On the discrepancies between the original Maddison dataset and more recent GDP reconstructions

Angus Maddison was one of the most cited economists of the 20th century.

I often get emails asking me about Maddison’s figures, because I have worked a lot on historical national accounts reconstructions (see here, here, or here), and I was for 2 years faculty in the University of Groningen, where Maddison worked, and this is a justifiably renowned place for this type of work.

As an example, I copy here the relevant bits from an email that a friend and colleague recently wrote me. This colleague is asking about Portugal’s figures, but notice that a lot of what I will have to say can be generalized to other regions:

“The updated Maddison dataset has Portuguese GDP per person at 985 in 1530 (I assume in 1990 dollars).  1239 in 1600.  1192 in 1700, 1614 in 1750, 1330 in 1800 and 1225 in 1850.

The original Maddison dataset has Portuese GDP per person as 425 in 1000, 606 in 1500, 740 in 1600, 819 in 1700, and 923 for 1820.

Do you have any insight into these discrepancies?”

I do. In short, for this and many other cases, Maddison’s figures are simply made up. Read more below.

(Also, do notice that the modern Maddison project database is using 2011 prices, not 1990 ones, so for instance Jaime Reis’ and my own figures for Portugal, 1527-1850, are presented in 1990 prices in our paper, but the Maddison project pushed this into 2011 prices, so the levels may look different but they are not; the underlying real indexes are the same; our figures are the ones included in the Maddison project.)

picture_Maddison

Nothing I say here means to disparage Angus Maddison. He was a pioneer in these things (especially in so far as putting together in one place many estimates which in fact were often produced by many people – this had a lot of merit, though he’d very frequently get the citations that truly belonged to those who did the original work, but that’s is another story). He did stimulate many of us to continue these lines of inquiry, and do a better job. He was great.

But do notice that in one of the first posts in this blog, I did write,

“Despite the importance that Maddison’s GDP and population figures had in stimulating our thinking about economic history and development, it is fair to say that his pre-1820 figures were less than solid.”

The best way to think about Maddison’s estimates, especially for the period before 1820, is: when estimates did not exist (and they usually did not), he’d make very strong assumptions.

If you are an economist working on long-run growth you need to be wary of these assumptions. For China, real income per capita was stuck at exactly $600 “international” 1990 GK dollars between 1300 and 1850, according to Maddison in his late “Chinese Economic Performance in the Long Run” book (p. 157). Does this show that China was Malthusian? no, because Maddison effectively assumed China was Malthusian, and this is why he chose these numbers. I once saw an economist present at a conference and looking frankly ridiculous by claiming he had found that China was Malthusian for most of its history by using Maddison’s numbers. And, by the way, we now know that Maddison was not quite right, as you can read here.

Part of the problem is that Maddison often presents Tables in a way that makes it difficult to trace the original sources (this, I dare say, often made scholars un-aware that those were not his actual estimates, but those of others, which let to the citations problem I mentioned above).

How did he get to those figures? If you look into Volume 2 of his “World Economy” books, which is where most figures are put in Tables together and hence where typically people look first, or if you look online, in this original page, there are none listed. The historical statistics excel file avaliable even says “Copyright Angus Maddison” once you open it…

If you dig deep in his book, however, you do find some info. For population, continuing with the Portugal example, you find in Volume 1, p. 230 that “Population of 13 small West European countries assumed to move parallel to the total for the 12 countries above“. The latter were bigger countries, but the source of most of those is given as Maddison (1991), so you’d have to keep tracking other references to see how even those were actually done. Once you find the answer you will see that the evidence is rather thin. Needless to say, modern reconstructions are much more careful than this.

As for GDP per capita, in Volume 1, p. 249 he writes:

“I assumed a growth rate of Spanish GDP per capita of 0.25 per cent a year from 1500-1600, no advance in the seventeenth century, and some mild progress from 1700 to 1820. I adopted a similar profile for Portugal”.

That’s it. He simply assumed the numbers — vaguely citing only the following as inspiration, though not using it directly and citing with the following caveats (notice his own choice of words): “Yun’s (1994) rough per capita GDP estimates for Castile (about three-quarters of Spain) … his indicators for secondary and tertiary activity are weak“.

So, Maddison’s Portugal GDP per capita numbers are hence simply assumed to behave similarly to those of Spain, which are simply assumed by himself.

By contrast with this, here are the primary, archival sources that Jaime Reis and I use in our recent reconstruction of Portuguese per capita GDP, 1527-1850, which is forthcoming in the June edition of the Journal of Economic History. I list these here as an example, so you can see how much more exhaustive modern work is that what Maddison did. You can see a picture of what these account books look like if you click in our appendix, here.

I hope this discussion has been sufficiently clear on the origins of the differences between modern estimates vis-a-vis those of Maddison. It should be by now clear why they are considerably different at times. And furthermore – if you read our article, you will see that Portugal does not behave like Spain at all…

Primary Sources

We have collected both prices and wages from account (receipts and expenditures) books of the institutions listed below. Almost all were purchasers both of commodities and labor services. Some of them were also sellers of certain commodities produced by them. The account books of these institutions always display: the date of the transaction, the gross and unit value of the commodity, the unit of measurement employed, the quality of the product (e.g., coarse or fine paper, mutton, pork, or beef), and particular features of the transaction.

In order to proxy missing values we sometimes used a similar product or labor type (e.g., tallow candles for wax candles or carpenters for masons, both being skilled workers) by adjusting its price using a price ratio with the original product at a nearby year. Furthermore, to complete our Linen series for Lisbon during 1766–1829, we relied on Madureira (1997), listed in the secondary sources section.

Lisbon and Its Hinterland

Casa da Congregação do Oratório, Casa da Saúde, Lº 1º Receita e Despesa (Arquivo Municipal de Lisboa)

Casa dos Contos: Archive of the Court of Auditors

Convent of Nossa Senhora da Luz: National Archive

Convent of Santa Marta de Jesus: National Archive

Convent of Santo Alberto: National Archive

Convent of São Domingos de Lisboa: National Archive

Convent of Carmo, Expenses of the Sacristy: National Archive

Hospital of S. José: National Archive

Hospital of All Saints: National Archive

Holy House of Mercy of Almada: Archive of the Holy House of Mercy of Almada

Holy House of Mercy of Lisbon: Archive of the Holy House of Mercy of Lisbon

Holy House of Mercy of Lisbon, Shelter: Archive of the Holy House of Mercy of Lisbon

Holy House of Mercy of Lisbon, Foundlings: Archive of the Holy House of Mercy of Lisbon

Monastery of Chelas: National Archive

Monastery of S. Dinis de Odivelas: National Archive

Convent of Santo António da Convalescença: National Archive

Fabric of the See of Lisboa: National Archive

Seminary of Santa Catarina: National Archive

Administration of the Royal Household, Kitchens: National Archive

Porto and Its Hinterland

For Porto, we rely on Godinho (1955) as a secondary source plus the following primary sources:

Casa Pia Orphanage (administration): Porto District Archive

The See of Porto (revenues and expenditure): Porto District Archive

Colégio dos Órfãos, Daily Expenditure: Porto Municipal Archive

Porto Holy House of Mercy, Jailhouse Expenditure: Archive of the Santa Casa da Misericórdia do Porto

Porto Holy House of Mercy, General Hospital: Archive of the Santa Casa da Misericórdia do Porto

Porto Holy House of Mercy, Interments: Archive of the Santa Casa da Misericórdia do Porto

Porto Holy House of Mercy, Hospice for the Homeless: Archive of the Santa Casa da Misericórdia do Porto

Porto Holy House of Mercy, D. Lopo Hospital: Archive of the Santa Casa da Misericórdia do Porto

Porto Holy House of Mercy, Foundling Home: Archive of the Santa Casa da Misericórdia do Porto

Municipality of Porto, Palace of the Municipality: Porto Municipal Archive Municipal Abattoir, Porto Municipality: Porto Municipal Archive

Coimbra and Its Hinterland

University of Coimbra, Refectory: Archive of the University of Coimbra

Hospital of the University, Accounts and Administration: Archive of the University of Coimbra

Hospital of Nossa Senhora da Conceição, Accounts: Archive of the University of Coimbra

College of São Pedro, Kitchen: Archive of the University of Coimbra

Colégio de São Pedro, Book of purchases: Archive of the University of Coimbra

Expenditure on the Churches of the Reverend Chapter of the See of Coimbra: Archive of the University of Coimbra

Chapel of S. João da Sé, Revenue and Expenditure: Archive of the University of Coimbra

Chapter of the See, register of expenditures: Archive of the University of Coimbra

Fabric of the College of São Pedro, Register of Expenses: Archive of the University of Coimbra

Municipal Council of Coimbra, Revenue and Expenditure: Archive of Coimbra

Municipality Works of the Church of the See of Coimbra, Expenses: Archive of the University of Coimbra

University of Coimbra, Receipts and Expenditure: Archive of the University of Coimbra

Hospital of São Lázaro, Receipts and Expenditure: Archive of the University of Coimbra

Holy House of Mercy of Coimbra, Income and Expenditure: Archive of the Holy House of Mercy of Coimbra

Episcopal Mitre of Coimbra, Expenses: Archive of the University of Coimbra

Register of the Granary of the Chapter of Coimbra: Archive of the University of Coimbra

Royal Hospital of Coimbra, Registers of Expenditure: Archive of the University of Coimbra

Évora and Its Hinterland

For Évora, we rely on Santos (2003) and Godinho (nd) secondary sources plus the following primary sources:

Royal Public Granary of Évora, Accounts: Archive of the District of Évora

Évora Aqueduct, Accounts of the Repairs and Maintenance: Archive of the District of Évora Repairs of Évora City Streets ,Wages and other Expenditure: Archive of the District of Évora

Casa Pia Orphanage, Revenues and Expenditures: Archive of the District of Évora

Casa Pia, Hospice of Nossa Senhora da Piedade, Accounts: Archive of the District of Évora

Holy House of Mercy, Books and Accounts: Archive of the District of Évora

Convent of Paraiso, Accounts: Archive of the District of Évora

Convent of the Saviour, Accounts: Archive of the District of Évora

College of Nossa Senhora da Purificação: Archive of the District of Évora

Secondary Sources

Costa, Leonor, and Jaime Reis. “The Chronic Food Deficit of Early Modern Portugal: Curse or Myth?” Análise Social LII (2017): 416–29.

Godinho, Vitorino Magalhães. Introdução à história económica. Lisboa: Livros Horizonte, n.d.

———. Prix et monnaies au Portugal 1750–1850. Paris: Librairie Armand Colin, 1955.

Madureira, Nuno. Mercados e Privilégios. A Indústria Portuguesa entre 1750 e 1834. Lisboa: Estampa, 1997.

Santos, Rui. Sociogénese do Latifundismo Moderno: Mercados, Crises e Mudança Social na Região de Évora, Séculos XVII a XIX. Lisboa: Banco de Portugal, 2003.

 

addendum: I originally wrote in this post, a few hours ago, that the Maddison project now corrects for inflation by expressing variables in 2015 dollars, but this should be 2011. I heard they were thinking of changing to 2015 in the next version.

3 thoughts on “On the discrepancies between the original Maddison dataset and more recent GDP reconstructions

  1. The Maddison Project series, revision 2018 includes not only the reference year change (to the year 2011) but also other very relevant and valuable methodological changes such as the inclusion of historical benchmarks in the calculation of the PPAs. However, as this correction is made only for some countries but not for all, some changes in the relative positions of the countries in the period 1870-1950 are not very clear yet. A summary of this is found on page 17 of Jutta Bolt, Robert Inklaar, Herman de Jong
    and Jan Luiten van Zanden (2018)

    Like

    1. Thanks for your comment. If you click in the third link right at the start of the post, you will see that Herman de Jong and I have written a book chapter just about the matters you are raising. I have not covered them here simply because there are not relevant for what I was discussing in this post. When there are no benchmarks further back, time-series are all we have:

      https://www.rug.nl/ggdc/historicaldevelopment/maddison/publications/wp12.pdf

      However, I thank you for making a small correction: I originally wrote in this post that the Maddison project now corrects for inflation by expressing variables in 2015 dollars, but this should be 2011. I do believe they are thinking of changing to 2015 in the next version.

      Like

Leave a Reply to npalma Cancel reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s