Small Arms Survey: Intelligence Source Guide

The Small Arms Survey is the reference research body for the weapons that do most of the killing: firearms holdings, licit and illicit small arms flows, and violent death. It publishes estimates rather than counts, and using it well means working with its uncertainty rather than around it.

At a glance

Source	Small Arms Survey
Category	Conflict, Crime & Human Security › Military, Weapons & CBRN
Homepage	https://www.smallarmssurvey.org/
Format	`HTML`
Access	Open — no account required
Disciplines	Weapons Intelligence, Open Source Intelligence
Mission domains	Weapons Trafficking

Illicit small-arms flows & holdings. — as catalogued in the platform’s own source registry.

The Small Arms Survey is an independent research centre based in Geneva and associated with the Graduate Institute of International and Development Studies. It is not a feed and not a live database in the operational sense; it is a research programme that publishes reports, briefing papers, handbooks and a small number of maintained datasets on small arms and light weapons and on armed violence. The datasets that matter most for analytical work are the Global Firearms Holdings estimates, which model civilian, military and law enforcement firearm holdings country by country; the Global Violent Deaths data, which brings conflict and non-conflict violent deaths into one comparable frame; and the Small Arms Trade Transparency Barometer, which scores major exporting states on how openly they report their small arms exports. Around these sit region-specific field research programmes, weapons identification and tracing reference material used by field documentation teams, and analysis of the authorised international trade built on customs data. Almost everything the Survey publishes is an estimate with a stated method and an acknowledged range, which is a very different object from a transaction record and has to be handled accordingly.

The analytical job it does that nothing else does is to put a defensible number on things nobody counts. Nobody counts civilian firearms in a country with no registration system, nobody counts the weapons circulating among non-state armed groups, and nobody counts the small arms that cross a border without a licence. The Survey's contribution is to build estimates for those quantities from whatever combination of registration records, survey data, expert judgement, seizure data, comparative modelling and field documentation is available, and to publish the method so that the estimate can be argued with. That is what makes it the WEAPINT foundation for the entire category that major-weapons databases exclude by design. Its second distinctive contribution is the harm side: bringing conflict deaths and criminal homicide into one comparable violent-death frame, which is the only way to see that in many countries the lethal violence problem is not the war. Its third is transparency assessment, which turns the reporting behaviour of exporting states into a measurable variable and thereby makes it a subject of analysis rather than a background complaint.

Who publishes it, and why that matters

The Survey is a donor-funded research centre rather than a public agency or a commercial provider. Swiss federal support has historically been central, with additional project funding from other governments, international organisations and foundations. This produces a specific and predictable pattern: research depth follows funding, so regional programmes flourish while a donor is interested and go quiet when the money moves, and the coverage map reflects donor priorities as much as analytical need. The incentive structure is scholarly and policy-facing rather than commercial, which is good for methodological honesty – the Survey publishes ranges, discloses weak evidence and revises estimates – and less good for continuity and cadence. Flagship products have changed form over the years, and outputs that were once annual have not always remained so. Treat the organisation as a highly credible producer of periodic research rather than as an infrastructure dependency. If a specific estimate is load-bearing for your work, download and archive the publication, because the web presence has been reorganised more than once and the stable thing is the document you hold.

Provenance is the first question to ask of any dataset and the one most often skipped. Who collects it, what their incentive is, whether they publish a methodology, and whether they correct the record when they get something wrong all bear directly on how much weight a finding drawn from it can carry.

What a record actually contains

The fields you will be working with, what each one means, and whether it is something you can pivot on. Read the meanings carefully — more analysis is wrecked by misreading a field than by failing to find one, and a field that looks like an observation is often an inference.

Field	Type	What it means	Pivot value
`country`	string	The unit of estimation for the holdings and violent death datasets. Sub-national estimates exist only in specific field research outputs and are not part of the global series.	Country dashboards, national firearms legislation, and conflict and crime data for the same state.
`estimate_type`	enum	Whether a holdings figure covers civilian, military or law enforcement stocks. These are estimated by different methods with different reliability, and civilian holdings carry by far the widest uncertainty.	Force size data for military holdings; registration and licensing statistics for civilian holdings where they exist.
`estimate_value`	int	The central figure. It is the midpoint or best estimate of a modelled range and is not a count. Quoting it without its range is the characteristic misuse of this source.	None safely without the accompanying range and method note.
`estimate_range`	string	The plausible interval around the central figure. For countries without registration systems this interval can span a factor of two or more, and that width is information rather than noise.	Sensitivity analysis; the honest presentation of any derived per-capita or comparative figure.
`base_year`	int	The reference year of the estimate. Holdings estimates are built for a specific year and do not update continuously, so an estimate is frequently older than the analysis quoting it by several years.	Subsequent import, production and seizure data that would move the stock since the base year.
`method_note`	string	How the figure was derived – registration data, household survey, expert estimate, comparative modelling, or a combination – and what was assumed. This is the field that determines whether the number can bear the weight you are putting on it.	The underlying source data, and comparable countries estimated by the same method.
`violent_death_count`	int	Deaths from violence for a country-year, distinguishing conflict deaths from intentional homicide and other categories. Comparability depends on the national reporting system, which varies from forensic registration to nothing.	Public health mortality statistics, criminal justice statistics, and conflict event datasets.
`firearm_share`	int	The proportion of violent deaths involving a firearm, where the reporting system records the instrument. Missing or unreliable in many countries, and its absence is often mistaken for a low share.	Weapon availability estimates; policy analysis of firearm regulation effects.
`transparency_score`	int	The Barometer's assessment of how completely and promptly a major exporter reports its small arms exports. It measures reporting behaviour, not trade volume, and is regularly misread as the latter.	National arms export reports; declarations to international transparency instruments.
`trade_value`	int	Authorised small arms and ammunition trade derived from customs statistics under the relevant commodity codes. Inherits every weakness of customs reporting, including transit attribution, re-export and classification error.	Customs trade databases and national export reports for the same flows.
`weapon_identification_reference`	string	Reference material for identifying types, markings, manufacturers and calibres in the field. It is documentation support for tracing work rather than a dataset, and it is what makes a recovered weapon into a traceable record.	National tracing requests, manufacturer records and international tracing mechanisms.
`region_programme`	string	Which field research programme produced a given output. Determines the depth and recency of coverage for a country and explains why two neighbouring states can have very different evidentiary support.	The programme's other outputs, which usually share fieldwork and sources.

Coverage — and what is not in it

Coverage is global in the modelled datasets and highly uneven in the field research. The holdings estimates aim at every country, with quality ranging from near-administrative accuracy in states with functioning firearms registries to broad modelled inference in states with none. Violent death data covers most of the world but rests on national reporting systems whose quality spans forensic death registration at one end and essentially no capability at the other, with the gaps filled from public health estimates and conflict datasets. The Transparency Barometer covers major exporting states only, by construction. Field research is where the depth is, and it is concentrated in regions where the Survey has run funded programmes – the Sahel and West Africa, North Africa, the Horn, parts of the Middle East, parts of Latin America and the Caribbean – producing granular work on trafficking routes, armed group holdings and craft production that has no equivalent anywhere else and no equivalent in the regions the programmes have not reached. Update rhythm is publication-driven rather than periodic. Datasets are refreshed when a research cycle completes, which can be years apart, and there is no expectation of an annual release for most products. The practical implication is that this is a structural source: it tells you about the standing situation, not about this month.

Known blind spots

Absence of evidence here is not evidence of absence. These are the conditions under which Small Arms Survey will not show you something that is nevertheless real:

Illicit holdings are the number everyone wants and the number least well supported. Weapons outside any register, in the hands of individuals, criminal groups and armed movements, are estimated by inference from seizures, surveys and comparison, and the resulting figures carry uncertainty that is often wider than the estimate itself.
Holdings estimates are pegged to a base year and do not track events. A conflict, a mass seizure, a disarmament programme or a surge of imports since the base year is not reflected, and nothing in the dataset signals that the estimate has been overtaken.
Craft-produced weapons and privately made firearms are structurally hard to estimate, unregistered by definition, and increasingly significant in several regions, so the estimates most likely understate exactly the category that is growing fastest.
Ammunition is far less well covered than weapons, despite being the binding constraint on the actual use of small arms, because it is consumable, poorly marked in many cases, and rarely counted anywhere.
Non-state armed group holdings are covered only where field research has reached them, which means the picture is detailed in a handful of theatres and absent elsewhere, and absence reflects programme funding rather than a low problem.
Violent death comparability breaks down precisely where violence is worst, because states with collapsed administration cannot register deaths, and the resulting undercount is systematic rather than random.
Trade analysis built on customs data cannot see transit, re-export and mislabelling, and the commodity code structure does not cleanly separate the categories that matter for policy, so authorised trade figures are indicative rather than accurate.
The publication cadence means the source cannot support current-awareness work. A trafficking route that opened this year will appear in the literature, if at all, after a research cycle that takes considerably longer than the route takes to establish itself.
Country coverage in the qualitative research is donor-shaped, and an analyst reading only the Survey's outputs will inherit that shape as an implicit map of where the small arms problem is.

Write the blind spot into the product. A statement that something “was not observed in Small Arms Survey” is defensible; a statement that it “did not happen” is not, and the difference is what survives cross-examination.

Access, licensing and what you may do with it

Access model: Open — no account required

Everything is free and public through the Survey's website, distributed as documents and downloadable datasets rather than through an API. There is no key, no account and no programmatic interface, so collection means retrieving publications and dataset files and holding them locally. The site has been reorganised over the years and links to older material have not always survived, which makes local archiving a practical necessity rather than good hygiene: if a specific estimate underpins your analysis, hold the PDF and the dataset file, with the retrieval date. Because release is event-driven rather than scheduled, the sensible collection design is to monitor the publications listing on a low frequency and pull new material when it appears, rather than attempting any regular sync. For field documentation teams, the weapons identification material is best held offline in the form it is published, since the environments where it is needed are frequently the environments with no connectivity.

Licence

The Survey publishes for public use with attribution and states its terms on the site; confirm the current wording for any commercial or redistribution use rather than assuming a research organisation's outputs are freely reusable in a product. Attribution should name the specific publication or dataset and its year, because the Survey's estimates are revised between research cycles and an unattributed figure is not reproducible. Where you build derived indicators on a holdings or violent death estimate, the method note is part of what you are relying on and should be reflected in your own documentation. The Survey's credibility is a public good in this field, and misattributed or decontextualised quotation of its figures – which happens constantly in public debate about firearms – damages it in ways that eventually reduce the resource available to everyone.

Rate limits and fair use

Not applicable in any meaningful sense. This is document retrieval from a research organisation's website, at a cadence measured in weeks or months. Monitor the publications listing infrequently, fetch new items once, and cache. There is no justification for any scripted activity beyond a low-frequency check for new releases, and a scraper hammering a small research institute's web infrastructure is both rude and pointless given that nothing changes between publication events.

Licensing changes, and it changes without warning. A dataset that was free for research this year may not be free for commercial or evidential use next year. Confirm the current terms before you build a dependency on it, and record the terms you relied on alongside the data — the licence in force at the time of collection is part of the provenance.

Collecting it

How Small Arms Survey is actually pulled, in the order you would set it up. Prefer the bulk or export interface over per-item lookups wherever one exists: it is kinder to the publisher, faster for you, and gives a reproducible snapshot rather than a series of point-in-time answers you cannot reconstruct later.

Method	Format	Cadence	Notes
Publication monitoring	HTML	weekly or monthly is ample	Watch the publications and news listing for new reports, briefing papers and dataset releases. Release is event-driven, so this is a trigger rather than a sync.
Dataset download	CSV	on release	The maintained datasets are distributed as files. Take the whole file, record the release identifier and base year, and parse locally rather than extracting the few rows you currently want.
Report archival	bulk	on release	Store the full document, not a citation. Web reorganisations have broken links to older material more than once, and the method notes you will need to defend a figure live in the report rather than the dataset.
Reference material for field use	bulk	as revised	Weapons identification and tracing reference material should be held offline in publishable form for documentation teams, since it is needed in exactly the places with no connectivity.
Structured extraction from reports	JSON	per report of interest	Much of the most valuable content – trafficking routes, seizure patterns, armed group holdings – exists only as narrative in reports and has to be read and coded by an analyst to become queryable.

Ingesting it into the platform

Every step below is idempotent and cursor-based: interrupt one and it resumes from where it stopped rather than duplicating rows or losing progress. Collection is recorded per source, so a feed that quietly stops publishing shows up as a stale timestamp instead of silently thinning your coverage.

Register as an event-driven source — Add the Survey in sources.php with a publication-triggered rather than scheduled cadence, and record explicitly that its outputs are estimates with ranges, so nothing downstream treats a figure as a count.
Monitor for releases — Configure collect.php to poll the publications listing at a low frequency under cron.php, treating a new item as a work trigger for an analyst rather than as data to be automatically merged.
Import datasets with their uncertainty — Bring holdings and violent death files through import.php with the range, base year and method note carried as first-class fields, so that any chart or export can present the interval alongside the point.
Resolve to country entities — Map every estimate to canonical country entities through resolve-everything.php and expose them on country.php and country-risk.php, where the weapons availability picture belongs alongside conflict and crime indicators.
Code the narrative findings — Have analysts extract trafficking routes, seizure patterns and armed group holdings from report text into structured records in cases.php, so that qualitative field research becomes searchable rather than sitting in a PDF nobody rereads.
Correlate with flows and events — Use correlate.php to place holdings estimates alongside recorded conflict events, seizure reporting and field-traced weapons documentation, which is where standing stock, active flow and observed use can be reconciled.
Attach to mission views — Surface the results in organized-crime.php and human-rights.php where weapons availability is a driver, and in country-dashboard.php as a structural indicator rather than a current-awareness signal.
Preserve provenance in output — Any figure exported through export.php or written into a product via reports.php must carry the publication, base year and range. The Summarise skill in copilot.php can describe what the records say; it does not produce estimates, and no holdings figure in the platform originates anywhere but a published source.

Registered sources and their last-collected state are listed in sources.php, and the scheduled chain that keeps them current is in automation.php.

How it is wrong, and how to tell

Every dataset is wrong in characteristic ways. Knowing which ways is the difference between using a source and being used by one, and it is the part of source evaluation most often skipped because it is the part that takes work.

This is high-quality research and low-precision data, and both halves of that sentence matter. The methodological standard is unusually good: the Survey publishes how it derived each estimate, states ranges, distinguishes strong from weak evidence, and revises when better information arrives. Its field research in the regions where it has operated is among the most rigorous documentation of illicit weapons circulation that exists in the open literature, built on interviews, physical inspection and seizure records rather than on secondary reporting. What it cannot do is give you accuracy about quantities nobody measures. The civilian holdings estimates for states without registration are inferences from comparison and survey, and they are honest about being inferences; the violent death series is only as good as national death registration, which in the worst-affected countries barely exists. The judgement to make is therefore not whether the Survey is reliable – it is – but whether the specific estimate you want to use was built from evidence strong enough to support your conclusion, and that is answerable only by reading the method note. Confidence should be high for the direction and rough magnitude of comparisons between well-documented countries, moderate for global aggregates, and low for any precise claim about illicit stocks in a specific country.

Characteristic false positives

Point estimates quoted without ranges. This is the dominant failure mode across the entire public discourse that uses this source, and it converts a careful modelled interval into a false fact within one citation hop.
Base year drift, where an estimate built for a year several years past is presented as the current situation, with no adjustment for imports, conflict, seizures or disarmament in the interval.
Reading civilian holdings as an ownership rate. Holdings are weapons, not owners, and firearm ownership is highly concentrated in most societies, so dividing holdings by population produces a number that describes nobody.
Treating the Transparency Barometer as a measure of trade volume. It scores how openly a state reports, so a high score can belong to a large exporter and a low score to a small one, and the inference usually drawn from it is backwards.
Customs-derived trade figures attributed to the wrong country because of transit, re-export or the way the reporting state records the transaction, compounded by commodity codes that do not align with policy categories.
Cross-country aggregation over incompatible national definitions of what counts as a firearm, which vary on calibre, action, antiquity and deactivation, and are not harmonised by the aggregation.
Violent death comparisons that read a low recorded rate in a state with collapsed administration as low violence, when it is an absence of death registration.
Attributing a Survey estimate to a later publication that merely cites it, so that a figure appears more recent and better supported than the research behind it actually is.

None of these make the source unusable. They make it a source that requires corroboration before an assertion built on it goes into a product, which is true of every source and admitted by few.

Ageing

Ageing here is publication-driven and slow, which is both the source's weakness and, correctly understood, an argument for how to use it. Holdings estimates are effectively frozen at a base year until the next research cycle, so their staleness is measured in years and is invisible in the data itself; an estimate for a country that has since fought a war or run a large disarmament programme is not marked as superseded, it simply becomes wrong without notice. Violent death series are revised as national statistics are restated, typically with a lag of a year or more, and can move materially for countries with weak reporting. Field research findings about trafficking routes and armed group holdings age fastest of all, because those arrangements adapt to enforcement within months while the research cycle takes considerably longer. A stale record in practice looks like a firearms holdings figure quoted for a current-year assessment, sourced to a report whose fieldwork predates a major conflict, presented as a single number with no range and no base year. That combination is common and it is a reporting error every time.

What this source feeds

A source is only worth what it lets you conclude. These are the disciplines that collect through it, the mission domains it serves and the data points it yields — every one is a tag, so you can follow any thread from here into the rest of the library.

Collected by these intelligence disciplines

Serves these mission domains

Weapons Trafficking

Yields these data points

Location / Coordinates

How each sector uses Small Arms Survey

The same dataset is worked very differently depending on who you are, what authority you hold, and what you are ultimately producing. A military analyst is supporting a commander’s decision; a journalist is meeting a publication standard; an NGO caseworker is protecting a person. The records are shared — the constraints, thresholds and outputs are not.

🎖 Military and defence

For stabilisation, security sector reform and post-conflict planning this is the standing reference on what weapons are already in the environment before any force arrives. Holdings estimates inform disarmament, demobilisation and reintegration planning and the scale of stockpile management problems, and the field research documents how weapons actually circulate in specific regions, which determines whether a collection programme will work or simply move the market. The category the Survey covers is the category that arms most non-state adversaries, so it is the necessary complement to major-weapons databases that exclude small arms entirely. Treat every figure as a planning range and pair it with current field reporting, because the estimates are structural and the tactical situation is not.

🕵 National intelligence

For proliferation and trafficking work this is where the analytical baseline lives for the weapons class that dominates irregular conflict. Its value is in establishing what is normal for a region – typical holdings, established routes, characteristic sources of diversion, the balance between state stockpile leakage and cross-border flow – so that a new report can be assessed as consistent or anomalous. The field research is the richest open material on diversion mechanisms and is worth reading in full rather than mining for figures. Use it as the interpretive layer over physically documented weapons recoveries, which supply the specific evidence the Survey's estimates cannot.

👮 Law enforcement

For firearms enforcement, organised crime and border work the practical value is twofold: the identification and tracing reference material that turns a recovered weapon into a traceable record, and the trafficking research that explains where the weapons in a jurisdiction are likely to have come from. Tracing itself runs through national systems and international mechanisms rather than through the Survey, and any specific weapon should be traced through the proper channel with the manufacturer and the mandated authorities. What the Survey adds is the pattern context that turns a series of individual seizures into a route hypothesis, and the standards framework that determines whether the marking and record-keeping in a jurisdiction can support tracing at all.

🔍 Private investigation and corporate security

For corporate investigators, risk consultancies and insurers the relevant use is environmental: how armed is a given operating environment, how are weapons distributed between state and non-state actors, and what does that mean for personnel security and duty of care. The holdings estimates support that judgement at country level with appropriate humility, and the field research supports it at regional level with real detail. Investigators working supply chain or export compliance questions will find the transparency and trade analysis useful for assessing whether a counterparty's stated export history is consistent with what its home state reports.

📰 Journalism and OSINT media

This is one of the most quoted and most misquoted research bodies in the field. The global civilian firearms figure in particular circulates as a bare number stripped of its range, its base year and its method, and each republication makes the error more entrenched. Reporters should cite the specific publication and year, state the range, and be clear about the distinction between weapons and owners. The Survey's briefing papers are written to be usable by non-specialists and generally state their own limitations more plainly than the coverage that follows them.

🌍 NGO, humanitarian and human rights

For humanitarian and human rights organisations this is the evidential backbone of small arms policy work: it supports arguments about the relationship between weapon availability and civilian harm, about the effectiveness of stockpile management and collection programmes, and about the reporting obligations that states have accepted and are not meeting. The violent death data is particularly important because it makes the case that in many countries the lethal violence burden is criminal rather than conflict-related, which reframes what interventions are appropriate. Use ranges rather than point estimates in advocacy, because a corrected figure costs more credibility than a sharper one gains.

🎓 University and research

The Survey's datasets are standard in armed violence and peace research and its method notes are unusually detailed for a policy-facing organisation, which makes them usable in ways most grey literature is not. Researchers should treat holdings estimates as modelled quantities with the model's assumptions inherited into any downstream analysis, archive the exact release used, and be explicit about base years when combining with annual panel data. The field research is best read as qualitative evidence with its own sampling limitations rather than mined for numbers, and the regional concentration of that research is itself a selection problem worth addressing in any comparative design.

Playbook: working Small Arms Survey end to end

A repeatable sequence from first pull to finished product. Each phase states what you are trying to establish, not merely what to click — the objective is a defensible chain of reasoning, not a completed checklist.

Phase 1 — Decide whether you need stock, flow or harm

Holdings, trade and violent death answer three different questions and the Survey publishes all three. Stock tells you what is present, flow tells you what is moving, harm tells you what is happening to people. Analysts routinely reach for a holdings figure when their actual question is about flow, and then find the estimate cannot bear the weight because it was never built for it.

Phase 2 — Find the primary publication, not the citation

Track any figure back to the report that produced it. Secondary citation strips ranges, base years and caveats, and a surprising proportion of confidently stated small arms numbers in circulation are third-hand versions of an estimate whose author would not recognise the use being made of it.

Phase 3 — Read the method note before the number

Establish whether an estimate rests on registration data, household survey, expert judgement or comparative modelling. These are not equally strong and the difference determines what conclusions the figure can support. This step takes ten minutes and prevents most of the errors that follow.

Phase 4 — Record the base year and test its relevance

Write down when the estimate was built and list what has happened in that country since – conflict, imports, seizures, collection programmes, legal change. If any of those is significant, the estimate is a historical statement and must be presented as one.

Phase 5 — Work in ranges from the start

Carry the interval through every calculation rather than doing the arithmetic on midpoints and adding a caveat at the end. Derived figures built on midpoints acquire a spurious precision that the caveat does not remove, and reviewers will find it.

Phase 6 — Separate licit and illicit deliberately

Be explicit about which side of the line each figure sits on and how the source drew it. Authorised trade data and illicit circulation estimates come from entirely different evidence and combining them without saying so produces totals that are not measurements of anything.

Phase 7 — Bring in physical documentation

Where field-traced weapons documentation exists for your theatre, use it to test the Survey's regional picture. Physically inspected weapons with recorded markings are the strongest available evidence about actual sources of supply, and they either corroborate the modelled picture or usefully complicate it.

Phase 8 — Reconcile with customs and national reporting

Compare the Survey's trade analysis against the exporting and importing states' own reports and against customs statistics. Divergence is normal and informative, and identifying whether it comes from transit, classification or non-reporting is often the analytically interesting part.

Phase 9 — Map to the policy framework

Locate your findings within the applicable instruments – the UN small arms programme of action, the international tracing instrument, the firearms protocol and the arms trade treaty – because that framework determines what a state is obliged to do and therefore what a shortfall means. Findings expressed outside the framework are harder for the people who could act on them to use.

Phase 10 — Assess the harm dimension separately

Run the violent death picture as its own line of analysis rather than as a consequence of the holdings figure. The relationship between availability and lethal violence is contested, varies by context, and is exactly the kind of claim that will be attacked if you assert it as an obvious link.

Phase 11 — Package the evidence for handover

Structure findings so that the specific claims, their sources, their base years and their uncertainty are separable, because this material frequently ends up supporting enforcement, policy or accountability processes run by someone else. Anything concerning trafficking that touches live criminal activity should go to the mandated national authority rather than into a public product.

Phase 12 — Publish with the uncertainty visible

State ranges, base years and methods in the body of the product. This source's greatest vulnerability is downstream simplification, and the analyst who preserves the uncertainty is the one whose figure survives contact with an informed critic.

The platform ships this as a step-checked workflow in playbooks.php, so progress is recorded against a case rather than held in someone’s head.

What to pair it with

No single source carries a finding. These are the datasets that corroborate, extend or contradict this one — and a source that contradicts is worth more than one that agrees, because it is the only thing that will tell you when you are wrong.

Source	Relationship	What it adds
Conflict Armament Research iTrace	corroborates	Physically documented and traced weapons from conflict zones, which is the hard evidence complement to the Survey's modelled estimates and regional narratives.
UNODC	extends	Global firearms trafficking studies and homicide statistics, providing a criminal justice frame and an independently constructed violent death series to test against.
UN Office for Disarmament Affairs	prerequisite	The institutional home of the small arms programme of action and the international tracing instrument, which define the obligations against which national performance is measured.
Arms Trade Treaty	extends	Treaty obligations and state party reporting covering small arms among other categories, and the framework in which transfer risk assessments are formally made.
UN Comtrade	corroborates	The customs statistics underlying much authorised small arms trade analysis, useful for reconstructing the flows independently and seeing where the reporting breaks down.
SIPRI Arms Transfers Database	contradicts	Covers the major conventional weapons the Survey excludes and excludes the small arms the Survey covers. Read together they describe the whole arms picture; read separately either one implies a false completeness.
INTERPOL	extends	International firearms tracing and records mechanisms available to law enforcement, which is where an individual recovered weapon is actually traced rather than through any research dataset.
GICHD	extends	Ammunition and stockpile management expertise, including the physical security and safety dimension of the stockpiles whose leakage the Survey estimates.
ACLED Conflict Events	corroborates	Geocoded violence events that provide the activity picture against which standing weapon availability estimates can be interpreted.

Legal, ethical and operational constraints

The Survey's own material is published for public use with attribution and presents no unusual licensing difficulty, though commercial redistribution should be cleared. The significant legal and ethical constraints attach to the subject matter rather than the source. Analysis of illicit weapons flows sits close to live criminal activity, and in most jurisdictions there are obligations attached to knowledge of specific trafficking offences that go beyond publishing a report; information about identifiable current trafficking should be routed to the competent national authority and, where appropriate, to the relevant international mechanism, not held in a private product. Field documentation of weapons in conflict settings raises safety obligations to sources and to the people documenting, and engages the international frameworks on marking, record-keeping and tracing that determine what can lawfully be requested from whom. Publishing regional analysis of trafficking routes carries a genuine risk of operational utility to the wrong reader, and responsible practice – which the Survey itself follows – is to describe patterns and vulnerabilities at a level that supports policy and enforcement without functioning as guidance. This library will not assist with acquisition, evasion or the sourcing of weapons in any form.

Operational security

Reading research publications is low-exposure and reading them from an attributable network is normally unremarkable. The residual considerations are specific rather than general. A concentrated pattern of downloads about one country's weapons picture is a legible signal to the site operator, which matters only if the interest itself is sensitive; if it is, retrieve broadly rather than narrowly. More importantly, staff who work with the Survey's field research on active trafficking should be aware that their interest, if it becomes visible, can put local researchers and sources at risk in ways that have nothing to do with digital tradecraft. Do not contact the Survey's regional researchers about live investigations through unprotected channels, and do not aggregate their published field findings with your own operational data in any document that could be shared outward without review.

Two rules that hold regardless of jurisdiction. Collection that is lawful is not automatically proportionate, and a dataset assembled for one purpose does not carry consent for another. Where the records concern identifiable people, the question is not only whether you may hold the data but whether holding it serves the purpose you are accountable for.

Is it earning its place?

Sources accumulate. Feeds get added during an incident and are never reviewed again, and a decade later the pipeline is carrying dead weight that nobody dares remove. These are the measures that show whether Small Arms Survey is contributing anything, and they are worth baselining now so the answer is available later.

Publication capture: whether every new release relevant to your countries of interest was detected and archived, since missing a research cycle means running on an estimate that has been superseded.
Base year currency: the average age of the holdings estimates in active use across your analytical products, which is a direct measure of how much of your weapons picture is historical.
Range preservation: the proportion of internal products that present intervals rather than point estimates, which is the practical measure of whether the uncertainty survived your pipeline.
Method attribution: the share of quoted figures traced to a primary publication rather than a secondary citation, which is what protects you from inheriting somebody else's simplification.
Corroboration rate: how often the Survey's regional picture agrees with physically documented weapons evidence for the same theatre, tracked as a running calibration of the modelled estimates.
Narrative coding yield: how many structured findings analysts extract per report, which measures whether the qualitative research is actually entering your knowledge base or being cited and forgotten.
Framework mapping: the proportion of findings expressed against the applicable international instruments, which determines whether the work is usable by the bodies that could act on it.

Beware of volume. Indicator counts rise easily and say almost nothing. Unique contribution — findings this source produced that no other source in your stack would have — is the measure that matters, and it is usually far lower than anyone expects.

Tradecraft notes

The distinctions that separate a competent analyst from a fast one:

An estimate is not a count, and a modelled interval reported as a single number is the most common way careful research becomes careless assertion. Carry the range or do not carry the figure.
Weapons are not owners. Concentration of firearm ownership is high in most societies, so per-capita holdings describe an average that corresponds to no actual person and support no conclusion about how many households are armed.
The base year is part of the number. An estimate built before a war and quoted after it is not an estimate of the present, and nothing in the dataset will tell you that.
Transparency scores measure reporting, not trade. A state can export heavily and score well, or export little and score badly, and reading the score as a volume indicator inverts the finding.
Ammunition is the binding constraint and the worst-covered variable. Analyses that treat weapon counts as capability ignore that a rifle without a supply chain is a short-lived asset.
Field research is evidence with a sampling frame, not a census. The regions the Survey has studied are the regions donors funded, and treating the resulting map as a map of the problem imports the funder's priorities into your assessment.
Craft production and privately made firearms break registration-based estimation entirely, and their significance is growing faster than the estimation methods are adapting.
Violent death undercounting is worst exactly where violence is worst, so cross-country comparisons of recorded rates systematically flatter the states with the least capacity to record.
Trace individual weapons through the mandated channels, not through research literature. The Survey tells you what patterns exist; it does not and cannot tell you where a specific serial number came from.

Questions analysts actually ask

Can I use the global civilian firearms figure in a report?

Only with its range, its base year and its source publication named. That figure is the single most decontextualised number in the field, and quoting it bare is the fastest way to signal that you did not read the underlying research.

Does the Survey tell me how many illicit weapons are in a given country?

It publishes estimates for some countries with explicit and often very wide uncertainty. The illicit component is the hardest quantity in the whole domain to estimate and the estimates should be used as orders of magnitude, not as figures to compute with.

How does this relate to SIPRI's arms transfer data?

They are complements with almost no overlap. SIPRI covers major conventional weapons and explicitly excludes small arms; the Survey covers small arms and light weapons. Using either alone and describing the result as the arms picture is incomplete in a way readers will not detect.

Can I trace a specific recovered weapon using Survey material?

No. The identification references help you describe a weapon accurately, which is a prerequisite for tracing, but tracing itself runs through manufacturers, national point of contact arrangements and international law enforcement mechanisms. Route any specific weapon through the mandated channel.

How often are the datasets updated?

On research cycles, not on a schedule. Some products have gone years between releases and the cadence has changed over time. Design collection around publication events and never assume an annual refresh.

Why do the violent death figures differ from national police statistics?

Because of definitional and source differences. Public health mortality data, criminal justice statistics and conflict datasets classify deaths differently, and the Survey's frame brings them together, which necessarily means it will not match any one of them. The method note explains the reconciliation.

Is the trade data reliable enough for compliance work?

It is indicative rather than authoritative. It is built on customs statistics whose commodity codes do not align with policy categories and which handle transit and re-export inconsistently. Use it to identify questions and use national licensing records to answer them.

Does the coverage map tell me where the small arms problem is?

No, it tells you where research has been funded. Depth of coverage correlates with donor interest, and treating a thin evidence base as a small problem is a mistake that has real consequences for resource allocation.

What does the platform add over reading the reports?

Persistence and integration: archived publications with their base years, estimates modelled with ranges intact, narrative findings coded into searchable records, and country-level weapons availability placed next to conflict and crime indicators. No estimate is generated in the platform; every figure traces to a published source.

Standards, formats and interoperability

What this source speaks natively, and what it has to be translated into before a partner can consume it. Work that arrives in a recognised format is easier to defend, easier to hand over and easier to automate against:

The UN Programme of Action on small arms and light weapons is the principal political framework, and national reporting against it is the benchmark most transparency assessments are constructed around.
The International Tracing Instrument sets the marking and record-keeping requirements that determine whether a recovered weapon can be traced at all, and is the standard against which a jurisdiction's tracing capability should be judged.
The UN Firearms Protocol supplementing the transnational organised crime convention provides the criminal law framework for illicit manufacturing and trafficking, which is the frame law enforcement users need.
The Arms Trade Treaty covers small arms among its categories and defines the transfer risk assessment obligations that state parties have accepted.
MOSAIC, the modular small arms control implementation compendium, provides the operational standards for stockpile management, marking, destruction and collection programmes.
Harmonised System commodity codes are the vocabulary of the customs data underlying trade analysis, and their imperfect alignment with weapon categories is a known and unfixed source of error.
Cause-of-death classification underlies the violent death series, and differences in coding practice between health and criminal justice systems are the main reason national figures diverge.
In the platform, estimates normalise to country-attached measures with explicit intervals and export as CSV, JSON and JSONL; they have no meaningful expression in indicator formats such as STIX or MISP.

References

Primary documentation and authoritative references for this source. Publishers revise and retire material, so treat the retrieval date as part of the citation and re-check before relying on any of it in a formal product.

Small Arms Survey — Small Arms Survey, Graduate Institute of International and Development Studies. The organisation's site and the route to all publications, datasets and regional programmes. The correct place to start and to check whether an estimate you hold has been superseded.
Global Firearms Holdings — Small Arms Survey. The civilian, military and law enforcement holdings estimates with their base years and method documentation. Read the methodology before quoting any country figure.
Global Violent Deaths — Small Arms Survey. Conflict and non-conflict violent deaths in one comparable frame, which is the dataset that reframes lethal violence as a broader problem than armed conflict.
Small Arms Trade Transparency Barometer — Small Arms Survey. Scores of major exporters on the completeness and timeliness of their small arms export reporting. A measure of reporting behaviour, not of trade volume.
UNODC — United Nations Office on Drugs and Crime. Firearms trafficking research and homicide statistics from a criminal justice perspective, and an independent series to test violent death figures against.
UN Office for Disarmament Affairs — United Nations. The small arms programme of action, the international tracing instrument and national reporting, which together define what states have committed to do.
Arms Trade Treaty — ATT Secretariat. Treaty text, reporting templates and state party submissions covering small arms transfers among other categories.
Conflict Armament Research — Conflict Armament Research. Field documentation and tracing of weapons recovered in conflict, the physical evidence counterpart to modelled holdings estimates.
GICHD — Geneva International Centre for Humanitarian Demining. Ammunition and stockpile management standards and practice, covering the physical security failures that turn state stockpiles into illicit supply.
INTERPOL — INTERPOL. International police cooperation on firearms trafficking and the mechanisms through which individual weapons are actually traced between jurisdictions.
IANSA — International Action Network on Small Arms. The civil society network working on small arms control, useful for understanding how research findings enter advocacy and policy processes.
UN Comtrade — United Nations Statistics Division. Customs trade statistics underlying authorised small arms trade analysis, with all the transit, re-export and classification limitations that implies.

Link integrity: every reference above was verified with a live request when this page was generated. Where a publisher had moved or withdrawn a document, the link was repointed at a preserved copy in the Internet Archive and marked as archived. Anything with no reachable copy anywhere had its link removed rather than left to rot — the source is still credited, it simply cannot be linked.

Put it into practice

The Quantus Intel threat intelligence platform operationalises this source: it archives each publication and dataset release with its base year and method note, carries estimate intervals through as structure rather than collapsing them to points, codes the regional field research into searchable findings, and places weapons availability beside conflict and crime indicators on the country view.. Browse the full source catalogue, or follow any tag above into the rest of the library.