Sources and methods

The SIPRI Military Expenditure Database has data for countries for the period 1949–2023. The data is subject to continuous revisions, which can be extensive, for example when new and better sources come to light.

Quick links:

1. Purpose of the data

The main purpose of the data on military expenditure is to provide an easily identifiable measure of the scale of resources absorbed by the military. Military expenditure is an input measure which is not directly related to the 'output' of military activities, such as military capability or military security. Long-term trends in military expenditure and sudden changes in trends may be signs of a change in military output, but such interpretations should be made with caution.

Military expenditure data measured in constant dollars is a trend indicator of the volume of resources used for military activities, which allow comparisons to be made over time for individual countries and between countries. The share of gross domestic product (GDP) is a rough indicator of the proportion of national resources used for military activities, and therefore of the economic burden imposed on the national economy.

2. Sources

The sources for military expenditure data are, in order of priority:

(a) primary sources, that is, official data provided by national governments, either in their official publications or in response to questionnaires;

(b) secondary sources which quote primary data; and

The first category consists of national budget documents, defence white papers and public finance statistics published by ministries of finance and ministries of defence, central banks and national statistical offices. It also includes government responses to questionnaires about military expenditure sent out by SIPRI, the United Nations or the Organization for Security and Co-operation in Europe (OSCE), and expert analyses of government budgets by members of the SIPRI Military Expenditure and Arms Production Expert Network (see below).

The second category includes international statistics, such as those produced by NATO and the International Monetary Fund (IMF). Data for most NATO countries is taken from NATO defence expenditure statistics as published in a number of NATO sources. Previously data for many developing countries was taken from the IMF's Government Financial Statistics Yearbook, which provides a defence line for most of its member countries, Country Reports by IMF staff and from the United Nations Statistical Yearbook (UNSY), which used to provide such data in many cases. This category also includes the publications of other organizations that provide proper references to the primary sources used. The three main sources in this category are the German Statistisches Jahrbuch für das Ausland, the Europa Yearbook, and Country Reports of the Economist Intelligence Unit.

The third category of sources consists of specialist journals and newspapers.

Secondary sources of data represent a much higher proportion of data for earlier years, in particular the pre-1988 data. The IMF GFS and UNSY, as well as NATO data, are particularly important. In more recent years, a lot more primary data from national authorities has become available.

3. Definition of military expenditure

Although the lack of sufficiently detailed data makes it difficult to apply a common definition of military expenditure on a worldwide basis, SIPRI has adopted a definition as a guideline. Where possible, SIPRI military expenditure data include all current and capital expenditure on:

(a) the armed forces, including peacekeeping forces;

(b) defence ministries and other government agencies engaged in defence projects;

(d) military space activities.

This should include expenditure on:

i. personnel, including:

a. salaries of military and civil personnel;

b. retirement pensions of military personnel, and;

c. social services for personnel;

ii. operations and maintenance;

iii. procurement;

iv. military research and development;

v. military infrastructure spending, including military bases. and;

vi. military aid (in the military expenditure of the donor country).

SIPRI’s estimate of military aid includes financial contributions, training and operational costs, replacement costs of the military equipment stocks donated to recipients and payments to procure additional military equipment for the recipient. However, it does not include the estimated value of military equipment stocks donated.

Civil defence and current expenditures on previous military activities, such as veterans' benefits, demobilization, conversion and weapon destruction are excluded.

In practice it is not possible to apply this definition for all countries, and in many cases SIPRI is confined to using the national data provided. Priority is then given to the choice of a uniform definition over time for each country in order to achieve consistency over time, rather than to adjusting the figures for single years according to a common definition. In the light of these difficulties, military expenditure data is most appropriately used for comparisons over time, and may be less suitable for close comparison between individual countries. Reference should always be made, when comparing data for different countries, to the footnotes and special notes attached to the data for these countries, which indicate deviations from the SIPRI definition, where these are known.

4. Methods

SIPRI data reflects the official data reported by governments. As a general rule, SIPRI takes national data to be accurate until there is convincing information to the contrary. Estimates are made primarily when the coverage of official data does not correspond to the SIPRI definition or when there are no consistent time series available that cover the entire period covered by the data.

In the first case, estimates are made on the basis of an analysis of official government budget and expenditure accounts. The most comprehensive estimates, for China and Russia, have been presented in detail in previous SIPRI Yearbooks. Other countries where the entire series is estimated are Israel and the United Arab Emirates.

In the second case, differing sources, covering overlapping time periods, are linked together to produce a longer series. Where these sources follow different definitions leading to systematic differences in their levels, one series (the ‘primary’ source) is used directly without adjustment, while the percentage change between years in the other source (the ‘secondary’ source) is applied either forwards or backwards from the primary source series, in order to make the trend as accurate as possible. In selecting the primary source, preference is given to:

(a) the source most closely matching the SIPRI definition,

(b) the most recent source, and

However, in many cases it is only possible to achieve two out of three of these criteria.

The use of these types of estimates has been particularly necessary for the older, pre-1988 data produced in the recent data extension exercise. This is because the sources of data used for more recent years, often primary sources, are not available for earlier periods, and have had to be combined with other, often secondary sources, which frequently follow a different definition of military spending. In particular, sources such as the IMF GFS, UNSY and Statistisches Jahrbuch für das Ausland tend to exclude military pensions.

In order not to introduce assumptions into the military expenditure statistics, estimates are always based on empirical evidence and never based on assumptions or extrapolations. Thus, no estimates are made for countries that do not release any official data. These countries are displayed without figures.

SIPRI stimates are presented in blue text in the tables. Red text is used when data is uncertain for other reasons, such as the reliability of the source or because of the economic context. Figures are more unreliable when inflation is rapid and unpredictable. Supplementary allocations made during the course of the year to cover losses in purchasing power often go unreported and recent military expenditure can appear to be falling in real terms when it is in fact increasing.

Data for the most recent years include two types of estimate which apply to all countries:

(a) figures for the most recent years are for adopted budgets, budget estimates or revised estimates, and are often revised in subsequent years; and

(b) the deflator used for the latest year in the series is an estimate.

Unless exceptional uncertainty is involved with these estimates, they are not marked in red text.

5. SIPRI estimates for China

In its estimates of Chinese military expenditure, SIPRI seeks to take into account a number of sources of military expenditure outside the official defence budget. Such sources of military expenditure include funding from other central government ministries (some of which is publicly available, some of which is not), funding from local government and funding from internal People's Liberation Army (PLA) sources—the latter probably represents a much smaller share of the total than in the past. SIPRI's estimate of China's military spending is based on a methodology used in a study published in SIPRI Yearbook 1999 by Professor S. Wang and revised in 2021 in a SIPRI background paper titled ‘A new estimate for China’s military expenditure’ by Tian, N. and Fei. S.

SIPRI estimate for Chinese spending are based on data from various editions of the China Public Finance Yearbook, the China Statistical Yearbook and other official publications, but also in some cases require additional estimation where the data series is not available.

The items outside the official defence budget that are included in the estimates are:

(a) spending on the paramilitary People's Armed Police (PAP);

(b) soldiers' demobilization and retirement payments from the Ministry of Civil Affairs;

(c) additional military research, development, testing and evaluation (RDT&E) funding outside the national defence budget;

(d) additional military construction expenses;

(e) commercial earnings of the PLA (zero as of 2015);

(f) subsidies to the arms industry (zero as of 2010);

(e) Chinese arms imports (zero as of 2020); and

(g) the Chinese Coast Guard (since 2013).

These figures and estimates are derived as follows:

Actual PAP spending can be found in the Chinese version of the published expenditure figures up to 2019, while the figures for 2020 is estimated based on the rate of change of the Public Security budget.
Official spending figures for demobilization payments also come from published expenditure figures up to 2019, with the figures for 2020 estimated based on the rate of change of the official budget.
Estimates for subsidies to the arms industry are based on a share of the total budget for industrial subsidies. From 2005, this share is assumed to have declined due to the increasing profitability of most of the arms industry in China, and to have been zero from 2010 onwards.
Estimates for additional military RDT&E from 2007–20 are based on a share of total Central Government appropriations for Science & Technology (S&T). The share is based on information for 2011–14 on the proportion of the S&T budget that is allocated to civilian agencies that disclose their spending in annual reports. The remainder is assumed to be allocated to the agencies that do not disclose annual reports, with military and security significance, and it is estimated that 90% of this is for military purposes. The estimates for 1997–2006 are based on a slightly smaller share of a previous series for Central Government S&T appropriations, which used a different classification system, giving somewhat higher figures than the new system. The estimates up to 1996 are Professor Wang’s estimates, and are based on a share of overall government Research and Development and Science and Technology budget.
Estimates for additional military construction are based on a share of the government's capital infrastructure budget. As these figures are not published beyond 2006, estimates for 2007–18 are based on the average economic growth rate;
Estimates for arms imports use figures provided by Russia for the value of arms transfers to China for the years where this information is available, as Russia accounts for the vast majority of Chinese arms imports. For the years where these figures are not available, the estimates are based on the rate of change of China's arms imports as measured by the SIPRI arms transfers Trend Indicator Value (TIV). See ‘A new estimate for China’s military expenditure’ by Tian, N. and Fei. S. for details on the phasing out of the category appropriations for arms imports outside the national defence budget.
Income from commercial activities of the PLA is assumed to have declined steadily since 1999, as a policy of divestment from such activities has been followed. This is assumed to have reached zero in by 2015. The figures for 1989–98 are Professor Wang’s, and are based on a share of the official defence budget.
Due to the lack of budgetary information of the Chinese Coast Guard (CCG), one way to assess changes in spending on the CCG is through the increase in the number of its ships. An estimate of spending by the CCG can then be based on the number of ships, as published each year in the Military Balance. The average annual increase in the number of ships operated by the CCG between 2015 and 2019 was 16 per cent. Based on the estimate that 50 per cent of CCG spending in 2015—that is 5.6 billion yuan ($0.9 billion)—counts as military spending, an increase of 16 per cent per annum would mean spending of 6.9 billion yuan ($1.0 billion) in 2016, 8.1 billion yuan ($1.2 billion) in 2017, 9.2 billion yuan ($1.4 billion) in 2018, 11.1 billion yuan ($1.6 billion) in 2019 and 11.8 billion yuan ($1.7 Billion) in 2020.

The resulting SIPRI estimates for Chinese military spending for recent years come to around 1.4 times the official defence budget for most years.

A 2006 report by the US-China Policy Foundation, based on a analysis of available Chinese-language sources, broadly concurs with the list of items included by SIPRI, but also adds various additional forms of funding to the PLA from local government, as well as some higher education expenses for PLA officers and compensation for disaster relief activities. The report concludes, however, that there is not at present enough information to make a reasonable estimate of total Chinese defence-related spending.

While details of some elements of Chinese military spending outside the official defence budget are publicly available (such as the PAP budget) others—most importantly R&D spending—are not, and can at present only be the subject of educated guesswork. Further research based on publicly available Chinese-language sources could provide improved estimates, but without greater transparency on the part of the Chinese Government, a completely accurate figure is not currently possible.

SIPRI does not publish estimates for Chinese military spending before 1989. While data for the official National Defense budget is available back to 1950, relevant data for other components of the SIPRI estimate are not. Moreover, it cannot be assumed that the methodology used could be reliably applied to data before 1989, or that the relationship between the official defence budget and total military spending remained the same. Further research would therefore be needed to produce estimates for Chinese military spending for earlier years.

6. The SIPRI Military Expenditure and Arms Production Expert Network

SIPRI's efforts to gather reliable military expenditure data (as well as data on the arms industry) are assisted by the SIPRI Military Expenditure and Arms Production Expert Network, an international network of experts associated with the SIPRI Military Expenditure and Arms Production Programme.

Network members provide SIPRI with regular data on the military expenditure or arms industry of specific countries on which they have a particular expertise (most often, though not always, their own). In addition, SIPRI will from time to time consult with members of the network on specific issues relating to their areas of expertise. Network members receive a small annual honorarium in return for the provision of data. This is typically $150–500, depending on the volume of data supplied. In addition, they receive a copy of each edition of the SIPRI Yearbook to which they have contributed data. They are also credited for their contribution in the relevant section of the Yearbook.

The SIPRI Military Expenditure and Arms Production Programme periodically seeks new members of the network for countries where new or improved sources of data are considered necessary, or where we become aware of relevant potential network members who could provide such improved data. Potential network members must fulfil the following requirements:

Education to Masters Degree level or equivalent.
Evidence (through relevant publications or otherwise) of a detailed understanding of the military budget or arms industry of the country for which they are to provide data.
Publications in relevant fields, including international security, military affairs, peace and conflict studies, defence and peace economics, etc.

If you are interested in becoming part of the SIPRI Military Expenditure and Arms Production Expert Network or want more information, please contact Dr. Lucie Béraud-Sudreau, Director of the SIPRI Military Expenditure and Arms Production Programme.

7. Calculations

The SIPRI military expenditure figures are presented on a calendar-year basis. The only exception is the United States, for which statistics report data on a financial-year basis. Calendar-year data is calculated on the assumption of an even rate of expenditure throughout the financial year.

The deflator used for conversion from current to constant prices is the consumer price index (CPI) of the country concerned. This choice of deflator is linked to the purpose of the SIPRI data—that it should be an indicator of resource use on an opportunity-cost basis.

Average market exchange rates (MER) for the relevant year are used to convert local currency figures into US dollars, or in some cases official fixed exchange rates. For the constant dollar figures, this means the MER for the base year (currently 2022), while for the current dollar figures given for the most recent year’s data, this means the MER for that year (currently 2023). Market exchange rates are determined by the supply and demand of currencies used in international transactions, and as such do not always accurately reflect differences in price levels between countries. Fixed official exchange rates may be even less reflective of the relative purchasing power of a currency.

The choice of base year (the year in whose prices the data is expressed) also has a significant impact on cross-country comparisons of expenditure data because different national currencies vary against the dollar in different ways. For the current edition of the SIPRI Military Expenditure Database, the base year has been updated to 2023.

The ratio of military expenditure to GDP is calculated by dividing military expenditure in nominal, local currency figures for each year by GDP for each year, also in nominal, local currency. This figure (the ‘military burden’) represents the most common measure of the economic burden of military expenditure. SIPRI also presents figures for military expenditure as a share of total government expenditure from 1988, using IMF figures for general government expenditure (including central, regional, local government etc.), where available.

SIPRI presents estimates of aggregate military expenditure, in constant (2022) US dollars for various geographic regions.

For the purpose of calculating aggregate totals, estimates have been made for countries for which data is lacking in some years. These estimates are made on the assumption that the trend for these countries is the same as for the geographical region in which they are located. However, in some cases where no data has been available for many years, it is not considered viable to make meaningful estimates in this way, as the degree of uncertainty is too large. These countries are therefore excluded from the world, regional and group totals. These countries are Djibouti (no data since 2008), Eritrea (2003), Laos (2013), Turkmenistan (1999), Uzbekistan (2003) and Yemen (2014). Other countries are excluded from the totals due to the lack of economic data allowing conversion into constant dollars. These countries are Cuba, North Korea, Somalia and former Yugoslavia. Syria is also excluded from all totals as there is no economic data beyond 2011, so conversion into constant (2022) prices is impossible.