ICEQC-R-2005-02 — Primary Completion Beyond Enrolment: A Comparative Measurement Framework

feasibility and course, completion, course in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 1. Completion is the operative international commitment to connect evidence with delivery, unequal effect and correction. The Dakar Framework similarly calls for free and compulsory primary education of good quality and completion by all children, with particular attention to girls, children in difficult circumstances and members of ethnic minorities. The formulation contains four elements: a universal population, an opportunity to participate, progression through a defined course and completion of that course.[REF-01] [REF-02]

Each element creates a measurement requirement. “Children everywhere” requires a population denominator and an account of children outside registered provision. “Able to complete” concerns opportunity as well as observed outcome; current non-completion may reflect access barriers, cost, disability, insecurity, language, institutional supply or poor quality. “A full course” requires an identified beginning and end, a classification of the programme and a rule for curriculum or grade completion. “By 2015” requires a time series capable of distinguishing progress from statistical revision.

An enrolment ratio addresses only part of this object. A net enrolment ratio estimates the participation of children in the official age group for a level. A gross enrolment ratio compares enrolment of all ages with the official-age population. Both can show the reach of current provision, subject to their definitions. Neither shows directly whether entrants remain through the final grade. A system can expand grade-one entry rapidly while losing many pupils before the end; its enrolment ratio may improve years before completion does. A high net enrolment ratio can coexist with late dropout, repetition or weak final-grade progression.

Completion is not merely a more advanced point on the same scale. It joins a stock measure of pupils present at a time with a terminal event or status. That change introduces questions of cohort, time and evidence. If the measure uses pupils in the last grade during one school year, it records a cross-section of final-grade flow. If it follows children who entered together, it records a cohort trajectory. If it asks young people whether they have completed primary education, it records attained status. These forms answer different questions.

comparability and nationally, relevant, circumstances in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 1. Completion is the operative international commitment to connect evidence with delivery, unequal effect and correction. A total rate can meet a numerical target while a material group remains excluded if another group is overrepresented in the proxy numerator or if the rate exceeds 100 per cent. Equality is therefore part of the construct, not an optional analytical supplement. The same reasoning extends, where data and safeguards permit, to location, poverty, disability, language, minority status, displacement and other nationally relevant circumstances.[REF-09]

equity and cycles, different, duration in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 2. What constitutes a full course of primary schooling to connect evidence with delivery, unequal effect and correction. ISCED 1997 defines the primary level as normally beginning between ages five and seven, lasting four to six years and providing a sound basic education in reading, writing and mathematics together with an elementary understanding of other subjects. National structures depart from the modal pattern. Some integrate primary and lower-secondary grades; some divide primary provision between institutions; some have cycles of different duration.[REF-03]

For comparative measurement, the “full course” is the complete nationally defined programme mapped to ISCED level 1 for the reference year. The map is to identify the official entry age, nominal duration, first and final grade, graduation or promotion rule, and any programme variants included. It should record whether special, accelerated, non-formal or private programmes are counted and on what equivalence basis. A title such as “basic school”, “elementary school” or “first cycle” is not sufficient classification evidence.

Completion can be evidenced in several ways. A pupil may be recorded as a new entrant to the last grade; may remain enrolled at a prescribed census date; may finish the final school year; may satisfy an attendance rule; may pass an examination; may receive a certificate; or may enter the next level. These events do not occur at the same time and are not interchangeable. A measure based on final-grade entry will be available earlier and more widely than one based on certified completion, but it will include some pupils who later leave or fail. A measure based on transition to secondary education will omit completers who do not continue and may be distorted by limited secondary places.

The national measurement standard should designate one event for each published indicator. Where final-grade new entrants are used as the international proxy, the publication should say so. The word “graduates” must be reserved for a result based on the national graduation rule. “Completed primary” in household data must be tied to a question and coding instruction that identifies the last grade or credential. The public meaning of the terms should follow their measurement basis.

A reform of cycle length requires explicit treatment. Suppose a system changes primary education from six to seven grades. During transition, pupils may complete under both structures, the official final-grade entrance age changes, and cohort exposure differs. Joining the old and new values without a break would imply a continuous measure of an unchanged object. The correct publication identifies the two structures, reports them separately during overlap, explains the denominator rule and establishes the first year in which the revised definition can support a comparable trend.

The term “completion rate” is used for at least four distinct measures. The first is the **gross intake ratio to the last grade**, the present international proxy. It relates new entrants to the last grade, regardless of age, to the population of the official age for entry to that grade. The second is a **cohort completion probability**, the share of a defined group of entrants who eventually complete. The third is an **age-group attainment rate**, the share of persons in an age band who have completed primary education. The fourth is a **graduation ratio**, newly certified completers relative to a specified population. A fifth measure, survival to a particular grade, is often discussed with completion but represents progression to that grade under a reconstructed or observed cohort.

The proxy is attractive because ministries commonly collect enrolment and repeater data by grade each year. Its denominator can be obtained from population estimates. It is comparatively timely and can be disaggregated by sex where both components permit. Its principal weakness is that numerator members are not necessarily the same age as denominator members and may not complete the final year. It is a gross rate in the statistical sense, even where the shorter title “primary completion rate” is used.

capacity and transfer, violate, assumptions in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 3. Four meanings commonly carried by the word completion to connect evidence with delivery, unequal effect and correction. In their absence, reconstructed cohort methods infer progression from two consecutive years of grade enrolment and repetition. The 2004 World Development Indicators notes the simplifying assumptions: pupils who leave do not return, promotion, repetition and dropout rates remain constant over the period, and the rates apply equally to all pupils in a grade. Migration, re-entry, grade skipping and transfer can violate those assumptions.[REF-04]

An age-group attainment rate addresses the population result directly. For example, a household survey may estimate the proportion of 15- to 19-year-olds who completed primary school. The chosen age band should allow most late entrants and repeaters sufficient time to finish while remaining close enough to current conditions to be policy relevant. The measure includes past experience rather than only the current school flow. It can be disaggregated by household wealth or parental characteristics, but it depends on sample size, accurate reporting and stable classification.

A graduation ratio is institutionally precise where certification is universal and records are complete. It becomes less comparable where completion does not require an external credential, where examination rules differ, or where certificates are withheld for fees or administrative reasons. Passing a selective examination can measure achievement under a national standard, but using it as the sole completion event may conflate completion of the course with success in selection for the next stage.

These measures form a family and do not compete for a single label. Agreement among them strengthens confidence. Divergence provides information. A high final-grade intake proxy with a lower young-person attainment rate may indicate recent expansion, an inflated denominator relationship, final-year loss or differences in definition. A high cohort survival estimate with a low population attainment rate may indicate exclusion at entry. A high graduation ratio with weak transition may identify limited secondary capacity rather than primary failure.

Source: ICEQC measurement synthesis based on the definitions and limitations reported in ISCED 1997 and World Development Indicators 2004.[REF-03] [REF-04]

An event measure records that completion occurred during a period. A status measure records that a person had completed by the date of observation. An opportunity measure examines whether the system provided the conditions necessary to enter, continue and finish. Confusion among these dimensions leads to claims that the data cannot support.

uncertainty and appropriate, organising, discipline in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 4. Completion as event, status and opportunity to connect evidence with delivery, unequal effect and correction. They can count final-grade pupils, repeaters, examination candidates and graduates. A population survey is strongest on status and participation among individuals, including those outside school. Evidence of opportunity requires a wider account: whether a place was available, direct and indirect costs were manageable, the route was physically and socially accessible, instruction was acceptable, and provision could adapt to a child's circumstances without lowering the entitlement. The right-to-education framework of availability, accessibility, acceptability and adaptability provides an appropriate organising discipline.[REF-10]

The target measure should remain an outcome indicator. It must not be redefined to include every condition of opportunity. Nevertheless, policy interpretation should ask whether a low or unequal result originates in supply, demand, progression, quality or measurement. An increase following abolition of fees may represent genuine access and flow improvement, but it may also bring large classes and high repetition if teacher and classroom supply do not expand. The completion measure records the eventual terminal flow; it does not reveal the mechanism without supporting evidence.

This distinction protects both accountability and diagnosis. Governments can be held to a clear result without implying that the result alone explains performance. A published rate must be accompanied by a bounded diagnostic profile capable of locating failure. The profile need not contain every education statistic. It should contain measures that connect the population to entry, entry to progression, progression to completion and completion to educational substance.

The numerator used in the current proxy is generally final-grade enrolment less repeaters. This subtraction is intended to count pupils entering the last grade for the first time. It is an administratively feasible approximation. Its accuracy depends on three records: the identification of the final primary grade, a complete count of pupils enrolled in that grade, and a valid distinction between new entrants and repeaters.

The first record can fail during structural reform or where programme variants end at different grades. If the national system includes a mainstream six-year cycle and an authorised accelerated programme, the numerator cannot simply combine their terminal grades without an equivalence and double-counting rule. If private or community provision follows a different grade structure, coverage must be declared. The international classification map should precede aggregation.

The enrolment count should use a specified reference date. Registers opened at the start of the year may include pupils who never attend or leave before the census. A late census may exclude early leavers but may also miss mobile pupils. Systems are to state whether the count is opening enrolment, enrolment on a census date, average attendance or another measure. Changes in the census date can alter the numerator independently of educational progress.

Repeater status should refer to the same grade in the previous school year, not simply to age or examination failure. A pupil returning after an interruption may be recorded as a repeater in one system and a re-entrant in another. A pupil who transfers and re-registers can be counted twice where records are institution-based. A pupil who repeats only part of a programme may not fit the administrative category. Instructions and validation should address these cases.

The subtraction method can produce biased results where repeater reporting is weak. Schools may have an incentive to under-report repetition if it is treated as failure, or to over-report it if resources follow pupil-years. Age-grade distributions, previous-year enrolment and promotion records can provide plausibility checks. Large year-to-year changes in the repeater share require explanation before the completion proxy is interpreted.

Final-grade entrants are not final-grade completers. Some will leave, repeat or fail before the end of the year. The size of that difference depends on the national programme and cannot be assumed constant across countries. Where end-of-year records are available, systems should publish the relationship between first-time final-grade entrants, pupils completing the year, examination candidates, successful candidates and certified graduates. The international proxy can then retain its established form while its local meaning becomes more precise.

The denominator is derived from the official entry age and nominal duration of primary education. If entry is officially at age six and the programme lasts six years without repetition, the official age for entering the final grade is ordinarily eleven. The completion proxy compares first-time final-grade entrants of all ages with the estimated population aged eleven. This is a reference-population device; it is not a count of the same children as the numerator.

Population estimates between censuses depend on fertility, mortality and migration assumptions. In settings with weak civil registration, an old census, rapid displacement or substantial migration, the single-year age estimate can be uncertain. Small errors matter because the denominator is a narrow age group. A revised census may change both the current rate and the historical series even though school records do not change.

The official-age rule may be a poor description of actual age progression. Late entry and repetition make final-grade pupils older than the denominator age. Early entry produces the opposite effect. A temporary enrolment campaign can bring several ages into grade one together; years later, the final-grade numerator may represent an unusually large mixed-age flow. The gross proxy correctly records that flow but must not be interpreted as the percentage of children of the official age who completed.

Denominator governance should include an identified official population series, the date and source of its base census, projection method, revision policy and geographic boundary. Education authorities are not to select among population estimates according to which produces the preferred rate. Where the national statistical office and an international source use different series, both values and the reason for the difference must be documented.

For subnational rates, migration and small populations can produce pronounced volatility. Boarding schools and cross-boundary attendance can place pupils in one jurisdiction while their household population is counted in another. The numerator is ordinarily assigned by residence if the denominator is residence-based; if school location is used, the mismatch must be explicit. Small-area results may require pooled years or counts alongside rates.

distribution and repeater, subtraction, effect in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 7. Why the rate can exceed 100 per cent to connect evidence with delivery, unequal effect and correction. The numerator includes new entrants to the final grade regardless of age, while the denominator contains one official-age population. Late entrants, repeaters earlier in the cycle and early entrants can cause the terminal flow to exceed the size of that age group. Population underestimation, different reference dates, migration and incomplete repeater subtraction can add to the effect.[REF-04]

The result is not evidence that more than all children complete. Nor does it necessarily demonstrate error. It indicates that the gross proxy has reached a boundary at which its percentage form is easily misunderstood. The underlying counts remain informative. Publication should retain the observed rate, mark it as a gross measure, and provide a diagnostic note.

Truncating the value at 100 removes information and can produce a false plateau. A system moving from 118 to 104 may be resolving an age backlog while maintaining broad access; truncation would show no change. A comparison that treats 103 and 100 as equivalent may conceal different age and data conditions. For target assessment, a supplementary rule may classify values near or above 100 as consistent with universal terminal flow, but the source value should remain visible and must not substitute for population attainment evidence.

An investigation is to examine the age distribution of final-grade entrants, repetition by grade, the date and coverage of enrolment, the population projection, recent census revisions, migration, changes in cycle structure and the inclusion of programmes outside the main system. Where age is missing for many pupils, the limitation itself must be reported. The investigation may confirm that the result reflects a temporary cohort bulge, identify a correctable data defect, or show that the denominator needs revision.

remedy and described, verified, completion in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 8. The upper-bound property to connect evidence with delivery, unequal effect and correction. This qualification has direct implications for reporting. The measure is more accurately described as entry to the last grade than as verified completion.[REF-04]

The size of the upper-bound difference is not known from the proxy itself. In a system with near-universal completion of the final year, it may be small. In a system with a high-stakes examination, seasonal withdrawal, conflict or substantial final-grade repetition, it may be material. A universal adjustment factor would create false comparability. National systems are to measure the final-grade flow where records permit.

The preferred flow account begins with first-time entrants to the final grade. It then identifies pupils present at the reference census, pupils completing the prescribed instructional period, pupils eligible for any terminal assessment, assessment participants, successful candidates, certified completers and entrants to the next level. Not every system uses each stage. The point is to identify the events that exist and prevent movement between them from disappearing inside one label.

A lower observed number of graduates does not automatically mean the proxy is invalid. Certification may cover only part of provision or may require a selective examination that is not itself the definition of completing the course. The relevant comparison depends on national rules. The statistical report should distinguish loss during the final year from failure to obtain a separate credential.

For year *t*, let (E_{g,t}) denote all pupils enrolled in the nationally defined last primary grade *g* at the official census date, (R_{g,t}) denote pupils repeating that grade, and (P_{a,t}) denote the population of official age *a* for entering grade *g*. The proxy is:

**PCRP(t) = 100 × [E(g,t) − R(g,t)] / P(a,t)**

The notation PCRP is used in this report to distinguish the proxy from observed cohort or population-attainment completion measures. It is not proposed as a replacement for established international terminology. Its purpose is to keep the method visible while the measures are compared.

The formula requires alignment of level, year, territory and population concept. If school data refer to the school year ending in 2003 and population data to mid-2002, the publication must follow the stated international reference convention and identify it. If the numerator covers the de facto population while the denominator uses usual residence, the mismatch requires examination. If the final grade changes within the year, the applicable structure must be recorded.

The numerator is to be published as a count, together with the final-grade enrolment and repeater count from which it is derived. The denominator must also be published. A rate without components cannot be independently checked and makes revision difficult to understand. Sex-specific rates require sex-specific numerators and denominators; a sex distribution of the numerator alone is not a rate.

Source: ICEQC measurement specification derived from the contemporaneous proxy definition.[REF-04]

Completion failure begins before entry for children who never enrol. The first supporting view should therefore examine access to grade one. Gross intake records new entrants of all ages relative to the official entrance-age population. Net intake seeks to identify new entrants of the official age relative to that age group. Administrative accuracy depends on identifying first-time entry and age; household attendance can test whether children recorded as enrolled are actually participating.

An entry rate below the completion proxy is possible during demographic change or where older cohorts entered under different conditions. It must not be treated as an automatic inconsistency. The measures refer to different stages and age groups. A decline in births can leave a final-grade cohort larger than the current entry cohort. Conversely, recent expansion can produce a high intake rate while completion remains low until those entrants progress.

Entry age distribution is essential. A gross intake rate above 100 per cent may reflect a drive to enrol previously excluded older children. That can be a legitimate and necessary expansion. It also changes later grade flows, teacher requirements and the interpretation of completion. Reporting only the rate would lose the distinction between timely entry and backlog entry.

The minimum entry profile comprises the number of new entrants by single year of age and sex, the official entrance-age population, the proportion entering at official age, and the number or proportion never recorded in school where household evidence permits. Geographic reporting should separate absence of a nearby place from non-participation despite nominal availability.

Net enrolment measures the proportion of the official primary-age population enrolled in primary education. Depending on the convention, children of primary age enrolled at another level may be excluded from the numerator even though they are in education. Gross enrolment includes pupils outside the official age range and can exceed 100 per cent. Attendance measures actual participation during a defined recent period and can differ from administrative enrolment.

These measures identify the population currently connected to education but not the stage reached. A child can be enrolled for several years without progressing at the expected rate. Another may attend irregularly and remain on the register. Participation should therefore be reported with age-for-grade and repetition evidence.

Household and administrative participation estimates can diverge for legitimate as well as problematic reasons. Their reference dates may differ; survey respondents may report attendance while schools report registration; one source may include non-recognised provision; population weights may rely on different projections. Reconciliation should begin by aligning concepts rather than declaring one source erroneous.

Progression measures connect entry with the terminal stage. Promotion is movement to the next grade; repetition is enrolment in the same grade in a subsequent year; dropout is departure from the programme without recorded transfer or completion. These are flow concepts. Their estimation requires consistent grade records across years and a balance equation that accounts for entrants, promotions, repeaters, transfers and exits.

Repetition extends the time and cost required for completion. It can also inflate gross measures because pupils are older than the official age when they reach later grades. A low reported repetition rate is not necessarily efficient where policy requires automatic promotion or where unrecorded dropout replaces repetition. Interpretation should examine the promotion rule and learning evidence.

Survival to grade five or to the last grade estimates the proportion of an entering cohort expected to reach that point. The reconstructed cohort method is useful where individual tracking is unavailable, but its assumptions must be published. A rate calculated to grade five is not equivalent to completion in systems with a longer primary cycle. Both the destination grade and programme duration must remain visible.[REF-04]

Retention should also be examined through age-specific attainment. If a large share of older primary-age children remains in early grades, a stable enrolment ratio may conceal weak progression. A distribution of pupils by age and grade provides more diagnostic information than a national average age or repetition rate.

The last grade requires its own flow account because the international proxy stops at entry to that grade. The account is to identify first-time entrants, repeaters, transfers in, pupils present at relevant dates, leavers before completion, completers under the national rule and unknown outcomes. Where examinations govern certification or secondary selection, candidates and results must be reported separately.

Loss during the final year must be calculated only where opening and closing populations can be reconciled. A simple subtraction of graduates from opening enrolment may misclassify transfers, deaths, migration or pupils completing without certification. The method is to state which events are observable and how unknown status is treated.

Final-grade repetition may reflect learning difficulty, examination policy, scarce secondary places or administrative practice. Where repeating is used to obtain a better selection result after the course has been completed, subtracting all repeaters from enrolment may not correspond exactly to first-time exposure to the grade. National metadata should explain such arrangements.

Population attainment provides the strongest direct evidence of the share of young people who report having completed primary education. The reference age must be chosen with care. Measurement at the official graduation age will classify many late entrants or repeaters as non-completers before they have had a reasonable opportunity to finish. Measurement much later improves finality but refers to older education conditions.

A useful publication can present attainment for successive ages or bands, such as the official completion age plus three to five years, subject to sample size. The resulting curve shows whether completion is merely delayed or remains permanently low. It can also reveal sex and household differences in late completion.

The household question should record the highest grade or level completed, not only the highest level attended. Coding should map national responses to the applicable programme structure. Respondents who attended the final grade without completing it must not be coded as completers. Unknown and inconsistent responses should remain identifiable.

Census data can provide small-area and small-group detail but are infrequent and may contain limited education questions. Sample surveys provide more frequent and richer household characteristics but have sampling error and may not support local estimates. A national system is to use each source according to its strength and should publish confidence intervals or sampling errors for survey results.[REF-11]

Completion measures educational attainment in a formal sense; it does not establish what was learned. National programmes differ in curriculum, instructional time, language and standards. Even within a system, learners who complete may have experienced substantially different opportunities. World Development Indicators 2004 explicitly cautions that a high completion rate does not necessarily mean high levels of student learning.[REF-04]

The response is not to redefine the completion rate as a learning score. That would combine two different constructs and make each harder to interpret. Completion and learning must be reported as separate but connected dimensions. A child can be excluded from learning because the child does not complete; a child can complete without achieving intended competencies; and a child can acquire knowledge outside the formal completion route. Policy needs to distinguish these cases.

Learning evidence should correspond to the national curriculum and the use proposed. Where a representative end-of-primary assessment exists, the completion profile can report the distribution of performance among assessed pupils and the proportion of the relevant population represented. A score among enrolled final-grade pupils excludes children who left earlier. It must not be presented as the learning level of all children of the age group without a method that addresses exclusion.

The Education for All quality framework also directs attention to enabling conditions. Teacher availability, instructional time, textbooks, language, safety and school climate are not substitutes for outcomes, but they help explain why entry does or does not lead to completion and learning. A compact set must be selected on evidence and national relevance rather than attached as an unlimited list.[REF-07]

The proposed architecture links the measures without creating a composite index. Each view answers a different public question.

1. **Population access:** Who enters primary education, at what age, and who remains outside any recognised route?
2. **Participation and progression:** Who is enrolled or attending, how do pupils move through grades, and where are repetition and exit concentrated?
3. **Terminal attainment:** Who reaches the final grade, who completes under the national rule, and what does the international proxy represent?
4. **Distribution:** Which groups and places receive different opportunities or outcomes, and where is evidence absent?
5. **Educational substance:** What learning and conditions accompany completion, and what limitations apply to the available measures?

The architecture is designed to prevent a headline rate from carrying unsupported meaning. It also limits reporting burden. Measures are included because they locate a break in the route to completion or qualify the interpretation of the result. Indicators without a defined decision use must not be made mandatory merely because data can be collected.

Source: ICEQC comparative measurement framework informed by the international target, current indicator definitions and the EFA quality framework.[REF-01] [REF-02] [REF-04] [REF-07]

The available international evidence establishes large differences in terminal primary participation while also demonstrating limits of coverage and comparability. World Development Indicators 2004 reports the most recent primary completion proxy for 2000/01 to 2002/03 in its education-efficiency table. Regional aggregates are 100 per cent for East Asia and the Pacific, 97 per cent for Europe and Central Asia, 87 per cent for Latin America and the Caribbean, 91 per cent for the Middle East and North Africa, 78 per cent for South Asia and 48 per cent for sub-Saharan Africa. The low- and middle-income aggregate is 86 per cent. The sub-Saharan African aggregate is explicitly marked as representing only 60 per cent of the population.[REF-04]

The Global Education Digest 2004 reaches the measurement question through a wider set of internationally standardised participation and progression measures, including survival to grade five and age-specific enrolment. It records continuing work to improve the comparability of primary completion and cautions that such progression measures remain indirect evidence of learning outcomes. This distinction supports the indicator-family approach adopted here.[REF-06]

These figures are not to be read as a league table. A value of 100 for East Asia and the Pacific is a population-weighted regional proxy, not evidence that every child in every country completed. Regional averages can conceal large national and within-country differences. They also inherit the age and denominator properties of the gross proxy. The result for sub-Saharan Africa carries a material coverage limitation; missing countries may differ systematically from reporting countries.

Sex-specific aggregates reveal further differences. In the same table, South Asia records 84 per cent for males and 71 per cent for females. The Middle East and North Africa records 94 and 87 per cent respectively. Sub-Saharan Africa records 51 and 44 per cent within the stated coverage. East Asia and the Pacific records 101 and 98 per cent, illustrating that the gross measure can exceed 100 for one group. These data identify an equality concern but do not explain whether it arises at entry, through progression or at the terminal grade.[REF-04]

The Global Monitoring Report 2004 uses an earlier baseline and a trajectory analysis. It reports population-weighted completion rates of 97 per cent for East Asia and the Pacific and 93 per cent for Europe and Central Asia in 2000. Its assessment places Latin America and the Caribbean on track and identifies slower-than-required progress in South Asia and the Middle East and North Africa. For sub-Saharan Africa, it contrasts observed annual improvement of 0.20 per cent during 1990–2000 with required improvement of 4.59 per cent during 2000–2015.[REF-05]

The difference between values in the two publications does not necessarily constitute a contradiction. They can reflect different reference periods, vintages, country coverage, weighting and revisions. A comparative framework should preserve those attributes. Copying a figure into a later table without its vintage and coverage removes the information needed to reconcile it.

Notes: WDI values refer to the most recent data available for 2000/01–2002/03 and are gross proxy estimates. GMR findings use their own published baseline and trajectory method. Values and assessments must not be merged into a single series. Source: World Development Indicators 2004 and Global Monitoring Report 2004.[REF-04] [REF-05]

A completion rate normalises the terminal flow by an age-group population, but it does not normalise the quantity or content of schooling completed. A child completing four years and a child completing six years are both counted under their national primary definitions. ISCED mapping creates a common level category; it does not erase the difference in nominal duration.[REF-03]

Every comparison must therefore display programme duration. Where the policy question concerns the capacity to deliver the nationally defined primary entitlement, the national completion rate is appropriate. Where the question concerns years of schooling acquired, completion must be accompanied by duration or a separate attainment measure. A ranking that treats identical completion rates as identical accumulated schooling would be invalid.

Cycle duration also affects timing. In a six-year programme, policy changes at entry take longer to appear in final-grade flow than in a four-year programme. A comparison of progress over a five-year interval may therefore capture different stages of reform. Trend analysis should consider the exposure period and avoid attributing a terminal result to a policy introduced after the measured cohort entered.

Rapid expansion creates a characteristic sequence. Gross intake rises first, often above 100 as over-age children enter. Primary enrolment grows and age-grade dispersion widens. Classroom pressure and teacher demand increase. Several years later, a larger and older group reaches the final grade, raising the completion proxy and sometimes pushing it above 100. The sequence can represent genuine progress while also exposing quality and capacity risks.

The measurement framework should identify the expansion cohort rather than treating each annual rate independently. Entry by age and grade enrolment by age allow analysts to trace its approximate movement. Resource planning should anticipate the later demand for upper-primary teachers, classrooms, materials and secondary places. A terminal increase without this preparation may be followed by higher repetition, dropout or restricted transition.

A declining gross completion proxy after a backlog has passed is not necessarily deterioration. If age patterns normalise and the rate moves from above 100 towards 100 while population attainment remains high, the change may indicate successful absorption. The explanatory record should distinguish this case from a decline in final-grade entrants caused by renewed exclusion.

Conflict and displacement affect both components and the underlying opportunity. School registers may exclude closed areas, mobile learners and temporary provision. Population estimates can become highly uncertain. Children may cross administrative or national boundaries, attend irregular programmes or lose documentation. A national rate can improve mechanically if the denominator changes or if the most affected areas disappear from reporting.

Coverage must therefore be treated as a result. Publication should identify territories and providers from which no valid return was received, the estimated population affected, any substitute data and the limits of national aggregation. A modelled value must not be presented with the same status as a complete administrative count.

Where data are insufficient for a national rate, counts and bounded ranges may be more honest. Household surveys may also be infeasible or unsafe. The absence of a precise estimate does not remove the duty to document educational access and restore records. Data collection must not expose children or households to additional risk.

Completion statistics often combine a school-location numerator with a residence-based population denominator. In areas with substantial commuting, boarding or migration, the rate can exceed or fall below the resident experience. Urban centres may count pupils from surrounding districts; sending districts lose them from the numerator while retaining them in the denominator.

The preferred subnational measure assigns pupils by usual residence where reliable and lawful. School-location measures remain useful for service capacity and finance. Both can be published if clearly labelled. National totals are less affected by internal movement but can still be influenced by cross-border migration and refugee populations.

Residence data must be collected sparingly and protected. Detailed geography can identify small groups or displaced households. Statistical disclosure controls must preserve confidentiality while allowing sufficient information for equitable planning.[REF-12]

Promotion rules affect progression indicators. Automatic promotion can reduce recorded repetition without improving learning or attendance. A high-stakes terminal examination can create final-grade repetition among pupils attempting to qualify for secondary education. In another system, pupils may complete primary school without any external examination. Direct comparison of repeater or graduate rates would conflate policy with performance.

Metadata should identify the promotion rule at each grade, the definition of repetition, any terminal examination, the relationship between examination success and completion, and whether secondary selection occurs at the same point. Changes in these rules create series breaks or at least interpretation changes.

Completion should normally refer to finishing the primary course under the national rule, while examination results must be reported as a separate educational or selection outcome. If the examination is legally constitutive of completion, that status must be explicit and comparison limited accordingly.

The universal target covers children, not only pupils in government schools. The numerator must therefore include all recognised primary provision within the national definition, subject to data quality. Excluding private, community, religious or other non-government schools understates terminal participation; including institutions without a verified level map can overstate it or introduce duplicates.

Provider coverage must be documented by sector and territory. A provider register, common pupil definitions and reporting calendar are necessary. If one sector reports enrolment but not repeaters, the combined proxy will be biased upward. Estimates used to fill the gap must be separately identified and revised when actual returns become available.

Non-recognised provision presents a policy and statistical question. Children may receive substantial primary learning outside the formal register, but the programme may not satisfy the nationally defined course or may be invisible to administrative data. Household attainment questions and equivalence procedures can identify the issue. The statistical system must not silently classify all such participation as either completion or non-completion.

Annual school returns are the principal source for the international proxy because they can provide grade-specific enrolment, repetition and sex distribution across an entire registered system. Their value lies in continuity and operational detail. Their limitation is equally clear: they describe the population known to institutions and can omit children, providers or territories outside the reporting frame.

A controlled return should define the census date, pupil status, grade, programme, age, sex, repeater status, transfer status and provider sector. Definitions must be common across reporting units. A form that permits each institution to interpret “repeaters” or “completion” locally does not produce an aggregable national measure. Instructions must be tested in the languages and administrative settings in which they will be applied.

The return is to be accompanied by a register of expected institutions. Response coverage cannot be calculated from the forms received alone. Authorities need to know which schools were expected, which responded, which were closed, which submitted partial data and which were estimated. Coverage by pupil enrolment is useful, but an enrolment-weighted response rate can hide missing small or remote schools. Both institution and pupil coverage must be reported.

Validation begins within the return. Final-grade repeaters must not exceed final-grade enrolment; male and female counts should reconcile with totals; ages must be plausible for the programme; the number of new entrants should correspond to the subtraction rule. Cross-year validation compares enrolment with the preceding grade and year, allowing for repetition, transfer and change. Large discrepancies require resolution or a documented qualification.

Administrative data should retain the submitted value, the validated value and the reason for any amendment. Unrecorded replacement of school returns by central estimates weakens trust and prevents improvement at source. Imputation is permissible where its method and scale are disclosed and where imputed records remain distinguishable.

Examination records can identify candidates, participants and successful candidates at the end of primary school. They are useful where the examination has stable coverage and a defined relationship to completion. They are incomplete measures where schools or pupils can complete the course without entering the examination, where candidates repeat for selection purposes or where access depends on fees or institutional nomination.

Linkage with school records should use a stable pupil identifier where available. Name-based linkage is vulnerable to spelling, order, transliteration and duplicate names. Aggregate comparison can still test whether candidate counts are plausible relative to final-grade enrolment, but it cannot identify which pupils are missing.

Results should separate absence, withdrawal, invalidated scripts, failure and unclassified status. Combining all non-success into failure obscures barriers to participation and administrative loss. Where an examination is used for selection, the pass threshold and available secondary places may shape results independently of primary curriculum completion.

Certification records can provide a precise count of issued credentials but may undercount entitlement where certificates are delayed or withheld. A count of certificates printed is not a count of pupils who satisfied requirements. The responsible authority should distinguish eligibility, issuance and receipt.

A population census can measure school attendance and educational attainment across the resident population and can provide local detail unavailable from sample surveys. It also establishes the base for the single-year age denominators used in administrative rates. Its influence on completion measurement is therefore dual: it supplies direct household evidence and anchors the population projection.

Education questions should distinguish attendance from enrolment and highest level attended from highest grade completed. National categories must be capable of mapping to the primary programme applicable to each respondent's schooling history. This can be difficult after structural reform; an older respondent may have completed a programme that no longer exists. Coding manuals should retain historical equivalence rules.

Census non-response, age misstatement and proxy reporting affect estimates. A household member may not know the exact grade completed by another member. Ages may be rounded, producing heaping around preferred digits and unstable single-year denominators. Post-enumeration assessment and age-distribution analysis should inform the limitations statement.[REF-11]

Census revision of population estimates can materially alter completion rates. When a new census reveals that the projected population of final-grade entrance age was too low, historical proxy rates may fall after revision. The education series should show the revised vintage and explain that the change is statistical. Previously published values should remain accessible in a revision record.

Household surveys can identify attendance and attainment among children and young people regardless of their connection to registered schools. They also support analysis by household wealth, parental education, language, work, disability or other characteristics, depending on the survey. This makes them indispensable for examining distribution and for identifying children absent from administrative records.

The sampling frame determines who can be represented. Nomadic populations, institutions, displaced settlements, remote areas or insecure territories may be excluded. A nationally labelled survey must not be described as complete national evidence where its frame omits material groups. The coverage statement should accompany the estimate.

Sampling error must be reported. Differences between groups or years that lie within the uncertainty of the estimates must not be presented as established change. Complex survey design, clustering and weighting must be reflected in standard errors. Small groups may require pooled years or qualitative evidence; suppressing an imprecise rate is preferable to false precision.

Question wording affects the construct. “Did you finish primary school?”, “What is the highest grade completed?” and “What certificate do you hold?” can produce different responses. The analytical record should preserve the questionnaire, coding rules, field dates and treatment of current final-grade pupils. Translation must be reviewed for the distinction between attendance, promotion and completion.

The completion proxy needs a credible estimate of the population at one age. Complete civil registration would support annual cohort estimates, but many systems rely on census projections because birth registration is incomplete. The population series is not a neutral background number; it is a measured and modelled component of the rate.

The statistical authority should publish the base population, fertility, mortality and migration assumptions relevant to the projection and should establish a revision schedule. Education analysts need not reproduce demographic models, but they should understand the uncertainty and avoid treating the denominator as exact where the underlying evidence is weak.

Comparison of school-age population with registered births can be informative only after allowing for registration completeness, deaths and migration. Direct substitution of registered births for a population estimate in low-registration settings would exclude unregistered children from the denominator and could reward the very administrative weakness the indicator should expose.

Subnational denominators require particular caution. Internal migration assumptions are often weaker than national projections. A school expansion programme can attract households or pupils and change both residence and attendance patterns. Rates must be supported by counts and, where uncertainty is material, ranges or alternative denominators.

Reconciliation is not the mechanical selection of one preferred source. It is a documented comparison of concepts, populations, periods and classifications. The purpose is to determine whether differences are expected, whether they reveal coverage gaps, and which result can support which decision.

The first step is to align the object. Administrative final-grade entry cannot be expected to equal household completion among 15- to 19-year-olds. A valid comparison might instead relate administrative cohorts from earlier years to current attainment, allowing for age and migration. The second step aligns territory and provider coverage. The third aligns timing. The fourth examines sampling and population uncertainty.

Residual differences must be retained. If household evidence shows materially lower attainment than administrative terminal flow would suggest, possible explanations include final-year loss, inflated new-entrant counts, population denominator error, migration, recall or different programme coding. The reconciliation note should rank these explanations by evidence and state what further test is required.

No source must be adjusted merely to produce agreement. Administrative data may be revised following verified duplication; survey coding may be corrected following questionnaire review; population estimates may change after demographic analysis. Each adjustment needs an evidential basis and a version record.

Source: ICEQC source-reconciliation framework drawing on official education and population-statistics principles.[REF-04] [REF-11] [REF-12]

Data quality must be tested in an order that follows the production process. **Definition tests** ask whether the indicator and programme are specified. **Coverage tests** ask whether the intended population and providers are represented. **Internal tests** ask whether components reconcile. **Temporal tests** compare the series with preceding and subsequent values. **External tests** compare related sources. **Use tests** ask whether limitations are compatible with the decision.

A technically accurate calculation can fail the coverage test. A complete provider return can fail the definition test if repeaters are interpreted inconsistently. A stable series can fail the external test if household evidence shows a large unexamined difference. No single validation percentage establishes fitness for use.

Thresholds should identify records for review rather than determine truth automatically. A year-to-year movement above a set number of percentage points may be plausible after fee abolition, return from conflict or structural change. The system is to require an explanation and supporting counts. The same rule must not delete the value because it is unusual.

The use test is decisive. A provisional national proxy may be sufficient to identify broad capacity needs but insufficient to allocate sanctions to districts. Survey estimates may support national inequality analysis but not local ranking. The publication is to state the highest authorised use of each measure.

Missing data are often concentrated where educational exclusion is greatest. Schools in remote, insecure or weakly administered areas may be least likely to report. Household surveys may omit unsettled populations. Treating missingness as random can produce an optimistic result.

Publication should distinguish item missingness, institution non-response, territorial exclusion, unavailable denominator and suppressed data. Each has a different implication. An unknown repeater count can bias the proxy; missing sex can prevent equality analysis; an excluded district limits geographic coverage; a weak population projection affects the rate itself.

Imputation should use a stated model and must not remove the missing-data flag. Results must be published both with and without material imputations where feasible. Sensitivity analysis can show the range under plausible assumptions. The aim is not to manufacture completeness but to establish what conclusion remains stable despite the gap.

Where no defensible national estimate exists, the authority should publish the available counts and the plan to close the evidence gap. Withholding all information until perfection is reached weakens accountability; publishing an unqualified estimate does so as well.

The international commitment explicitly includes girls and boys. A total completion proxy must therefore be accompanied by sex-specific numerators, denominators and rates. Reporting the percentage of final-grade entrants who are girls is insufficient because it does not relate either group to its population.

The sex-specific proxy inherits all properties of the total measure. A male or female rate can exceed 100 per cent; age patterns and repetition may differ; population estimates by single year and sex can be unstable in small populations. The difference in percentage points and the ratio between rates can both be informative, but neither should replace the component values.

Analysis should locate the stage at which disparity emerges. Entry, attendance, repetition, survival and final-year completion may show different patterns. A narrow final-grade gap can coexist with large exclusion at entry if only a selected group of girls enters and those who enter progress strongly. Conversely, parity at entry can be lost through work, safety, early marriage, pregnancy, cost or institutional practice.

comparability and parity, evidence, complete in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 32. Sex-specific measurement to connect evidence with delivery, unequal effect and correction. Available global evidence indicates that the target will not be achieved everywhere. The measurement response must be precise about the level, indicator and direction of disparity and should avoid treating national parity as evidence that all girls and boys complete.[REF-04] [REF-05]

National averages can conceal regions with low school supply, weak progression or unreliable data. Geographic reporting should use administrative areas meaningful for policy while preserving comparability over time. Boundary changes must be recorded and, where possible, historical data recalculated to stable units.

Rates need residence-consistent numerators and denominators. Where pupil residence is not available, school-location results must be labelled as service-area measures. Maps can be useful but must not imply precision beyond the data and should show missing or low-coverage areas distinctly rather than colouring them as average.

Counts matter alongside rates. A small district with a very low rate and a populous district with a moderately low rate require different scale responses. Travel distance, population density, language and conflict exposure may explain different mechanisms. Geographic inequality is not a diagnosis by itself.

Administrative school records rarely contain a reliable household-income measure. Household surveys can estimate completion or attainment by wealth proxies, consumption or another economic classification. These measures identify distribution but require careful interpretation because wealth indices are relative to the survey population and may not be comparable across countries or survey rounds.

The age band should allow for delayed completion. Poorer children may enter later and repeat more often; measuring at the official completion age can exaggerate permanent inequality by counting delay as final non-completion. Reporting attainment across successive ages can distinguish delayed from absent completion while recognising that delay itself imposes costs and increases risk of exit.

uncertainty and should, reflect, interdependencies in Primary Completion Beyond Enrolment: A Comparative Measurement Framework require ### 34. Household economic circumstance to connect evidence with delivery, unequal effect and correction. It must not infer household preference from non-attendance without examining these constraints. The United Nations Millennium Project places education within a wider investment framework that includes health, nutrition, water, transport and public capacity; completion policy should reflect those interdependencies.[REF-08]

Many administrative systems do not identify children with disabilities consistently, and household questions may rely on narrow medical labels. Absence of a disaggregated rate cannot be interpreted as absence of inequality. The measurement plan should first examine whether the population is visible and whether classifications support rather than obstruct educational access.

School records must not require a diagnosis as the sole basis for identifying support needs. Household data may provide a broader population estimate, but question design and stigma affect reporting. Rates based only on children already placed in special provision will exclude those outside education and those included in mainstream schools without recorded status.

Publication should state the definition, data source and coverage and should avoid comparing categories that differ materially between systems. Qualitative and service-access evidence may be necessary until a valid rate is available. Confidentiality is particularly important in small communities.

Language of instruction and minority status can affect entry, comprehension, repetition and completion. Data may be necessary to identify inequality, but collection can create risk where identity is contested or where small groups can be recognised. National law, consultation and statistical confidentiality should govern the approach.

Administrative language categories may reflect the language of the school rather than the learner. Household self-identification can provide a different perspective. Neither must be converted automatically into an explanation of performance. Curriculum access, teacher language capacity, location, poverty and discrimination may interact.

Where a category cannot be collected safely, geographic, school-language or qualitative evidence may provide a limited alternative. The limitation is to be explicit. An aggregate described as universal must not silently exclude populations for whom no classification or return exists.

Distance, seasonal movement and low population density can make a conventional graded school difficult to provide. Children may enter late, attend intermittently or move between sites. School-census definitions designed for stable institutions can lose both pupils and provision.

The completion framework should identify valid alternative routes and their equivalence to the national primary course. Flexible calendars or mobile provision may support completion, but reduced instructional entitlement must not be accepted merely because delivery differs. Records must follow learners where possible and reconcile transfers.

Household surveys need frames capable of representing mobile and remote populations. Where this is not possible, targeted enumeration or administrative mapping should supplement the national estimate. Coverage limitations must be stated beside the result.

Single-dimension averages can conceal compounded exclusion. Rural girls from a minority-language group may face a different completion pattern from the national female or rural average. Analysis should examine intersections where there is a policy reason and sufficient evidence.

Small cells create statistical and confidentiality risks. A highly disaggregated table can produce unstable rates and identify individuals. Minimum reporting thresholds, pooling and controlled access to microdata must be established before publication. Suppression must not erase the existence of the group; a note can state that evidence was collected but cannot be published at that level.

Intersectional analysis must be selective. The purpose is to identify a mechanism or distribution requiring action, not to generate every possible cross-tabulation. A pre-specified analytical plan reduces the risk of selecting only striking results.

A national completion target must not be declared achieved solely because the aggregate proxy reaches 100 per cent. At minimum, the result must be examined by sex and territory and reconciled with population attainment. Material groups or areas with no data must be identified. The standard is a universal opportunity, not an average offset.

This does not require every subgroup estimate to equal precisely 100. Sampling error, population estimates and gross-rate properties make such a rule unsound. It requires evidence that no material disparity remains unexamined and that differences are within a stated tolerance or subject to a funded corrective plan.

The target-assessment statement should separate three judgements: whether terminal flow is consistent with universality; whether population evidence supports near-universal attainment; and whether material distributional gaps remain. A country can meet one judgement and not the others. This form is more informative than a binary label.

Source: ICEQC equity-reporting specification based on the universal and non-discriminatory character of the education commitment.[REF-01] [REF-02] [REF-09]

The policy choice is not between enrolling more children and improving quality. Poor quality contributes to non-completion through weak learning, discouragement, repetition and household judgement that the costs of attendance exceed its value. Rapid expansion without teachers, materials and space can weaken conditions and place later completion at risk. The two objectives must be planned together.[REF-07]

Completion evidence should therefore be read with a small group of condition and outcome measures. Their selection should follow the mechanisms most likely to affect progression in the national context. A general international list can guide but cannot replace the diagnosis.

A quality companion measure should never be used to discount the right of a child to access or complete. Nor should completion be treated as evidence that an inadequate learning environment is acceptable. The framework presents the dimensions separately so that failure in one cannot be compensated statistically by strength in another.

Expansion raises the number of pupil-years that the system must provide. Completion requires teachers not only in grade one but throughout the cycle. Shortages in upper grades can create large classes, restricted promotion or multigrade arrangements unsupported by training. A cohort-flow projection should therefore inform teacher and classroom plans.

Pupil-teacher ratios are broad capacity measures and do not show distribution, absence, subject competence or effective instructional time. National planning should examine teacher allocation by grade and area, vacancies, attendance and the preparation required for the primary curriculum. The measure is to use consistent definitions of teachers and pupils.

Where emergency recruitment is used, policy should identify preparation, supervision and progression to a sustainable workforce. A temporary increase in access must not be financed by an indefinite reduction in instructional entitlement for disadvantaged pupils.

Official school days do not equal time received. Closure, teacher absence, pupil absence, late start, early finish and non-instructional activity can reduce exposure. Completion of the formal grade sequence may therefore represent substantially different learning time across schools.

Attendance measures should use a defined period and distinguish enrolment from presence. School registers can provide daily records but may be completed retrospectively. Household reports can identify recent attendance but are affected by recall and season. Sample observation can validate both at higher cost.

The completion profile should include instructional time where loss is a documented mechanism. It must not assume that increasing formal calendar days alone will change learning or retention. Implementation evidence must show that time reaches pupils.

Textbooks and other essential materials can affect participation and progression, particularly where households must purchase them. A national measure should specify the material, usability, pupil access and reference date. Counting books in a store does not establish classroom availability.

Language can determine whether pupils understand early instruction and whether families can support learning. Measurement should record the instructional language and relevant learner language evidence without reducing quality to a single language policy. Teacher capacity, materials and transition arrangements are part of the implementation question.

Curriculum opportunity concerns whether the intended primary content was taught with adequate time and sequence. Completion certification must not be assumed to establish uniform opportunity where disruption or local shortage has shortened provision. Sampled work and school records can provide evidence, subject to clear criteria.

An end-of-primary assessment can qualify the interpretation of completion, but its population coverage is decisive. If only enrolled final-grade pupils sit the assessment, the result describes a selected group. In low-completion systems, it may exclude a large share of the relevant population. Reporting should show participation relative to eligible pupils and, where possible, to the age-group population.

The assessment construct, language, administration and scoring must support the claim. National curriculum assessments can inform system improvement. Cross-national assessments can provide comparative evidence for participating populations. Neither must be presented as a complete measure of educational quality.

A combined “quality-adjusted completion rate” is not recommended at this stage. Multiplying a completion proxy by the proportion meeting a learning threshold would compound uncertainties, treat excluded children through assumptions and conceal whether change arose in access or learning. A two-dimensional profile is more transparent.

Policy response must be directed to the point at which children are lost. Low entry requires analysis of school supply, cost, legal access and household barriers. High entry with low survival requires attention to repetition, attendance, quality and progression. Strong final-grade entry with lower graduation requires examination of final-year conditions or certification rules. High administrative flow with low household attainment requires data reconciliation and possible final loss.

The diagnostic sequence should use counts before rates. A grade-flow table can show where enrolment falls, but change between grades combines promotion, repetition, transfer and demographic differences unless cohorts are linked. The analysis is to state whether it is longitudinal, reconstructed or cross-sectional.

Interventions should have a corresponding measure. A school-fee policy should monitor entry and household cost; a teacher-deployment policy should monitor vacancies, class conditions and progression in affected areas; an attendance programme should monitor presence and grade continuation; an assessment reform should monitor learning and unintended exclusion. Completion remains the long-term result, not the only implementation evidence.

The cost of universal completion is not the cost of enrolling one cohort in grade one. It includes the full pupil-years required for entrants to progress, the additional cost of repetition, support for children at risk of leaving, teacher and classroom expansion, and the demand created at subsequent levels. A rate can improve temporarily without sustainable finance.

Budget analysis should distinguish allocations, releases and expenditure and should examine distribution to the places with the largest completion gaps. Equal per-pupil expenditure may be insufficient where distance, language, disability or small-school scale raises costs. Household expenditure must be considered because nominally free education can still impose barriers.

The Millennium Project report available in January 2005 argues for country-based investment planning across interdependent sectors. Completion strategies should identify education costs while recognising that health, nutrition, water, sanitation and transport affect attendance and progression. This does not justify attributing every development expenditure to education; it requires coherent planning and a clear account of responsibility.[REF-08]

The public profile should state the headline completion proxy, reference year, component counts, programme duration, sex-specific rates, geographic range, coverage and a concise limitations statement. It should include the most recent comparable trend and distinguish provisional from revised data. A reader is not to need to consult a technical annex to learn that the rate is a gross proxy or that a region is missing.

The diagnostic account should present entry, age, survival, repetition, final-grade flow, population attainment and learning or condition evidence selected for national relevance. It should explain major divergence and identify the policy action taken. Tables should preserve counts and denominators.

The technical record should contain the full indicator specification, ISCED map, provider frame, collection dates, validation rules, imputation, population source, sampling errors, revisions and contact for statistical clarification. It must be stable enough for independent reproduction.

Official language should match the measure. “The primary completion proxy was 87 per cent” is preferable to “87 per cent of children completed” where the numerator counts new entrants to the last grade. “Administrative records covered 92 per cent of expected schools” is preferable to “national data show” where coverage is incomplete.

Trend claims should identify the interval and comparability. “The revised proxy increased by six percentage points between the school years ending in 1999 and 2003 under the current population series” is more defensible than “completion improved rapidly”. Causal language requires a design capable of attribution.

Target language should distinguish achieved, consistent with achievement, on track and insufficient evidence. A gross rate at or above 100 may be consistent with universal terminal flow but does not by itself establish universal population completion. An “on track” judgement depends on the trajectory method and must not be treated as a measured outcome.

Education and population data are revised. Late school returns, corrected repeaters, provider-frame changes and new censuses can alter the series. Revision is a normal part of official statistics when governed transparently.[REF-12]

Each published value should carry a vintage and status. A revised release should state which years and components changed, why, and whether the interpretation changed. Historical tables should use one coherent vintage while preserving access to earlier releases.

Silent revision damages comparability. It can also create apparent disagreement between official reports issued at different dates. A reconciliation table should show previous and revised values for material changes and separate statistical revision from educational change.

Education ministries ordinarily own administrative school records and use them for planning. National statistical offices ordinarily hold responsibility for official population estimates and may conduct censuses and surveys. Examination bodies hold candidate and result records. Effective completion measurement requires agreed definitions and exchange, but it should preserve the professional responsibilities of each body.

The statistical function must be able to publish methods, limitations and corrections without pressure to produce a preferred target result. Policy authorities may explain action and context, but must not suppress an inconvenient estimate or change a denominator without demographic grounds. The Fundamental Principles of Official Statistics provide the relevant standards of impartiality, scientific method, transparency and confidentiality.[REF-12]

A national completion committee, where established, must be technical and decision-focused. Its tasks are to approve definitions, resolve classification, review coverage, reconcile sources, authorise revisions and refer policy findings to responsible authorities. It must not become a parallel producer of unversioned statistics.

Pupil-level or household-level data can improve cohort analysis and equity diagnosis, but collection must be proportionate and protected. Identifiers must be used only for defined purposes, access must be controlled, and public tables should prevent identification of individuals.

Small-cell suppression rules must be established before analysis. Geographic maps and group tables require particular care where a condition or identity is sensitive. Suppression must be recorded so that missing published cells are not mistaken for zero.

Data sharing for research should use de-identified files and binding conditions appropriate to national law. The public interest in completion evidence does not remove the rights of children and households.

The primary completion proxy can support annual national monitoring, broad comparative assessment, projection of terminal-grade demand and identification of unusual trends requiring diagnosis. With adequate disaggregation and components, it can contribute to sex and territorial analysis. It can also provide one element of evaluating progress toward the international target.

It must not alone determine school sanctions, teacher performance, individual entitlement or the quality of learning. It must not be used to rank small areas with weak denominators. It must not be converted into a causal estimate of a policy. It must not be treated as an observed cohort probability.

The authorised use should appear in the indicator specification. Where decision stakes increase, stronger evidence is required. An allocation formula may use completion gaps if population and coverage are adequate and if the formula does not reward under-reporting. High-stakes institutional judgement requires direct evidence within institutional control.

International tables are to preserve the national programme duration, reference year, coverage and proxy status. Regional aggregates should state weighting and country coverage. Values above 100 should remain visible with explanation. Missing values must not be replaced by regional averages without an imputation flag.

Country profiles should present the indicator family rather than a rank. A compact profile can show entry, primary-age participation, survival, completion proxy, population attainment, sex gap and available learning evidence. The source year may differ by measure and must be displayed.

Comparison should focus on questions that remain meaningful across systems: whether children reach the nationally defined terminal grade, the scale of age distortion, the distribution of attainment, and the quality of evidence. Differences in curriculum and duration should limit stronger claims.

A trajectory to 2015 requires a baseline, a target, a comparable series and a rule for expected progress. Linear percentage-point change, proportional annual improvement and gap closure produce different judgements. The 2004 Global Monitoring Report publishes observed and required annual improvement under its method; users should retain that method with the result.[REF-05]

Projection should use more than the latest two observations where possible and should test the effect of revision and exceptional years. A policy shock can change entry immediately but completion only after the length of the cycle. The model is to reflect cohort timing and capacity constraints.

“On track” is not a forecast that the target will be reached. It is a conditional classification based on the method and available observations. Publication should state the assumptions and identify regions or groups for which data are inadequate.

Before release, the responsible statistical authority should confirm that the programme map is current, component counts reconcile, coverage is known, the population denominator is authorised, revisions are documented, disaggregation is safe, and the language matches the measure. Unresolved material issues should appear in the release, not only in internal correspondence.

The policy authority should respond separately. It should identify what the evidence shows about access, progression, terminal flow, distribution and quality; what it does not show; and which action follows. Statistical uncertainty is not a reason for inaction where the direction of need is clear, but action must not convert uncertainty into an unsupported claim.

Source: ICEQC release-control framework consistent with official-statistics principles.[REF-12]

Implementation should begin with a written mandate identifying the purpose of the completion series, the responsible producers, the population and programme in scope, the publication schedule and the decisions the evidence may inform. The mandate is to connect the international commitment with domestic statistical authority. It must not imply that an international definition replaces national law concerning completion or certification.

The responsible education authority should own the operational records; the responsible statistical authority should approve or publish the population denominator under national arrangements; and the body releasing the combined rate must be named. Where responsibilities are decentralised, the mandate should allocate data collection, validation, correction and publication explicitly.

The initial mandate should authorise a diagnostic indicator family rather than a single target number. This prevents later disputes when the headline measure cannot explain an apparent change. It also creates authority to obtain the minimum supporting counts without establishing an unrestricted data collection programme.

The first technical task is to inventory every programme that may fall within primary education. The inventory is to include public and non-government provision, standard and accelerated routes, special arrangements, programmes operating under other ministries, and provision affected by territorial exceptions. Each programme is to be mapped to ISCED 1997 and the national structure.[REF-03]

For each programme, the inventory records official entry age, duration, grade names, final grade, promotion and completion rules, census date, provider count, enrolment coverage and available pupil attributes. Programmes not included in the national proxy require a reason. Equivalent programmes must not be merged until the counting and grade correspondence are established.

The inventory also lists population sources, household surveys, censuses, examinations and longitudinal records. It should identify their field dates, coverage, questions, units and access conditions. This creates the basis for reconciliation and prevents an indicator from being designed around one convenient source while other material evidence is ignored.

A baseline normally includes several preceding years so that volatility, revisions and structural breaks can be seen. Reconstructing a historical series requires the definitions and population vintages applicable to each year. Recalculation under a current definition is useful only if source data support it and the result is marked as revised.

The baseline package should contain final-grade enrolment, repeaters, proxy numerator, population denominator, proxy rate, provider coverage and sex distribution. Entry and progression measures must be added for diagnostic use. If an age-group attainment estimate is available, its reference population and survey precision must be included.

The baseline narrative should identify the largest uncertainties. A country may have complete public-school returns but weak non-government coverage; reliable enrolment but poor repeater identification; or strong administrative records but a population projection based on an old census. Improvement priorities should follow the uncertainty that most affects interpretation.

Revision of a school census must be limited to fields with a defined use and a feasible reporting basis. The final-grade module should collect total enrolment, repeaters, first-time entrants if directly observable, age, sex, transfer status and provider identifiers. Where end-of-year data are collected, completers and leavers should follow national rules.

Definitions must be accompanied by examples addressing re-entry, transfer, partial-year attendance and reform grades. Training should reach local officers and institutions and should include validation, not only form completion. Pilot testing should examine whether different reporters classify the same cases consistently.

Electronic processing can reduce arithmetic error, but the reporting environment in 2005 varies widely. Paper and electronic systems should use the same controlled definitions and validation logic. Technology must not be made a condition of including remote schools. The quality requirement concerns the record, not the medium.

A stable pupil identifier can support direct cohort measurement, transfer control and targeted support. Its introduction requires a clear purpose, coverage plan and safeguards. An identifier that operates only in well-resourced schools may initially worsen comparative coverage and must not displace aggregate reporting before it is sufficiently complete.

The identifier is to remain stable across schools and years, avoid embedding sensitive characteristics and include a process for duplicate resolution. Records should distinguish confirmed transfer from unexplained absence. Access to identifiable histories must be restricted to authorised educational purposes.

Where identifiers are not feasible, sample cohort studies can estimate progression and investigate mechanisms. A representative sample of schools and pupils can follow entry cohorts, verify movement and compare administrative classification with household status. Findings must not be generalised beyond the design.

Education and population authorities should meet before each annual release to confirm the official age, population vintage and known revision. A new census should trigger a planned transition rather than an unannounced denominator replacement. The transition is to compare old and new projections and identify the years requiring revision.

If the census produces a sharp difference at the official final-grade entrance age, analysts should examine age reporting, under-enumeration and migration before accepting the full change. The national statistical authority remains responsible for the demographic judgement. Education officials should explain the effect on the completion rate.

For years spanning a structure change, the denominator may need more than one age or separate programme calculations. Aggregation must be weighted from component numerators and denominators, not by averaging rates.

A household validation programme should use existing census or survey opportunities where possible. The core questions are current attendance, age, highest grade attended and highest grade completed. Additional questions must be introduced only where they address a defined ambiguity, such as equivalence of an alternative programme.

The analysis is to pre-specify age bands and disaggregation. It should compare administrative results with population attainment after allowing for cohort timing. A one-year discrepancy must not be overinterpreted when the sources describe different events.

Survey documentation, weights and standard errors should accompany the estimate. A finding that cannot be reproduced from the archived questionnaire and coding rules must not be used to revise an official series.

The production calendar should align school returns, validation, population estimates, examination data and publication. A late but more complete return may be preferable to a premature “final” figure; a timely provisional estimate can still be released if its status and revision date are clear.

Suggested stages are: confirm classifications before the school year; update the provider frame; collect at the common census date; conduct local and central validation; obtain the authorised population denominator; calculate component and rate tables; reconcile external evidence; review confidentiality; release provisional data; resolve outstanding returns; and issue the final or revised series.

Each stage is to have an accountable office and completion evidence. The calendar is to allow schools time to correct records and must not impose repeated uncoordinated submissions for the same pupils.

Statistical improvement requires personnel able to define, collect, validate, analyse and explain the measure. Short training limited to software operation is insufficient. National and local teams need knowledge of education structure, demographic denominators, survey evidence, flow analysis and public communication.

The recurrent cost of school returns, provider-frame maintenance, survey modules and publication must be budgeted. A donor-supported pilot must not become the only source for a core indicator without a financing and transfer plan. Common definitions and documentation are assets that should remain under national stewardship.

Success must be measured by reduced unresolved discrepancy, stronger coverage, timely correction and more defensible use—not by the number of indicators collected. A smaller system with reliable components can provide greater public value than a large dashboard of uncertain rates.

Consider a system with a primary net enrolment ratio near the target but a substantially lower completion proxy. The first inference is that current participation does not translate fully into terminal flow. It is not yet possible to conclude that dropout is the sole cause. Recent enrolment expansion may mean that large early-grade cohorts have not reached the final grade; demographic change or a revised denominator may also contribute.

The diagnostic sequence examines entry cohorts, grade enrolment, repetition, age-for-grade and survival. If losses concentrate between particular grades, school supply, curriculum transition, examination or household cost at that stage must be investigated. Household attainment can determine whether the pattern persists among older youth.

The policy response must not be a general completion campaign without a location of loss. Additional grade-one places will not correct an upper-primary bottleneck. A survival intervention will not reach children who never entered. Measurement distinguishes the problem before resources are assigned.

Assume a national proxy of 108 per cent after rapid expansion. The publication retains 108, labels it gross, and reports numerator and denominator. Analysts examine the age distribution of new final-grade entrants. If many are older than the official age and correspond to an earlier enrolment expansion, the rate can be explained as backlog progression.

The population series is then reviewed for census vintage and migration. Repeater classification and provider duplication are tested. Household attainment among an older age group is consulted. If attainment remains below universality, the system cannot claim that every child completed merely because terminal flow exceeded the age-group denominator.

The following year may produce 103 and later 99 as the backlog passes. The trend is to be interpreted with age and attainment, not as a decline from 108 to 99 in the share of children completing.

Suppose the numerator is unchanged but a revised population estimate lowers the official-age denominator, raising the rate by five percentage points. The release is to classify the movement as statistical revision. Historical values must be recalculated under the new series where appropriate, and the previous vintage retained in the revision table.

No claim of educational improvement must be made for the revision. Policy need may still change because the estimate of the population requiring provision has changed. The statistical and planning consequences must be explained separately.

Suppose school records show a proxy of 90 per cent while a household survey estimates primary attainment of 72 per cent among an age band expected to have completed. The difference is too large to dismiss, but the measures are not directly equivalent.

The review aligns cohorts and definitions, checks whether survey respondents completed under an older structure, examines sample coverage and confidence intervals, and tests school coverage, age and repeater data. Migration may move completers into or out of the surveyed population. Final-year loss and certification rules may also matter.

The release is to present both results with a reconciliation status. Selecting 90 because it is official administrative data or 72 because it is household-based would be premature. The difference identifies a priority study.

Suppose girls and boys have the same completion proxy, but girls enter later, repeat more and take longer to finish. A parity statement based only on the terminal rate would conceal unequal burden and exposure to risk. Entry age, age-for-grade and cohort duration should accompany the result.

Alternatively, the same terminal parity may follow lower female entry combined with stronger survival among girls who enter. The policy problem then remains at access. Stage-specific measures prevent the success of one selected group from masking exclusion of others.

Suppose a public report describes an 80 per cent examination pass rate as primary completion. If only 70 per cent of final-grade pupils took the examination, the published percentage represents 80 per cent of candidates, not children or final-grade entrants. Its implied success among final-grade pupils is at most 56 per cent before other adjustments.

The corrected publication reports candidate coverage, pass rate among participants and the national rule relating the examination to course completion. It keeps the completion proxy and examination outcome separate. If non-participation is concentrated by sex or region, that distribution is reported.

A system may report high survival and completion after restricting repetition. The change can reflect improved flow, removal of arbitrary retention or administrative promotion without adequate learning. Completion data alone cannot choose among these explanations.

The review examines attendance, instructional conditions and curriculum-linked learning evidence. It also tests whether promotion policy increased the number of pupils entering the final grade without support. The policy response may require targeted learning assistance rather than restoring repetition as an automatic remedy.

Suppose non-government schools provide 20 per cent of final-grade enrolment but do not report repeaters. Treating all their pupils as new entrants biases the numerator upward. Excluding the sector biases coverage downward. The authority is not to merge the incomplete component silently.

The publication can report the public-sector proxy, total final-grade enrolment and the unresolved non-government component, or use an explicit estimate with sensitivity bounds. The priority is to extend the common repeater definition and provider return.

A district containing several boarding schools may report a proxy far above 100 per cent because the numerator uses school location while the denominator counts resident children. Neighbouring districts may appear low for the opposite reason. The result measures service location, not resident completion.

Where residence data are available, resident rates must be calculated. School-location counts remain useful for staffing and finance. If residence cannot be established, district ranking must be suspended and the national aggregate retained with a note on internal movement.

A national return may cover all functioning reporting schools but omit areas where administration has collapsed. Describing the result as 100 per cent school response would be technically correct only relative to a narrowed frame and substantively misleading.

The provider frame should retain closed, inaccessible and status-unknown institutions. Coverage should state the territories and estimated population not represented. If no valid denominator exists, the release should avoid a precise national rate or present a bounded estimate with assumptions.

Indicator production should lead to a specified decision. Low official-age entry may require place creation, fee removal or outreach. High age distortion may require late-entry strategies and flexible support. Concentrated repetition may require curriculum, teaching or promotion review. Final-year loss may require examination, cost or transition measures. Geographic inequality may require needs-based finance.

The strength of evidence should match the decision. A national trend can justify broad planning; institution-level intervention requires local verification. A household survey can identify a poverty gradient; it cannot determine which individual pupil should receive a benefit without a separate eligibility process.

The authority is to record the decision, evidence, affected population, responsible office, resource commitment and review date. This closes the link between monitoring and public action without making the statistical producer responsible for policy delivery.

The first priority is a complete programme and provider frame. Without it, the numerator's scope is uncertain. The second is reliable identification of final-grade repeaters. The third is documentation of the official-age denominator and its census base. The fourth is sex-specific component publication. The fifth is a household attainment benchmark.

Where these foundations are secure, systems should improve age-grade data, end-of-year flow and subnational residence. Longitudinal identifiers must be introduced only with coverage and safeguards. Learning evidence must be strengthened separately and linked at aggregate level before any attempt at pupil-level integration.

Priorities will differ by country. The sequence is to follow the largest threat to interpretation, not an externally fixed technology model.

The upper-bound property of the proxy creates a clear research need: how many first-time final-grade entrants complete the year under different national rules, and which pupils do not? Studies should distinguish transfer, temporary absence, repetition, examination non-participation, failure, migration and permanent exit.

A sample design should represent varied completion levels, provider sectors and locations. Administrative linkage must be validated through school and household follow-up. Findings should report unknown outcomes and must not assume that non-certification equals non-completion.

The result can inform whether a national adjustment or separate observed-completion measure is feasible. It must not be converted into a universal international correction factor without evidence across systems.

Late entry and repetition are central to the proxy's interpretation. Research should map age at entry, age-for-grade and age at completion and should examine differences by sex, location and household circumstance. The purpose is to distinguish temporary backlog, persistent delay and permanent exclusion.

Population attainment curves by age can show when completion stabilises. Cohort or retrospective data can estimate time to completion. Analysis should account for structural reform and should avoid judging older pupils against a current age rule that did not apply when they entered.

Countries with both sources should establish a reconciliation study. It should compare definitions, cohorts, territories, provider coverage and uncertainty; reproduce each estimate; and trace plausible sources of difference. The study is to be commissioned independently of any need to defend one series.

Publication of agreement alone is insufficient. The study is to identify the remaining difference and its policy consequence. A small average difference can still conceal large disparity for a particular group.

The relationship between completion and learning must be examined without treating one as a substitute for the other. Studies can compare grade progression, opportunity to learn and curriculum-linked performance among cohorts and can investigate whether low achievement predicts repetition or exit.

Assessment coverage must include or explicitly account for pupils who leave before the final grade. Studies confined to completers can describe variation among those who remain but cannot estimate learning for the full cohort. Household-based assessment may offer broader coverage but introduces its own design and language challenges.

The research question is not whether completion “causes” learning in the abstract. It is which educational conditions allow broader participation to produce both sustained progression and meaningful learning.

Repetition and dropout consume pupil-years without producing timely completion, but a simple efficiency ratio can be misleading. Repetition may reflect policy, learning support or examination practice; automatic promotion can improve measured efficiency while leaving learning weak. Cost studies should connect financial records with flow and educational outcomes.

The analysis is to estimate the recurrent cost of current flows, the additional cost of universal progression under realistic teacher and infrastructure requirements, and the distribution of expenditure. Household cost and opportunity cost must be examined where they affect exit.

Research protocols should state the question, population, method, data access, confidentiality, analytical plan and publication commitment. Studies must be registered within the national evidence programme to reduce duplication and selective reporting.

Negative and inconclusive findings have public value. A pilot that cannot link pupils reliably or a survey question that produces inconsistent completion coding must be reported so that the method is not repeated uncritically.

Research should preserve the evidence cut-off and policy context of each release. Later improved methods may revise the series, but they must not be projected backwards without data and documentation.

International education data often use a single year label for a school year that spans two calendar years. Population estimates ordinarily refer to a date or mid-year point. The release must identify the convention used to bring them together. A rate labelled 2002 must not leave the reader to determine whether enrolment was counted at the beginning of 2001/02, the end of that school year or in calendar year 2002.

The convention is to remain stable. If it changes, overlapping values must be calculated where possible and the series marked. A presentational change in the year label can otherwise appear to shift progress forward or backward by one year.

Cross-national tables should retain the national source period and the harmonised reference year. This permits comparison while allowing a country to reconstruct the value from its own release.

The official final-grade entrance age is derived from the legal or administrative entry age and the nominal programme sequence. Actual age is not used to restrict the numerator. Where entry is permitted across an age range, the specification should designate the reference age used internationally and should publish the actual age distribution.

If different regions or programmes have different official ages, separate rates may be required. Selecting one age denominator for a mixed numerator can create a result with no clear population meaning. A combined national rate should aggregate programme numerators over their corresponding denominators.

Age should ordinarily be recorded at a common date, such as the beginning of the school year, rather than at an unspecified point. Date-of-birth records may be incomplete; the proportion of missing or estimated ages must be reported.

Children can leave school and later return. A reconstructed cohort assumption that dropouts never return will not describe such systems accurately. Administrative definitions should distinguish a pupil repeating the same grade continuously from a pupil re-entering after an interruption, while preserving the information needed to identify first entry to the final grade.

Re-entry is a positive route to eventual completion and must not be treated as statistical contamination. It does, however, widen the age distribution and change cohort timing. Population attainment by age and observed longitudinal records are better suited to measuring eventual completion.

Policy reporting should present both timely and eventual completion where delay is substantial. A system can improve eventual attainment while still imposing long interruptions that increase risk and cost.

A pupil moving between schools during the reporting period may appear in two registers. If both schools include the pupil in their census count, final-grade enrolment is overstated. A common census date reduces but does not remove the risk. Transfer records and identifiers support resolution.

The national instruction should specify which institution reports a pupil on the census date. Late returns must not add a pupil already included elsewhere. Aggregate anomaly tests can compare transfers in and out, but unmatched counts require local review.

Cross-border and cross-sector transfers are harder to reconcile. Data exchange must be proportionate and should protect personal information. Where linkage is unavailable, the likely scale of duplication must be estimated through a sample study rather than assumed to be zero.

Grade skipping and accelerated learning can allow pupils to reach the final grade without the nominal number of pupil-years. If the programme is officially recognised and mapped to primary education, its completers belong in the universal target. The measurement record should identify their route and prevent duplicate counting when pupils move into the mainstream programme.

The gross proxy may include first-time entrants to the final grade from accelerated provision. Programme duration and entry age differ, so a separate component calculation is preferable. Equivalence should concern completion of the nationally approved primary entitlement, not mere placement in a grade bearing the same name.

Learning and transition evidence may be needed to evaluate the programme, but should remain separate from the completion count. An accelerated route must not be judged automatically inferior or equivalent without evidence.

The universal commitment does not cease to matter when a child passes the official age. Late completers must be counted in gross terminal flow and in eventual population attainment. At the same time, policy should monitor timely completion because delay exposes children to additional cost and may restrict progression.

An age-specific completion profile can report the percentage completing by the official age plus one, plus three and plus five years, subject to data. This profile distinguishes systems that achieve similar eventual attainment through very different time paths.

The selected ages should reflect the national entry distribution and programme duration. International comparison should avoid a single rigid age band where it would classify expected late completion in one system as failure and not in another.

Observed cohort rates require a rule for pupils who die, emigrate or cannot be traced. Excluding all unknown pupils assumes their outcomes resemble those retained; counting them all as non-completers assumes the opposite. Both can bias the rate.

The study is to publish the original cohort, confirmed completions, confirmed non-completions, deaths, transfers or emigration, and unknown status. A primary estimate and sensitivity bounds can then show the effect of alternative assumptions. Record linkage must not be pursued beyond lawful and proportionate means.

The denominator for a right-based population target may differ from the analytic denominator for school effectiveness. A school-flow study may exclude confirmed emigration from its risk set; a national population account still needs to address resident children. The purpose determines the rule.

When a school closes, pupils may transfer, interrupt or leave education. Administrative systems should preserve the institution's records and track the disposition of pupils. Removing the school from the provider frame without a status record can make non-response disappear.

Completion analysis should identify cohorts exposed to closure and compare their subsequent registration and attainment. The effect is not to be inferred from the closure count alone. A planned consolidation with transport and transfer may differ materially from abrupt closure.

An annual release often precedes complete validation. Provisional data should use a coverage threshold and a declared estimation method. A later final release should incorporate returns and corrections and should explain material movement.

Late returns are not necessarily representative of timely reporters. Applying the average of responding schools can bias an estimate if late institutions are remote or weakly administered. Imputation should use available provider characteristics and must be tested against prior years.

Repeated lateness is a system-quality issue. The response plan should address reporting capacity and burden rather than merely increasing sanctions that may encourage fabricated on-time returns.

Fee abolition, compulsory-age change, curriculum reform, decentralisation and provider recognition can all alter both education and measurement. A trend is to distinguish real effects from new reporting coverage or classification.

The publication record should identify the date each reform took effect, the cohorts exposed and the data fields affected. A fee abolition can produce immediate entry change, while completion effects take the duration of the cycle. Recognition of previously uncounted schools can raise enrolment and completion in the release year without a corresponding immediate change in attendance.

Interrupted-series or cohort analysis may support stronger inference where enough observations exist. A simple before-and-after comparison must be described as descriptive and should account for concurrent population and data changes.

National and regional rates must be calculated from summed numerators and summed denominators for compatible components. An unweighted average of district or country rates gives small units the same influence as large ones and answers a different question. A weighted average should state its weights and coverage.

Aggregation can hide heterogeneity. The total is to be accompanied by distribution or range where policy requires. A region with a high weighted average can contain low-completion small countries; a national rate can conceal sparsely populated districts with severe exclusion.

If some components lack denominators or repeater data, they must not be inserted through an unexplained average. The aggregate coverage and estimation method must remain visible.

Completion rates are often published to one decimal place or as whole percentages. Precision should reflect the components. A survey estimate with a wide confidence interval must not be presented to two decimal places. A gross administrative rate can be arithmetically precise while its denominator remains uncertain.

Component counts must be retained so that published rounding does not prevent recalculation. Target judgements must not turn on a rounding boundary. A value of 99.6 rounded to 100 is not stronger evidence than its underlying uncertainty and distribution.

Categories such as on track, off track or seriously off track can support communication if their rule is published. They must not replace the underlying value or uncertainty. A country near a threshold can change category through minor revision without a material change in conditions.

The classification is to specify baseline, target, trajectory and minimum observations. Systems with insufficient data must be labelled accordingly rather than assigned to the favourable or unfavourable category through assumption.

Combining completion with enrolment, literacy, gender or learning into one index can provide a broad summary but reduces diagnostic clarity. Weighting embodies value judgements, missing data can change country coverage, and strength in one dimension can compensate for weakness in another.

For primary completion policy, a profile is preferable. If a composite is published for another purpose, its components, weights, normalisation, missing-data rule and sensitivity must be available. The completion measure should remain accessible in its original form.

An estimate describes an unobserved or incompletely observed current or past value; a projection describes a possible future value under assumptions. They must not appear in the same table without status labels. Modelled values must be distinguishable from reported administrative or survey data.

Projection scenarios can show the annual improvement required to reach 2015 and the implications for pupil numbers and resources. They are not promises or forecasts. Changes in entry, repetition, population and capacity must be represented rather than assuming a constant rate where evidence does not support it.

International comparability depends on national records and classifications. Harmonisation should clarify definitions and support quality; it must not lead countries to publish an international proxy as if it were the only valid domestic measure. National graduation and cohort measures can provide stronger evidence within the system.

Where an international series adjusts national data, the adjustment and source must be returned to national authorities for review and published. Differences between national and international values must be reconciled through metadata, not concealed by choosing one.

Every additional field imposes work on schools and local offices. Burden can reduce accuracy in core data. The measurement programme should identify which fields are essential annually, which can be collected periodically and which belong in sample studies.

Final-grade enrolment, repeater status, age, sex and provider coverage are core to the proxy. Detailed household circumstance should generally come from surveys rather than school forms. Learning and instructional-condition evidence may use representative samples.

Where evidence remains limited, the minimum defensible claim states the measure, source, period, coverage and limitation. For example: “Registered institutions reporting for the school year ending in 2003 recorded 84,200 first-time entrants to the last primary grade. Relative to the official-age population estimate, the gross terminal-grade intake proxy was 76 per cent. The result excludes the identified non-reporting territories and does not measure final-year completion.”

This form is not weakened by precision. It tells the public what is known, what population is represented and what further evidence is required. It also creates a stable basis for improvement.

Where the proxy is consistently near 100 per cent, the measurement priority changes from national expansion to distribution, denominator quality and educational substance. Authorities are to test whether small groups or territories remain below the national pattern, whether rates above 100 reflect age distortion, and whether young-person attainment confirms broad completion.

Stable high flow does not justify discontinuing the indicator. It provides an early warning if migration, fiscal pressure or institutional disruption affects particular cohorts. The annual public profile can be concise, with periodic deeper household and learning analysis.

Rapidly expanding systems need cohort-sensitive planning. Entry by age, grade enrolment and teacher or classroom supply must be projected together. Completion targets should reflect the time required for new entrants to traverse the cycle and must not reward premature claims based on grade-one growth.

The statistical priority is accurate age, repetition and provider coverage. Household evidence should assess who remains outside the expansion. Quality evidence should examine whether capacity reaches later grades as the cohort advances.

Where many children never enter, a completion proxy among terminal-grade pupils describes only the output of a narrow route. The first public table should give equal prominence to entry and never-attendance evidence. A high survival rate among entrants cannot be represented as broad system success without the population context.

Policy analysis should examine physical supply, cost, legal and administrative exclusion, safety, language, disability and household constraints. The measure of progress should show both new access and later continuation.

High repetition affects age, cost, classroom capacity and the completion proxy. Analysts are to identify grades and groups with concentrated repetition and should distinguish formal repetition from re-entry after interruption. The age distribution at the final grade is essential.

Reducing repetition by rule alone can improve flow statistics without addressing learning. Reform should combine valid assessment, early support and monitoring of subsequent learning and dropout.

Common national publication requires a minimum definition and transfer protocol while respecting subnational authority. Programme variants must be mapped separately and aggregated from components. A central average must not erase jurisdictional differences in duration or graduation rules.

The national framework should specify which level corrects school data, which supplies population estimates and which publishes. Subnational profiles should use consistent residence and boundary rules. Fiscal analysis should connect responsibility for completion with the resources assigned.

Single-age denominators and annual numerators can fluctuate sharply in small populations. A few pupils can move the rate by several percentage points, and migration can dominate the trend. Counts must be prominent; multi-year averages may support interpretation but must not replace annual records.

Confidentiality limits disaggregation. Administrative cohort review may provide better evidence than survey estimates because national household samples can contain too few observations. External comparison should acknowledge volatility.

National aggregates in populous systems can remain stable while millions of children are excluded. Subnational and household distribution is therefore central. Provider-frame completeness and consistent definitions across administrative units require continuing audit.

Sampling can validate records and study mechanisms, but a small national sample must not replace administrative counts for allocation. Publication should combine scale—numbers of children—with rates.

After conflict, disaster or administrative breakdown, restoration of the provider frame and pupil records is a priority. Provisional counts, rapid school mapping and population scenarios may be more defensible than a precise rate. Returning and displaced pupils must be admitted without requiring unavailable documentation while new records are established.

Historical comparison should identify the period during which coverage failed. A later rise can reflect restoration of reporting as well as education. Both are important and must be separated.

Changes in entry age, programme length or institutional boundaries require parallel reporting. Old-structure and new-structure cohorts must be tracked separately until the transition ends. A bridge table should map grades and completion events.

Targets must not be recalculated retrospectively unless historical source data can support the new definition. Public communication should explain why two valid completion measures may coexist temporarily.

The common standard is not a common policy prescription. Every system is to identify the primary programme, publish the proxy components, disclose coverage, preserve revisions and report distribution. The diagnosis and policy response depend on context.

This balance permits international comparison without false uniformity. It also protects national ownership: the international measure informs the shared target while domestic evidence determines how children are being excluded and what action is lawful, feasible and effective.

The uncertainty surrounding primary completion does not arise from one source and must not be reduced to a general caution. The school numerator may contain classification error, omission, duplication and reporting delay. The population denominator is an estimate subject to age error and demographic assumptions. Household attainment carries sampling and response error. A reconstructed cohort adds model assumptions. Each uncertainty affects interpretation differently.

An administrative total does not have a conventional sampling margin, but this does not make it exact. Its uncertainty must be described through coverage, validation exceptions, revision history and sensitivity to uncertain components. A survey estimate requires a confidence interval as well as non-sampling qualifications. A projection requires scenarios or an interval based on its assumptions. These forms are not to be combined into an unexplained single margin.

The public release should identify the dominant uncertainty. If provider response is complete but the population base is old, denominator uncertainty deserves prominence. If the denominator is strong but repeater status is missing for a sector, the numerator construction is the principal weakness. Generic language that “figures may be subject to error” is not sufficient.

Sensitivity analysis asks whether a decision changes under plausible alternative values. For the proxy, analysts can vary the repeater count, provider imputation and population denominator within evidenced bounds. The aim is not to choose the most favourable result but to show which conclusions remain stable.

If every credible scenario places completion far below the target, policy urgency does not depend on a precise point estimate. If the result lies near a target boundary and changes category under modest revisions, the authority should avoid a definitive achievement claim and prioritise data improvement. Precision matters most where it could alter a consequential decision.

Sensitivity should use documented alternatives. An arbitrary ten per cent range has little meaning. Bounds can derive from audit findings, alternative population series, prior-year response, survey confidence intervals or observed differences between provisional and final returns.

When the proxy exceeds 100, the usual percentage intuition fails. An uncertainty interval centred above 100 does not imply that the underlying probability exceeds one. It describes uncertainty in a gross flow ratio. Reporting should keep the observed components and shift attention to age composition and denominator adequacy.

The authority may state that terminal flow is consistent with full coverage only when household attainment, timely entry and distribution support that judgement. It must not transform the rate into a bounded probability by truncation. A footnote alone is inadequate where the headline can be misread.

A time series can move because the education process changed, because measurement changed, or both. Provider registration, repeater instructions, census revision, school-year convention and programme structure can all create discontinuity. A graphical line joining values across a break implies comparability that may not exist.

The publication is to classify each material movement as educational, statistical, mixed or unresolved. This is an analytical judgement supported by component records. A statistical change can still have policy significance—for example, discovery of a larger school-age population changes capacity needs—but it must not be claimed as improvement or decline.

Where a bridge is possible, both definitions must be calculated for an overlap year. Where it is not, the graph should show a break and narrative should begin a new series.

Population-weighted regional values describe the aggregate relationship between summed terminal flows and age-reference populations for reporting countries. They do not describe the typical country and cannot establish that every country shares the trend. A large system can dominate the average.

Coverage compounds the issue. The 2004 sub-Saharan African aggregate in World Development Indicators is marked as representing only 60 per cent of the population. A regional claim should retain that limitation and must not assume that missing systems match those reporting.[REF-04]

Alongside a weighted aggregate, publications can report the number of countries with valid data, the population coverage and a distribution of country values. No single summary removes the need for country profiles.

Country examples can show that rapid progress is possible or illustrate a policy mechanism. They do not establish that the same intervention will produce the same effect elsewhere. Selection often favours cases with clear data and visible improvement.

A comparative case should state why it was selected, the starting conditions, the measure used and the competing explanations. Policy transfer requires evidence that the institutional, fiscal and demographic conditions are sufficiently similar. An exemplar is a source of questions and design options, not a universal estimate.

Completion can rise after fee abolition, school construction, teacher recruitment or social protection, but temporal sequence alone does not establish attribution. Other reforms, demographic movement, economic conditions or changes in reporting may occur concurrently. Strong causal claims require a design that addresses plausible alternatives.

Routine monitoring should describe association and timing. An evaluation can use phased implementation, comparison groups, longitudinal data or another suitable design where feasible and ethical. The measure of exposure should reflect actual implementation, not policy announcement alone.

The public authority can act on a plausible and urgent diagnosis without claiming causal certainty. It should state the evidence, reasoning, residual uncertainty and review plan.

Different denominators answer different questions. Population at official final-grade entrance age supports the gross proxy. First-grade entrants support a cohort survival question. Examination candidates support a pass rate. Eligible final-grade pupils support examination participation. The full resident age group supports population attainment.

A high percentage can be produced by narrowing the denominator. Public tables must therefore place the denominator in the column title or immediate note. Changing the denominator between years or groups without disclosure invalidates comparison.

The following claim test must be applied: identify the persons described in the sentence, locate the same persons or a defensible reference population in the denominator, and verify that the numerator represents the stated event. If any element differs, the language must be revised.

Statistics can make exclusion visible, but collection and publication can also expose people. Identity, disability, migration or minority status may be sensitive. The decision to collect a characteristic must be justified by a public purpose, lawful, accurately defined and accompanied by protection.

An unknown category must not be reassigned for convenience. Missingness may reveal fear, weak administration or an unsuitable question. Public reporting should state the extent and consider whether a safer method is required.

The duty of visibility also applies to children outside registered provision. A system that publishes only school records can make those children statistically absent. Household and population evidence must be used to restore the population perspective without creating personal risk.

A delayed statistic can lose operational value; a premature statistic can mislead. The appropriate balance is a controlled provisional release followed by a scheduled final revision. Provisional status must be visible in headings and machine-readable metadata.

The provisional method should specify expected coverage, imputation and the date of the next release. A final release must not be called final if material territories remain unresolved unless the limitation is structural and explicitly recorded. Historical values may still be revised after a census, but their prior status must be preserved.

Timeliness targets should account for school capacity. Repeated demands for rapid resubmission can reduce accuracy and instructional time. A coordinated calendar and automated arithmetic checks can improve speed without weakening source verification.

The producer of the completion series should publish findings and corrections according to professional standards. A policy authority must not delay an unfavourable result to protect a target claim. Equally, statistical staff must not convert a methodological judgement into an unmandated policy prescription.

The two functions meet in a formal response. Statistics state the result, method and limitations. Policy explains the action, resources and timetable. Research tests mechanisms where evidence is insufficient. This separation supports accountability because each claim has an identifiable owner.

Uncertainty should qualify the strength of a numerical claim; it must not imply that no conclusion is possible. If all available sources show large exclusion, small differences among estimates do not remove the duty to act. If evidence is genuinely insufficient to distinguish progress from revision, the authority should improve the data before declaring success.

Official communication should answer three questions: what is known, what remains uncertain and what action follows. Technical detail supports these answers but must not be used to bury them. A concise limitations box near the headline can direct readers to the full record.

The releasing authority should sign an assurance statement confirming that the measure is defined, component data have passed declared validation, coverage and revisions are disclosed, uncertainty is proportionately represented, and the public wording does not exceed the evidence. Exceptions must be listed rather than overridden.

Assurance is not a guarantee that every record is error-free. It is an accountable statement that the evidence has been produced and presented under the approved method. Later correction remains possible and must be treated as a feature of responsible official statistics.

No conclusion about primary completion must be stronger than the weakest material link in the measurement chain. A complete numerator with an uncertain denominator supports a count more strongly than a rate. A precise proxy with poor population coverage supports registered terminal flow more strongly than universality. A high completion measure with no learning evidence supports formal attainment more strongly than educational quality.

This rule does not require every dataset to serve every purpose. It requires the publication to select a purpose consistent with the evidence and to preserve the distinction for readers and decision-makers.

Two completion values are comparable only to the extent that their constructs, programmes, populations, periods and component quality are sufficiently aligned for the proposed conclusion. Comparison is therefore a decision made against a stated purpose, not an inherent property of two percentages.

For descriptive international monitoring, programmes mapped to ISCED level 1 and proxy rates produced under the common gross construction may be compared with visible duration, reference year and coverage. For a claim that one system retains pupils more effectively, the comparison additionally needs entry and cohort-flow evidence. For a claim about educational quality, completion rates are insufficient and separate learning evidence is required.

The test proceeds through five questions. Are both numerators first-time entrants to the nationally defined last primary grade? Are both denominators the official-age population under an authorised series? Do the reference periods and territories align? Are provider and repeater data sufficiently complete? Does the proposed sentence stay within the common construct? A negative answer does not always prohibit comparison, but it requires narrowing or qualification.

International classification places programmes within a shared level, but the completed educational experience may differ in duration and organisation. The structural-equivalence test records those differences before rates are interpreted.[REF-03]

At minimum, the comparison states nominal years, entry age, final grade and whether primary is institutionally integrated with lower secondary. If one system's primary programme is four years and another's is six, the rates can compare completion of each national primary programme but cannot support an unqualified claim about equal years of schooling.

Structural reform introduces a cohort issue. Pupils completing under the former programme must not be mixed with pupils entering the terminal grade under the new programme unless component calculations are shown. A country profile may need two rows during transition.

The rate is to be reconstructed from published components. Final-grade enrolment minus repeaters must equal the proxy numerator. Dividing by the stated population and applying the declared rounding rule must reproduce the rate. Sex and territorial totals should reconcile, allowing for unknown categories.

Failure can arise from a simple transcription, use of different vintages or a hidden adjustment. The authority is to resolve it before interpreting the result. If an international publication applies an estimation not present in the national release, that status should appear in the component record.

Component consistency does not establish validity, but inconsistency disproves reproducibility. It is therefore an early release gate.

Policy interpretation requires evidence that the pupils represented by the completion measure were exposed to the policy. A national reform announced in 2004 cannot explain the entire 2004 final-grade flow in a six-year primary programme. Most pupils entered years earlier.

The exposure test maps the policy's start, implementation coverage and the grades or cohorts affected. It then identifies the earliest terminal result that could plausibly reflect full exposure. Intermediate indicators—places created, teachers deployed, fees removed or attendance changed—can be monitored sooner.

Partial exposure can still matter. A final-grade support programme introduced during the measured year may affect course completion even though it did not affect earlier survival. The claim is to identify the stage and avoid attributing the whole terminal rate.

A population denominator is adequate when its authority, age, territory, reference date and vintage correspond to the numerator and when known uncertainty is acceptable for the use. Adequacy is not established merely because the number came from an official table.

The test examines census age, projection horizon, migration, boundary changes and the relationship between official and actual age. For a national monitoring estimate, an authorised projection with disclosed weakness may be adequate. For district sanctions, the same uncertain single-age projection may be inadequate.

Where two official population series exist, the statistical authority should designate the release series or publish a reconciliation. Education authorities are not to average them without demographic justification. Sensitivity shows whether policy classification depends on the choice.

Coverage must be evaluated against the intended population, not the convenience of the responding frame. The test records expected and valid providers, pupils represented, territories included and the likely association between missingness and completion.

A 95 per cent pupil-weighted response may be adequate for a provisional national estimate if missing institutions are distributed and imputation is stable. It may be inadequate for an equity claim if the omitted five per cent consists mainly of remote or conflict-affected schools. Coverage adequacy is therefore purpose-specific.

The release is not to rely on one response percentage. Institution count, pupil share, territorial scope and imputed numerator share provide a fuller account.

An aggregate result supports a universal claim only when material distribution is examined. The test requires sex-specific components and geographic evidence. Other dimensions follow national context and safe data availability.

The objective is not to demand identical point estimates in every small group. It is to determine whether the aggregate hides systematic exclusion, whether uncertainty prevents a conclusion and whether the responsible authority has a corrective plan. Rates, counts and confidence intervals or coverage qualifications are read together.

A national rate rising while the lowest-performing areas stagnate must be reported as uneven improvement. A narrowing gap caused by decline in the higher-performing group is not equitable progress and must be identified.

Formal completion has intrinsic importance, but a public claim that completed education is of good quality requires more. The educational-substance test asks whether pupils had access to the intended curriculum, sufficient instructional conditions and valid evidence of learning.

No single companion measure is mandated globally. The evidence must suit the national curriculum and the population represented. Assessment among final-grade pupils must disclose that earlier leavers are absent. School-condition measures must not be treated as learning outcomes.

The test can yield distinct judgements: completion evidence is strong; learning evidence is limited; or both indicate a concern. This prevents a favourable rate from cancelling weak learning and prevents a learning score among selected pupils from cancelling exclusion.

A trend is acceptable when the definition, programme, territory, provider coverage and population vintage are comparable or when differences have been bridged. The test reviews metadata before calculating change.

If a new census revises the entire population series, trend must be recalculated under one vintage. If historical school components cannot be revised, old and new series remain separate. If provider-frame expansion adds institutions, the release should quantify the coverage effect where possible.

The test also considers volatility. A two-point line is not a robust trajectory. Several observations and component analysis are preferred, with exceptional years identified.

Near a universal target, small statistical differences can determine a categorical label. The target-proximity test requires stronger evidence when the estimate lies close to the boundary. It reviews denominator sensitivity, coverage, distribution and household attainment before declaring achievement.

A gross proxy of 99, 100 or 103 per cent can all be consistent with broadly universal terminal flow under different age conditions; none alone proves that every child completed. A proxy of 60 per cent, by contrast, establishes a substantial shortfall under any modest uncertainty. The required precision depends on the decision.

Authorities are not to exploit uncertainty selectively—emphasising it to minimise a low result while ignoring it for a favourable result. The same release standard applies across outcomes.

A projected path to universal completion implies pupils, classrooms, teachers, materials and recurrent finance across the cycle. The feasibility test translates the trajectory into service demand. It considers the larger cohorts already in early grades and the cost of reducing repetition and dropout.

The test is to use grade-specific enrolment and demographic projections, not simply multiply the completion gap by a national unit cost. Marginal expansion can be more expensive in remote areas or for learners requiring additional support. Existing capacity may also be used more efficiently, but savings must be evidenced.

A target classification without a feasible delivery path is not a complete policy assessment. Conversely, fiscal difficulty does not lower the statistical standard or the educational entitlement; it identifies a need for prioritisation and support.

An indicator is actionable when the public authority can identify a stage, population and intervention within its responsibility. The national completion proxy alone may show the scale of a problem but not its location. Supporting entry, flow and household evidence supplies that diagnosis.

The test asks whether the proposed action addresses the measured mechanism, whether resources and authority are assigned, and whether implementation can be monitored before the long-term completion result changes. A general commitment to “raise completion” fails the test if no operational decision follows.

Actionability does not justify overclaiming causation. A prudent response can be based on converging descriptive evidence and reviewed as stronger findings emerge.

Data demands must be proportionate to the decision. Annual school counts are justified for system monitoring. Detailed household characteristics are better obtained through periodic surveys. Pupil-level longitudinal data may be justified for transfer and cohort analysis where safeguards and capacity exist.

Collecting sensitive variables without an analysis and protection plan creates burden and risk. Conversely, failing to collect core sex, age or provider information can make universal claims unverifiable. The proportionality test balances public value, quality, burden and rights.

When an error is found, the authority determines its scope, cause, affected releases and consequence. It corrects components and derived rates, publishes the revision and preserves the prior vintage. A correction is not limited to silently replacing a number online.

If the error changes a policy conclusion or target status, the public explanation should say so. If it is immaterial after rounding, the technical record still notes it according to the revision policy. Recurrent errors should lead to a source-system response.

An informed analyst must be able to reproduce the rate from the published component record and understand every adjustment. Replication does not require release of identifiable pupil data. Aggregate inputs, definitions, formulas, vintages and estimation methods are sufficient for the headline series.

For survey attainment, replication requires access to the documented public-use or controlled data under appropriate conditions, questionnaire, weights and variance method. For a modelled trajectory, code or complete equations and parameters must be retained.

The final test asks whether the publication helps the public understand who completes, who does not, what the evidence can establish and what action follows. A technically correct but opaque table can fail this test. So can a persuasive narrative that omits a material limitation.

The public-interest test favours clear measure names, visible denominators, counts, distribution, correction and accountable response. It rejects rankings that cannot support decisions, composite scores that conceal failure and celebratory language that exceeds the evidence.

The tests are applied in sequence before public interpretation. Construct, structure and components come first; without them, later policy analysis rests on an unstable measure. Denominator and coverage establish the population meaning. Distribution and educational substance bound universality and quality. Trend, target and exposure govern temporal claims. Feasibility and actionability connect evidence to public decision. Correction, replication and public interest govern release.

Not every failure has the same consequence. A missing learning measure prohibits a quality claim but may not invalidate the completion proxy. An irreproducible numerator invalidates the rate. A weak district denominator may restrict subnational ranking while leaving the national series usable.

Every national programme should maintain an annual account from grade-one entry to terminal-grade flow. The account reports counts before rates and preserves grade, sex, age and provider distinctions. It identifies where enrolment change cannot be reconciled through promotion, repetition, transfer or known exit.

The account is not a longitudinal cohort unless pupils are linked. It is an operational balance used to detect pressure and loss. Its principal public outputs are entry, age distortion, repetition, survival and the completion proxy. Detailed exceptions remain in the quality record.

Where a suitable household survey or census exists, the authority should produce an attainment analysis at intervals supported by sample and cost. The analysis reports primary completion by age, sex, residence and approved national dimensions, with uncertainty and frame coverage.

Its purpose is to restore the population perspective and test whether administrative terminal flow corresponds with eventual attainment. It must not be forced into an annual cycle if the underlying survey cannot support annual change.

At least one representative study should quantify what happens after first entry to the final grade. It distinguishes course completion, repetition, transfer, examination participation, certification and unknown outcome. The study establishes whether the upper-bound gap in the proxy is negligible or material under national conditions.

The result may support a separate national completion series. It must not be used to adjust every year indefinitely if final-grade conditions change. Periodic replication is necessary.

The age study combines school age-grade records with population attainment by age. It estimates the prevalence and distribution of late entry, repeated grades and delayed completion. It identifies whether rates above 100 arise from backlog, data weakness or both.

Policy use includes timing of entry campaigns, flexible pathways, support for over-age pupils and projection of later-grade demand. The study is to avoid treating age deviation as pupil failure without examining institutional causes.

The equity diagnostic selects a limited number of disparities with material evidence and public responsibility. For each, it traces access, progression, terminal flow and attainment, then examines cost, location, language, safety, disability access or another plausible mechanism.

The output is a decision table, not only a set of group rates. It identifies the affected population, responsible authority, planned measure and review evidence. Statistical uncertainty and confidentiality remain visible.

The learning profile reports end-of-primary evidence alongside the share of the population represented. It examines curriculum opportunity, instructional time, teacher capacity and essential materials where these conditions explain learning or progression. It does not create a combined completion-learning score.

If assessment reaches only final-grade pupils, the report identifies the excluded population. Where no representative outcome measure exists, a research plan is preferable to a national learning claim based on examination candidates alone.

The projection follows actual cohorts through the remaining grades under alternative promotion and repetition assumptions. It estimates teacher, classroom, material and recurrent finance needs and identifies where capacity could restrict completion.

Scenarios should include the current path, a feasible improvement path and a path consistent with the target where analytically possible. Assumptions, unit costs and distribution are published. The model does not convert a desired target into an assertion that resources or implementation are secured.

The quality report summarises provider response, item missingness, audit findings, imputation, population vintage, revision and reconciliation. It identifies improvements completed since the previous release and unresolved risks.

This report permits the headline completion publication to remain concise without hiding material limitations. A quality issue that affects interpretation is repeated beside the result, not confined to the separate report.

The implementation review examines whether decisions responding to completion evidence were carried out. It distinguishes adoption, financing, delivery and reach. A policy document or budget line is not evidence that the intended children received the service.

The review occurs on a timetable appropriate to the mechanism. Teacher deployment can be checked during the year; completion effects require cohort exposure. Findings can lead to redesign before the terminal indicator changes.

Periodic independent review should examine definitions, calculation, source quality, revisions and public language. Independence can be provided through national statistical institutions, audit bodies or commissioned research under domestic arrangements. The reviewer requires access to aggregate components and controlled source evidence.

Review must not duplicate routine data production or claim authority over education policy. Its purpose is to test whether the measure is reproducible, limitations are complete and uses remain within evidence.

Every release is to preserve the published reader document, component tables, metadata, reference register, population vintage and correction record. Machine-readable files should retain stable identifiers so that a later series can trace revisions.

Archiving does not freeze the method. When definitions improve, the archive permits users to understand the transition and prevents later knowledge from being projected silently into earlier publications.

A senior statistical committee should approve definitions, source changes and revisions; operational teams should validate data; policy authorities should respond to findings; and research functions should investigate unresolved questions. These responsibilities are to be written and publicly identifiable.

The programme is to publish an annual forward schedule and a list of material methodological work. Consultation can improve definitions, but final statistical decisions should follow professional standards and documented authority.[REF-12]

The minimum annual products are a public completion profile, component data table, sex and geographic distribution, methodological note, revision log and policy response. Where household or learning evidence is not updated annually, the latest valid result and its year are shown without implying current measurement.

The cycle is complete only when evidence has been produced, interpreted within its limits, connected to an accountable decision and reviewed. Publication without action leaves the public problem unresolved; action without valid measurement can misdirect resources; evaluation without preservation prevents learning.

The national programme should therefore maintain a small number of linked products rather than an expanding collection of disconnected indicators. Their common reference is the child’s route from population access to a completed and credible primary education.

Some systems can produce complete annual school returns but have no recent census or household survey capable of testing attainment. The proxy may be suitable for current flow and internal planning, provided its population denominator is credible. It cannot establish who never entered registered provision or whether administrative completion corresponds with population attainment.

The release is to use the administrative series, classify the absent population perspective as a material evidence gap and schedule an attainment module in the next suitable population collection. It must not reduce the value of strong administrative data, nor elevate those data beyond their frame.

A representative survey may provide a sound estimate of young-person attainment while school returns cannot distinguish repeaters or cover all providers. In this condition, the survey supports population and equity analysis; the administrative proxy must be withheld or classified as diagnostic.

Household evidence is not a substitute for annual operational data. Authorities still need grade flow and provider coverage to locate current loss and plan resources. The two improvement tracks should proceed together: publish valid attainment and rebuild the administrative components.

A national proxy may be stable and near the target while sex or geographic components are incomplete. The headline can describe aggregate terminal flow, but it cannot support a universal or equitable achievement claim. Missing distribution is more consequential near the target because small excluded populations can disappear within the average.

The release is to show unknown sex, non-reporting areas and the coverage of subnational denominators. Data work should focus on the missing components rather than creating additional national indicators.

Administrative and household estimates can both be well produced yet differ because they measure current flow and past attainment. The proper response is cohort alignment. Analysts identify which school years correspond approximately to the surveyed age group, account for migration and structural reform, and compare an expected range rather than identical dates.

If the difference remains, both values are published with an unresolved status and a study of final-year outcomes or historical coding. Credibility is not strengthened by suppressing one source.

In a crisis or severely under-administered setting, no defensible completion percentage may be available while large educational need is evident. Action need not wait for a precise rate. School mapping, population scenarios, community evidence and verified counts can support emergency provision.

The public statement should separate the needs assessment from formal completion monitoring and specify how records will be restored. An invented precision can misallocate scarce resources and later obstruct evaluation.

A system may have a proxy and attainment evidence consistent with near-universal completion but still show poor learning, delayed completion or a small persistently excluded population. Statistical achievement of a broad target does not end the public duty to address those conditions.

The release is to recognise the attained dimension while naming the remaining one. This avoids two errors: denying genuine progress because quality remains incomplete, and declaring the education obligation fulfilled because one indicator is favourable.

Across all six conditions, the governing judgement is fitness for the stated claim. Strong evidence is used fully but not extended to an unmeasured construct. Weak evidence is described without concealing need. Divergence prompts reconciliation. Achievement is acknowledged without allowing averages to close unresolved equity or quality questions.

This discipline makes comparison more credible and policy more specific. It also keeps the child, rather than the rate, as the object of the international commitment.

Counts remain indispensable when a rate cannot carry the full claim. The number of first-time final-grade entrants indicates the scale of current terminal demand. The number of children absent from a provider frame indicates the potential scope of omission. Counts of repeaters and unknown outcomes identify operational burdens even where the population denominator is disputed.

Counts do not solve comparability by themselves; population size and programme structure still matter. Their value is that they expose the components from which a rate is built and permit policy planning while demographic work proceeds. Every percentage in the completion profile should therefore be traceable to published or controlled counts.

Interviews, school observation and documented case studies can explain how cost, language, safety, teaching conditions or administrative practice affect progression. They are especially useful where aggregate categories conceal different mechanisms. Such evidence must be sampled, recorded and analysed under a stated method.

Narrative evidence cannot establish national prevalence from selected examples. Its proper role is to interpret patterns, identify omitted mechanisms and test whether an intervention operated as intended. Public reports are to distinguish representative qualitative research from illustrative cases.

A stronger claim becomes permissible when additional evidence closes the relevant gap. Observed cohort records can replace an inferred survival claim; end-year linkage can move from terminal entry to confirmed completion; household attainment can strengthen a population statement; valid assessment can support a learning judgement.

The progression is to be explicit. Authorities are not to retain cautious language after evidence becomes strong merely to avoid accountability, and must not adopt strong language before the required evidence exists. The claim-to-evidence register provides the transition rule.

The same principle governs comparison across publication years. If a later release contains stronger source coverage, a revised denominator or observed final-year outcomes, it should identify the improvement and, where possible, restate earlier years on the new basis. Where restatement is impossible, the series break remains. Apparent continuity is not more valuable than an accurate account of changed evidence. This discipline allows public confidence to grow with the statistical system rather than depending on the preservation of an outdated claim.

Release authorities should also retain the explanation given at the time of each value. A later revision may correct the number, but it must not erase the institutional conditions, known limitations or policy decisions recorded in the original period. Historical integrity requires both the corrected series and the contemporaneous decision record.

That record is part of accountable statistical stewardship.

This glossary controls the meanings used in the framework. National systems may use different established labels, but a published crosswalk should preserve the underlying distinction. Definitions describe statistical constructs and do not alter domestic legal rules.

“Completion rate” must be accompanied by “gross proxy” where the numerator is final-grade entrants. “National” requires a territorial and provider coverage statement. “Universal” requires population and distributional evidence. “On track” requires the trajectory method. “Improved” requires a comparable series and must not describe a denominator revision.

PCRP denotes the primary completion proxy only within this report and associated technical records. NER and GER may denote net and gross enrolment ratios where defined in the publication. Abbreviations must not appear before definition in a public release.

**Purpose:** estimate current terminal-grade entry relative to the population expected to enter that grade at the official age.

**Numerator:** pupils enrolled in the final grade at the reference date minus pupils repeating that grade, or directly recorded first-time entrants where the record is demonstrably stronger.

**Denominator:** official resident population of the age expected to enter the final grade after progression without repetition.

**Unit:** percentage. The result may exceed 100 and is not truncated.

**Disaggregation:** sex; territory or provider sector only where numerator and denominator concepts align. Additional categories should ordinarily be analysed through household or linked evidence.

**Interpretation:** gross terminal-flow proxy and upper-bound estimate of actual completion where final-year dropout is not removed.

**Prohibited use:** direct statement of the percentage of an age cohort that completed; learning claim; high-stakes institutional ranking without stronger local evidence.

**Purpose:** assess timely access to grade one.

**Numerator:** new entrants to grade one who are of the official entrance age under the common age-reference rule.

**Denominator:** population of official entrance age.

**Limitations:** age documentation and population estimate; excludes late entrants who nevertheless obtain access. Publish with the gross intake rate and age distribution.

**Purpose:** measure total new entry flow relative to the official-age population.

**Numerator:** all new entrants to grade one regardless of age.

**Denominator:** official entrance-age population.

**Interpretation:** capacity and backlog entry. Values above 100 are possible and must be explained.

**Purpose:** identify late entry, repetition, acceleration and age distortion through the cycle.

**Measure:** distribution of pupils in each grade by single year of age, with missing age separately shown.

**Limitations:** date-of-birth evidence, census date and structural variation. A summary “over-age” threshold must not replace the full distribution in technical analysis.

**Purpose:** identify pupils repeating a grade and the associated flow and resource burden.

**Numerator:** pupils in the grade who were enrolled in the same grade in the preceding school year under the controlled rule.

**Denominator:** enrolment in the grade at the reference date.

**Limitations:** re-entry, transfer, promotion policy and reporting incentive. It does not measure learning difficulty directly.

**Purpose:** estimate holding power from grade one to the final primary grade.

**Method:** observed cohort where valid longitudinal records exist; otherwise reconstructed cohort with formula and assumptions disclosed.

**Limitations:** migration, transfer, re-entry, grade skipping and stability assumptions. Destination grade must be stated.

**Purpose:** estimate eventual completion in the resident population after allowing for delayed progression.

**Numerator:** persons in the selected age band coded as having completed the nationally applicable primary programme.

**Denominator:** all persons in that age band represented by the survey or census.

**Limitations:** age selection, historical programme mapping, recall, proxy response, frame coverage and sampling error.

**Purpose:** quantify movement from first entry to the final grade through the national end-of-year completion event.

**Numerator:** first-time final-grade entrants confirmed as completing the course under the stated rule.

**Denominator:** first-time final-grade entrants in the corresponding cohort.

**Limitations:** linkage, transfer, emigration, death and unknown outcome. It must not be estimated by subtracting graduates from unrelated opening enrolment.

**Purpose:** describe access to and outcome from a terminal assessment.

**Measures:** candidates as a share of eligible pupils; participants as a share of candidates; successful candidates as a share of participants; and, where useful, successful candidates as a share of eligible pupils.

**Limitations:** eligibility, selection, absence, repeated candidacy and relationship to completion. The denominator must appear in the label.

**Purpose:** report curriculum-linked learning separately from formal completion.

**Requirements:** defined construct, target population, participation coverage, administration, scoring, language treatment and uncertainty. Results among final-grade pupils must not be generalised to all children without an exclusion method.

Every institution is to have a stable identifier, provider type, location, operational status and programme map. The annual expected frame is frozen for response calculation, with later discoveries recorded as frame amendments. Closed and inaccessible institutions remain visible with status.

Totals must equal valid component sums. Repeaters cannot exceed enrolment. First-time final-grade entrants derived by subtraction cannot be negative. Age-specific counts plus missing age must equal the relevant total. Sex components plus unknown sex must reconcile. Programme and grade codes must exist in the classification register.

Final-grade enrolment is compared with the preceding year's penultimate-grade enrolment, allowing for repetition, transfer and structural change. Large deviation is flagged, not automatically rejected. Repeater counts are compared with prior-year final-grade enrolment. Provider opening, closure and expansion are considered.

The official age and denominator source are checked against programme duration. Sex-specific denominators must sum to the stated total subject to rounding and unknown classification. The population vintage and any census revision are recorded. Subnational denominators are checked for boundary alignment.

Values above 100, changes beyond the national review threshold, reversal of a long trend and large sex or territorial divergence are referred for substantive review. The reviewer examines components before the rate. A valid unusual result is retained with explanation.

The reconciliation record names both measures, their values, vintages and apparent difference. It does not begin with a conclusion about which source is correct.

Analysts determine whether each value represents final-grade entry, graduation, survival or population attainment. Measures of different constructs are not expected to agree numerically. A valid relationship or cohort lag is specified before comparison.

The administrative provider frame is compared with the survey or census population. Residence, school location, migration, institutional populations and excluded areas are documented.

National grades and credentials are mapped for the relevant historical cohorts. Alternative and non-government programmes are reviewed. Structural reform is treated explicitly.

School year, survey field date, age band and expected completion year are recorded. The comparison recognises that current terminal flow and older-youth attainment refer to different cohorts.

Administrative duplication, repeater quality, population projection, survey sampling error, recall and coding are assessed. Sensitivity is calculated for material uncertain components.

The difference is classified as expected conceptual difference, explained data difference, correctable error, unresolved material discrepancy or immaterial residual. Evidence supporting the classification is cited.

Material discrepancies appear in the release. Correctable errors are revised with a record. Unresolved differences receive a research or data-improvement action, responsible body and review date.

Each subgroup rate requires a numerator and denominator defined on the same basis. Distribution of the numerator is not a population rate. Survey subgroup estimates require weights and design-based uncertainty.

Sex and geography form the minimum. Other dimensions are selected from national evidence of exclusion and the ability to collect information safely and validly. The choice and omission are documented.

For each selected dimension, analysts compare entry, participation, progression and terminal attainment. This locates disparity rather than assuming it arises at the final grade.

Attainment is examined across ages where late entry or repetition differs by group. Eventual completion does not erase the burden of delay; both are reported.

Unknown status is shown. Small or sensitive groups are protected through aggregation, suppression or controlled access. Suppression is not reported as zero.

Differences are presented with uncertainty and context. A statistical association does not establish cause. Policy response is based on additional evidence concerning barriers and institutional responsibility.

Consider a hypothetical six-year primary programme with entry at age six. The official age for entering the final grade is eleven. At the census date, registered final-grade enrolment is 92,400. Of these pupils, 7,800 are valid repeaters. The proxy numerator is therefore 84,600. The official population aged eleven is 100,000. PCRP is 84.6 per cent.

This result means that first-time entrants to the final grade were equivalent to 84.6 per cent of the official-age population. It does not mean that 84.6 per cent of eleven-year-olds entered the final grade, because the numerator contains all ages. It does not show how many entrants finish the year.

If audit finds that 1,500 pupils classified as new entrants were repeaters, the corrected numerator is 83,100 and the rate is 83.1 per cent. The 1.5-point revision is statistical. It may also reveal a policy issue if repetition was rising, but that requires separate evidence.

If a census revision increases the age-eleven population to 104,000, the original numerator produces 81.3 per cent. If it lowers the estimate to 96,000, the rate is 88.1 per cent. The range demonstrates why the population vintage belongs beside the result.

Suppose linked end-year records confirm 79,524 of the 84,600 first-time entrants as completing, 2,100 as transferring with unknown completion status, 1,476 as leaving and 1,500 as unresolved. The confirmed completion ratio among original entrants is 94.0 per cent. If all unknown and transferred pupils completed, the upper sensitivity value is 98.2 per cent; if none did, it remains 94.0 per cent. The population completion proxy based on confirmed completers would be 79.5 per cent, but it must not replace the established international proxy without a defined national series.

If the numerator contains 45,200 boys and 39,400 girls and the age-eleven population contains 50,500 boys and 49,500 girls, the rates are 89.5 and 79.6 per cent. Girls constitute 46.6 per cent of the numerator, but the relevant disparity is the 9.9 percentage-point difference in population-related rates. Both components must be published.

Note: All values are hypothetical and illustrate the framework; they are not observations for any country.

The first table should contain no more than the measures necessary to interpret terminal flow: proxy numerator and denominator, proxy rate, programme duration, provider coverage, sex-specific rates, comparison-year value and revision status. A limitations note appears directly below it.

The second table should move from entry to the final grade. It contains new grade-one entrants by age, primary-age participation, repetition by grade, survival to the final grade, first-time final-grade entrants and confirmed end-year completers where available. Measures derived by different methods are labelled.

The third table should report household or census attainment for declared age bands, with standard errors for survey estimates, field date, frame coverage and historical programme mapping. Administrative proxy and household attainment must not occupy a single unlabeled trend column.

Column headings should contain units. Counts must not include decorative symbols. Missing, not applicable, suppressed and estimated values require distinct codes explained in a note. Wide tables should preserve row labels and support horizontal reading without merging cells. Source and limitation notes should remain associated with the table when displayed or printed.

The record identifies the indicator, version, owner, producer, publication status and language. The public title and technical name are both retained where they differ.

The record states the concept, purpose, numerator, denominator, formula, unit, programme scope and authorised uses. It explicitly distinguishes the proxy from observed completion.

The ISCED level, national programme, entry age, duration, final grade, completion rule and provider variants are recorded for every reference period. A classification version links all values produced under the same structure.

The school census date, expected provider frame, response, territorial exclusions, imputation and population source are recorded. Household and examination sources have their own field and coverage metadata.

Validation rules, exception counts, sensitivity, provisional status and revision history are retained. Each published value can be reconstructed from its component vintage.

Available disaggregation, unknown status, suppression rules and disclosure review are stated. The metadata is not to expose protected microdata.

The schedule begins with first-time final-grade entrants and repeaters at the reference date. Transfers arriving after the date are recorded separately. Pupils must not appear simultaneously as first-time entrants and repeaters.

The schedule records confirmed transfers, deaths where lawfully available, emigration or movement outside scope, temporary withdrawal, permanent exit and unknown status. Categories are mutually exclusive for the reporting period.

Course completion, examination eligibility, candidacy, participation, success and certification are separate fields. A system uses only the events that apply, but does not merge them into one count.

Opening pupils plus valid transfers in should equal confirmed course completers, repeaters remaining, transfers out, other known exits and unknown outcomes, subject to the declared timing rule. A residual becomes a quality exception, not an unnamed adjustment.

The baseline uses the earliest comparable value within the approved period. A value reconstructed under a later definition is labelled revised. Missing intervening years do not become observed through interpolation.

The target is 100 per cent completion by 2015 under the international commitment, but the statistical proxy can exceed 100. Trajectory assessment should use a bounded target classification while retaining the unbounded gross source values.

The method may calculate annual proportional improvement or annual closure of the remaining gap. It should state the formula. A linear percentage-point path and a proportional path are not interchangeable, especially from a low baseline.

Observed change should use comparable vintages and enough observations to reduce the influence of one anomalous year. Census revision and structural break are treated before the slope is estimated.

On-track status means the measured trend meets the published rule under current assumptions. Off-track status identifies a gap. Insufficient-data status is mandatory where the series cannot support the calculation. None is a guarantee of the 2015 outcome.

Policy introduced in year *t* affects completion after the remaining duration of schooling for exposed pupils. A trajectory note should distinguish current flow from projected effects of recent entry reform.

The register prevents a finding from disappearing between statistical release and policy planning. It also prevents the report from assigning action to a body without authority or resources.

Each entry contains the finding, evidence class, affected population, plausible mechanism, decision, responsible authority, resource implication, implementation date, monitoring measure and review date. Causal status is stated.

An item closes only when the decision has been made and implementation or a reasoned decision not to proceed is documented. Improvement in the headline rate alone does not close a specific unresolved data or equity issue.

The school census should cover every institution providing a programme mapped wholly or partly to primary education. The reference date is common across providers unless an approved calendar exception is recorded. Institutions report pupils meeting the controlled enrolment definition on that date, not cumulative registrations during the year.

The expected provider register is established before collection. It contains continuing, newly opened, temporarily closed, permanently closed, inaccessible and status-unknown institutions. Response rates use this register as the denominator. Removing an institution because it did not respond is prohibited.

Each return contains the stable institution identifier, provider authority, location, programme, grades offered, operational status and contact responsible for certification of the data. A name change does not create a new identifier. Merger, division and relocation are recorded as events so that trends can be reconstructed.

The final-grade module contains total enrolment; first-time entrants where directly recorded; repeaters; transfer entrants; age by single year; sex; and unknown values. If the system collects end-year data, it adds pupils completing the course, pupils repeating, confirmed transfers, other known exits and unknown outcomes.

The form is not to ask schools to calculate the national completion rate. Schools supply controlled counts; the responsible statistical function applies the authorised population denominator. This division reduces inconsistent calculation and protects the common method.

The instruction asks whether the pupil was enrolled in the same grade during the preceding school year. It gives examples for continuous repetition, return after interruption, transfer from another institution and repeat examination without grade re-enrolment. Where the national rule treats one of these cases separately, the separate field is retained.

Schools are not to infer repetition solely from age, test performance or teacher judgement. If previous status cannot be established, it is recorded as unknown and not automatically counted as a first-time entrant.

Age is calculated at the common age-reference date from date of birth where available. If only declared age is known, the record is marked accordingly. Unknown age remains a separate category. Local reporters must not alter age to fit the official grade range.

Age validation tests for impossible dates, concentrations on a default date and abrupt shifts from the prior year. These tests identify records for review; they do not authorise deletion of over-age pupils.

A pupil enrolled in more than one institution on the census date is counted once under the national rule. The sending and receiving institutions report transfer status consistently. Where no pupil identifier exists, local offices reconcile documented transfers and record the residual duplication risk.

Transfers after the census date belong to the end-year flow account, not the opening proxy numerator, unless the national method defines and documents another reference basis.

The institution head or authorised officer certifies that the return follows the definitions and is complete to the best of the institution's records. Certification does not remove the duty of central validation. It identifies responsibility and creates a route for correction.

Local offices document amendments separately. They must not replace a school value without recording the original, reason, evidence, officer and date. The institution is informed of material correction and given an opportunity to provide source evidence.

Follow-up is prioritised by the likely effect on coverage and equity, not only by institution size. Small remote schools can represent populations otherwise absent from the release. The authority records contact attempts and reasons for non-response.

Imputation is a last stage for a provisional aggregate. The method may use prior-year verified data adjusted for known change, matched institutions or another tested approach. Imputed values retain a flag and are replaced by valid returns when received. The release shows the share of the proxy numerator that is imputed.

Training should use actual classification problems, not only definitions read aloud. Reporters practise identifying a repeater, re-entrant, transfer and final-grade completer under national rules. Local officers practise reconciliation and exception documentation.

Support channels must be available during collection, with decisions on ambiguous cases added to a controlled clarification register. A clarification that changes the definition materially is issued to all reporters and may require resubmission.

After each cycle, the authority records completion time, fields with high missingness, repeated clarification and fields unused in analysis. Unnecessary fields are removed. A field is not retained merely because it was collected previously; continuity must be balanced against data quality and purpose.

The module estimates primary attainment and current participation in the resident population and supports distributional analysis. It may be included in a population census or a representative household survey. The design must reflect the host instrument's frame, respondent rules, language and field calendar.

The minimum sequence establishes whether the person ever attended school; whether the person is currently attending; the highest level and grade attended; whether that grade was completed; and the highest level or grade completed. Where credentials determine national completion, the applicable certificate can be recorded without replacing grade history.

Questions should use national programme names familiar to respondents and must be mapped during processing. A single question asking “Are you educated?” or “Did you go to primary school?” cannot establish completion.

A responsible adult may answer for children, but proxy response becomes less reliable for older household members or complex programme histories. Where feasible, young people answer their own attainment questions. The protocol records respondent relationship and handles “do not know” separately.

Field staff must not infer completion from current age or literacy. A person who attended the final grade but did not complete it remains distinguished from a completer. Historical programme names and reformed structures are included in interviewer guidance.

Analysis should begin with single-year ages around the expected completion point and then use pre-specified bands with adequate precision. A principal youth band may start several years after the official completion age to allow for delayed entry and repetition. The exact choice is justified for the national structure.

Older groups provide evidence of historical attainment, not current system performance. Combining a wide age range can conceal recent change. Tables should show field year and the approximate schooling period represented.

Survey estimates use final person weights reflecting selection probability, non-response adjustment and calibration under the survey design. Variance estimation accounts for stratification and clustering. Point estimates without design-based uncertainty are insufficient for subgroup comparison.

Frame exclusions are documented. If institutional, nomadic, displaced or insecure populations are outside scope, the national label is qualified. A high response rate within the frame does not remove frame exclusion.

Responses are coded first to national programme and grade and then mapped to ISCED 1997. Coding preserves “level attended but not completed”. Alternative programmes require an approved equivalence rule based on their authorised content or credential, not on a name resembling primary school.

Structural reform requires a historical crosswalk. The same grade number can have different level meaning before and after reform. Age, year attended and programme title may be needed to code the response correctly.

Sex and residence are core analytical variables. Household economic classification, language, disability or other characteristics are included only where the host survey measures them validly and safely. The education module must not create ad hoc sensitive classifications without appropriate design and authority.

Wealth or consumption groups are defined by the survey methodology. Their relative nature and comparability limits are stated. Household circumstance at survey date may differ from circumstance when schooling occurred, especially for older youth.

Checks identify impossible grade-level combinations, completion reported below the first grade, current attendance below a completed higher level, age inconsistency and excessive unknown response. Editing rules should correct only logically demonstrable errors. Probabilistic reassignment must be documented.

Interview re-contact or supervisor review can assess a sample of records. Translation and interviewer effects must be examined where response patterns differ sharply without substantive explanation.

The release states the questionnaire, field dates, frame, sample, weights, standard errors, age bands and coding. Estimates are not reported where precision or confidentiality fails the approved threshold. Suppressed values remain identified as unavailable rather than zero.

The denominator record identifies the national statistical authority or other authorised source, release date, base census, projection version and revision status. Education publications must not copy an undated population figure into successive releases.

The official final-grade entrance age is derived explicitly from entry age, programme duration and progression rule. If a system defines age at the start of the school year while population estimates refer to mid-year, the alignment convention is stated.

The denominator's geographic boundaries and residence concept are documented. Subnational school-location numerators are not divided by residence denominators without a qualification or adjustment.

The record summarises the fertility, mortality and migration assumptions material to the relevant cohorts. It need not reproduce the full demographic model, but it should direct users to the official methodology and identify unusual uncertainty.

When a new census changes the age population, the record shows old and new denominators, resulting completion rates and years revised. The education numerator remains separately identified so readers can see whether school flow changed.

The annual design calculates the proxy from validated school and population components and presents the indicator family. It is suited to tracking system flow and distribution. It does not by itself identify the cause of change. The analytical file retains institution-level components under controlled access and publishes aggregate counts.

The design is to pre-specify comparison years, revision handling and subgroup outputs. Unexpected results are investigated before release, but valid adverse findings remain. A narrative explains the principal movement through components rather than attributing it automatically to policy.

Where individual identifiers are absent, aggregate enrolment and repeater data can reconstruct a notional cohort. The design estimates promotion, repetition and dropout under the assumptions that rates remain stable during the simulated passage, leavers do not return and flows apply similarly within grades.[REF-04]

The resulting survival estimate is useful for locating loss but must not be labelled an observed cohort. Sensitivity can vary promotion and re-entry assumptions. Structural change during the period may make reconstruction inappropriate.

A prospective design enrols a defined sample or population of grade-one entrants and follows them through the primary cycle. It records transfer, repetition, interruption, re-entry and completion. The design offers direct evidence but requires durable identity and long-term follow-up.

Loss to follow-up must be reported. A cohort restricted to pupils present on a single school census excludes children who never enter. Its completion probability is conditional on entry and must be combined with population access evidence.

A census or survey asks young people about completed education and can estimate population attainment. It represents those in the household frame at the field date, including many no longer in school. It is well suited to distributional analysis.

The design is less precise about the timing and institutional route of completion. Recall and historical programme coding matter. Linking respondents to childhood circumstances can be difficult because current household status may have changed.

A final-grade panel follows first-time entrants from the school census through the end of the year and, where relevant, examination and certification. It estimates the difference between the international proxy numerator and confirmed course completion.

The panel is to include varied provider sectors and settings, validate baseline repeater status and follow transfers. It can produce a national adjustment study but must not alter the international series until the method, coverage and continuity are approved.

A tracing study follows a sample of pupils recorded as leaving or of unknown status and determines whether they transferred, migrated, worked, re-entered or remained outside education. It provides mechanism evidence but raises privacy and safeguarding requirements.

Contact procedures must be proportionate, voluntary where applicable and designed not to expose households. Unknown outcomes remain part of the result. The study is not to treat inability to locate a pupil as confirmed dropout.

A policy evaluation defines exposure and comparison before examining completion. A phased programme may permit comparison of early and later implementing areas; a rule based on eligibility may support another design. Random assignment may be possible for a service component but is not a general requirement and must be ethically justified.

Completion timing should match cohort exposure. An intervention introduced in grade five cannot explain earlier progression, and a grade-one policy must not be judged by national completion in the following year. Intermediate measures establish whether the intended mechanism operated.

An audit samples institutions and compares submitted counts with registers and pupil histories. It estimates error in enrolment, repeater status, age and provider coverage. The sample is to represent known risk, but a purely risk-based sample cannot produce an unbiased national error estimate; a probability component is needed for that purpose.

Findings should distinguish misunderstanding, weak source records, processing error and deliberate misreporting. Corrective action should improve definitions and systems, not only amend the audited year's total.

The report is confined to evidence available by 13 February 2005. It does not use later revisions to international indicator definitions, later census results or subsequent progress toward the 2015 target. A later reader may know that methods changed; those changes cannot be attributed to the evidence position examined here.

World Development Indicators 2004 describes the World Bank and UNESCO Institute for Statistics as working jointly on development of the primary completion indicator. This report therefore treats the proxy as an important emerging comparative measure, not as a final solution to every completion question.[REF-04]

The report does not reconstruct country rates from raw microdata. Regional and country examples are drawn from official publications and retain their reported periods and limitations. The framework's hypothetical calculations are explicitly non-observational.

International tables contain missing countries, varied reference years and provisional values. Regional aggregates can have incomplete population coverage. The report uses these data to demonstrate measurement and policy questions, not to produce a new ranking.

ISCED 1997 supports classification but does not make programme duration, content or graduation rules identical. The framework therefore limits comparative claims and requires national metadata. It does not propose a universal primary-course length.

No internationally complete end-of-primary learning dataset covers all children and systems represented in completion statistics. The report consequently recommends separate learning profiles and rejects an unsupported quality-adjusted composite. It does not define a common global learning threshold.

The report specifies a household approach but does not privilege one survey instrument. National questionnaires, frames and programme histories differ. Each application requires testing, sampling documentation and historical coding.

The policy discussion identifies plausible mechanisms connecting cost, supply, quality and completion. It does not estimate causal effects. Country action should use evaluation designs proportionate to the intervention and available evidence.

The rights instruments establish obligations and principles. Statistical indicators provide evidence relevant to their implementation but do not determine legal compliance in isolation. This report does not offer a legal judgement on any country.

Children absent from school and omitted from household frames can remain statistically invisible. The proposed coverage rules reduce this risk but cannot guarantee complete observation. Public reporting must preserve the limitation and pursue safer, stronger evidence.

Some systems cannot immediately collect pupil identifiers, end-year outcomes or detailed age records. The framework distinguishes minimum annual components from progressive improvements. Lack of advanced systems does not justify misleading use of the data that are available.

The framework is designed for the 2005 evidence position. If a country later changes structure or method, its series needs a new classification and bridge. The present specifications must not be preserved merely for continuity where better evidence becomes available; revision must instead be transparent.

The headline names the measure and year without claiming observed individual completion from the proxy. It may state direction only where the comparison is valid. Examples include “First-time entry to the final primary grade increased under the revised population series” or “Primary completion proxy remains below the path required for the 2015 commitment”.

The opening paragraph states numerator, denominator, rate, programme duration, coverage and revision status. The next sentence gives the sex distribution and principal geographic finding. The third states the main limitation.

The release explains whether change arose through more final-grade entrants, fewer repeaters, a revised population estimate, improved provider coverage or a combination. It identifies relevant entry and progression evidence without turning correlation into cause.

Sex-specific results are mandatory. Geographic and household results are included where valid. Missing groups or suppressed small cells are acknowledged. National average language does not imply uniformity.

Learning and educational-condition evidence is presented separately. If no comparable learning measure exists, the release says so rather than inferring quality from completion.

The responsible policy authority identifies the response, affected population, resource or administrative decision and review date. Statistical producers describe evidence and corrections; they do not claim implementation success in advance.

The note contains formula, source periods, programme classification, provider response, population vintage, imputation, survey precision, revision and contact. Tables and machine-readable data use the same values and labels as the narrative.

The record states publication, reference period, evidence class, responsible statistical body and policy authority. It links to the exact data vintage considered.

The finding is written in measure-specific language. It identifies scale, distribution and confidence. It separates educational and statistical movement and lists unresolved discrepancy.

Options are directed to the diagnosed stage: entry, progression, terminal year, data system, equity or quality. Each option states authority, cost, implementation time, expected mechanism and risk.

The selected action records why it is proportionate to the evidence and how it protects the universal entitlement. A decision to gather further evidence includes interim protection where delay could harm children.

The monitoring plan identifies implementation, intermediate and completion measures. It sets a review date consistent with cohort timing and states what evidence would justify revision or discontinuation.

The decision and its evidence are published in accessible form, subject to confidentiality. Later review records whether implementation occurred and whether the original assumption remained valid.