Labour productivity is defined as output per unit of labour input (persons employed). Labour productivity growth may be due to either increased efficiency in the use of labour, without more of other inputs, or because each worker works with more of the other inputs, such as physical capital, human capital or intermediate inputs. More sophisticated measures, such as “total factor productivity”, which is the output per combined unit of all inputs, are not included. Estimated labour productivity may also show an increase if the mix of activities in the economy or in an industry has shifted from activities with low levels of productivity to activities with higher levels, even if none of the activities have become more productive. For a constant “mix” of activities the best measure of labour input to be used in the productivity equation would be “total number of annual hours actually worked by all persons employed”. In many cases, however, this labour input measure is difficult to obtain or estimate reliably.
The limitations to the international and historical comparability of the estimates are summarized under the following three headings.
1. Output measures in national currencies
Output measures are obtained from national accounts and represent, as much as possible, GDP at market prices for the aggregate economy and value added at basic prices for the individual sectors. However, despite common principles that are mostly based on the United Nations System of National Accounts, there are still significant problems in international consistency of national accounts estimates, in particular for economies outside the Organization for Economic Co-operation and Development (OECD). Such problems include:(a) different treatment of output in services sectors; (b) different procedures in correcting output measures for price; and (c) different degree of coverage of informal economic activities in developing economies and of the underground economy in developed (industrialized) economies in national accounts.
2. Purchasing power parities
The International Comparison Program (ICP) price surveys to obtain PPPs are carried out for selected benchmark years only. Not all estimates are for the same year, so that it was necessary in Maddison (1995: Monitoring the World Economy, 1820-1992) to carry some data forward to 1990 with the use of national price indices. The precise nature of the ICP price surveys can differ across economies, principally for non-OECD countries. The ICP pricing procedures have been criticized for lack of comparability and reflection of the specified items between economies. Furthermore, the multilateral character of the estimates is affected by the fact that the PPPs are, in fact, estimated for six different regions, and “globalized” with particular interregional (binary) links. Finally, within each of the regions, the aggregation procedures of the PPPs differ. For example, for 1990 the country PPPs within the European Union are unweighted for size of GDP (using the so-called EKS procedure), whereas the PPPs for non- European OECD countries are combined with those for the European Union and weighted for size of GDP. Even though the industry by origin PPPs for manufacturing, transport and communication and wholesale and retail trade are assumed to be a proxy of relative producer prices, the comparability of these measures suffers from biased sample coverage. Moreover, due to the “unit value” characteristics of part of the information, the method takes, in many cases, insufficient account of quality differences across economies.
Estimates of employment are, as much as possible, for the average number of persons with one or more paid jobs during the year. Particularly for low- and middle-income economies in Asia and Latin America, statistics on the number of self-employed and family workers in agricultural and informal manufacturing activities are probably less reliable than those for paid employees. As in the case of output estimates, the employment estimates are sensitive to under-coverage of informal or underground activities, which harbour a substantial part of labour input. In some cases, informal activities are not included in the production and employment statistics at all. In agriculture the labour force estimates include a substantial part of (part time and seasonal) family workers. However, the estimates presented for the economies in this data set are meant to cover all economic activity.
ILO gathers data to estimate the indicators from international data repositories managed by various international organisations. It rarely collects information directly from national sources.
The estimates for the aggregate economy are derived from the World Development Indicators (WDI) Database of the World Bank. Complete documentation of sources and methods by country and underlying documentation on the use of PPPs, etc. can be downloaded from the website of the World Bank.
The aggregate economy estimates for OECD countries, most of which are included in the tables under the headings of “Developed Economies & European Union”, and GDP (after 1990) are mostly obtained from OECD: National Accounts, Volumes I and II (annual issues) and The Statistical Office of the European Communities (Eurostat) New Chronos database. A. Maddison: The World Economy: Historical Statistics (Paris, OECD Development Centre, 2001) has been extensively used to cover the period 1980-1990. Employment estimates for the aggregate economy are mostly taken from OECD: Labour Force Statistics (annual issues), Eurostat and the Bureau of Labor Statistics (BLS): Comparative Civilian Labour Force Statistics.
For other countries outside of the OECD, the national accounts and labour statistics assembled from national sources by international organizations such as the World Bank, the Asian Development Bank, the Food and Agriculture Organization (FAO), the ILO and the United Nations Statistical Office, are mostly taken as the point of departure. In individual cases use has also been made of national accounts statistics. The total economy series are linked to a benchmark estimate of GDP at market prices in US dollars for 1990 from Maddison (2003, op. cit.). Maddison’s dollar estimates are based on purchasing power parities for GDP. The original PPPs were obtained from the ICP.
The PPPs for the total economy used by Maddison represent multilaterally weighted PPPs. Multilateralization implies that the weights of all economies are used to obtain the aggregate PPPs, which makes comparisons between economies fully transitive, i.e. comparisons between economies A and B and economies B and C equal a comparison between economies A and C. The year 1990 was chosen because it is still the latest for which a reasonably comprehensive and reliable set of PPPs can be obtained for a largest possible range of economies in the world economy. The multilateral weighting system for the aggregate economy is the Geary-Khamis system, which essentially weighs PPPs for each country on the basis of its relative size in terms of GDP.
Whenever information was available from more than one repository, the information and background documentation from each repository was reviewed in order to select the information most suitable for inclusion, based on an assessment of the general reliability of the sources, the availability of methodological information and explanatory notes regarding the scope of coverage, the availability of information by sex and age, and the degree of historical coverage. Occasionally, two data repositories have been chosen and presented for a single country; any resulting breaks in the historical series are duly noted.
For countries with less-developed labour market information systems, such as those in the developing economies, information may not be easily available. Many of these countries, however, do collect labour market information through household and establishment surveys, population censuses and administrative records.
Limitations to comparability are often indicator-specific; however, there are standard issues that require attention with every indicator. For example, the precision of the measurements made for each country and year, and systematic differences in the type of source, related to the methodology of collection, definitions, scope of coverage and reference period, will certainly affect comparisons. In order to minimize misinterpretation, detailed notes are provided that identify the repository, type of source (household and labour force surveys, censuses, administrative records, and so on), and changes or deviations in coverage, such as age groups and geographical coverage (national, urban, rural, capital city and so on).