SIPRI Military Expenditure: Intelligence Source Guide

SIPRI's military expenditure series reconstructs what states actually spend on their armed forces, on a consistent definition, across most of the world and back to the middle of the twentieth century. It is the economic denominator for every capability argument, and almost every misuse of it comes from picking the wrong one of its six presentations.

At a glance

Source	SIPRI Military Expenditure
Category	Conflict, Crime & Human Security › Military, Weapons & CBRN
Homepage	https://www.sipri.org/databases/milex
Machine interface	`https://milex.sipri.org/sipri`
Format	`HTML`
Access	Open — no account required
Disciplines	Economic Intelligence, Measurement & Signature Intel
Mission domains	Military & Defense

Global military spending database. — as catalogued in the platform’s own source registry.

The SIPRI Military Expenditure Database is an annually updated compilation of estimated military spending for the great majority of the world's states, assembled by the Stockholm International Peace Research Institute and published free of charge alongside a fact sheet on world military expenditure trends. Its defining characteristic is a single consistent definition applied to every country, derived from a NATO-style specification: spending on the armed forces including peacekeeping contributions, on defence ministries and other government agencies engaged in defence projects, on paramilitary forces where those are judged to be trained, equipped and available for military operations, and on military space activities. Within that scope it counts personnel costs including military pensions and family services, operations and maintenance, procurement, military research and development, and military construction and infrastructure, and it attributes military aid to the donor rather than the recipient. It excludes civil defence and, so far as possible, current spending on past military activities such as veterans' benefits, demobilisation and weapon destruction. The same national total is presented in several ways: local currency at current prices, current US dollars, constant US dollars at a fixed base year, share of gross domestic product, share of general government expenditure, and spending per capita. The data is distributed as a downloadable workbook with extensive country notes, alongside a web query interface.

The analytical job this dataset does that nothing else does is to make national military effort comparable when the underlying national accounts are not. Every state publishes a defence budget, and no two states mean the same thing by it: some hide procurement in other ministries, some carry paramilitary forces in the interior budget, some fund nuclear programmes through civilian agencies, some report appropriations rather than outturn, and some publish almost nothing. SIPRI's contribution is a single definition imposed on all of them, with the adjustments documented, so that a comparison between two countries or between one country and its own past is meaningful rather than an artefact of accounting convention. That is what makes it the ECONINT foundation for defence analysis. It also serves as the reconciliation target for physical observation: when overhead imagery, procurement records and construction activity indicate more capability than the recorded expenditure could plausibly buy, the discrepancy is itself the finding, and identifying it requires an expenditure baseline you trust. The long time series is the second unique property. Very few security datasets extend across seven decades on a stable definition, and structural questions about rearmament cycles, post-conflict demobilisation and the relationship between economic growth and military effort are only answerable at that length.

Who publishes it, and why that matters

The same institutional considerations apply as to SIPRI's other databases: an independent Swedish-based research institute, principally funded by a Swedish government grant with additional project money, publishing free of charge as a matter of policy rather than as a loss-leader. What distinguishes the expenditure work specifically is that it is estimation as much as compilation. For opaque states SIPRI constructs figures from partial budget documents, national accounts, published procurement, and inference about what sits outside the published defence line, and those estimates are the institute's analytical judgement rather than a reported number. Its willingness to publish an estimate with an uncertainty note, rather than a blank or a false precision, is the single most useful editorial decision in the dataset, and it is only trustworthy because SIPRI has no stake in the answer. The continuity risk is the same as elsewhere: a decades-long public good maintained by a mid-sized institute dependent on sustained public funding. The practical mitigation is the same too, which is to archive every annual edition you use rather than assuming the series will always be there in the form you remember.

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 state the figure describes, with historical entities handled by explicit rows and notes. Successions, dissolutions and unifications break naive time series and are documented in the workbook notes rather than in the data cells.	Country profiles, alliance membership, and the arms transfer record for the same state.
`year`	int	Calendar year. Where a state's fiscal year does not align, SIPRI applies a stated adjustment convention, so a figure attributed to a year may span two national budget periods.	Alignment with conflict onset dates, elections and economic shocks.
`milex_local_current`	int	Spending in national currency at current prices. This is the closest field to the underlying national source and the one least contaminated by conversion assumptions, which makes it the right starting point for any within-country analysis.	National budget documents and audit reports, where those are published.
`milex_usd_current`	int	The same figure converted at the year's market exchange rate. Useful for describing a single year's international position and dangerous for time series, because it moves with the currency rather than with the defence effort.	None safely. Prefer the constant-price series for any comparison over time.
`milex_usd_constant`	int	Converted and deflated to a fixed base year. This is the standard series for cross-time comparison, and the base year is rebased periodically, which changes every number in the column between editions.	Long-run trend analysis, provided you never mix editions.
`share_of_gdp`	int	Military spending as a proportion of gross domestic product, the so-called military burden. It measures the economic effort a state is making, not the capability it is buying, and the denominator comes from external economic sources that are themselves revised.	Economic indicators, alliance spending commitments, and fiscal capacity analysis.
`share_of_govt_spending`	int	Military spending as a share of total general government expenditure. Often more revealing than the GDP share for questions about political priority, because it compares defence against the state's other choices rather than against the whole economy.	Public expenditure data and social spending comparisons for guns-versus-butter analysis.
`milex_per_capita`	int	Spending divided by population. Interpretable within similar force models and misleading across them, since a conscript force and an all-volunteer professional force convert money into personnel at completely different rates.	Population and demographic data; force size where an order-of-battle source is available.
`estimate_flag`	enum	SIPRI marks figures that are its own estimates, figures subject to significant uncertainty, and years where no data could be produced, using notation explained in the workbook. Ignoring these markers is the fastest route to false precision.	The country notes, which explain what was estimated and on what basis.
`country_note`	string	Per-country documentation of definitional deviations, excluded or included items, data breaks, fiscal year handling and the reasoning behind estimates. For opaque states this text is more important than the number it accompanies.	Everything. Any cross-country comparison that has not been checked against both countries' notes is provisional.
`series_break`	enum	Points at which a country's series is not continuous because of definitional change, state succession or a change in what the national source reports. Trends drawn across a break are artefacts.	Historical events explaining the break; alternative sources covering the discontinuity.
`base_year`	int	The reference year of the constant-price series in the current edition. It is a property of the edition, not of the data, and it is the single most common cause of figures that will not reconcile between two analysts.	Edition metadata; your own archive of prior editions.

Coverage — and what is not in it

Coverage is close to global for recent decades and thins as you go back. A consistent series covering the great majority of states exists from the late 1980s onward, with substantial coverage extending back to around 1949 for countries whose records permit it. The update is annual, conventionally in April, accompanied by a fact sheet summarising world and regional trends. Coverage is by state and by year only: the database gives national totals, not a breakdown by personnel, procurement, research and operations, which for most countries must be obtained from national publications or from alliance reporting where that exists. Some states are absent entirely for lack of any usable information, some have gaps in particular periods, and several major spenders carry figures that are explicitly SIPRI estimates rather than reported numbers. The country notes disclose which is which, country by country and often year by year. Because the constant-price series is rebased and because national accounts and GDP figures are revised by their own producers, the whole historical series moves between editions in ways that have nothing to do with any change in military spending, and this is the property analysts most consistently fail to account for.

Known blind spots

Absence of evidence here is not evidence of absence. These are the conditions under which SIPRI Military Expenditure will not show you something that is nevertheless real:

Off-budget and extra-budgetary military spending is invisible where it has not been publicly documented, and states with the most reason to conceal spending are precisely those where the estimate carries the widest uncertainty.
The paramilitary boundary is a judgement call applied to forces that differ enormously between states, so two countries with similar internal security architectures can be treated differently depending on how their forces are constituted and reported.
There is no disaggregation by spending category in the main series, so the database cannot distinguish a state that spends on personnel from one that spends on procurement, which is the difference between a large army and a modern one.
Military nuclear programmes administered by civilian agencies, national space programmes with military components, and dual-purpose infrastructure are handled by documented rules that inevitably produce cross-country incomparability at the margin.
Arms acquired through credit, barter, resource-backed arrangements or grant aid do not necessarily appear in the recipient's spending in the year the equipment arrives, so expenditure and capability can diverge sharply.
Market exchange rate conversion systematically understates the resources of states with low domestic price levels, conscript personnel and indigenous defence industries, which is a large and non-random distortion in exactly the comparisons analysts most want to make.
In high-inflation economies the deflator becomes the dominant term, and the constant-price series for those countries is carrying more assumption than measurement.
Appropriation is not outturn. Where a national source reports a budget rather than actual spending, over-execution and under-execution are invisible, and both are common in states under fiscal stress or at war.
The dataset is annual and lags by roughly a year, so it cannot describe a rearmament under way now, only one that was under way when the last national accounts were published.

Write the blind spot into the product. A statement that something “was not observed in SIPRI Military Expenditure” 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

Access is free and requires no key. The primary artefact is a downloadable workbook containing the several presentations of the data on separate sheets, together with the country notes and the definitional documentation, and there is a web query interface for building targeted extracts. There is no public programmatic API, which does not matter much because the update cadence is annual. The correct collection design is to download the complete workbook at each release, store it permanently with its edition identifier and base year, and serve all internal queries from your own parsed copy. The notes are not an appendix; they are the part of the release that tells you which figures are estimates and where series break, and a pipeline that parses only the numeric sheets will produce a clean dataset that silently discards the uncertainty information. Parse the notes into your record model even if you have to do it semi-manually once a year.

Licence

SIPRI provides the data free for research, educational and non-commercial use with attribution, and asks that reproduction and commercial use be cleared with the institute. As with its other databases, the terms have been restated over time and should be confirmed against the current statement rather than a summary. Citation must include the edition or access date because the series is revised and rebased annually; a figure attributed simply to SIPRI is not reproducible and, in any adversarial setting, is not defensible. Internal analysis and derived indicators are uncontroversial; republishing the workbook, embedding the series in a commercial product, or presenting SIPRI's estimates as your own analysis are not. Where you build a public index on top of the series, name the edition and base year in the methodology so that others can reconstruct your figures.

Rate limits and fair use

Not a meaningful constraint, because there is nothing to poll. One workbook download per annual release is the entire collection requirement. If you use the query interface interactively, ordinary courtesy applies: serialise requests, identify yourself, and do not script the interface when a single bulk download would give you the same data. Trigger collection on publication of the annual fact sheet rather than by polling.

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 SIPRI Military Expenditure 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
Annual workbook download	bulk	once per release, conventionally April	The correct primary method. One file contains every presentation of the data plus the notes, and downloading it whole avoids the selection bias of pulling only the sheet you think you need.
Notes and methodology parse	HTML	annual, alongside the workbook	Extract the country notes, estimate markers and series breaks into structured form. This is the step most pipelines skip and the reason most downstream analyses overstate precision.
Web query extracts	CSV	ad hoc	Convenient for a single analytical question. Record the parameters and the edition, and do not use it as a substitute for holding the full workbook.
Fact sheet ingestion	HTML	annual	SIPRI's own summary of what moved and why, including flags on methodological changes and rebasing that would otherwise be discovered by an analyst wondering why the numbers changed.
Edition archival	bulk	every release	Keep every workbook indefinitely with its base year recorded. Since the constant-price series is rebased and history is revised, your archive is the only route back to a figure you previously published.

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 with edition semantics — Add the database in sources.php as an annual source and record base year and edition as first-class properties of every collection, so that two ingests are never silently merged into one incoherent series.
Schedule the release-window collection — Have collect.php attempt the workbook download during the release window with cron.php owning retries and failure alerting, since a missed annual collection is a year-wide gap that nothing else will fill.
Parse all presentations, not one — Import local-currency, current-dollar, constant-dollar, GDP share, government-spending share and per-capita series as distinct measures through import.php, so that analysts choose the right one for their question instead of using whichever one the pipeline happened to keep.
Carry uncertainty as structure — Model estimate flags, series breaks and unavailable years as explicit values rather than nulls or blanks, so that a chart drawn from the data can show a discontinuity rather than interpolating across it.
Attach the country notes — Store the per-country documentation alongside the numeric series and expose it wherever a figure is displayed, because a spending number for an opaque state without its note is a number with unstated error bars.
Resolve to country entities — Map every row to canonical country entities through resolve-everything.php, with an explicit historical mapping for states that changed name or ceased to exist, and surface the result on country.php and country-dashboard.php.
Correlate with the materiel record — Use correlate.php to place expenditure alongside arms transfer deliveries and conflict event data for the same state and period, which is where the analytically interesting discrepancies between money, materiel and activity become visible.
Version derived indicators — Any composite score built on this series – burden indices, regional aggregates, country risk inputs feeding country-risk.php – must carry the edition and base year through to export.php, or it will be irreproducible within twelve months.

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.

As an estimation product this is about as good as the field gets, and the reason to trust it is the documentation rather than any claim of accuracy. SIPRI states its definition, states where a country deviates from it, marks its own estimates as estimates, and explains the basis for each. That is a far stronger epistemic position than a national defence budget presented without commentary, even though the national budget is a reported number and SIPRI's figure may be a construction. Reliability varies enormously across the dataset and the variation is disclosed: figures for states with published, audited, comprehensive defence accounts are close to exact; figures for states with opaque budgets, large off-budget flows or active concealment are analytical judgements with wide and sometimes unstated uncertainty. The systematic weaknesses are conversion and deflation rather than compilation. Market exchange rates and national deflators do a great deal of work in the headline dollar series, and for several major spenders that work is more consequential than any error in the underlying local-currency figure. Treat existence and order of magnitude as high confidence, within-country trends in local currency as high confidence, cross-country dollar comparisons as moderate, and cross-country comparisons of capability inferred from dollars as low.

Characteristic false positives

Reading a current-dollar spike or collapse as a change in defence policy when it is a currency movement. This is the single most common error and it affects reporting on every state with a volatile exchange rate.
Using share of GDP as a proxy for military capability. It measures economic effort, so a poor country making a large effort and a rich country making a small one can look identical while fielding incomparable forces.
Comparing figures across editions after a rebasing of the constant-price series, which changes every number in the column without any underlying change, and produces confident findings about trends that do not exist.
Treating SIPRI estimates for opaque states as measured values, then building precise-looking ratios on top of them. The estimate flag exists to prevent exactly this and is routinely stripped in downstream pipelines.
Drawing a trend line across a documented series break caused by state succession or a definitional change, which manufactures a rise or fall out of an accounting event.
Comparing per-capita spending between a conscript force and a professional volunteer force, where personnel costs per soldier differ by an order of magnitude and the ratio measures employment model rather than defence effort.
Confusing budget with outturn where the national source reports appropriations. Under-execution during fiscal stress and over-execution during war are both common and both invisible in the series.
Reading absence as zero. Countries with no usable data are marked as unavailable, and a pipeline that coerces those to zero will produce regional aggregates that are wrong in a direction nobody notices.

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

The series ages on an annual clock with a systematic lag, and the most recent covered year is always the softest, because national accounts, budget outturns and GDP figures all mature over subsequent years and SIPRI revises accordingly. Expect the latest year to move in later editions, sometimes substantially for states with weak statistical capacity. In the other direction, historical figures are stable in local currency and unstable in constant dollars, because rebasing and revised deflators move them without any change in the underlying reality. The practical consequence is that a constant-dollar figure has a shelf life of about one edition. GDP shares age badly for a third reason, which is that the denominator is revised by economic authorities on their own schedule, so a burden figure can change because the economy was restated. A stale record here looks like a figure quoted from an edition two or three years old, in constant dollars from a superseded base year, for a country whose GDP series has since been revised, cited without an access date. It will not reconcile with the live database and there will be no way to explain the difference.

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

Military & Defense

Yields these data points

Location / Coordinates

How each sector uses SIPRI Military Expenditure

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 defence planning staff the series is the affordability check on every capability argument. It establishes what a state has historically been able to sustain, which constrains what any announced programme can realistically deliver, and it exposes the difference between a headline procurement commitment and the recurring cost base that would have to support it. Alliance burden-sharing debates run on the GDP share, which planners should handle carefully because it is an effort measure rather than a capability measure. Use the local-currency series for judging a single state's trajectory, the constant-dollar series for regional balance questions, and always pair the money with the materiel record, since expenditure that does not show up as delivered equipment is either sustainment, personnel or a question worth asking.

🕵 National intelligence

For economic intelligence this is the primary open baseline against which reported and observed military activity is reconciled. The analytically productive move is not to read the figures but to find the gaps: capability that appears without expenditure to explain it, expenditure that appears without observable procurement or construction, or a state whose spending pattern diverges from its declared posture. Each of those is a lead. The country notes are the most valuable part of the release for this work because they tell you where SIPRI itself could not see, and those are the areas where other collection has to carry the weight. Treat estimates for closed states as hypotheses to be tested against physical and financial observation rather than as inputs.

👮 Law enforcement

The law enforcement relevance runs through procurement corruption and illicit finance. National military spending is one of the largest discretionary flows in many state budgets, frequently exempted from ordinary procurement transparency on national security grounds, and the gap between recorded expenditure and delivered equipment is a well-established corruption signal. Investigators working defence procurement fraud, sanctions circumvention financing or state-linked money laundering can use the series to size the flow and to identify periods of anomalous growth worth examining against contract records. It is context for building a case, not evidence in one, and substantive matters belong with the mandated financial intelligence and anti-corruption authorities.

🔍 Private investigation and corporate security

For corporate intelligence and market analysis this is the standard sizing tool for defence markets and the standard input to country risk work. It supports judgements about whether a market can sustain a programme, whether announced spending is consistent with historical fiscal behaviour, and whether a counterparty's claims about a national defence budget are plausible. Due diligence teams should note that the absence of category breakdowns means the series cannot tell you whether growth is going into procurement or into payroll, which is usually the commercially decisive question and has to come from national sources.

📰 Journalism and OSINT media

This dataset generates more misleading headlines than almost any other in the security field, and the errors are avoidable. Report the constant-price series for trends and say so, avoid current-dollar comparisons between countries with moving currencies, do not treat GDP share as a measure of military strength, and read the country note before writing about any state whose figure is a SIPRI estimate. The annual fact sheet is written to be quoted and contains SIPRI's own framing of what changed, which is a safer starting point than the workbook for a reporter working to deadline.

🌍 NGO, humanitarian and human rights

For advocacy on arms control, development financing and public spending priorities the share-of-government-expenditure series is usually the more honest instrument, because it compares military spending against the state's other choices rather than against the size of the economy. It supports arguments about opportunity cost that are factual rather than rhetorical, and its documented methodology makes those arguments harder to dismiss. Organisations should be careful to use a single edition consistently across a campaign and to acknowledge estimate uncertainty for opaque states, since the credibility cost of a corrected figure exceeds the value of the sharper number.

🎓 University and research

This is a canonical dataset in political science, economics and security studies, and the methodological hazards are well documented in the literature: rebasing, exchange rate conversion, definitional deviation and the treatment of estimates. Good practice is to archive the exact edition, work in local currency where the research design permits, model the estimate flags rather than discarding them, and state explicitly how series breaks were handled. Researchers combining this with conflict or transfer data must reconcile the year conventions, since fiscal-year adjustments here and delivery-year distributions there are not the same temporal object.

Playbook: working SIPRI Military Expenditure 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 — Choose the presentation before you choose the country

Six presentations of the same underlying figure answer six different questions. Local currency for within-country trajectory, constant dollars for cross-time international comparison, current dollars for a single-year international snapshot, GDP share for economic effort, government-expenditure share for political priority, per capita for burden on the population. Decide which question you are answering and commit, because switching presentations mid-analysis is how contradictory findings get published.

Phase 2 — Download the whole edition and archive it

Take the complete workbook rather than an extract, record the base year and the release date, and store it permanently. Everything downstream depends on being able to reconstruct exactly which numbers you used, and the live database will not help you a year from now.

Phase 3 — Read the country notes for every state in your analysis

The notes disclose definitional deviations, excluded items, fiscal year handling, series breaks and the basis for estimates. For opaque states they are more informative than the figure. An analyst who has not read the note for a country in their comparison does not know what they are comparing.

Phase 4 — Map the uncertainty landscape first

Before analysis, classify your countries into those with reported, audited figures, those with reported but partial figures, and those where SIPRI is estimating. Your conclusions will have completely different confidence in each group, and mixing them into one chart with one visual weight is a misrepresentation even when every number is correct.

Phase 5 — Establish the trajectory in local currency

Work out what each state's own spending has done in its own money before touching any conversion. This isolates the defence decision from the exchange rate and the deflator, and it frequently shows that a dramatic dollar movement corresponds to no change in national behaviour at all.

Phase 6 — Convert deliberately and say how

Only then move to constant dollars, and state the base year in the product. Where purchasing power differences are material to the argument – conscript armies, indigenous industry, low domestic price levels – say explicitly that market exchange rate conversion understates real resources and do not pretend the dollar figure settles the question.

Phase 7 — Test against the materiel record

Place the expenditure series next to arms transfer deliveries and, where available, observed procurement and construction. Money without materiel points to personnel, sustainment, off-budget flows or waste; materiel without money points to grant aid, credit, concealment or an estimate that is too low. Both are findings.

Phase 8 — Look for the breaks and explain them

Identify every discontinuity in your series and attribute each to a cause: state succession, definitional change, war, currency reform, statistical revision. Unattributed breaks left in a chart will be read as events by your audience, and you will be the one who has to defend them.

Phase 9 — Set the regional and alliance context

A single state's spending is nearly meaningless without its neighbours and its alliance commitments. Build the regional aggregate and the peer comparison, using consistent presentations, and note where a regional total is dominated by one or two spenders, which is usually the case and usually unmentioned.

Phase 10 — Reconcile with independent sources

Compare against national budget documents, alliance reporting where the state is a member, and international financial institution government finance statistics. Divergence is normal and definitional; understanding the specific reason for each divergence is what separates an assessment from a citation.

Phase 11 — Express the finding with its uncertainty

State ranges rather than point estimates for countries carrying estimate flags, and avoid decimal precision that the underlying data cannot support. A conclusion that survives being stated as a range is a conclusion; one that requires precision to hold is an artefact.

Phase 12 — Stamp the edition into the product

Name the edition, base year and access date wherever a figure appears. This is not bureaucratic hygiene. Rebasing and revision guarantee that a reader checking your number against the live database will find a different one, and only the edition stamp explains why.

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
SIPRI Arms Transfers Database	extends	The materiel counterpart. Expenditure says what was spent; transfers say what arrived, and reconciling the two is where most of the analytical value in both datasets is realised.
SIPRI Arms Industry Database	extends	Where the procurement share of national spending goes, at company level, and the route from national budgets to industrial capacity.
NATO	corroborates	Publishes defence expenditure figures for its members on its own definition, including category breakdowns that SIPRI does not provide. Divergence from SIPRI for the same country is definitional and instructive.
World Bank open data	prerequisite	GDP, population and government finance series that form the denominators for burden, per-capita and priority measures. Their revisions move SIPRI-derived ratios independently of any defence decision.
UN Comtrade	extends	Trade statistics that can indicate imported defence-related goods in monetary terms, providing a partly independent read on procurement flows.
ACLED Conflict Events	extends	Conflict intensity data for the same country-years, which is the behavioural variable most expenditure analyses are ultimately trying to explain or predict.
Our World in Data	corroborates	Republishes and visualises military expenditure alongside economic and social indicators, useful for sanity-checking your own aggregations against an independently constructed presentation.
UNIDIR	extends	Research on military spending transparency and its relationship to arms control and confidence-building measures, which is the policy context in which these figures are contested.

Legal, ethical and operational constraints

Use of the data is governed by SIPRI's terms, which permit free non-commercial research and educational use with attribution and require permission for reproduction and commercial exploitation; confirm the current wording before building a product on it. Beyond licensing, the ethical constraints are about representation rather than restriction. These are estimates presented with documented uncertainty, and stripping that uncertainty to produce a cleaner chart is a form of misrepresentation even though no individual number was altered. Where the figures are used in public advocacy, litigation or policy argument, the obligation to state the edition, base year and estimate status is not optional, because the counterparty will check and the credibility cost of an unexplained discrepancy falls on you. Analysts should also be conscious that military expenditure estimates for closed states are politically contested and are used in arguments about threat and rearmament; presenting a SIPRI estimate as a measured fact in that context is both inaccurate and, in its consequences, not a small error.

Operational security

Downloading a public workbook is close to the lowest-exposure collection activity in this catalogue and needs no special handling. The residual exposure sits in the query interface, where a pattern of narrow, repeated queries about one or two states discloses an analytical interest to the site operator and to anyone with access to those logs. The mitigation is the same one that produces better analysis anyway: take the full workbook and query your own copy, so that no question you are asking is ever visible outside your infrastructure. If your interest in a specific state is itself sensitive, do the download from infrastructure not attributable to the interest, and avoid creating an account or subscribing to notifications with an identifying institutional address.

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 SIPRI Military Expenditure is contributing anything, and they are worth baselining now so the answer is available later.

Annual collection success and completeness: whether the workbook was retrieved in the release window and whether the parsed row count and country count match the edition's own documentation.
Note capture rate: the proportion of country notes and estimate flags successfully parsed into structured fields, which determines whether your uncertainty handling is real or decorative.
Revision magnitude: how much the last five years of figures moved between editions, per country, which is the empirical measure of how provisional recent data is for your area of interest.
Estimate share in analysis: the fraction of the countries in any given product whose figures are SIPRI estimates rather than reported values, reported alongside the finding.
Break handling coverage: the number of series discontinuities identified and explicitly attributed in your data model versus the number present in the source notes.
Reconciliation gap: the difference between SIPRI figures and independent national or alliance figures for the same country-year, tracked over time as a measure of definitional divergence rather than error.
Edition stamping compliance: the share of published internal products carrying edition, base year and access date, which is the single best predictor of whether a figure can be defended later.

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:

Local currency first, always. The exchange rate and the deflator are assumptions layered on top of the national decision, and stripping them back is how you find out whether anything actually changed.
Share of GDP measures effort, not strength. A state can double its burden and field a worse army, and the two most common defence spending arguments in public debate both confuse these.
The base year is a property of the edition. Two analysts using constant dollars from different editions will never reconcile, and neither will be wrong, which is why the edition stamp is load-bearing.
Estimate flags are data. A pipeline that drops them produces a dataset that is technically complete and epistemically dishonest, and the people consuming it downstream will have no way to know.
Absence is not zero, and a regional aggregate built by summing across countries with missing years understates the region by an amount nobody will notice until someone checks.
Money and materiel should be read together. The gap between what a state spends and what it visibly acquires is one of the more productive questions in defence economics and is only visible if you hold both series.
Fiscal year adjustment means a SIPRI year is not necessarily a national budget year, which matters whenever you are aligning spending against a dated event such as a conflict onset or a policy change.
For states with conscription, indigenous production and low domestic price levels, the dollar figure understates real military resources, and a comparison that does not say so is making an argument by omission.
The country note is the analytical content for opaque states. If your finding about such a state does not engage with what SIPRI says it could not see, you have not finished the work.

Questions analysts actually ask

Why do SIPRI and my national defence ministry report different figures for the same year?

Because they use different definitions. SIPRI applies one specification to every country, which usually means including items a national budget places elsewhere, such as military pensions or certain paramilitary forces, and excluding items the national figure includes. The country note explains the specific adjustments for that state.

Which series should I use for a ten-year trend?

Local currency at current prices if the question is about one country's own decisions, and constant dollars at a stated base year if you need international comparability. Never current dollars, which mostly measure the exchange rate over that period.

Are the figures for closed or opaque states usable?

Yes, with explicit uncertainty. SIPRI publishes estimates for several major spenders where reported data is incomplete, and marks them as estimates with notes explaining the basis. Use them as ranges and hypotheses, not as measurements, and never build a precise ratio on top of one.

Why did the whole historical series change since I last used it?

Almost certainly rebasing of the constant-price series, revision of GDP data by the economic authorities that produce it, or SIPRI's own revision of past estimates as better information emerged. All three are normal, which is why editions must be archived and cited.

Can I get a breakdown by procurement, personnel and research?

Not from this database, which publishes national totals. Alliance reporting provides category breakdowns for member states, and some national budget documents do the same, but there is no globally consistent disaggregated series.

Does this include nuclear weapons programmes?

Handled according to SIPRI's stated definition, which is documented in the methodology, and the treatment of programmes run through civilian agencies varies with how a state organises them. Read the definition and the relevant country note rather than assuming either inclusion or exclusion.

How do I compare a conscript force with a professional one?

Not on per-capita spending, and cautiously on total spending. Conscription converts money into personnel at a completely different rate, so a like-for-like comparison needs force size, personnel cost structure and equipment holdings, none of which come from this dataset.

How quickly does new spending appear?

Slowly. The database is annual, published roughly a year in arrears, and the most recent year is provisional. It is a structural source, not a current-awareness one, and any question about spending decisions taken this year must be answered from national budget documents.

What does the platform add over the workbook?

Scheduled annual collection with edition and base-year retention, uncertainty markers modelled as structure rather than lost in parsing, country entity resolution across historical state changes, and correlation against transfer and conflict data. No figure is generated or inferred; the estimates remain SIPRI's.

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 scope definition is derived from a NATO-style specification of military expenditure but is SIPRI's own, so figures are not interchangeable with NATO's published series for member states even though the two are related.
Country identification should be normalised to ISO 3166 on ingest, with an explicit historical mapping for states that have dissolved, merged, been renamed or come into existence during the covered period.
Constant-price conversion follows a stated base year that changes between editions; any downstream system must treat base year as part of the measure's identity rather than as presentation metadata.
GDP and government expenditure denominators derive from international financial institution statistics and inherit those bodies' revision cycles and definitional conventions, including the distinction between general government and central government.
The UN standardised instrument for reporting military expenditures is the international transparency mechanism the data speaks to conceptually, and state submissions to it are the nearest thing to an official comparator.
In the platform the series normalises to time-series measures attached to country entities and exports as CSV, JSON and JSONL; it has no natural expression in indicator-exchange formats such as STIX or MISP.
Currency codes should be recorded as ISO 4217 alongside local-currency figures, since currency reforms and redenominations within the covered period will otherwise corrupt long series.

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.

SIPRI Military Expenditure Database — Stockholm International Peace Research Institute. The database landing page with the current release, access terms and links to the workbook. The correct citation target and the correct starting point.
Military Expenditure Database sources and methods — SIPRI. The definition of what is counted, the treatment of paramilitary forces, aid and pensions, the conversion and deflation approach, and the notation used for estimates. Non-optional reading.
SIPRI databases — SIPRI. The full index of SIPRI resources, several of which answer questions analysts wrongly bring to the expenditure series.
SIPRI arms and military expenditure research programme — SIPRI. The programme page linking the expenditure, transfer and industry work, and the route to the annual trends fact sheets that summarise each release.
SIPRI publications — SIPRI. Yearbooks, fact sheets and background papers, including the methodological discussions that explain why particular countries are treated as they are.
NATO — North Atlantic Treaty Organization. Publishes defence expenditure for member states on its own definition with category breakdowns, providing the clearest available demonstration of how much definition drives the headline number.
World Bank Open Data — World Bank. GDP, population and government finance series underlying the burden and per-capita measures, with their own revision cycles that propagate into any SIPRI-derived ratio.
UN Office for Disarmament Affairs — United Nations. Home of the international military expenditure transparency instrument and the policy machinery around military spending reporting.
UNIDIR — United Nations Institute for Disarmament Research. Research on transparency, confidence-building and the political uses of military expenditure data, which is the context in which these numbers get argued over.
Our World in Data — Global Change Data Lab. Independently constructed presentations of the same series alongside economic and social indicators. Useful for checking your own aggregation logic against someone else's.

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 collects each annual edition with its base year and country notes intact, models estimate flags and series breaks as structure rather than losing them in parsing, resolves changing state names to stable country entities, and places spending alongside delivered materiel and recorded conflict so the discrepancies are visible.. Browse the full source catalogue, or follow any tag above into the rest of the library.