Methodology
How this data is collected
JuveLaw is a comparative reference, which means almost every figure it publishes is a simplification of something more complicated. This page explains which simplifications we make, how a figure earns its place, and where the record is thin. Read it before citing anything here.
What is covered
Two layers. The first is a worldwide layer: one row per country recording the minimum age of criminal responsibility, the upper age of juvenile jurisdiction and ratification of the Convention on the Rights of the Child. The second is a depth layer: 10 jurisdictions across 6 regions carry full profiles — court structure, sentencing limits, transfer to adult court, diversion, detention alternatives and records confidentiality — together with up to 10 landmark judgments each.
The depth layer is not a ranking of importance. It reflects where the primary materials are accessible to us and verifiable in a language we read. Jurisdictions outside it are under-represented here for reasons of our own capacity, not because less is happening in them.
Where the figures come from
Sources are ranked, and a lower tier never silently overrides a higher one. Every structured field carries at least one source; where a field is supported by a source specific to it, the link is attached to that field rather than to the profile as a whole.
| Tier | What it is | How it is used |
|---|---|---|
| 1. Primary | Statutes, criminal and juvenile justice codes, law reports, constitutional texts. | Preferred for every structured field. A figure sourced here is recorded with the provision that sets it. |
| 2. Government | Ministry and judiciary publications, official statistics, state reports to treaty bodies. | Used where the primary text is unavailable in a language we can verify, or where the statute delegates the figure. |
| 3. Treaty body | UN Committee on the Rights of the Child concluding observations and general comments. | Authoritative for the international standards themselves; used for national figures only as corroboration. |
| 4. Academic | Peer-reviewed comparative law scholarship. | Used to interpret a provision, and to flag a figure that needs re-checking. Rarely the sole source for a value. |
| 5. NGO and news | Monitoring organisations, legal charities, reported journalism. | Used to detect that something has changed, never as the final authority for what it changed to. |
What “as of” means
An “as of” date is the date a human last checked that value against its source. It is not the date the law changed, and it is not the date the page was rebuilt. A row verified two years ago is shown with a two-year-old date rather than refreshed by the act of serving it.
This has an uncomfortable consequence we accept deliberately: an old date on a row is a signal that we have not re-checked it, and you should treat that row as a pointer to its source rather than as a statement of current law. We would rather publish a visibly stale date than an invisibly stale figure.
How conflicts are resolved
Where sources disagree, the higher tier wins and the disagreement is recorded in the row's notes rather than erased. Three patterns recur and each has a fixed treatment:
- A graduated minimum age. Where a jurisdiction sets one age generally and a lower one for grave offences, the recorded figure is the general minimum, and the exception is stated in the notes. Recording the lower figure would make the country look like an outlier it is not; omitting the exception would hide the thing a researcher most needs.
- A rebuttable presumption. Where capacity above the minimum age must be proved before a child can be convicted, the minimum age is recorded as the statutory figure and the presumption is noted. The two are different mechanisms and collapsing them into one number loses the distinction.
- Federal variation. Where the age is set sub-nationally, the federal or most widely applicable figure is recorded and the variation is noted. Countries where this matters most are flagged in their profile.
How changes are recorded
A tracked field is never overwritten in place. When a value changes, the previous value is retained with its own date and a dated entry is added to the reform timeline recording what changed, when, and from what to what. The history of a figure is part of the dataset, not an artefact of editing it — which is why the timeline can be downloaded alongside the current values.
Children are not named
We publish anonymised cases only. Where a court withheld a child’s identity, so do we — our data model has no field in which to store it.
In practice this means case names appear exactly as the official law report published them. Where a court used initials or a pseudonym, so do we, including where journalism has since published a name. Decided landmark authority is cited by its reported caption because that caption is the citation; it is not an exception to the rule, it is the rule applied to a case the courts themselves published by name.
Case summaries describe the procedural posture and the legal question. They do not describe offences in detail and do not carry victim detail. Where a judgment's facts cannot be summarised without identifying a child, the case is not published.
Known limitations
These are the weaknesses we know about. There will be others.
- Language bias. Our verification is strongest where primary materials are published in English. Elsewhere we lean more heavily on government translations and on secondary sources, and the gap is widest exactly where scrutiny is scarcest.
- Law as written, not law as practised. A minimum age in a statute says nothing about how often children below it are detained in practice, how long remand lasts, or whether diversion is funded. This dataset describes the formal position only.
- Narrative fields are summaries. The descriptive fields on a jurisdiction profile are written from the sources; they are not the text of the law and should not be quoted as if they were.
- Uneven currency. Rows are re-verified as capacity allows, not on a fixed cycle, so verification dates vary widely across the table. Sorting by date is the fastest way to see which parts of the record are thin.
- Absence is not zero. Where no figure has been verified, the value is recorded as absent and displayed as “Not recorded”. It is sorted last rather than treated as a low number. An absent value is a statement about our research, never about the jurisdiction.
- Landmark selection is editorial. Which judgments count as landmark is a judgement call made by us, informed by how often a case is cited and whether it changed the law. Reasonable comparative lawyers would choose a different set.
Corrections
If something here is wrong, tell us and cite the source that shows it. A correction is applied to the record, the previous value is retained in the history, the verification date is reset and — where the change is substantive — a dated entry is added to the reform timeline. We do not make substantive corrections silently.
Write to [email protected].
Licence and reuse
The datasets are published for reuse with attribution, including commercially. Every download page carries the licence and a citation string that includes a retrieval date. Because figures change, a citation without a retrieval date cannot be checked, and an unverifiable citation of a comparative dataset is worth very little.
Reference information, not legal advice. Juvenile law varies by jurisdiction and changes often.