Maddison Project

From Devec
Jump to: navigation, search
This article is about an economic statistics project. View a list of all economic statistics projects

The Maddison Project, also known as the Maddison Historical Statistics Project, is a project to collate historical economic statistics, such as GDP, GDP per capita, and labor productivity.[1][2][3] It was launched in March 2010 to continue the work of the late economic historian Angus Maddison. The project is under the Groningen Growth and Development Centre at the University of Groningen,[2] which also hosts the Penn World Table, another economic statistics project.[4]

This page describes the data and methods both produced by the explicit Maddison Project and produced by Angus Maddison before his death (since the Maddison Project is continuation of his work).


Item Value
Start date March 2010 for the explicit Maddison Project[1]
1960s for the original work by Angus Maddison that was the genesis of the project.[5]:3
Data versioning Only one update released as Maddison Project, published January 2013 with data till 2010.[5][6]
Multiple versions by Angus Maddison, the last of which was published in February/March 2010.[7]
Focus Historical: identify general ballparks and trends in living standards and economic growth over long time periods.
Provide better insight into the timeline of the Great Divergence between Western Europe and other regions that were historically similarly situated, such as China and India.
PPP comparison method used The method developed by Kravis, Heston, and Summers for the International Comparison Program.[7] The same method is used by the Penn World Table.


The list of versions is incomplete.

Version Formats available Year of publication Month of publication Most recent year till which data is present Dimensions (inputs) Metrics (outputs)
Maddison 2010 (version published at start of Maddison Project)[7] Excel spreadsheet, both horizontal (years are columns)[8] and vertical (years are rows)[9] 2010 March (for explanatory note), February (for data) 2008 for GDP and GDP per capita
2009 for population; also 2030 projections for population
Country (or geographical region) and year Population, real GDP, real GDP per capita, population growth, real GDP growth, real GDP per capita growth
Maddison 2013 (first update by Maddison Project)[5] Excel spreadsheet, only vertical (years are rows)[6] 2013 January 2010 Country (or geographical region) and year Real GDP per capita

Data description

Data dimensions and metrics

The data presented in the Maddison Project database is a partial function where:

  • The inputs (the dimensions) are region (usually, country) and year.
  • The metrics include:
    • Population: Included in
    • Real GDP
    • Real GDP per capita, expressed in 1990 international Geary–Khamis dollars. For simplicity, we will refer to this as GDP per capita.

Year dimension

While calendar years are the finest granularity at which data is presented, not all calendar years have data. Here is a description of how the granularity changes over time. Note that we count a year is present if there is data for at least one country for that year.[6][9]

Year range Data granularity for Maddison 2010 (data till 2008) Number of regions with data in Maddison 2010 Data granularity for Maddison 2013 (data till 2010) Number of regions with data in Maddison 2013
2009 to 2010 Not present 0 Every year 137
1993 to 2008 Every year 181 Every year 182
1991 to 1992 Every year 181 Every year 179
1990 Every year 181 Every year 183
1974 to 1989 Every year 157 Every year 165
1973 Every year 172 Every year 180
1952 to 1972 Every year 157 Every year 165
1950 to 1951 Every year 157 Every year 158
1820 to 1949 Every year Variable Every year Variable
1801 to 1819 Not present 0 Every year Variable (5 to 7)
1700 to 1800 1700 42 1700, 1725, 1750, 1775, 1800 Variable (4 to 18)
1400 to 1699 1500, 1600 (once every 100 years, but not 1400) 42 and 39 1400, 1450, 1500, 1550, 1600, 1650 (once every 50 years) Variable (4 to 9)
Before 1400 1, 1000 41 and 39 1, 730, 1000, 1150, 1280, 1300, 1348 Variable (5 to 16)

Region dimension

The region dimension includes most modern countries, but also includes historical countries (such as the former USSR) and regions within countries (such as centre-north Italy) for which it is easier to get historical data than their modern country equivalents. The working paper says:[5]

A related issue is that historical estimates often refer to different territorial entities than the countries within the borders of 1990, the basic unit of account used in the Maddison framework. He made many corrections for (minor) changes in borders (an overview will be provided in future work). However, moving back in time sometimes means that we have only estimates for Northern Italy (instead of Italy as a whole), for Holland (Netherlands) or for the Cape Colony (South Africa). When those smaller regions represented less than two-third of the population and/or the GDP of the modern country (within current borders), we have presented the estimates in italics to warn users.

In addition, data is also presented on aggregate regions (such as Western Europe) and the whole world. Data on aggregate regions and former countries is presented in bold.

Other information

As mentioned in connection with the region dimension, italics and bold are used for some cells.

Notes are added for some estimates, that are visible by hovering over the cell in spreadsheet software. A triangle at the top right of a cell indicates that there are notes for that cell.

Although the working paper mentioned that the estimates would be color-coded (see #Accuracy and precision), this color-coding has not been implemented in the data.



Data is not present for every combination of country and year. This could be because of the absence of reliable sources that could be used to construct the data. See the table above (in #Year dimension) for more information on completeness.

Data on former countries continues to be calculated for years after that country ceases to exist as a geopolitical entity.

Accuracy and precision

Data is fairly inaccurate and imprecise, with the accuracy and precision getting worse the farther back in time we go, and also varying significantly across geographical regions. However, the database does not provide error bars or any other explicit quantification of uncertainty.

The working paper uses Steve Broadberry's classification of data sources, listing four broad types of data sources in decreasing order of reliability, along with the color coding that should show up in the database.[5]:4 Unfortunately, the color coding was not implemented in the dataset.[6]

Rank Type of estimate Color coding in database
1 official estimates of GDP, made by national statistical offices or by international agencies (UN, for example) black
2 historical estimates based on the same methods and broad range of data blue
3 historical estimates based on indirect proxy variables orange
4 guesstimates red

Design decisions that might lead to systematic biases

Maybe something about the purchasing power parity marking China as too expensive?[10]

Differences between versions

Both the 2010 version and the 2013 version use 1990 international dollars for the GDP per capita estimate, so the numbers are comparable between the versions.

Fill this in later

Data sources

Methods of estimation


Person Affiliation Qualification Opinion
Branko Milanović[3] World Bank Development economist Only source for long-run national GDPs going back to 1920s
Also, differing conclusions about Chinese GDP and growth rates due to higher estimates of their price levels
Morten Jerven[11] Norwegian University of Life Sciences Development economist One of three main sources of GDP numbers
Bill Gates[12] Bill & Melinda Gates Foundation Smart well-read person Mostly echoes Jerven
Max Roser[13] Our World in Data Economist Lists the database as one of the data sources for GDP.
Paul Krugman[14] New York Times Economist, columnist Data source for historical debt, growth, and labor output and productivity data.

Usage in debates

See also

External links


Note: For references to papers, a page number is included after the reference number in the original text. This page number is the page number as labeled on the page, rather than the page number in the PDF (these could differ due to the presence of a cover page).

  1. 1.0 1.1 "Maddison Project". Retrieved October 3, 2017. 
  2. 2.0 2.1 "The Database". Groningen Growth and Development Centre. Retrieved October 3, 2017. 
  3. 3.0 3.1 Milanović, Branko (July 19, 2013). "The end of a long era". World Bank. Retrieved October 3, 2017. 
  4. "The Database. Penn World Table version 9.0". Groningen Growth and Development Centre. Retrieved October 3, 2017. 
  5. 5.0 5.1 5.2 5.3 5.4 Bolt, Jutta; van Zanden, Jan Luiten (January 2013). "The First Update of the Maddison Project Re-Estimating Growth Before 1820" (PDF). Retrieved October 3, 2017. 
  6. 6.0 6.1 6.2 6.3 "The Maddison Project 2013 data". January 1, 2013. Retrieved October 7, 2017. 
  7. 7.0 7.1 7.2 "Background notes on "Historical Statistics" in" (PDF). March 1, 2010. Retrieved October 7, 2017. 
  8. "Horizontal file". February 1, 2010. Retrieved October 7, 2017. 
  9. 9.0 9.1 "Vertical file". February 1, 2010. Retrieved October 7, 2017. 
  10. Jan Luiten van Zanden; Debin Ma (September 1, 2017). "What Makes Maddison Right?" (PDF). Retrieved October 5, 2017. 
  11. Jerven, Morten. "Why Do GDP Growth Rates Differ?". Retrieved October 3, 2017. 
  12. Gates, Bill (May 8, 2013). "Bill Gates: how GDP understates economic growth. GDP may be an inaccurate indicator in sub-Saharan Africa, which is a concern for those who want to use statistics to help the world's poorest people". The Guardian. Retrieved October 3, 2017. 
  13. "Economic Growth § Data Sources". Our World in Data. 2017. Retrieved October 21, 2017. 
  14. Krugman, Paul (April 26, 2013). "Debt and Growth Data". New York Times. Retrieved October 3, 2017.