administrative transactions establish earlier implementation; operating and outcome evidence establish later states.

A proxy can be accepted where the relationship is clear, the direct measure is unavailable and the limitation alters the wording. The proxy is named rather than concealed under the target concept.

Every quantity traces to numerator, denominator, unit and inclusion rule. Recalculation uses source values before rounding. Ratios are aggregated through matching totals where appropriate.

Definitions are compared with prior periods and comparison systems. Material changes create a break or bridge. The reviewer does not approve a trend merely because the values share a label.

The review considers whether missing units differ systematically and whether the conclusion changes under simple bounds or plausible adjustment. Response is examined across relevant regions and provider types.

Unknown values are not coded favourably. For a critical service, the unknown can trigger direct verification or precaution without being statistically classified as failure.

The reviewer asks whether the average conceals a group, institution or place without the required service. Relevant disaggregation follows the claim and known risk.

Safety, non-discrimination, required curriculum and award integrity are treated as non-compensable. A favourable aggregate cannot close a verified exception.

Sampling, population, measurement, classification and model uncertainty are recorded. Formal intervals are used only where the design supports them and are not described as covering every source of error.

Published precision follows evidence quality. If plausible values cross a decision threshold, the claim remains provisional or further evidence is obtained.

The intended use is stated. Screening, operational allocation, public monitoring, international comparison and causal evaluation require different thresholds.

Where evidence is adequate for a narrower use, the approval states that use. Evidence sufficient to send support to a reported school need does not become a national prevalence estimate.

The final wording includes the value or finding, population, period and material qualification. Preliminary, estimated or projected status appears beside the result.

Limitations are written plainly. They do not rely on generic language such as “data should be interpreted with caution” without stating why and how interpretation is affected.

Source: ICEQC minimum-evidence decision review.

The analyst prepares the record; a competent reviewer challenges claim, definition, calculation and limitation; the authorised publisher or decision-maker accepts the disposition. One person can hold more than one role in a small system, but material high-consequence claims receive an additional informed check.

The subject ministry, district or institution can correct fact and explain context. It does not remove a supported limitation merely because the qualification is unfavourable.

The approved claim retains the evidence version and date. If source values or definitions change, the decision review is reopened. Dependent public statements and targets are reassessed.

The review record remains separate from personal case information. It contains enough detail to reproduce the public inference without unnecessary disclosure.

The final question is whether a reasonable informed reader could infer more from the wording than the evidence establishes. If so, the wording is narrowed or the limitation moved closer to the result.

Approval means that the evidence and statement correspond for the intended use. It does not signify endorsement of the observed condition, and it does not close any educational service gap revealed by the evidence.

A publication containing several indicators maintains a claim register. The register links every headline, chart conclusion and material policy statement to the indicator profile and decision-review disposition. This prevents a valid technical table from supporting broader narrative language elsewhere in the publication.

The register identifies claims that share a denominator or source. A correction to population projections, provider coverage or assessment participation can then be traced to every dependent result. Related claims are not assumed to require the same revision: an enrolment count may remain valid while the enrolment ratio changes because its denominator is corrected.

The register also records negative statements. “No material regional difference was found” requires evidence that regions were represented and the design could detect a policy-relevant difference. Absence of statistical significance or missing disaggregation is not translated into evidence of equality.

When a claim fails the threshold, the authority selects among four responses: narrow the wording; obtain a more direct source; reduce the decision consequence; or defer the claim while protecting affected learners. The choice and due date are recorded.

Deferral is not suitable where the available evidence indicates immediate harm. In that case, provisional protective action proceeds while verification establishes scale and cause. Conversely, a weak national estimate does not justify individual adverse action without case evidence.

The response considers burden and opportunity cost. A costly new census is not commissioned when a representative sample can resolve the policy question. A sample is not used when every institution requires an individual safety decision. Evidence design follows the unresolved decision rather than institutional preference for one method.

At least once within the relevant reporting cycle, the authority compares language across policy documents, statistical releases, budget reports and public summaries. Terms such as school opened, teacher appointed, learner enrolled, programme completed and quality improved are checked for consistent evidentiary meaning.

Inconsistency is corrected at source and in the public account. The purpose is not uniform prose. It is to prevent the same educational state from appearing differently according to the interests of the reporting unit, and to prevent different states from being merged under one favourable label.

The consistency review also examines omissions. If expansion is reported through enrolment and expenditure while instructional time, school operation or assessment participation is absent, the publication states that the quality conclusion remains incomplete. Addition of a missing domain follows evidentiary need, not a requirement to make every annual release identical. This preserves comparability without allowing routine format to determine which educational conditions become visible.