Frequently Asked Questions

The Global Macro Database (GMD) is a free dataset that brings together macroeconomic statistics for 239 countries and territories in a single, consistent format. It integrates 160 historical and contemporary sources, which makes it the most comprehensive dataset of its kind. It is updated quarterly, and is introduced in our paper (Müller, Xu, Lehbib & Chen, 2025; NBER Working Paper 33714).

46 variables across six groups (national accounts, consumption and investment, the external sector, government finances, money and interest rates, and prices and labor markets) for 239 countries from 1086 to 2025, with forecasts to 2030. Because not every country has every variable for the whole period, it is an unbalanced panel, and the availability depends on the underlying sources.

It is free for academic and non-profit research, provided you attribute and cite it. Commercial use is not permitted, which includes products and services, trading and investment research, consulting deliverables, paid reports, training commercial models, and reselling values derived from the data. If you are unsure, treat your use as commercial and email kmueller@globalmacrodata.com. See the full terms of use.

Cite Müller, Xu, Lehbib & Chen (2025), The Global Macro Database: A New International Macroeconomic Dataset, NBER Working Paper 33714. Ready-to-use BibTeX is on the home page and the research paper page.

On the Data page you can download the full dataset as CSV, Excel (.xlsx), or Stata (.dta). Alternatively, you can access the data directly through our Stata, Python, or R functions.

All three packages accept version, country, and variable filters and have full feature parity:

• Stata: ssc install gmd, then gmd (e.g. gmd rGDP, country(FRA))
• Python: pip install global_macro_data, then from global_macro_data import gmd; gmd()
• R: install from GitHub, then gmd(...)

Full instructions and examples are on the Data page.

Yes. The download page and packages offer the option to return the harmonized raw data from each underlying source (for example gmd rGDP, raw in Stata). These data are only harmonized to make units and currencies comparable, but are otherwise reported as-is. In fact, many of our users find the raw data useful even where they do not use the combined GMD series.

Quarterly. Each release is versioned (e.g. 2026_06). Values can change between versions as we correct errors, add additional data, change the priority of sources, among other reasons. Full release notes are on the Data page.

The GMD prioritizes sources based on our experience regarding their reliability and availability. We differentiate sources along two lines: update frequency and country coverage. Regarding updates, current sources such as national statistical offices or international organizations frequently post new data that we incorporate with every GMD release. In contrast, historical sources are not or infrequently updated. Regarding country coverage, aggregators provide data on several countries, while country-specific sources only offer information for individual nations. The combined GMD series combine these different sources into a single series using the methods outlined in the technical documentation. The results can thus sometimes differ from the data of individual sources, but they reflect our best efforts to combine the available information.

Because each series is built by splicing many sources together (ratio chain-linking), the national-accounts identity Y = C + I + G + NX won't hold exactly: there is an implicit statistical discrepancy, just as official sources like Eurostat report. The GMD currently does not force components to reconcile (no hierarchical rescaling), so that each series stays as close to the raw data's growth rates as possible. We plan to offer reconciled versions in future releases.

Units depend on the type of variable, and every variable's unit is shown on the Explore page and in the documentation. In brief:

• Level variables — GDP, consumption, investment, trade, government revenue, expenditure and debt, and the monetary aggregates (M0–M4) — are in millions of local currency. Nominal series are in current prices; real series (e.g. rGDP) are in constant 2015 prices.
• US-dollar variants (the _USD series, e.g. nGDP_USD) are in millions of US dollars.
• Per-capita GDP is in local currency (rGDP_pc) or US dollars (rGDP_pc_USD).
• Ratio variables (the _GDP series such as govdebt_GDP) are in percent of GDP.
• Rates — inflation, unemployment and the interest rates — are in percent.
• Index variables (CPI, the GDP deflator, REER and the house-price index) are rebased so that 2015 = 100.
• The exchange rate (USDfx) is local currency per US dollar, and population (pop) is in thousands of persons.

Real and index variables (real GDP, the consumer price index, and the GDP deflator) are rebased to a common base year of 2015, using the IMF World Economic Outlook as the reference, so levels are comparable across countries and over time. For sources with no data in 2015, we use the most recently-available year.

The GMD uses an anchor year of 2018. Historical data (before 2018) is chain-linked backward from the anchor and stays fixed across releases unless an error is found; recent data (after 2018) is chain-linked forward and absorbs new revisions and forecasts. This keeps the long-run history stable while letting recent years update with each release. The anchor year may shift in future major releases.

Long-run series span eras of currency reform, so the GMD converts everything to a single consistent unit using official rates. Examples include applying the ECB's irrevocable euro conversion rates or stepping through Brazil's many currency changes. Conversions are documented in the country notes.

All ratio variables (e.g. govdebt_GDP, exports_GDP, CA_GDP) are expressed in percent: a value of 50 means 50% of GDP.

Some variables are constructed directly, in that the GMD harmonizes and splices the observations that underlying sources actually report for that series. Others are derived: they are computed from those series through standard accounting identities. Which is which depends on how most sources report the data:

• Constructed directly: nominal and real GDP, consumption, investment, exports and imports, the CPI and house-price index, inflation, population, the monetary aggregates, interest rates, the exchange rate, and the central- and general-government revenue, expenditure and debt series reported as a share of GDP.
• Derived: US-dollar versions of the level series (converted with the exchange rate USDfx; real GDP in USD instead uses a chain method that preserves real growth — see below), per-capita series (rGDP_pc = real GDP ÷ population), the GDP deflator (nominal ÷ real GDP), the national-accounts ratios such as cons_GDP and exports_GDP, the consolidated government aggregates, and the government levels (e.g. govdebt), back-computed from the share-of-GDP series.

Put differently, what is derived or directly-constructed depends on the type of variable. Consumption is collected as a level and its share of GDP is derived, whereas the current account and government finances are collected as shares of GDP and their levels are derived. On the Explore page, directly-constructed variables let you compare the underlying sources, while derived ones show a note pointing to the underlying variable(s); the exact status of each variable is recorded in the documentation.

Real GDP in U.S. dollars is not a year-by-year conversion of local real GDP at the current exchange rate. Instead, the GMD instead follows the World Bank's method, which preserves each country's real (volume) growth:

1. Anchor in the base year (2015). Nominal GDP in 2015 is converted to USD at the 2015 average market exchange rate. Because current and constant prices coincide in the base year, this is also real GDP in USD for 2015.
2. Apply the real growth rates from the constant local currency real GDP series (rGDP) backwards and forwards.

The resulting series reflects the true volume growth of the local economy, abstracting from inflation and exchange rate fluctuations. This is different from nGDP_USD (nominal GDP in USD), which is a straight year-by-year conversion at the market rate (nGDP ÷ USDfx). Full detail is in the Technical Appendix.

cons is total consumption and remains the headline series, because historical sources usually report it. As of version 2026_03, hcons (household consumption) and gcons (government consumption) provide a finer breakdown where the data allows.

Government-finance variables come in central-government and general-government versions, plus consolidated aggregates that maximize data availability. General government is broader (it includes state and local government and social security), while central government covers only the central state. Choose the concept that matches your question. Note that, because we chain-link series, central government values can sometimes exceed general government values. We are working on more advanced splicing methods that will retain variation in the raw data as much as possible while enforcing that central government numbers never exceed those of the general government.

Every series is inspected visually. For each variable and country, all available sources are plotted on a single chart (over 4,000 plots in total) and manually reviewed for level shifts, disagreements between sources, outliers, unit or currency errors, and bad splices. Issues are documented and corrected, and remaining uncertainties are flagged.

To build one long series from many sources, the GMD chain-links the series for many variables: where two sources overlap, it matches their growth rates (ratio splicing) rather than pasting raw levels together, which avoids artificial jumps. Where sources do not overlap, a representative growth rate is used. Every break and method is recorded in the country notes.

The GMD flags two kinds of doubtful points: outliers (implausible jumps in a series, cross-checked against other sources and known economic events) and definitionally impossible values. Confirmed errors are corrected or removed, while genuine economic events are kept.

The Documentation page links to per-country and per-variable PDFs that show the raw underlying sources, the final GMD series, the splice points between sources, and notes on adjustments for each country-variable combination. You can also explore the same data directly on the Explore page.

The team actively fixes issues reported by users and welcomes new data sources. Email kmueller@globalmacrodata.com or open an issue on the GitHub repository.

The email you provide when downloading is used to notify users about critical updates. Release notes for every version are published on the Data page.