3. Cross-country comparisons of labour productivity levels
Productivity is a key source of economic growth and living standards. In this chapter the focus is on labour productivity levels, which are widely used to assess convergence across countries.
The two main components of labour productivity: Gross Domestic Product (GDP) or Gross Value Added (GVA) and hours worked (or when the latter is not available, employment) are discussed in turn.
Gross Domestic Product (GDP) is a widely used measure of output in the compilation of productivity indicators. It measures the value added generated by an economy, i.e., the value of goods and services produced during a given period, minus the value of intermediate consumption used in the production process. Countries measure GDP in their own currencies. To compare these estimates across countries, they have to be converted into a common currency. The conversion is often made using nominal exchange rates, but these can provide a misleading comparison of the volume of goods and services produced across countries. A better approach is to use Purchasing Power Parities (PPPs), which are currency converters that control for differences in price levels between countries and so allow for correct international comparisons of the volume of GDP and of the size of economies (Eurostat-OECD, 2024[1]).
Key findings
When using PPPs rather than exchange rates as currency converters to US Dollars (USD), the OECD economies together accounted for about 45% of the world GDP in 2022 (Figure 3.1). China (around 18% of world GDP) and India (around 7%) were the largest non-OECD countries.
The United States accounted for the largest share (around one third) of PPP converted GDP in 2022 in the OECD area, followed by Japan, Germany, France, the United Kingdom, Türkiye, Italy and Mexico. The top 3 OECD countries together accounted for about a half of OECD total.
How to read the indicators
The compilation of GDP is based on harmonised accounting concepts and definitions that ensure its comparability across countries. In practice, however, the measurement of GDP can be affected by three main issues:
The measurement of the non-observed economy. An exhaustive coverage of production activities can be difficult to achieve in some countries and national estimates may differ in their coverage of non-observed activities. The size of the non-observed economy is generally larger in emerging-market and developing economies reflecting, in part, the higher degree of informal activities and employment.
International production arrangements. In the last decades, globalisation has led to a fragmentation of production processes across countries. In some cases, national accounts record output in the country where intellectual property (IP) assets are located rather than in the country where output is physically produced (e.g., in the case of contract manufacturing). This can lead to a disconnection between GDP and production factors, as well as to changes in GDP due to the relocation of IP assets from one country to another. Moreover, some of the income generated by IP assets may be ultimately transferred abroad. This can happen, for example, when IP assets are located in the balance sheets of affiliates of multinational enterprises which ultimately transfer the related benefits to their parent company (UNECE, 2015[2]). Gross National Income (GNI) is a measure reflecting total income of agents (excluding capital gains and losses) residing in a country, i.e. it accounts for income received by resident agents from abroad and deducts income generated by local production that is transferred to agents residing abroad.
The measurement of the digital economy. The digital transformation also poses many challenges to the measurement of the production of goods and services and hence GDP. The emergence of new digital services, the increasing scale of peer-to-peer interactions through digital intermediary platforms, the development of “free” services blurring distinction between consumers and producers, are only a few examples of the challenges currently faced by national accountants (Ahmad and Schreyer, 2016[3]) (Ahmad, Reinsdorf and Ribarsky, 2017[4]) (UNECDE, 2023[5]). Moreover, shorter cycles of market entering and exiting of ICT products exacerbate long standing challenges on the distinction between price movements and quality increases (Aeberhardt et al., 2020[6]).
When it comes to the measurement of GDP in volume or real terms (i.e. excluding the impact of inflation), the 2008 System of National Accounts (2008 SNA) recommends the production of estimates based on annually chain-linked volume indices. Most countries covered in the report derive annual estimates of real GDP using annually chain-linked volume indices (i.e. updating every year the prices used to measure volume indices). The United States and Canada use chain-linked Fisher indices while other OECD countries use the chain-linked Laspeyres ones. However, Mexico and South Africa currently produce fixed-base volume indices (i.e. measuring volume indices at the prices of a fixed given period) with the base year updated less frequently.
For further methodological information, consult the OECD Productivity Statistics – Methodological notes at https://www.oecd.org/sdd/productivity-stats/OECD-Productivity-Statistics-Methodological-note.pdf.
In productivity analysis, the volume of labour input is most appropriately measured by the total number of hours actually worked, i.e. hours effectively used in production, whether paid or not. The use of total hours worked accounts for variations in working time patterns (e.g. part-time or full-time employment) and employment legislation (e.g. statutory working hours) across countries and over time that can affect the comparability of total employment figures. However, total employment (i.e. the number of persons employed) is often used as a proxy for labour input, particularly when data on total hours worked cannot be estimated.
The relevant concept for measuring labour input is hours actually worked, as opposed to hours paid, contractual hours, or usual hours worked. Hours actually worked reflect regular hours worked by full-time and part-time workers, paid and unpaid overtime, hours worked in additional jobs, excluding time not worked for reasons such as public holidays, annual paid leave, sick leave, maternity leave, strikes, bad weather and economic conditions.
Key findings
The United States accounted for about one quarter of both total hours worked and total employment in 2022, the largest shares in the OECD area (Figure 3.2 and Figure 3.3). However, the ranking of countries in terms of their share in total labour input depends on the measure of labour input used, i.e. hours worked or employment.
Estimates of average hours worked per worker differ substantially across countries. While some countries recorded more than 2000 hours worked per worker in 2022 (such as Colombia, Mexico, Costa Rica and Poland), others recorded less than 1500 hours (Denmark, Germany, Iceland, Luxembourg, the Netherlands and Norway) (Figure 3.4).
Differences in average hours worked per worker across countries partly reflect structural differences in the organisation of labour markets. Differences in the method used to measure hours can also play a role in explaining these differences (Ward, Zinni and Marianna, 2018[7]) (see How to read the indicators for further details).
How to read the indicators
The use of different sources may affect the comparability of labour productivity levels, but comparisons of labour productivity growth are less likely to be affected. In most countries, the main source to construct measures of hours actually worked is the labour force survey. However, many countries rely, only or in addition, on establishment surveys and administrative sources.
Computing estimates of hours worked also implies adjusting the activities covered by employment and hours worked to those covered by the output measure. This requires excluding resident persons working in non-resident production units and including non-resident persons working in resident production units in geographical and economic boundaries of employment and hours worked.
In practice, countries adopt one of two methods to estimate actual hours worked for productivity analysis:
the direct method, which takes actual hours worked self-reported by respondents in surveys, generally labour force surveys (LFS);
the component method, which starts from contractual, paid or usual hours per week from establishment surveys, administrative sources or the LFS, with subsequent adjustments for absences and overtime, and other adjustments to align hours worked with the concepts of hours actually worked and the concept of domestic output.
The direct method is relatively simple, but it depends heavily on respondent recall, cannot account for response bias, and assumes perfect alignment of measures of workers and output. The component method systematically attempts to address these issues, though it is more complex. Response bias and insufficient adjustments to align with the concept of domestic output can lead to systematic upward biases in estimates of average hours worked per worker based on the direct approach, compared to the component approach (Ward, Zinni and Marianna, 2018[7]).
The OECD simplified component method assumes that workers in all countries take on average all the leave to which they are entitled. However, actual take-up leave rates are likely to reflect differences in working cultures across countries. In addition, the national statistics offices may have access to a wider variety of national data sources. As a result, the OECD simplified component method estimates can be considered only as a stopgap for those countries currently using a direct approach with minimal or no adjustments, while these countries work towards improving their methodologies.
The effective quantity of labour input depends not only on the total number of hours actually worked, but also on the education, working experience, business functions and other workers’ characteristics. The measure of labour input used in this publication, i.e. total hours worked, does not account for the composition or “quality” of the workforce and likely underestimates the effective contribution of labour to production.
For further methodological information, consult the OECD Productivity Statistics – Methodological notes at https://www.oecd.org/sdd/productivity-stats/OECD-Productivity-Statistics-Methodological-note.pdf.
Labour productivity is the most frequently computed productivity indicator. It represents the volume of output produced per unit of labour input. The ratio between output and labour input depends to a large extent on the presence of other inputs, such as physical capital (e.g. buildings, machinery and transport vehicles) and intangible assets used in production (e.g. intellectual property assets), technical efficiency and organisational change.
Intangible assets play an increasingly important role in economic growth and productivity. Several important intangible assets are part of measured capital, in particular research and development, software and intellectual property products. There are measurement challenges related to the recording of capital services from intellectual property assets consistent with the location where output is produced. Intellectual property assets may also give rise to large income transfers between the countries where they are registered, and those of their ultimate owners, thus leading to a large gap between GDP and GNI (Gross National Income; see the section on the Size of output). In such cases, measures of GNI per hour worked can complement measures of GDP per hour worked.
Key findings
There are large disparities in labour productivity levels across countries, including within the OECD area. Measured as GDP per hour worked in PPP terms, average labour productivity in the OECD area was slightly above USD 67.5 per hour in 2022, with a standard deviation across countries of about USD 32. Labour productivity was more than twice the OECD average in Ireland and Norway, and about one third of the OECD average in Mexico and Colombia (Figure 3.5).
Labour productivity levels across OECD countries have converged since 2000, especially among the catching-up countries. Most of the countries with labour productivity levels below the OECD average in 2000 have caught up considerably since then. Labour productivity levels in Canada, France, Italy, the Netherlands and the United Kingdom were still above average, but closer to the OECD average in 2022 than in 2000. However, the gap with the OECD average increased for Greece, Israel, Japan, Mexico and New Zealand over the last 20 years (Figure 3.6).
In most countries, GDP per hour worked and GNI per hour worked are similar, as the underlying income flows are relatively small or offset each other. Ireland, Luxembourg and Norway, on the other hand, show significant differences between measures based on GDP and GNI, reflecting the important role of multinationals in output and income transfers. In such cases, measures using GNI are useful complements to measures based on GDP (Figure 3.7).
How to read the indicators
Following national accounts standards, and consistently with the measure of output, the measure of labour input in an economy includes the contribution of cross-border workers working in resident production units. Conversely, it excludes all persons working in non-resident production units. Depending on the original data sources used to estimate employment (e.g. labour force survey, administrative data, business statistics), various adjustments are needed to ensure consistency between labour and output measures.
In the above charts, national accounts data on hours worked for Austria, Estonia, Finland, Greece, Latvia, Lithuania, Poland, Portugal, Sweden and the United Kingdom have been replaced with estimates obtained with the OECD simplified component method described in the section on Hours worked. However, the impact of this correction on labour productivity growth rates is marginal (Ward, Zinni and Marianna, 2018[7]).
Some countries can be classified as investment hubs (with a relatively high stock of foreign direct investment). In this case, the difference between GDP and GNI of the hub country depends on whether enterprise headquarters are located in the country or not. If an affiliate is established in an investment hub but headquarters remain abroad, GNI should not be affected by profit shifting behaviour. Conversely, if headquarters are set up in the investment hub whose profits are artificially inflated, GNI will remain high, in line with GDP, unless profits are actually transferred abroad as dividend payments – then GNI would be reduced (Deaton and Schreyer, 2021[8]).
For further methodological information, consult the OECD Productivity Statistics – Methodological notes at https://www.oecd.org/sdd/productivity-stats/OECD-Productivity-Statistics-Methodological-note.pdf.
OECD Economic Outlook: Statistics and Projections (database), https://doi.org/10.1787/eo-data-en.
OECD Employment and Labour Market Statistics (database), https://doi.org/10.1787/lfs-data-en.
OECD National Accounts Statistics (database), https://doi.org/10.1787/na-data-en.
OECD Productivity Statistics (database), https://doi.org/10.1787/pdtvy-data-en.
References and further reading
[6] Aeberhardt, L. et al. (2020), “Does the Digital Economy Distort the Volume-Price Split of GDP? The French Experience”, INSEE, Economics and Statistics N° 517-518-519, 2020, https://www.insee.fr/en/statistiques/4770160?sommaire=4770271.
[4] Ahmad, N., M. Reinsdorf and J. Ribarsky (2017), “Can Potential Mismeasurement of the Digital Economy Explain the Post-Crisis Slowdown in GDP and Productivity growth?”, OECD Statistics Working Papers, OECD Publishing, Paris.
[3] Ahmad, N. and P. Schreyer (2016), “Measuring GDP in a Digitalised Economy”, OECD Statistics Working Papers, No. 2016/7, OECD Publishing, Paris, https://doi.org/10.1787/5jlwqd81d09r-en.
[8] Deaton, A. and P. Schreyer (2021), “GDP, Wellbeing, and Health: Thoughts on the 2017 Round of the International Comparison Program”, Review of Income and Wealth, Vol. 68/1, pp. 1-15, https://doi.org/10.1111/roiw.12520.
[1] Eurostat-OECD (2024), “Eurostat-OECD Methodological Manual on Purchasing Power Parities”, 2024 Edition.
[5] UNECDE (2023), “BPM7 Chapter 16/2025 SNA Chapter 22. Digitalization: Annotated Outline, in Towards the 2025 SNA”, https://unstats.un.org/unsd/nationalaccount/aeg/2022/M21/SNA_AO_Ch22_BPM_Ch16.pdf.
[2] UNECE (2015), “Guide to Measuring Global Productio”, https://unece.org/fileadmin/DAM/stats/publications/2015/Guide_to_Measuring_Global_Production__2015_.pdf.
[7] Ward, A., M. Zinni and P. Marianna (2018), “International productivity gaps: Are labour input measures comparable?”, OECD Statistics Working Papers, No. 2018/12, OECD Publishing, Paris, https://doi.org/10.1787/5b43c728-en.