This chapter focuses on the data and indicators available in Thailand for productivity analysis, including their production, dissemination and use to conduct analysis. It focuses separately on macro- (economy and sectoral) and micro-level (firm) data and analysis. While Thailand has already a broad set of productivity indicators at the macro-level, the chapter provides recommendations to further enhance their measurement, dissemination and analysis. Some pro-productivity institutions also conduct productivity analysis based on micro-data. However, data limitations hinder the analysis, including the lack of a common business identifier across sources, and the limited exchange and use of administrative data. The chapter outlines steps to further strengthen Thailand’s micro-data capabilities, considering institutional responsibilities and data ownership. It provides recommendations to facilitate data exchanges between institutions and with external researchers, enhance data linkages, and develop a Statistical Business Register.
Strengthening Productivity Analysis for Policymaking in Thailand
Abstract
Productivity captures how efficiently inputs of production are combined to produce output. It is the main source of increasing living standards in the long run. The measurement and analysis of productivity are, therefore, crucial for informing policy, but they face two main challenges.
First, productivity is not an observable quantity. Its measurement relies on the availability and quality of data on both output and inputs to production. In addition to data, methodological choices are required to surmount the problem of non-observability and compute reliable productivity statistics. Second, productivity is a multifaceted phenomenon determined by a host of factors, which are challenging to disentangle. Productivity is commonly thought to reflect technological change. However, depending on its measurement, productivity can also capture organisational efficiency and management, market structures, the composition and quality of input factors and the intensity with which they are deployed. Aggregate productivity also depends on the ability of the economy to shift resources to their most productive use.
Assessing a country’s productivity performance and its drivers requires combining the full array of available data and theory to address these two core challenges. It calls for insights at different levels of granularity. Monitoring aggregate productivity performance and trends at the whole economy and sectoral levels is key for diagnosing the potential for long-term economic growth and understanding the influence of a country’s economic structure and industry composition. Additionally, it is essential to analyse more granular data to understand how productivity dynamics are shaped by firm-level productivity, the allocation of resources and market shares, and the process of creative destruction.
This chapter reviews the data and indicators available in Thailand for productivity analysis, the public institutions that produce and disseminate them, and their uptake in analytical work. The purpose is to inform the availability and quality of data and analysis at aggregate and granular establishment- and firm-levels and discuss the potential of current productivity data and analysis to guide policy. This assessment is used to provide guidance for enhancing both the data infrastructure and its analytical use, building on international practices.
When discussing the data, measurement and analysis, this chapter looks at productivity on three levels: (i) aggregate/economy-wide, (ii) sector/industry-level, (iii) firm-level. The chapter is organised as follows.
The first and second sections focus on data and analysis at the aggregate and sector/industry levels. More specifically, the first section outlines the Thai institutions engaged in the production and dissemination of productivity indicators at the macro-level. It then assesses the relevant indicators available in Thailand for analyses at the aggregate (total economy) and sector/industry levels. Finally, it outlines ways to leverage the existing data to enrich productivity analysis, as well as avenues for improving and expanding data dissemination and productivity measurement. The second section provides recommendations (derived from the assessment presented in the first section) to enhance the uptake and analysis of productivity indicators at a macro-level.
The third and fourth sections focus on data and analysis at the more granular firm or establishment level. More specifically, the third section first reviews existing data sources to study productivity at a micro-economic level. It then provides suggestions to address data limitations and foster the dissemination and use of micro-data. Finally, the third section of this chapter reviews the concepts and indicators that allow firms’ performance and dynamics to be linked to industry-level and aggregate productivity. To conclude this chapter, the fourth section provides recommendations (derived from the assessment presented in the previous section) enhancing the data offered for productivity indicators based on firm- or establishment-level data, their analysis, and their dissemination.
This section first reviews the landscape of institutions providing productivity indicators at the macro-level (economy, sectoral/industry and, to some extent, regional levels) and the solutions they use to share and disseminate data. Then it reviews the indicators available for productivity analysis at the aggregate (whole economy) level as well as more disaggregated sector/industry levels. Finally, it discusses current productivity analyses in light of a framework which links aggregate performance to sector/industry dynamics.
Several institutions provide productivity indicators and dissemination tools
National Accounts data, that build on a large set of primary and secondary sources, are usually the main source for productivity measurement at the macro-level (economy-, industry- and regional- level). National Accounts data are reliable, readily available and provide coherent and consistent statistics.1 They are based on international standards and recommendations on how to compile measures of economic activity (SNA, 2009[1]; APO/OECD, 2021[2]).2
Mainly three Thai institutions engage in publishing key productivity indicators at the macro-level.3 The National Economic and Social Development Council (NESDC) releases productivity measures in its annual Capital Stock report. The Thailand Productivity Institute (FTPI) does the same in an online dashboard, and the Bank of Thailand (BOT) makes productivity statistics available on its website.
These institutions primarily produce and disseminate productivity indicators, but some also conduct analysis. The NESDC provides capital stock and GPD data and analyses the capital stock development and supply-side contributions to growth (labour, capital and MFP). The BOT compiles and disseminates several indicators of labour productivity.4 FTPI does not directly conduct macro-level analysis but rather provides a repository for productivity data through its dashboard, collecting information from national sources (NESDC and the National Statistical Office) but also from international databases (such as the APO and the International Institute for Management Development (IMD)).5
Note: The Bank of Thailand shares data through the Government Data Catalogue, but does not share productivity indicators.
Source: OECD elaborations.
The government has recently established the Government Data Catalogue (GDC) to help disseminate the data collected by Thai institutions. The GDC is an online data portal, that provides information (in the form of a catalogue) about existing resources from several public institutions. The GDC enables the dissemination and access of publicly available data in several ready-to-use formats (for instance csv, JSON, etc., also accessible through an Application Programming Interface (API)). The GDC also contributes to the diffusion of metadata and some institutions use it to publicise their reports. The GDC currently includes data collected from almost 300 organisations in Thailand (including ministries and NSO). However, some key productivity indicators are currently not available on the GDC, such as the multifactor productivity indicators from the NESDC and labour productivity indicators from the Bank of Thailand.
Several Thai institutions also allow users to download publicly available data through their websites, including the BOT, which also provides an API, the NESDC, although MFP measures are not available, and the NSO.
Table 2.1 summarises the main options for sharing macro-data currently used by Thai institutions.
Among non-Thai institutions, the Asian Productivity Organisation (APO) is a key source of productivity indicators and analysis for Thailand. The APO produces and disseminates productivity data through the annual APO Productivity Databook. It provides all the key measures for an in-depth analysis of aggregate productivity, together with methodological details and metadata, and also conducts relevant cross-country analysis (APO, 2023[3]).6 In addition to standard productivity measures, the 2023 edition of the Productivity Databook analyses specific thematic issues and factors related to productivity (such as global value chain participation, trade and foreign direct investments, or the role of structural changes for energy productivity improvements). The APO data therefore provides useful resources that Thai pro-productivity institutions could use for their own productivity analyses. Some indicators from the APO are already presented on the FTPI dashboard.
While Thailand generally adheres to and meets international dissemination standards, Thai institutions could at times enhance the availability, dissemination and documentation of productivity indicators.7,8 This would offer simpler, cost-effective improvements with potentially significant benefits for research and analysis on productivity.
In particular, the accessibility of productivity indicators could be enhanced. For instance, further to missing important productivity indicators, the GDC does not offer easy identification and search of productivity indicators (e.g. through a tag specifically for productivity metrics, following the functionalities offered for other indicators). Other data repositories provide useful but incomplete dissemination of productivity indicators. For instance, the FTPI dashboard could enhance the time coverage and timeliness of the data series provided (currently mostly limited to 2017-2021) and broaden the range of productivity statistics made available. The FTPI dashboard displays relevant indicators from other external institutions but could integrate additional data, such as productivity indicator for services from BOT or the decomposition of productivity growth by the APO. Additionally, important indicators, such as multifactor productivity measures from the NESDC, are only summarised in annual reports and are not available as downloadable time series.
In addition, Thai institutions could enhance their communication of metadata and methodology used to compute the productivity indicators. For instance, data providers, such as the NESDC or the Bank of Thailand, could further expand the description of data sources and methodologies, particularly regarding the measurement of key variables used to compute productivity (e.g. labour input) and model parameters used to compute MFP (e.g. income share).
Key indicators are already available for productivity analysis, but relevant ones are missing
Several measures of productivity are used by international organisations and national statistical agencies. Productivity is computed as a residual after subtracting the influence of different factors for a given level of output. The number of production inputs considered and the refinements in the measurement of the required series determine the complexity, use and interpretation of these measures. Box 2.1 provides a brief overview of various productivity measures used internationally. More detailed presentations of concepts, methodology and measurements can be found in (OECD, 2001[4]) and (APO/OECD, 2021[2]). The discussion below provides an overview of various productivity measures available in Thailand, as well as their use and interpretation, building on (OECD, 2001[4]) and (APO/OECD, 2021[2]). Table 2.3 summarises their availability.
Labour productivity computed as value-added per hour worked is generally considered the most relevant measure of how productively labour is used to generate value added. It is particularly relevant to control for cyclical fluctuations in hours, which may play a critical role during recessions and recoveries. It also enables more relevant international comparisons given the large differences in the average number of yearly hours worked per person across countries. It is computed for Thailand by both the BOT, which publishes a quarterly labour productivity index per hour worked for 21 disaggregated industries, and by the APO at the aggregate and sector levels. FTPI also disseminates these data, displaying data on labour productivity per hour worked by sectors and regions. A possible refinement to labour productivity measures based on the volume (number of workers or hours worked) of labour inputs, consists in accounting for the heterogeneity of skills among workers. Indeed, the volume of employment or the number of hours worked per worker does not fully reflect the total contribution of labour inputs to output. Increases in labour quality over time may also contribute significantly to labour productivity growth. To this aim, Quality-Adjusted Labour Input (QALI) indexes (labour services) have been developed and provide a more accurate picture of the contribution of the accumulation of human capital to economic growth (see Box 2.3). Importantly, Thailand has significantly improved the level of education of its citizens over the past 50 years, like other APO economies (OECD/APO, 2022[5]), making these QALI indexes a particularly relevant metric.
Capital productivity measures how productively capital is used to generate value added. It is symmetrical to labour productivity and reflects, after subtracting the contribution of capital, the joint role of labour, as well as technical and organisational efficiencies, but also economies of scale, and the intensity of use of input factors. The NESDC presents detailed information on the capital stock and reports capital productivity growth at the level of sector aggregates (agriculture, manufacturing and services). Currently, the approach of the NESDC is to measure capital stocks in an income and wealth perspective. Possible refinements for productivity analysis would be to measure productive capital stocks and capital services (see Box 2.2, OECD (2001[4]), OECD (2009[6]), APO/OECD (2021[2])).
Multifactor productivity (MFP), a more sophisticated indicator, relates a measure of output to several inputs and typically considers how productively an economy or firm combines labour and capital inputs jointly to generate value added. While labour productivity and capital productivity are obtained simply as a ratio of the output and input measures, computing multifactor productivity requires combining the two (or more) measures of inputs to compute MFP as a residual, i.e. after subtracting the contribution of both labour and capital.9 Such measures of value-added-based MFP are available for analysis in Thailand from the NESDC. The Capital Stock report of the NESDC uses a standard accounting framework (see also Box 2.1) breaking down GDP growth into the contribution of factor inputs, namely labour, capital (as well as land for agriculture), and the residual contribution of MFP (NESDC, 2022[7]). Such analysis is performed at the macro-sector level, notably for agriculture, manufacturing, and services. The APO also provides MFP indicators at the economy level, with corresponding information for other Asian countries allowing international comparisons.
Productivity generally refers to the part of output which is not directly explained by the quantity of inputs used for production. The OECD Manual on Measuring Productivity (2001[4]) provides a key reference for computing and interpreting productivity at aggregate (economy-wide) and industry levels. This methodology is the basis of the OECD Compendium of Productivity Indicators (OECD, 2019[8]).
While productivity is defined as a ratio of the volume of output to the volume of input used for production, analysts commonly define and use several productivity measures. These measures differ according to the input factors considered as well as the measure of output. These are summarised in Table 2.2 (OECD, 2001[4]). Statistical agencies generally publish productivity measures in the form of (chained) indices or growth rates reporting the evolution of productivity, usually focusing on value-added labour productivity, capital-labour MFP based on value-added and/or KLEMS-MFP.
MFP is generally computed by statistical agencies using a growth accounting methodology. In the case of value-added-based MFP, a change in MFP is computed as follows:
Where VA refers to real value added, K to real capital input, L labour input and and are share of costs of inputs, averaged over two periods and refer to one period changes.
Despite well-established concepts for measuring productivity indexes, differences in the quality and refinements of measures still exist. These differences are linked to challenges in measuring individual components of productivity indicators. The APO/OECD report on productivity statistics (2021[2]) reviews the most pressing challenges.
Different methodologies in the measurement of output and inputs (labour, capital) thus affect MFP estimates. For instance, large informal sectors or the growing importance of the digital economy affect the measurement of output. Challenges and refinements for the measurement of production inputs also influence the accuracy of productivity statistics. Common measures of labour inputs may rely on the total volume of employment or hours worked, but the measurement of employment in terms of jobs or persons, discrepancies with National Accounts, or the measurement of self-employment and informal employment can affect productivity measurement. In addition, more sophisticated measures of labour input may also account for changes in the composition of labour to compute a more accurate index of productivity (see Box 2.3). Regarding capital inputs (APO/OECD, 2021[2]), two measures of MFP exist based on capital stocks or capital services. Productivity analysis should ideally be based on capital services (see Box 2.2).
Depending on the purpose, capital can be seen as a measure of wealth or as an input factor in production. This leads to two distinct and widely used methodologies to calculate capital: the “income and wealth” perspective and the “production and productivity” perspective. For productivity analysis, OECD guidelines recommend using the second methodology, which consists of measuring the flow of capital services as a measure of capital stock.
Wealth capital stock. This approach focuses on measuring capital as the stock of assets surviving from past periods, accounting for depreciation. It is a measure of the value of the capital stock accumulated (net of depreciation) and therefore a measure of the “wealth” of the assets’ owner at a point in time. A measure of aggregate capital stock is obtained by summing estimates for different types of assets, using market prices as aggregation weights.
Capital services. This approach focuses on measuring the productive potential of capital and estimating the flow of capital services, which is the actual capital input used in production. To illustrate the difference to the wealth approach, OECD (2019[8]) offers the example of a taxi. The capital services provided by the taxi relate to the number of trips, distance driven, and comfort of the taxi, and would be measured at the rental price (which is often not observed for many types of assets) rather than the value of the vehicle, which would instead relate to the net wealth capital stock concept (OECD, 2019[8]). Measuring capital services is challenging, because flows of the quantity of capital services are not usually directly observed, and have to be estimated. The assumption is that the flow of capital services is proportional to the productive stock of an asset class. The productive stock measures the stock of assets, corrected for efficiency losses over time and retirements.
The aggregation of capital stock is more challenging in the production perspective than in the wealth approach. Jorgenson (1963[9]) and Jorgenson and Griliches (1967[10]) were the first to develop aggregate capital service measures that take the heterogeneity of assets into account. They defined the flow of quantities of capital services individually for each type of asset, and then applied asset-specific user costs as weights to aggregate across services from the different types of assets. While for some assets the user cost corresponds to the rental prices of capital goods, for others the rental transactions are not observed. In the latter case rental prices have to be imputed. For more information about the methodologies of calculating capital services see the OECD manual on measuring capital (OECD, 2009[6]) and the APO/OECD report for productivity statistics (APO/OECD, 2021[2]).
Enhanced measures of multifactor productivity can be obtained when refining the measurement of input factors. For instance, a possible refinement of multifactor productivity consists of using the quality-adjusted measures of labour inputs, as previously discussed for labour productivity. Accounting for these changes in the workforce is particularly relevant to assess the sources of growth and further disentangle the role of technology-driven productivity from other factors, notably the role of human capital accumulation. Such refined measures are available from the APO at the economy level, which provides the basis for a deeper analysis of multifactor productivity, as also presented in the report by the Ministry of Labour (MOL, 2022[11]). Another key refinement in the measurement of MFP consists in measuring the contribution of capital to production through estimates of capital services, based on measures of the productivity stock of capital (see Box 2.2). Finally, another refinement of MFP measure consists of computing KLEMS MFP, based on gross output and accounting for intermediate inputs (such as energy, materials and services used for production) in addition to labour and capital.
Overall, it appears that Thai institutions are producing relevant measures of productivity. Notably, labour productivity (not only based on the number of workers but also on hours worked, which is one of the preferred options according to international practice), capital productivity and basic measures of multifactor productivity using a standard accounting framework widely used across countries and international organisations.
Note: time periods refer to the latest available reports or analysis accessed at the time of writing. IMD stands for International Institute for Management Development. APO for Asian Productivity Organization.
Source: OECD elaborations based on information available online from the Office of the National Economic and Social Development Council (NESDC), the Thailand Productivity Institute (FTPI) and the Bank of Thailand (BOT).
Possible avenues for refinements include, among others, developing a measure of capital services to compute multifactor productivity (instead of a measure based on a wealth approach of capital stock relevant to National Accounts), as well as an index of quality-adjusted labour input index. While some of these refined measures are already available from other institutions, such as the APO, integrating these measures into the set of indicators produced by Thai statistical agencies still represents a key target for Thailand in the medium term. A recent analysis by the NESDC (2024[12]) demonstrates that Thailand is further developing its methodology for measuring productivity in a way that aligns with international standards and best practices. The NESDC is exploring refined productivity measures for Thailand, including measures that account for labour quality, as well as a KLEMS measure that accounts for intermediate inputs. Their study further clarifies the methodology used in current productivity measurement and discusses current challenges to the broad implementation of KLEMS measures, particularly concerning the availability of data on intermediate inputs. The NESDC analysis (2024[12]) serves as a key reference that aligns well with the diagnostics and recommendations of this chapter.
Next, the report explores how, over a shorter time horizon, Thai pro-productivity institutions can leverage existing indicators more effectively to enhance productivity analysis.
One of the most appropriate and internationally comparable measures of labour input in productivity measurement is the total number of hours worked by all persons engaged in production, including employees as well as self-employed workers. However, such measures only account for the volume of input, ignoring the heterogeneity across workers in terms of skills, education, and occupations.
A more refined approach is to use quality-adjusted labour input (QALI, also called CALI- composition adjusted labour input) measures which differentiate between hours worked by different types of workers. Accounting for the composition/quality of labour input is particularly important from a policy perspective, as it allows to better disentangle productivity growth due to increases in human capital from other factors captured by multifactor productivity.
The APO productivity database provides a breakdown of labour input into hours worked and quality of labour for Asian countries, which allows to decompose the growth of labour productivity and disentangle the role of changes in the quality of labour from multifactor productivity growth.
This exercise is particularly important for Thailand as the country has significantly improved the education levels of its citizens over the past 50 years (OECD/APO, 2022[5]). Thus, while hours worked have remained mostly constant since 1990, the quality of labour has contributed to most of the increase in labour input over the past decades (Figure 2.1, Panel A).
Evidence from the APO database shows that labour quality has been a large contributor to the labour productivity growth in Thailand since 1975 (green bar in Figure 2.1, Panel B). Without accounting for the quality of labour separately, the contribution of such increases in labour quality would instead be attributed to MFP, resulting in higher MFP estimates.
Source: APO Productivity Database 2023.
Methodology and necessary data for measuring quality-adjusted labour input
The main feature of quality-adjusted indexes of labour inputs is to aggregate hours worked by different types of workers to account for their heterogeneity in terms of skills and characteristics, in particular, related to productivity. Different approaches co-exist nowadays, and they differ from each other regarding the characteristics of workers they account for. Prominent examples of quality-adjusted series include the ASIA QALI indicators, the Eurostat QALI measure, the OECD CALI index, as well as indices implemented in national contexts, for instance in Australia, Canada, Denmark or the US (see Ward and Zinni (2024[13]) for a review, as well as APO/OECD (2021[2])). The ASIA QALI database developed by the Keio Economic Observatory (Keio University) represents a key implementation of this measure, used notably in the APO Productivity Databook. It contains data for 25 Asian countries, including Thailand.
Quality (or composition) adjusted indicators consider workers' characteristics that may be related to differences in skills and productivity (“quality”), and typically rely on observable differences in terms of occupation, educational attainment, age (a proxy for experience), employment status (employed or self-employed) but also gender (which, however, does not necessarily capture productivity differences across workers).
To aggregate hours worked by different types of workers, their relative productivity should be used as weights, and this is commonly done by relying on wage differentials. More specifically, following OECD recommendations (OECD, 2001[4]) changes in hours worked by different workers can be aggregated using an index (namely a Törnqvist index based on a geometric average with time-varying weights defined as the average weights over two consecutive periods), using the labour cost shares of the different groups in total labour cost as weights.
The data on hours worked and wages by different groups of workers required to produce these adjusted measures of labour input may be demanding to compile. Hours worked and labour compensations need to be available across groups of workers according to the different characteristics considered by year and ideally industry. Such information is usually constructed from labour force surveys, census of population and National Accounts, but also other surveys and censuses (see Ward and Zinni (2024[13])).
Thailand can improve the analytical use of macro-level productivity indicators
Productivity indicators currently available in Thailand form a strong base to monitor aggregate, sector and detailed industry productivity. This report now assesses the analysis of Thai productivity in light of a framework linking aggregate and industry levels.
The rest of this section discusses the scope for extending and enhancing the use of available indicators through further analysis following the analytical framework illustrated in Figure 2.2. More specifically, further to the analysis of aggregate productivity dynamics, the report discusses possible analyses along two dimensions:
Industry composition: monitoring “between-sectors” improvements, by assessing the contribution of structural changes (shift of resources and activity across sectors) to aggregate productivity growth.
Industry productivity: monitoring “within-sectors” improvements, by assessing productivity levels and trends of individual sectors and their contribution to aggregate productivity growth.
As a starting point, productivity analyses at the aggregate level should systematically provide information on productivity developments at the economy level over long periods and, if possible, in an international perspective. The purpose is to gauge aggregate developments of productivity, highlighting long-term trends and recent short-term evolution of productivity, comparing the performance of Thailand over time and across countries. Existing annual and ad-hoc reports already provide valuable insights into productivity analysis at aggregate levels. This is the case for instance of the NESDC analysis in its Capital Stock report as well as the ad-hoc reports from the World Bank and Bank of Thailand (World Bank and Bank of Thailand, 2020[14]) and the Ministry of Labour (MOL, 2022[11]). Additional evidence from international organisations, notably in reports from the APO (APO, 2023[15]) or the OECD (OECD, 2023[16]; OECD, 2024[17]), complement the assessment of Thailand’s productivity. These reports from Thai institutions and international organisations provide an overview of the evolution of labour and multifactor productivity in the form of indexes or growth rates, yearly or on average over sub-periods. Such indicators have revealed the slowdown in Thai productivity growth and remain key to assess future developments. Taking advantage of existing indicators, changes in labour productivity can be further decomposed into the contribution of changes in labour quality, capital deepening (possibly focusing on different types of capital), as well as the contribution of MFP growth, as also recently implemented by the NESDC (2024[12]).
Source: OECD elaborations.
While such economywide productivity dynamics are key metrics to assess the performance of the Thai economy and the drivers of economic growth, a more refined analysis should also investigate the role of industries in driving aggregate productivity changes, through the productivity growth of individual industries and changes in their economic weight.
Examining the performance and economic weight of sectors and their contribution to productivity can help tailor policies and make informed decisions on where to spur the allocation of resources. It allows identifying strengths and weaknesses across sectors, but also assessing the effect of long-term changes in the structure of the economy and the shift of resources across sectors with different levels of productivity (i.e. structural change). In other words, it sheds light on whether within-sector changes (within-effect) or structural changes (between-effect) drive aggregate productivity growth.
This analysis can be implemented at a broad sectoral level, but also at a more detailed industry level (for instance at a two-digit TSIC-2009 level, or more disaggregated if feasible, generally focusing on labour productivity).10 This focus on within- and between-effects is an integral part of the indicators that international organisations and national institutions monitor.11 Box 2.4 offers an example of an analysis examining the role of different industries in the productivity slowdown of the United Kingdom.
Productivity growth declined across many advanced economies over the past few decades, but the United Kingdom experienced particularly low productivity growth after the 2008 Great Financial Crisis, causing productivity to diverge relative to both pre-crisis trends and G7 countries. Coyle and Mei (2022[18]) investigate this productivity puzzle and decompose aggregate labour productivity growth into the contributions of different sectors of the economy, looking both at within-industry productivity growth and labour reallocation between sectors.
Firstly, the authors find that the UK aggregate productivity slowdown is mainly explained by a within-industry component, that is, by slow productivity growth within individual sectors, rather than the reallocation of labour to low-productivity sectors over time. The authors find that the reallocation term explains little of the aggregate productivity change over the period 1998-2015, although this contributes negatively in the US and France (and to most countries once the real estate sector is excluded).
Coyle and Mei (2022[18]) further identify more detailed industries which are responsible for the slowdown. Starting from higher levels of aggregation and dividing the economy into 20 industries (SIC07 sections A-T) and analysing pre- and post-2008 dynamics, the authors identify that the aggregate productivity slowdown can be to a large extent traced to the performance of the manufacturing and the information and communication sectors. They further refine the analysis at a more granular level (using 97 industries) and find that the transport equipment and pharmaceuticals industries contribute the most to the decline within manufacturing, while computer software and telecommunications contribute the most within information and communication. These results challenge common perceptions that sectors such as pharmaceuticals or ICT are strengths of the UK economy.
Coyle and Mei (2022[18]) also provide an insightful discussion of the relevant methodology to decompose aggregate productivity growth into the contributions of sectors. The main methodology is based on a Tornqvist decomposition, which allows output prices to differ across sectors/industries. The authors also test the robustness of the findings using a shift-share analysis and Generalised Exactly Additive Decompositions. Appendix II of their study details the differences and strengths of these approaches.
Structural changes in the sectoral composition of the economy can be a powerful driver of productivity growth that warrants close observation. To illuminate this, the report conducted jointly by the World Bank and the Bank of Thailand (World Bank and Bank of Thailand, 2020[14]) uses a “shift-share analysis” that disentangles within- and between- contributions to productivity growth. The authors estimate that the shift away from the low-productivity agriculture sector towards higher-productivity manufacturing was the main driver of productivity growth over the period 1980-98 in Thailand. Moreover, the changing composition of the economy contributed half of the labour-productivity gains over the period 1998-2008. However, this contribution declined after 2008. Warr and Suphannachart (2022[19]) and Klyuev (2015[20]) further suggest that structural change played a particularly important role for Thailand compared to other countries. The World Bank (2021[21]) uses data across 94 countries, of which 60 are emerging markets and developing economies (EMDEs), across nine sectors to study the sectoral sources of productivity growth over the period 1995-2017. Their results show that productivity gaps across sectors are still large, implying that reallocation could significantly boost productivity further.12 The large gaps point to the relevance of decomposing within- and between-effects to assess the potential benefits of policies aimed at supporting reallocating capital and labour across sectors, developing human capital and skills, or reducing market distortions and regulatory complexity that hinder firm entry, exit and growth.
Policymakers are also interested in evidence on the performance, in terms of productivity growth, of individual sectors and how this contributes to aggregate productivity growth (within-sector improvements). The decomposition of aggregate growth into the contribution of individual sectors shows that, over the most recent period, the within-sector effects have been the main driver of aggregate productivity growth.13 Such an exercise, readily implemented with the data available for Thailand, is illustrated by the decomposition of labour productivity into the contributions of nine sectors in the Thailand Country Profile of the APO Yearbook (APO, 2023[3]).14 The World Bank (2021[21]) discusses opportunities to pursue transformations of individual sectors that would support aggregate growth through a within-effect. This points to the necessity of implementing analyses of within-effects to support the monitoring of some objectives set by the 13th National Economic and Social Development Plan – for instance to support changes to the agricultural sector, or to identify and invest in high potential high value-added service industries. In this context, the analysis of the productivity growth in individual industries conducted by the Ministry of Labour (2022[11]) can serve as a basis. It could be extended to account for the weights of individual industries and compute their contribution to aggregate productivity growth.
In addition to the analysis of productivity indicators, monitoring a broad range of economic factors related to productivity such as R&D, digitalisation, human capital, financial development and competition can broaden the understanding of determinants and provide relevant and actionable policy insights (see Box 2.5).15 This aspect is further developed in Section 1 of Chapter 3, which presents an OECD framework linking pro-productivity policies to productivity determinants.
Productivity is tightly related to a host of factors. Providing evidence on an economy’s performance regarding these factors can raise the attention of policymakers towards key issues. This serves as a diagnostic of the strengths that may foster aggregate productivity performance or the weaknesses that may hinder it.
The joint work by the APO and the OECD (OECD/APO, 2022[5]) reviews possible drivers of productivity growth. It provides a framework to scope available data (or relevant gaps thereof) and analyse them in a structured way to reveal pressing policy challenges. Drawing also from the discussion of a policy agenda for higher productivity growth in OECD (2015[22]), OECD/APO (2022[5]) distinguishes three key dimensions:
Drivers that boost innovation and new knowledge and technologies (for instance research and development, digitalisation, and investment in intangible assets);
Drivers that contribute to the diffusion of existing knowledge and technologies (for instance human capital, public infrastructure);
Drivers that facilitate the allocation of resources within or between sectors and firms (for instance competition and business dynamics, globalisation, financial development).
OECD Economic Surveys illustrate how the analyses of complementary economic indicators can feed into policy recommendations. For instance, the OECD Economic Survey of Thailand (OECD, 2023[16]) identifies low foreign direct investment (FDI) relative to peer countries, more stringent FDI restrictions and tighter restrictions for services trade as important barriers to productivity improvements in Thailand. Based on these results, the Economic Survey of Thailand advocates for improvements in the conditions for foreign trade and investments.
This chapter finds that Thailand can already rely on a broad set of productivity indicators at the macro-level (at aggregate economy-wide level, broad sectors and detailed industry levels). Mainly building on National Accounts data (which in turn build on a set of primary and secondary sources), Thai institutions produce relevant measures of productivity. These measures include labour productivity (not only based on the number of workers, but also based on hours worked, which is one of the preferred options according to international practices), capital productivity and basic measures of multifactor productivity using a standard accounting framework widely used across countries and international organisations.
Despite the rich set of indicators available, the chapter uncovers a need for extending and enhancing the analysis of existing productivity indicators, their dissemination and measurement. Recommendations 4-6 propose improvements that Thai agencies can aim to implement at different time horizons. In the short term, Thai agencies may focus resources on consolidating and extending recurrent analyses of existing indicators (recommendations 4-5), while in the longer-term statistical agencies can plan further work on addressing productivity measurement issues (recommendation 6).
First, Thailand could strengthen its analysis of productivity at the macro-level. Occasional bulletins and annual reports (see recommendation 3 in the previous chapter) could more thoroughly integrate and investigate the full range of available indicators, in order to provide a more systematic and comprehensive assessment of productivity dynamics (recommendation 4). The methodology for a thorough assessment of productivity could draw on existing analyses from Thai institutions and international organisations, as well as examples from other pro-productivity institutions across OECD countries. In particular, analyses can aim to provide further international comparisons of Thailand’s performance and to understand how the performance of individual industries and the economic structure jointly determine aggregate productivity.
Second, this chapter identifies a need to strengthen the availability, accessibility and communication of productivity indicators to further support productivity analyses. Currently, Thai institutions disseminate productivity indicators on their websites and through the Government Data Catalogue (GDC). Thai institutions also sometimes disseminate rich information on productivity produced by international institutions (e.g. the Asian Productivity Organization). However, this information remains scattered and sometimes incomplete. Building on this foundation, Thai institutions should strengthen, harmonise and coordinate the dissemination of productivity indicators to ensure that users can easily access timely, updated and documented information (recommendation 5).
Third, in the medium-term Thailand can focus on enhancing the measurement of productivity (recommendation 6). Possible avenues for refinements go beyond the immediate scope of productivity and include improving the measurement of economic activity and production inputs, such as labour and capital, to address common challenges. This entails developing a measure of capital services to enhance the measurement of multifactor productivity (instead of a measure based on a wealth approach of capital stock relevant to National Accounts), as well as an index of quality-adjusted labour input index to enhance both labour productivity and multifactor productivity metrics (see also Box 2.2 and Box 2.3). Some of these refined measures are already available at aggregate levels from other institutions such as the Asian Productivity Organisation. Thailand may build on these to develop its own official measures, also at a more disaggregated level, following a methodology which is consistent with international best practices on productivity measurement.
The following sub-sections will delve into the specific recommendations.
Providing a comprehensive assessment of productivity at the macro-level
Several Thai institutions – mainly the National Economic and Social Development Council, the Bank of Thailand’ Economic Research Institute and the Thailand Productivity Institute – produce and disseminate productivity indicators at the macro level. The National Economic and Social Development Council mostly focuses on producing capital and multifactor productivity measures at the broad-sectoral level (agriculture, manufacturing, services), while the Bank of Thailand disseminates labour productivity indices (based both on the number of workers and the number of hours worked) at a detailed industry level. The Thailand Productivity Institute offers a productivity dashboard collecting information from various national and international sources, including labour and multifactor productivity measures available at economy, regional and industry levels, mainly for the manufacturing sector.
However, this relatively rich source of indicators remains somewhat scattered and reports or analyses generally do not provide a comprehensive overview of all indicators (labour, capital and multifactor productivity) across aggregate, sectoral and detailed industry levels, except some ad-hoc reports. Current annual reports also do not comprehensively capture productivity dynamics. Some reports emphasise recent productivity growth, and others present long-term developments through averages over sub-periods, which may not fully reflect the broader dynamics. Additionally, productivity metrics are rarely contextualised with international comparisons, as these are typically absent from annual reports and only appear in ad-hoc publications. As a result, while Thai agencies provide relevant evidence on aggregate productivity dynamics, there is a need to take stock of such evidence and update and extend it in a structured way and regularly.
Similarly, analyses could better link aggregate productivity to performance at more disaggregated levels, focusing on the performance of individual industries and the role of Thailand’s economic structure (i.e. the economic weight of sectors). Currently, only ad-hoc publications provide such analyses. Structuring productivity analyses with an analytical framework, as discussed in this chapter, can help provide updated evidence on the contribution of structural changes (i.e. shift of resources and activity across sectors with different levels of productivity, the between/reallocation-effect) and the contribution of individual sectors to productivity growth (within-effect). Although the integration of industry-level analysis is a key priority, extending the assessment of productivity at the rarely explored regional level could also be beneficial.
Finally, productivity analyses could further assess the performance of Thailand, at the macro level, regarding the macro-determinants or enablers of productivity growth, such as human capital, investment in R&D and financial development. This aspect is further developed in the analytical framework for pro-productivity public policies developed in Chapter 3.
Analyses conducted in the framework of the programme of work of the recommended productivity council (see recommendation 3) can address these gaps by articulating productivity analyses along four key dimensions presented below.
Recommendation 4: Integrate a thorough investigation of labour, capital and multifactor productivity indicators, in an international perspective, and at different levels of aggregations, along with an investigation of productivity determinants.
The analytical framework and programme of work developed by the secretariat could integrate four dimensions to the analysis of productivity at the macro-level. These dimensions can guide the work of individual Thai pro-productivity institutions and can structure the output (bulletins and annual reports) of the productivity council. Eventually, the annual report of the council should cover, to the extent possible, all these dimensions (see below the implementation). The purpose would be to provide, on a regular basis, a broad, comprehensive and updated assessment of Thailand’s structural developments of productivity.
Dimension 1: Monitor various indicators of productivity. To the extent possible, productivity analyses should comprehensively cover all indicators of productivity, including labour, capital and multifactor productivity. Productivity reports could routinely monitor straightforward, easily interpretable labour productivity metrics (such as value added per worker or hours worked) and capital productivity, alongside more sophisticated measures like multifactor productivity. When available, these reports should also analyse multifactor productivity based on a quality-adjusted labour input measure (see Box 2.3), and capital services (see Box 2.2), but also decompose the contribution of different types of capital (e.g. ICT vs non-ICT, tangible vs intangible). This approach allows for a more nuanced understanding of economic growth, with decompositions that help distinguish the impact of technology-driven productivity from other factors such as physical and human capital accumulation.
Dimension 2: Examine trends from a comparative perspective. Productivity analyses should examine productivity levels (when feasible) and trends over different time horizons, also in an international perspective. Thai pro-productivity institutions should inform policymakers about long-term developments of productivity as well as short-term dynamics. Productivity trends, especially when compared to relevant benchmark countries, provide key information on the structural productivity performance of Thailand. Short-term dynamics offer complementary insights into recent productivity developments. These are particularly relevant during periods of economic turmoil (e.g. the recent COVID-19 crisis and its aftermath), or times of heightened economic and social transformations (e.g. related to the imperatives of the twin digital and green transitions).
Dimension 3: Link aggregate performance to more disaggregated levels. Productivity analyses should dissect aggregate productivity performance based on more granular insights, in particular linking aggregate economy and industry-level performances. This should integrate the examination of Thailand’s changing economic structure and industry dynamics into a cohesive framework (see Figure 2.2). The framework should be centred on decomposing aggregate productivity growth to understand the industry origins of Thailand’s productivity performance. This involves investigating how structural changes and sector-specific dynamics individually contribute to overall productivity and economic growth, through an analysis of within- and between-industry effects. By comprehending these factors, policymakers can effectively assess the potential impact, on aggregate productivity growth, of strategies aimed at reallocating resources (such as employment and capital) and activity across sectors or driving the transformation of underperforming industries and consolidating the performance of productive ones. For this reason, the productivity council could link these analyses of sectoral reallocation and industry performances to the objectives of current and future National Economic and Social Development Plan.
While the industry-level focus is crucial, a comprehensive analysis could also benefit from exploring productivity at additional levels of disaggregation, particularly through assessments of regional productivity.
Dimension 4: Examine drivers of productivity. Further to evidence on the evolution of productivity, analyses can also examine the drivers of productivity growth. Productivity is tightly related to a host of factors. Providing evidence on an economy’s performance regarding these factors can raise the attention of policymakers on key issues (for instance investment in R&D and intangibles, human capital, public infrastructure, foreign direct investments, competition and framework conditions of the business environment). This aspect is further developed in the analytical framework for pro-productivity public policies developed in Chapter 3).
To implement this recommendation, Thailand can initially rely on Thai pro-productivity institutions’ existing work, to eventually produce a more comprehensive analysis covering all the above dimensions in the annual reports of the council. To this aim, reports from international organisations, such as the OECD's Compendium of Productivity Indicators and the APO’s Productivity Databook, along with annual reports and bulletins from various OECD countries' productivity commissions and boards (see Box 2.6), serve as key examples of effective practices in publishing productivity indicators and analyses. To improve its assessment of productivity, Thailand could implement the recommendation in two main phases:
Phase 1: Enhance current analyses of pro-productivity institutions. Initially, each pro-productivity institution (particularly, NESDC, Bank of Thailand, Ministry of Labour, Ministry of Agriculture, FTPI and NXPO) should assess, in coordination with the council, their productivity analyses to map their contributions with the dimensions presented above. This stock-taking exercise will be key to informing the compilation of the council’s bulletins (cf. phase 1 of recommendation 3) and annual reports (cf. phase 2). Furthermore, while it is not necessary for all institutions to analyse each dimension and cover all indicators, they should also aim to enhance their analyses within this framework, aligning them with international best practices and examples from other countries previously mentioned.
At the same time, the council can, when combining evidence from various institutions for the compilation of bulletins and annual reports, assess the extent to which current indicators and analyses accurately cover the key dimensions presented in this recommendation. On this basis, it can develop a list of missing indicators and analyses to be further developed in the second phase of implementation. Table 2.4 – which summarises the main relevant productivity indicators and analyses at the economy, industry and regional levels and examples of their application – can also be a starting point to assess current gaps in productivity measures and their desirable frequency. Table 2.2 also provides key information on the availability of indicators to support this stock taking.
Phase 2: Develop new indicators and extend analyses. Based on the previous assessments, the council should ensure that current gaps are addressed through the development of new indicators or analyses. For example, it may ensure that its annual report provides a deeper assessment of productivity in the service sector, decomposes labour productivity growth into the contributions of changes in human capital, accumulation of different types of capital (e.g. ICT vs non-ICT, intangible vs non-intangible) and MFP growth, apply the decomposition of aggregate productivity growth into within- and between-effects and possibly analyse regional productivity. A recent publication by the NESDC (2024[12]) aligns well with the recommendation, as it broadens the scope of productivity analysis in Thailand following some of the dimensions outlined above (notably providing international comparisons, investigating the role of accumulation of different forms of capital). This publication serves as a cornerstone for further advancements in productivity analysis. To strengthen the implementation of this objective, the council can build its own analytical capacity (i.e. including academics or experts in the secretariat, see recommendation 2) but could leverage existing skills, knowledge and evidence available from other institutions, notably the NESDC, academia and international organisations. The examples of implementations of indicators listed in Table 2.4 can also help identify relevant expertise within Thai institutions and applications of suitable methodologies. While annual reports should regularly cover all four dimensions presented above, they do not need to investigate all indicators each year. Following practices from other countries and the suggested frequency of indicators presented in Table 2.4, the council may present some indicators annually, such as productivity trends, productivity decomposition and productivity by industry. Different thematic analyses could be selected for annual reports, such as regional analysis or drivers of productivity and key challenges, such as human capital or the twin digital and green transitions. Ideally, Thailand should cover all indicators presented in Table 2.4 during the implementation period of the National Economic and Social Development Plan.
To support the analysis of productivity, and the implementation of recommendation 4, Thailand could pursue two complementary objectives. First, Thailand could strengthen, harmonise and coordinate the dissemination and documentation of existing productivity indicators (recommendation 5). Second, Thailand may enhance the measurement of productivity at the macroeconomic level through ambitious investments in official statistics (recommendation 6).
In 2016, the Council of the European Union issued a recommendation to establish National Productivity Boards. In this context, the Council recommendation provides guidance on the scope of productivity analyses and annual reports from the European National Productivity Boards provide various examples of such productivity analyses (EU Council, 2016[23]). These reports use a broad range of data from the National Statistical Offices, Central Banks, and international organisations to measure and assess productivity and competitiveness at the economy, sectoral and regional levels. Common analytical components of the reports include:
Monitoring long-term trends in productivity in a comparative perspective (cf. dimensions 1 and 2 of recommendation 3). Most annual productivity reports published by European National Productivity Boards assess developments in aggregate productivity over the long term using labour, as well as multifactor productivity, and benchmark productivity performance vis-à-vis the EU and/or peer countries (European Commission, 2021[24]). This is often a key starting point of a broad macro-economic diagnostic, and several reports also further disentangle the contribution of multifactor productivity growth from the contribution of labour input, its quality, as well as the contribution of ICT and non-ICT capital, or tangible and intangible capital (German Council of Economic Experts, 2023[25]; Belgium National Productivity Board, 2021[26]).
Performing analyses at disaggregated levels, especially sectoral and industry analyses, but also at regional levels (cf. dimension 3 of recommendation 3). The sectoral analysis of productivity is also well integrated into European National Productivity Boards’ annual reports. Among the 2019 and 2020 annual reports, around 2/3 and 1/3 also include respective data on labour and total factor productivity at the sectoral level (European Commission, 2021[24]). Some reports, such as the 2023 Belgium productivity report not only document annual growth rates of hourly labour productivity by broad and detailed sectors, but also measure the contribution of intra-sectoral dynamics and reallocation effects, all in an international perspective (Belgium National Productivity Board, 2023[27]). Although the regional dimension is not present in all European productivity reports (European Commission, 2021[24]), it still represents an important component of some reports. Belgium’s 2023 report, for example, includes a regional diagnostic using data from National Accounts (Belgium National Productivity Board, 2023[27]). It assesses the regional labour productivity performance over time, as well as productivity by 2-digit NACE industries and regions.
Examining drivers of productivity growth (cf. dimension 4 of recommendation 3). As detailed by the EU Council’s recommendations (EU Council, 2016[23]), Productivity Boards should consider the long-term drivers and enablers of productivity and competitiveness, including innovation, and the capacity to attract investment, businesses and human capital. Thus, several reports analyse these complementary factors, specifically investments in R&D, education and skills (Belgium National Productivity Board, 2022[28]; Finnish Productivity Board, 2023[29]).
Note: Reports from European National Productivity Boards are accessible at https://economy-finance.ec.europa.eu/economic-and-fiscal-governance/national-productivity-boards_en
Note: This table provides a summary of indicators that can be considered in recurrent analyses of productivity in Thailand. The list of indicators is not exhaustive and focuses on indicators directly related to productivity metrics. Related factors are not covered in this table and are for instance discussed in OECD/APO (2022[5]). 1 The list of references provided is not exhaustive. It focuses on references discussed in the main text of the report, focusing primarily on applications to Thailand, as well as other complementary examples (readers should also refer to discussions and references herein). 2 The suggested frequencies are “very frequent” (e.g. annual), “regular” (e.g. analysis conducted at least for each sub-period of the National Economic and Social Development Plan) or “frequent” (in between the very frequent and regular frequencies). The suggested frequency is purely indicative and can be adapted depending on the feasibility, the analytical needs and topics of particular interest, for instance, while the decomposition of growth into within-industry and reallocation (between) effects may not need to be conducted very frequently as the industry structure may be relatively stable on a year-to-year basis, episodes that may be associated with increased reallocation (e.g. crises) may increase the need for such analyses.
Source: OECD elaborations.
Strengthening, harmonising and coordinating the dissemination of productivity indicators
Thailand has achieved notable advancements in the dissemination of macro-data for productivity analysis. One achievement is the development of the Government Data Catalogue, a centralised online data portal to access data from various organisations. This is complemented by the dissemination of relevant indicators by Thai institutions (such as the NESDC or the Bank of Thailand) through their channels, including their annual reports or websites. The Thailand Productivity Institute also provides a relevant dashboard for productivity indicators, collecting information from several national and international institutions (including the Asian Productivity Organization).
However, the dissemination of productivity statistics and methodology is still incomplete and remains scattered. Efforts towards data dissemination should be continued and strengthened. For instance, it appears that key productivity metrics, such as the measure of MFP computed by the NESDC, are not accessible for downloads but only available through its annual reports, while the Bank of Thailand’s labour productivity measures are easily accessible through its website, but not available on the GDC. Additionally, some existing data repositories or dashboards (for instance from the FTPI) do not integrate the full scope and latest data available from other institutions. Finally, available data and reports would also benefit from providing additional details on the data and methodology used to measure productivity.
Recommendation 5: Better coordinate the dissemination of productivity indicators and ensure compliance with high-quality and harmonised dissemination standards.
To support analytical work on productivity, Thailand may improve the dissemination of productivity indicators along three dimensions:
Dimension 1: Ensure high-quality dissemination standards. All Thai institutions (notably the NESDC, the Bank of Thailand, the FTPI, the National Statistical Office, the Ministry of Agriculture, the Ministry of Labour and NXPO) must ensure that the dissemination of productivity indicators always meets stringent high-quality standards, also when disseminated through their websites or reports. Key considerations include timeliness, accessibility and transparency. Timeliness: Provide updated data and long-time series including the most recent period possible; Accessibility: Facilitate online data access in multiple user-friendly formats (e.g. csv, JSON, etc.) and with appropriate tools (including machine-readable information within Application Programming Interface). Transparency: Consolidate the communication of metadata and methodology (for instance on concepts, scope, classifications, basis of recording, data sources and statistical techniques).
Dimension 2: Centralise access to productivity indicators. All Thai institutions could integrate relevant productivity indicators in a common repository (the Government Data Catalogue) to enhance the accessibility and visibility of existing indicators available across institutions. The purpose of such a repository would be to aggregate and disseminate productivity indicators from various Thai and possibly also non-Thai institutions. This could also serve to facilitate the compilation of annual reports on productivity based on the full range of available indicators (see recommendation 4).
The council could coordinate Thai institutions' efforts to meet common dissemination standards for productivity statistics and to integrate these indicators into the common repository.
Dimension 3: Enhance the visibility of indicators. The council could strengthen the dissemination of productivity indicators by creating a productivity dashboard to accompany its annual reports.
While the implementation of appropriate dissemination of productivity statistics would fall under the responsibility of the respective institutions, the productivity council can coordinate the contribution of the different Thai institutions to a common repository and promote the exchange of good practices among them. The implementation can rely on a sequencing based on two phases.
Phase 1: Stock taking exercise. This initial phase would consist of taking stock of dissemination practices and consolidating the metadata and description of methodologies for productivity measurement. The council could coordinate a stock-taking exercise, according to which all participating institutions would review the indicators they compute and/or use to analyse productivity at the macro level. Each institution would be responsible for preparing detailed metadata and methodological notes, and also for assessing the accessibility of these indicators for external users. To support the implementation of this phase, the council could organise a workshop to share good practices among participants and/or issue guidelines to help harmonise dissemination standards. International guidelines can support this exercise (see UN (2022[32]) and references herein). Expertise also exists in Thai institutions, with the Bank of Thailand serving as an example of good practices in this respect.
Phase 2: Integrate all productivity indicators in the Government Data Catalogue and develop a productivity dashboard. Building on phase 1, Thai institutions should consolidate the dissemination of their productivity statistics through a coordinated effort to publish indicators, detailed metadata and methodological notes on a common repository, notably the Government Data Catalogue. This would ensure that all relevant indicators, including multifactor productivity from the NESDC, labour productivity from the Bank of Thailand and possibly indicators from the APO, are easily identifiable, accessible and understandable. To facilitate access to productivity data, the Government Data Catalogue could also enhance its search functionalities integrating a tag to identify all productivity indicators. Finally, the council can monitor the implementation of this phase with one or several progress reports, listing the indicators available on the GDC and assessing the quality of dissemination based on previously agreed standards.
To further enhance the dissemination and visibility of these indicators, the council could oversee the creation and maintenance of a productivity dashboard using the set of productivity and related indicators available. For instance, the French National Productivity Board publishes a statistical dashboard along with its annual reports.16 To go further, the productivity council could also oversee the creation of an online interactive dashboard, building on and extending the tool created by the Thailand Productivity Institute. This would provide user-friendly and up-to-date visualisations of productivity indicators at aggregate, sectoral and regional levels.
Enhancing the measurement of productivity at the macro-sector level
In addition to strengthening the analysis and dissemination of productivity indicators, Thailand could also pursue enhancements to the measurement of productivity. Measuring productivity accurately is challenging. It requires an appropriate measurement of production inputs and outputs, relying on a combination of various data collected by different agencies. While relevant data exist to compute productivity, limitations and challenges remain in Thailand similarly to other countries. These challenges are related for instance to the measurement of capital and labour inputs, but also of GDP. Addressing these challenges requires an ambitious measurement agenda and significant investments in official statistics.
Recommendation 6: Set up a productivity programme to guide the development of official statistics.
Thailand could continue to develop metrics that are useful for productivity analysis. To this aim, Thailand could set up a productivity programme to continue to develop and enhance productivity statistics to ensure quality data that reflects international standards and best practices. However, publishing productivity indicators complying with international standards hinges on the adequate measurement of production output and inputs and requires continued investments in official statistics and National Accounts. A better measurement of productivity implies addressing relevant challenges that emerge in maintaining harmonised and long timeseries for the data that are necessary to measure productivity. The guidelines and recommendations formulated by the APO and the OECD (APO/OECD, 2021[2]) – which provide an in-depth discussion of measurement challenges and potential improvements – could serve as a guide for this endeavour and a benchmark against which Thai agencies may assess their current practices. The recent publication by the NESDC (2024[12]) also discusses the opportunities and challenges to further develop productivity measurement in Thailand and serves as a foundation for implementing this recommendation.
Overall, Thailand may rely on two pillars to enhance productivity measurement. First, Thailand could set up a productivity programme directly aimed at maintaining and enhancing productivity statistics according to international best practices. Second, this programme should be supported by continued investments in official statistics.
A productivity programme, possibly under the responsibility of the National Statistical Office, would aim at providing reliable and up-to-date productivity statistics for the economy. The programme could be articulated around two dimensions:
Dimension 1: Ensure that labour and multifactor productivity indexes are computed according to best practices. The first component of the programme would be to ensure that productivity statistics are computed and released following the most advanced methodologies. For instance, it could integrate information on quality-adjusted labour input and capital services when available (see further discussion next) or compute measures of multifactor productivity using the KLEMS (capital, labour, energy, materials and service) approach.
Dimension 2: Publish indicators at more disaggregated levels. To support more refined productivity analyses according to recommendation 4, the productivity programme could also contribute to extending the measurement of productivity for more detailed industries, or also at a more detailed regional level.
However, a more accurate measurement of productivity also relies on the availability and quality of official statistics and measures derived from National Accounts. Improvements in three main dimensions represent key targets for Thailand17:
Dimension 1: Better measurement of output and value-added. Key statistical recommendations from APO/OECD (2021[2]) on this include a better measurement of the informal economy. Informal activities are generally often absent from traditional data sources and remain unobserved despite their relevance according to National Account standards. This is particularly relevant for Thailand, which still features high levels of informality. Additionally, APO/OECD (2021[2]) also discusses the need to address challenges related to capturing the digital economy in the measurement of GDP (for instance related to the growing importance of platforms or the difficulty in measuring price deflators for ICT products). Challenges related to digitalisation were notably at the root of debates on the role of measurement errors in the observed slowdown in productivity across countries. In this respect, the System of National Accounts 2025 provide a framework that could help Thailand address some of these challenges.
Dimension 2: Better measurement of labour inputs. APO/OECD (2021[2]) discusses how the measurement of the volume of hours worked can be improved using the full range of available sources instead of relying only on direct reporting from labour force surveys. Another key improvement would consist in producing an index of the quality and composition of labour inputs to support the measurement of multifactor productivity by disentangling the contribution of MFP and changes in the composition and quality of the labour force over time.
Dimension 3: Better measurement of capital inputs. Computing multifactor productivity requires a reliable measurement of capital inputs. The standard measurement of capital stocks for National Accounts (following the “net wealth approach”) already involves addressing several challenges summarised in APO/OECD (2021[2]). However, a better measurement of capital for productivity consists in measuring productive capital stocks and capital services. This measure will reflect the source of productive services provided by capital goods to the production process (capital services) instead of net wealth capital stock (see also Box 2.2).
Tackling these measurement challenges is complex, as each represents a specialised statistical and economic field with ongoing advancements. Addressing them requires significant coordination across agencies and may also involve trade-offs. As such, these are medium- to long-term objectives that go beyond the immediate scope of productivity measurement. The NESDC and the NSO should take the lead in these efforts, based on international guidelines and cooperation (see APO/OECD (2021[2])). However, the productivity council could play a key role in facilitating necessary inter-agency coordination and promoting the importance of investments in Thailand’s statistical system to drive improvements in productivity measurement. To help develop and promote this productivity measurement agenda, Thailand could implement the recommendation in two phases.
Phase 1: Promote knowledge sharing and develop expertise on productivity. In the short term, the council could facilitate knowledge sharing and the development of further expertise on productivity. This would involve all Thai pro-productivity institutions, experts from academia and international cooperation with relevant stakeholders such as international organisations. Workshops dedicated to productivity measurement could provide a suitable format for knowledge exchange. This step is complementary to phase 1 of recommendation 5, which involves a review and presentation of methodologies currently employed by Thai institutions.
Phase 2: Centralise the production of productivity indicators with a productivity programme and continue improvements in official statistics and National Accounts. In the medium term, Thailand could centralise in a single institution, within an expert team at the NSO or the NESDC, the production of the different productivity metrics, including labour, capital and multifactor productivity at different levels of aggregation, relying on the inputs from other agencies – see Box 2.7 presenting the case of the United States, Canada, the UK and Australia, and see also the report produced by the London Economics and DIW Econ (2017[33]) on productivity measures published by NSOs across several countries. This institution would also be responsible for communicating methodologies and disseminating data. This would help to further build expertise on productivity and facilitate the development and harmonisation of productivity methodology and indicators and their dissemination (complementing recommendation 5 related to strengthening dissemination). It could further help to shape the priorities of statistical agencies in terms of data development to support productivity measurement, by creating a demand for improved statistics on GDP, employment, capital and other statistics. Supported by the productivity council, it could also raise awareness on the importance of investing in statistical systems to support evidence-based policymaking to boost productivity. On this basis, the work of the NSO can continue to improve Thailand’s official statistics and National Accounts to enhance the measurement of output, labour, capital and, as a result, productivity. The transition towards the 2025 System of National Accounts represents a key building block for this endeavour.
The Office of Productivity and Technology (OPT) – The Bureau of Labour Statistics (BLS), United States
The United States provides a good example of a centralised and identified institution responsible for developing productivity statistics. The Office of Productivity and Technology (OPT) within the Bureau of Labour Statistics (BLS) is responsible for the productivity statistics and the BLS productivity programme ensures that productivity measures remain accurate, timely and relevant. The OPT webpage provides a reference point to access data on productivity and detailed methodological explanations. The OPT constructs measures of labour productivity for major sectors but also at a detailed industry level (at 6-digit NAICS level) and at state and regional level. OPT also publishes indexes of multifactor productivity, based on quality-adjusted labour input and capital services, at the aggregate and industry level (for 4-digit NAICS manufacturing industries).
The Canadian Productivity Accounts – Statistics Canada
The Canadian Productivity Accounts (CPA) are responsible for producing, analysing and disseminating Statistics Canada's official data on productivity and for the production and integration of data on employment, hours worked and capital services consistent with the System of National Economic Accounts. To this end, the CPA comprises three programmes. First, a quarterly programme provides current estimates on labour productivity and labour costs at the aggregate level for 15 industry groups on a timely basis. Second, the annual multifactor productivity programme provides yearly estimates on multifactor productivity and related measures (output, capital input, labour input and intermediate inputs) as they apply to the major sectors of the economy and to industries at the national and provincial levels. Third, the annual provincial programme provides estimates of employment, hours worked, labour productivity and labour costs at the industry level for each province and territory.
Productivity statistics – UK Office for National Statistics (ONS)
The UK Office of National Statistics (ONS) also provide rich productivity measures, including at industry and subnational levels. These include quarterly multifactor productivity (MFP) estimates, as well as quarterly capital services and quality-adjusted labour input estimates. The ONS routinely publishes a breakdown by industry division (two-digit level) for labour productivity and regularly produces productivity statistics at the regional level.
Productivity statistics – Australian Bureau of Statistics (ABS)
The ABS has been producing aggregate MFP statistics since 1985. The ABS produces annual indexes of labour productivity, capital productivity and MFP for the market sector, and since 2007, for each industry division within the market sector. Capital and multifactor productivity are published annually, while labour productivity is produced both quarterly and annually. The ABS also produces experimental industry-level KLEMS multifactor productivity (based on Capital (K), Labour (L), Energy (E), Materials (M) and Services (S)). Finally, the ABS produced both capital stock and services data at the total economy and industry level (1-digit), and it also compiles a quality-adjusted labour input (QALI) measure.
Sources: US Bureau of Labour Statistics – Office of Productivity and Technology (OPT); Statistics Canada – Labour Productivity Measures - National (Quarterly) (LPM) and Productivity Measures and Related Variables - National and Provincial (Annual) (MFP); UK Office for National Statistics – Productivity development plan: 2021 to 2023; Australian Bureau of Statistics – Interpreting ABS productivity statistics and Estimates of Industry Multifactor Productivity methodology.
Finally, some of the enhancements related to the availability of the micro-data discussed in the next sections (see recommendations 8-9) may directly contribute to the improvement of National Accounts and other aggregated data thereby enhancing productivity measurement at the macroeconomic level in the medium term.
Understanding the economic forces behind productivity performance requires more granular data to analyse micro-level productivity. A country’s (or an industry’s) productivity is fundamentally an aggregation of the performance of individual units (e.g. firms or establishments). Moving the analysis to the micro-level helps identify the causes of productivity performance, including firm productivity, the allocation of resources and market shares across firms and creative destruction through firm demography (entry and exit). These factors are outlined in Figure 2.4 and further discussed in this section of the report. Micro-data (at the level of firms or establishments) allow disentangling the role of such forces and are thus crucial for moving beyond aggregate- and industry-level analysis.
First, this section assesses the micro-data landscape for productivity analysis in Thailand. To this aim, it details the features of the data required to effectively study productivity at the micro-level. It also reviews the main sources of micro-data for granular productivity analysis in Thailand, highlighting their respective strengths and limitations against a benchmark of an ideal dataset. Second, the section then highlights the need for Thailand to further integrate administrative data and promote a sound dissemination of micro-data for productivity analysis. Third, the section reviews the use of available micro-data for granular productivity and assesses the potential to strengthen the analysis according to an analytical framework.
Rich micro-data are available but with significant limitations for productivity analysis
The report now discusses the essential features of the micro-data that productivity analysis requires. It then reviews the main available data for such analysis in Thailand, highlighting their strengths and limitations considering these essential features.
Essential features of micro-data for productivity analysis
Several measures of labour and multifactor productivity can be computed at the level of a production unit, typically an establishment or a firm. The feasibility of their implementation depends on the information collected in the micro-data. The characteristics of the microdata discussed below condition productivity measurement and an in-depth analysis of its drivers.
The main variables required to compute productivity at the unit level are value-added, employment (the minimum, to compute labour productivity), a measure of investment and/or capital (either focusing on tangible or total fixed assets, also information on intangible assets is highly valuable), gross-output (or sales/turnover) and intermediate inputs (covering, raw materials, energy).
The definition and identification of the observation units and the possibility to follow them over time (i.e. longitudinal data) are important features of the micro-data. The analysis of productivity at a micro-level is typically conducted at the establishment or the firm level and requires a unique identifier of these units, so that observations correspond to single unit-year pairs. Both the computation of productivity and its analysis largely benefit from a longitudinal structure of the dataset in which the same units can be followed from year to year.18 This longitudinal structure is preferred to a repeated cross-section as it allows tracking firm births and deaths or computing within-business productivity growth. Ideally, the analysis is conducted based on relevant statistical units, i.e. units defined for statistical purposes, but can also rely on legal or administrative units.19 The main statistical units used internationally are establishments, enterprises and business groups (UN, 2024[34]).20 While one or more establishments constitute an enterprise, enterprise groups result from a combination of enterprises. They are identified as the set of legal units bound together by legal and/or financial links under the same control (UN, 2024[34])
The availability of a detailed industry-level variable recording the main activity (at the two-digit, or more disaggregated, industry level) represents a key element for two reasons. First, the computation of multifactor productivity assumes that all units in a given industry have the same production technology. This assumption is more likely to hold within detailed industries. Second, this enables a more detailed analysis, either by computing statistics for detailed industries or by controlling for unobserved heterogeneity and time trends across industries in econometric analyses.
The broad coverage of the business population is another crucial dimension of the data. The micro-data should be at least representative of the targeted business population. Ideally, the data should cover the full population of units and correctly capture their entry to, and exit from, the market. This permits an analysis of firm demography and its contribution to productivity growth through creative destruction, as well as of the evolution of the productivity of units as they age and eventually exit, i.e. their firm life cycle.
Micro-data need to be collected at regular frequency and ensure sufficient time coverage. Sufficient frequency of data collection enables timely evidence supporting timely decision-making. Data for productivity analysis are ideally collected annually. Ad-hoc surveys may also provide more timely information when needed (e.g. during the COVID-19 pandemic). Moreover, extensive time coverage allows for the analysis of long-term productivity trends. Annual data are preferable to compute productivity according to state-of-the-art methodologies.
The availability of additional information on factors that influence productivity can enhance its understanding, although it is not necessary for monitoring productivity. For instance, valuable information includes the ownership structure of the unit, innovative activities (e.g. R&D expenditures), engagement in international trade and participation in global value-chains, financial information (e.g. short- and long-term debts), skill composition of the workforce, management quality and use of ICT. Such information can stem from data integration, by linking different micro-data sources, and does not need to be available from a single “primary source” used for productivity analysis.
Finally, data providers need to provide comprehensive metadata related to the above characteristics and information related to the accessibility of these data. The metadata should specify the definition of variables and units and inform about the time consistency of the data (for instance documenting potential changes in coverage and collection procedures). It should also inform about the statistical treatment of the data provided (such as the imputation of missing values).
The following sub-section discusses the Thai micro-data suitable for productivity analysis, benchmarking the data against the essential features discussed above Table 2.5 summarises the main takeaways discussing the strengths and limitations of productivity analysis for each data.
Strengths and limitations of available micro-data for productivity analysis
The NSO and other Thai agencies collect various micro-data relevant for productivity analysis, including censuses, surveys and administrative data. The analysis of resources available in Thailand identifies a range of data that can directly, or indirectly, contribute to productivity analysis at a granular level. The report focuses on the data that are identified as the main sources for existing studies of productivity in Thailand and the building blocks for the set of indicators related to productivity analysis discussed later.21 These data are the Thailand Economic Censuses from the NSO, the OIE Manufacturing Survey from the Ministry of Industry and the Corporate Profile and Financial Statement database from the Ministry of Commerce. Additionally, the report presents other micro-data which can serve as key inputs for further analysis, namely the Labour Force Survey and the Annual Survey of Research and Development and Innovation (RDI Survey). The rest of this sub-section presents these micro-data and highlights their strengths and limitations for productivity analysis.
Thailand’s Economic Censuses – National Statistical Office
The National Statistical Office of Thailand (NSO) conducts two economic censuses on Thai businesses in manufacturing and services: the Industrial Census and the Business Trade and Service Census.22
The Industrial Census is a repeated cross-sectional data collecting information for establishments in the manufacturing sector. The NSO initially collected these data in 1964 and now gathers them every five years.23 The census collects information on the number of establishments and their key characteristics. Researchers have heavily relied on the Industrial Census to study productivity in the Thai manufacturing sector, including the BOT and the PIER which have taken a leading role in advancing sound analysis at the establishment level.
The Business Trade and Service Census is a repeated cross-sectional data collecting information on establishments in the service sector. The NSO initially collected these data in 1966 and now gathers it every ten years.24 However, research on the service sector using these data is limited due to the lower frequency of data collection compared to the Industrial Census.
For the 2012 and 2022 editions, the NSO conducted joint censuses for establishments engaged in manufacturing and business trade and services, combining the Industrial Census and the Business Trade and Services Census into the Business and Industrial Census.
For the 2022 edition, the NSO further improved the data collection with the use of a common frame based on administrative records to construct the sampling frame of the surveys. The NSO constructs this common frame using data from registration systems of various agencies combined with an NSO database on natural and legal entities. This frame could serve as the basis for a possible data integration discussed in this chapter.25
The large coverage of the business population and its representativeness are key strengths of this dataset for productivity analysis. The sample of the Censuses is constructed as follows:
Establishments with one to ten persons engaged are covered by a representative sample. The NSO uses a systematic stratified sampling methodology, using location and industry as stratum.26
Establishments with more than ten persons engaged are covered by a complete enumeration.27
The use of the common frame represents a significant advance in collecting business information in Thailand. It reduces the burden on officials and businesses.28 However, challenges remain in terms of data integration, in particular related to the definition and identification of units across agencies.
The collection of a rich set of information on establishments is another key strength of this dataset. The census contains all the necessary information to compute measures of labour and multifactor factor productivity and extensive information to explore factors that can influence productivity, such as ICT adoption, obstacles to operation and innovative activities (e.g. R&D expenditures). It also contains relevant information on labour compensation and workers’ characteristics to explore important aspects of productivity studied in OECD countries, such as the “human side of productivity” (Criscuolo et al., 2021[35]), the link between productivity and employment (Calligaris et al., 2023[36]) as well as inclusiveness (OECD, 2018[37]). Additional information is collected for establishments with more than ten persons engaged constituted as a juridic partnership or limited liability company. This includes information on trade, innovation and use of ICT, such as payment to entities abroad, charges for use of intellectual property received (or paid) from clients abroad (to entities abroad) and connection to the high-speed internet.
However, the frequency of the dataset may be a limitation for the analysis. Its frequency – currently every five and ten years for the Industrial Census and the Trade and Business Census respectively – does not provide timely information for the analysis of productivity developments.29 It also limits the possibility to disentangle structural productivity developments from cyclical dynamics and major economic shocks. For instance, the 2012 wave includes data for the 2011 calendar year, when Thailand experienced major floods, and the 2022 wave focuses on 2021 when COVID-19 was still largely affecting economic activity. This latter point is however mitigated by the collection of information on the effects of COVID-19.
In addition, challenges in following firms over time may limit both the measurement of productivity (using state-of-the-art estimation of production functions) and the scope of the analysis. First, conditioning the analysis on the availability of the longitudinal linkages significantly reduces the number of observations, especially among micro-establishments.30 Moreover, the longitudinal identifiers to match firms across waves do not seem available for researchers outside the NSO – according to information collected during interviews of representatives of Thai agencies conducted by the OECD during a virtual “fact-finding mission”.
Manufacturing survey – Office of Industrial Economics (OIE), Ministry of Industry31
The Office of Industrial Economics (OIE) within the Ministry of Industry regularly conducts an annual survey on manufacturing establishments to produce its annual TFP report.
The survey is designed to be representative of the manufacturing sector. It covers around two to three thousand establishments per year in 22 manufacturing industries and around 73% of sales in manufacturing (for the 2022 wave) when compared to the Industrial Census.
The survey collects rich information on establishments supporting the analysis of productivity growth and its determinants. The data contain basic information on establishments (name, registration number, location and main industry), but also all the relevant information for a sound measurement of productivity and an analysis of influencing factors. This includes rich information on labour inputs (hours worked, trained labour and composition of the workforce in terms of skills and gender), capital and investment, intermediary inputs (such as raw materials) and production output. It also contains information on the use of robots and automation technologies and environmental impacts (such as water and air pollution and industrial waste).
Despite the apparent rich set of information, some authors report that employment is missing for a large share of observations (around 50% of the observations between 2017-20 according to Sangsubhan et al. (2023[38])). Merging the data with supplementary information on labour input from other datasets (e.g. social security data from the Ministry of Labour) could help fill the gaps and improve the coverage of productivity measures. The OIE may assess the feasibility of implementing such linkages, in coordination with other agencies.32
The composition of the sample may be another limitation of these data. Large firms make up around half of the sample. This implies that these data may not be fully suitable for the analysis of young firms or of the least productive ones (laggard firms), which tend to be smaller.33 This may significantly restrict the scope of within-industry analysis, notably the monitoring of the heterogeneity of productivity across units and its evolution over time. By design, this dataset covers only manufacturing, and no similar data seem to exist for the service sector.
Corporate Profile and Financial Statement (CPFS) – Department of Business Development, Ministry of Commerce
By law, all firms are required to register at the Department of Business Development (DBD) from the Ministry of Commerce and submit financial statements (Apaitan et al., 2020[39]). Thus, for most of the firms in the business population, the Ministry collects firms’ financial statements yearly since 2004 (the Corporate Profile and Financial Statement (CPFS) dataset).34 The dataset contains information for around 500 thousand firms per year, registered as partnerships, private limited companies, and public limited companies (Limjaroenrat, 2016[40]).35
The purpose of the data is to collect firms’ financial statements together with general information including registration year, registration type, operating status (for instance active or dissolved) and main industry (5-digit TSIC 2009 classification).
The key strengths of the CPFS data are its broad coverage and its annual frequency, combined with a panel structure allowing to track firms over time. This enables an analysis of firm demography and market structure – as in Apaitan et al. (2020[39]) – and an analysis of firm investment. These features make it highly relevant for the understanding of productivity and business dynamics. Thai institutions can notably use these data as a backbone for monitoring indicators of creative destruction, competition and reallocation. Researchers have also used it to assess the impact of public policies (Muthitacharoen, Paweenawat and Samphantharak, 2024[41]).
However, the data lacks information on labour input (e.g. employment) required to measure productivity. This prevents a broader use of these data for productivity analysis, a limitation that further data integration can address.
Labour Force Survey (LFS) – National Statistical Office
The NSO has continuously conducted the Labour Force Survey (LFS) since 1963, providing a better understanding of labour force characteristics at the national and regional levels.36 The survey is representative of the Thai population aged 15 years and over.
Among others, the survey collects information on workers’ occupation, education, work status (e.g. employee, self-employed), weekly hours worked, wages and whether workers are covered by social security.
The data are not a primary source to study productivity at the firm-level. In fact, only little information is collected about the establishments in which workers are employed. The LFS is primarily designed to collect information at the household level. Information on establishment only includes the economic activity (industry following the TSIC 2009 classification) and the size of the establishment.
However, the LFS is a useful source to extract aggregate information on the labour input that can be used for the macro-analysis of productivity discussed before. That information includes data on employment and hours worked. It can be further categorised by gender and education level of workers. Information from the LFS is used to compute measures of labour productivity at the industry level (see for instance Ministry of Labour (2022[11])), and could be a primary source to compute a labour quality index, as previously discussed (see Box 2.3).
Concerning the design of the survey, the NSO has adopted a stratified two-stage sampling approach to run the 2024 survey. The NSO constituted 77 strata corresponding to provinces and then divided each stratum into municipal areas and non-municipal areas.37
The survey has a panel structure, which allows longitudinal analysis, for instance tracking changes in workers’ skills or occupations over time.38 Such information from the LFS can contribute indirectly to productivity analysis, given the necessity to improve education, skills, as well as worker mobility to address skills shortages and mismatch identified by the Ministry of Labour (2022[11]).39
At the time of writing, matched employer-employee data are not available for a joint analysis of workers and firms. In these type of data, employer data (i.e. firm-level data) are paired with worker data via jobs and allow measuring simultaneously firm characteristics (including productivity) and worker characteristics. These data can shed light on the role of management, worker skills and diversity for firm productivity (Criscuolo et al., 2021[35]). They can also provide timely information on productivity and the role of government support during crises (Calligaris et al., 2023[42]). Establishing such linked employer-employee data may represent a long-term goal for Thailand.
Research and Development and Innovation survey (RDI) – Office of National Higher Education Science Research and Innovation Policy Council (NXPO) and National Research Council of Thailand (NRCT)
Thai agencies conduct annually the Research and Development and Innovation survey (RDI) since 2008 to study the status and operation of RDI of establishments in Manufacturing, Services and Wholesale and Retail trade Sector.40
The sample of around 5 000-6 000 establishments is derived from more than 95 000 establishments with an income above THB 12 million that are included in a database of high potential establishments. Mostly large firms compose the sample. The data are partly constructed as a panel, with the possibility to follow some firms over time. Around 60% of firms are surveyed randomly instead.
Among other variables, the survey collects general information on the establishment (such as the identification number, name, address, activity, foreign/domestic), the number of persons engaged, R&D expenditure and information related innovation activity.
With the information available, Thai institutions use this survey to inform policy aimed at strengthening the potential of manufacturing establishments to participate in RDI activities. Agencies occasionally use the survey to estimate the gross domestic expenditures on R&D (GERD). Researchers also investigate the link between R&D and productivity using such data. For instance, Charoenporn and Choksawatpaisan (2024[43]) study the link between R&D activities, innovation output, and productivity in the service sector. However, some limitations (e.g. related to the industry classification used) remain concerning the use of these data to provide evidence on policy relevant questions. These are in particular related to the Thailand’s science, technology and innovation (STI) agenda focusing on driving forward new and innovative target industries, such as electronic vehicles (EV) or Industry 4.0 and digitalisation, as well as “S-Curve Industries”.41
Thai agencies are further developing an integrative database system that links data on science, technology and innovation, with data on higher education (e.g. number of students and graduate at the firm-level) and expenditure on research and development projects.42 To date, this linked dataset integrates data from the National Research and Innovation Information System (NRIIS), the Higher Education’s UNICON and the National Science and Technology Information System (NSTIS). 43
Note: **Before 2012 the Business and Service Census cover all establishments with five persons engaged, after 2012 it is harmonised to the data collection of the Industrial Census.
Source: OECD elaborations.
Integrating administrative data and setting data-sharing mechanisms can address data limitation and foster the use of micro-data
The available micro-data offer a solid foundation for studying productivity in Thailand. However, they are not without limitations. The report identifies the insufficient statistical use of administrative data and incomplete data linkages and integration as important limitations. It also outlines a pathway to improve existing data and foster their use for analytical purposes. Moreover, it discusses international practices in repurposing and integrating administrative data and in facilitating micro-data sharing for analysis and research. The discussion highlights the needs to address gaps in Thailand’s availability of micro-data sources for productivity analysis.
Fostering the integration and use of administrative data for economic analysis
Administrative data have emerged as a key source of information for statistical and economic analysis. While originally collected by public entities for administrative purposes - including firms’ registration procedures, tax purposes and listing applicants to policy support - administrative data can be valuably integrated into statistical frameworks. Reusing administrative data may be more cost-effective than collecting ex-novo the data, for instance through costly surveys. Using administrative data thus reduces the burden on statistical agencies and businesses. They are often suitable for covering the entire target population and for observing units over time. Extensive use of administrative data enhances official national statistics (IMF, 2018[44]) and has become a cornerstone for analysing productivity and business dynamics.
Government and National Statistical Offices (NSOs) in several OECD countries have responded to the increasing demand and use of administrative data by integrating sources into their statistical information systems. Statistical authorities use administrative data to improve the timeliness, coverage and quality of National Accounts data. For instance, several countries use VAT and income tax records to compile National Accounts and GDP (e.g. Canada, Chile, Denmark, Finland, and UK within the OECD, but also other countries like Rwanda (IMF, 2018[44])).
Statistical authorities also widely use administrative data to construct Statistical Business Registers (SBRs), which are a centrepiece for business statistics. An SBR serves several purposes, including: (i) structuring the information of all registered firms of a country (ii) accurately identifying the target population for business surveys, enabling random sampling of units for investigation, and broadly ensuring the coherence and consistency of statistics, (iii) being the backbone of firm-level data linkages and (iv) being a reliable source for business demography statistics. Statistical Business Registers are generally built from a number of different administrative sources and surveys, including tax registers (e.g. for value added tax, corporate or personal income tax), compulsory registration systems (e.g. for limited liability businesses or those quoted on stock markets), social security and other sources from public or private sector data holders (see Box 2.14 in the next section).
Thailand’s statistical information system would largely benefit from further integration. This would enable more in-depth and comprehensive productivity analysis at a microeconomic level. As previously discussed, individual data sources could be linked to complement information that may be missing in individual datasets. For instance, integrating complementary information on employment would enable the Corporate Profile and Financial Statement (CPFS) database to become a primary source for computing annual productivity statistics for the population of incorporated and active firms. More generally, pursuing efforts to set up a Statistical Business Register following international guidelines would be a key target for improving the data landscape in Thailand. The creation and maintenance of a Statistical Business Register and the further integration of administrative data and surveys would not only support micro-level analysis but could also enhance National Accounts data through high-quality and consistent business statistics. As such, it would also contribute to a better measurement of productivity at aggregate, sectoral and industry levels.
However, integrating administrative data to information systems involves several considerations.First, it requires designing administrative data in a way that facilitates both their integration through linkages and their statistical use. Administrative data are not primarily collected for research but for administrative purposes, resulting in possible limitations such as inconsistencies, limited availability of relevant variables and lack of coherent business identifiers across data sources. Involving Thailand’s NSO in the design of administrative data to improve their usability for analysis can help overcome these limitations (see the case of France in Box 2.9).The use of common and unique business identifiers across data sources facilitates data integration (see Box 2.12).
Second, it requires the exchange of administrative data between public institutions including ministries and NSOs. At best, such exchanges may be facilitated by law, to ensure that statistical authorities can access and use these administrative data for statistical purposes (see Box 2.13). The OECD Recommendation on Good Statistical Practice deals with the right of statistical authorities to access administrative sources (see Box 2.8).
The OECD Council adopted the Recommendation of the Council on Good Statistical Practice in November 2015. This legal instrument, applicable to all OECD Members (and open to non-members), provides a set of twelve recommendations that serve as key references for assessing and benchmarking national statistical systems and establishing a sound and credible national statistical system. The recommendations cover different areas including (i) the institutional, legal and resource requirements for statistical systems, (ii) the methods and quality of processes of statistical collection, production and dissemination and (iii) coordination, cooperation and statistical innovation.
The fifth recommendation is to “ensure the right to access administrative sources to produce official statistics” and is further complemented with the following set of good practices:
The Statistical Authorities (SA) are authorised by law to use administrative records for the regular production of official statistics.
Administrative sources are used whenever possible and cost-effective to avoid duplicating requests for information and reduce reliance on direct surveys.
National Statistical Authorities are involved in the design of administrative data in order to make administrative data more suitable for statistical purposes.
National SA cooperate with owners of administrative data in assuring data quality.
Agreements are made with owners of administrative records which set out their shared commitment to the use of these data for statistical purposes.
Recommended practices are available for the reporting and presentation of administrative data.
Linking administrative data with survey data is encouraged by National Statistical Authorities with the aim of reducing the burden on respondents, reducing the costs in producing official statistics, and increasing the analytical value of official statistics.
The report monitoring the implementation of the OECD recommendations indicates that countries consider this recommendation on the use of administrative data very relevant. However, the capacity to access administrative sources for statistical purposes varies across countries and agencies, and some Adherents still face difficulties in accessing administrative sources. While statistical legislation usually guarantees to producers of official statistics (in particular NSOs) the right to access administrative data, challenges remain. Countries report the following issues (in order of importance):
Statistical law is insufficiently explicit
Legal restrictions
Conflict between different legal acts
Lack of cooperation by data owners
No legal basis to access data held by private entities
Delays in relation to access to administrative data sources
Administrative and financial obstacles
Countries generally recognise that there is a need to reinforce the legislation, in order to improve the cooperation between statistical authorities and owners of administrative data (see Box 2.13 on this).
Source: https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0417#monitoring; Report on the Implementation of the OECD Recommendation on Good Statistical Practice (2020).
Thailand has facilitated the exchange of data across agencies through digital technical solutions for data exchange, notably with the Government Data Exchange established by the Digital Government Development Agency.44 Despite such advances in solutions for secure data sharing, Thai statistical agencies may face barriers to accessing administrative data due to insufficiently explicit statistical law and conflicting legislation (see also Box 2.13 on common issues in accessing administrative data, discussed below). For instance, during interviews conducted by the OECD, officials from various agencies mentioned difficulties in accessing corporate and personal income tax and VAT data from the Revenue Department, despite their large potential for statistical purpose and economic research and the establishment of agreements between the NSO and the Revenue Department to facilitate such data sharing. The main barriers to data sharing seem related to confidentiality concerns and requirements of the Thailand Data Protection Act. In this context, the Thai statistical authority may evaluate the need to refine the legislation so that all relevant statistical agencies can benefit from the cooperation with owners of administrative data which are relevant for business statistics. To this aim, Thailand could follow international practices in modernising the statistical legislation to overcome issues in exchanging administrative data with the NSO for statistical purposes (see Box 2.13 discussed below). The NSO is currently drafting a new Statistical Master Plan and revising the National Statistical Act. These initiatives aim to enhance the country's statistical infrastructure, ensuring data quality and coherence across various sectors. The updated Master Plan is expected to provide a strategic framework for statistical development and the revised law will likely address legal and institutional aspects to support effective data collection and dissemination. These changes may help overcome existing barriers in data sharing across institutions.
Third, data integration requires a sound implementation of statistical standards to repurpose them for statistical use. Countries can rely on guidelines from international organisations and international collaboration to overcome potential challenges. For instance, international organisations provide clear guidelines that can help establishing SBRs (UN, 2024[34]; OECD/Eurostat, 2008[45]; Eurostat, 2021[46]; UNECE, 2018[47]).
Finally, it may require supplementing information from administrative data with surveys collecting data for specific analytical purposes. Surveys can fill information gaps and they can also provide timely information and circumvent potential delays in the collection and statistical treatment of administrative data. For instance, in economies with a high level of informality like Thailand, the use of administrative data may not be sufficient to cover the entire business population. Surveys of the informal sector may complement such administrative data and can help measure the informal contribution to economic output and identify or describe informal firms. Countries have implemented different, non-mutually exclusive, solutions to achieve this goal. Firm-level surveys can identify informal businesses which are not registered with the relevant administrations. The Egyptian Economic Census, for example, includes a question on whether the establishment is registered and holds the necessary licences to perform its activity. This enables the monitoring of informality in the Egyptian economy and the assessment of productivity and performance of informal establishments in comparison to formal ones. Additionally, household surveys, such as labour force surveys, may help identify workers whose jobs are not covered by social security schemes. Brazil provides a good example for the latter approach. Its quarterly household survey asks respondents whether they work in the private sector without a formal contract, employ others without company registration, are self-employed without a registered business, or are a family worker without a formal contract.45 In Thailand, the NSO is conducting the annual Thai Informal Employment Survey that comprehensively includes information both on demographic and labour-related characteristics of Thai informal workers.46 The data has been used to examine informal workers in Thailand in a PIER discussion paper (see Korwatanasakul (2021[48]) and Box.3.2).
The ESANE information system from the French National Institute of Statistics and Economic Studies (INSEE, the French NSO) is an example of successful data integration that addresses all these considerations and includes administrative data together with relevant survey data to produce business statistics (Box 2.9).
The French National Statistical Office (the National Institute of Statistics and Economic Studies – INSEE) has built the ESANE system of information, combining a large array of administrative and survey data to produce business statistics. The purpose of the system is to produce every year structural business statistics on several economic aggregates: turnover (and its breakdown by activities), value added, fixed assets, sales margin, total employment, wages, and gross margin rates. According to the INSEE, the statistics collected answer the numerous and varied needs of the political and administrative authorities, the European Commission (Eurostat), French statisticians and in particular National Accountants, professionals and social partners from the business world, the media or finally the general public.
Several aspects facilitate the data integration and its use for analysis. First, France ensures by means of law that statistical authorities can access administrative data, promptly and free of charge in order to use them for statistical purposes (Statistical Law No 51-711 of 7 June 1951). Second, INSEE oversees the administrative business register, Sirene, and the use of a Siren-id is compulsory in all administrative procedures which enables different data sources to be linked. Third, INSEE is also involved in the design of administrative data.
The ESANE information system provides a rich set of information, based on integrated administrative data and surveys. The administrative files provide accounting information on the company, on employees and payroll. Additional surveys allow to determine the different activities carried out by enterprises, to detect some legal restructurings, to complement the tax data on some investment or employment variables. These surveys also inform the main features of each economic sector, unavailable in the administrative data – such as data on the commercial equipment, the kind of customers or the nature of the work carried out in construction.
Finally, the information from ESANE can be complemented by additional information, for instance using surveys on the use of ICT in businesses, on their innovation activities, and others, also thanks to the data integration and use of common business identifiers.
Source: authors’ elaborations based on information available on INSEE’s website: https://www.insee.fr/en/metadonnees/source/serie/s1188
Fostering micro-data sharing across organisations and with external researchers
In Thailand, the sharing of micro-data across public institutions and with external researchers from academia and research centres appears less systematic and more challenging than macro-data dissemination.
Thai institutions currently share micro-data based on ad-hoc agreements or memoranda of understanding (MoUs). A more systematic approach would benefit all Thai institutions. This would allow the NSO to access a broad array of data covering the business population, which can help the design of the common frame for Censuses, and ideally the construction of an SBR. More generally, this would help build a comprehensive and integrated data infrastructure that pro-productivity institutions and researchers could use for productivity and economic analysis.
External researchers from academia, international organisations and to some extent the private sector play a key role in producing economic analyses. Policymakers should leverage this resource as a low-cost way to increase technical capacity and expertise. Letting external researchers access and use the data has several advantages. The research output is useful for policymakers and advances knowledge on specific issues. Moreover, providing technical experts with access can contribute to data improvements and documentation.47 Several researchers have already used Thailand micro-data, including the Industrial Census and the CPFS, for productivity analysis. However, accessing the data generally requires ad-hoc agreements.
Thailand could offer a centralised solution for the dissemination of multiple micro-data to enhance its use. This approach can reduce costs and bureaucratic hurdles by standardising data-sharing procedures. Numerous institutions across OECD countries have implemented centralised solutions, such as research data centres (RDC) to institutionalise micro-data access. Box 2.10 summarises different international practices.
When implementing these solutions, data providers need to address two main considerations, discussed below and summarised in Figure 2.3.
Degree of anonymisation. The degree of anonymisation aims to preserve the confidentiality of sensitive information. Data providers may consider four types of data anonymisation: fully anonymised (de-identification of data and deletion of other key variables allowing identification), partially anonymised (only de-identification), non-anonymised (i.e. raw data) or perturbed data (data modified using statistical methods maintaining statistical properties). While anonymisation safeguards confidentiality, it also suppresses relevant information for research. Usually, files provided to external researchers entail a higher degree of anonymisation compared to internal-use files.
Modality of data-access. International practices lay out three possibilities for accessing micro-data: on-site access, remote access and remote execution. These options are not mutually exclusive and several countries provide multiple options at the same time. On-site access involves providing physical facilities (with access to IT infrastructure) where researchers can directly access the data but cannot extract results into their own computers. Such facilities are usually placed in safe rooms which prevent data leaks (for instance, these rooms have no internet connection and are equipped with video-surveillance). The remote execution consists in the data provider executing the statistical codes prepared by researchers, while researchers do not directly see the data. They only receive the non-confidential outputs extracted from the code. This method eliminates the need for researchers to travel to specific locations but also restricts direct beneficial interaction with the data. Outputs are consistently checked by data providers (or an external organisation) before sending them back to researchers. Finally, a remote access provides a solution for researchers to access the data directly and safely from their own computers using a VPN connection or other end-to-end access mechanisms. Figure 2.3 below summarises the main benefits and drawbacks of these solutions and provides some examples of countries using them.
Internationally used frameworks and practices that defined standards and methods, also related to these decisions, can help Thailand implement a safe micro-data access for research. For instance, several countries are following the Five Safe Framework which builds data access around five areas: safe people, projects, settings, data outputs (OECD, 2023[49]).48
In addition, several countries have established international collaborations and networks to share experiences in granting micro-data access for research and to facilitate data sharing between neighboring countries. The INEXDA Network among central banks, Eurostat and international organisations serves to share experiences and practices among countries. Similarly, collaborations like the Nordic Microdata Access Network (NMAN), the International Data Access Network (IDAN) and the Nordic Institute for Interoperability Solutions aim to increase data sharing across countries (OECD, 2023[49]).49
Implementing relevant solutions for data access incur some costs that are largely offset by the benefits previously discussed. Implementing a centralised solution involves acquiring and managing multiple databases (potentially both external and internal) and ensuring the safeguarding of their confidential information. As such, providing micro-data access for research requires extensive processing of data and thus digital and analytical capabilities. However, the benefits are large as it significantly contributes to enhancing the supply and quality of analyses supporting evidence-based policymaking.
Note: the list of countries should be considered as examples and not exhaustive.
Source: OECD elaborations based on OECD (2023[49]).
Several OECD countries facilitate the use of micro-data by defining standardised procedures that allow the preservation of confidentiality and by simplifying the data access process.
Remote Access – CASD France
CASD in France represents an exemplary approach in gathering and providing access to several administrative and survey micro-data to researchers and projects that receive accreditation. CASD provides a dedicated secure space to remotely access confidential data from several data sources, such as INSEE, different French Ministries (Justice, Education, Agriculture and Finance Ministries) and private entities. CASD is a public interest group (consortium) bringing together INSEE, GENES, CNRS, École Polytechnique and HEC Paris. The access is provided via a secure terminal called SD-BOX by means of an individual smartcard (issued at the enrolment session) or via biometric authentication. The box serves solely as terminal for accessing the secure CASD server, while all connections are made via a virtual server dedicated per project. Indeed, researchers can access only a limited set of data specific for each project. Output can be exported as automatic or manual exports, in the latter case exports are examined by CASD beforehand to ensure statistical disclosure control is correctly applied. Automatic exports are transmitted without prior review by CASD but may undergo a retrospective check by the data depositor.
Remote Execution – Research Data Centre (RDC) of the Bank of Italy
The RDC of the Bank of Italy was established in 2019 to provide access to several granular data to external researchers, including household data, firm-level data. The access to firm-level data is provided through a remote execution solution, where researchers do not directly access the data but provide the statistical codes to the Bank of Italy and receive the subsequent aggregate outputs. The staff of the RDC consistently check all outputs to avoid statistical disclosure and send the researcher any data error message.
Virtual Laboratory for remote access – StatCan
Building on the foundation of the Research Data Centre and the Canadian Centre for Data Development and Economic Research, the National Statistical Office of Canada (StatCan) launched in 2021 the Virtual Data Lab Project (VDL). The laboratory provides remote access to detailed, de-identified micro-data for research through a secure cloud-based interface. The shift to cloud-based IT infrastructure has significantly improved access to micro-data for external users in Canada, by eliminating geographical constraints to data use, and providing modern and flexible platforms (Tumpane, 2023[50]). Access to the VDL is based on an agreement between StatCan and the accessing organisation, each user receiving accreditation, and relies on an evaluation of the sensitivity of the data. Data deemed too sensitive remains in StatCan premises access locations. Projects that do not qualify for remote access, or those involving highly sensitive data in micro-data research, will still require access through a physical RDC.
Source: StatCan
Micro-level productivity analysis exists but can be improved
As discussed before, a granular analysis using micro-level (establishments or firms) is crucial to dissect aggregate productivity and identify the drivers of productivity growth.
The analysis of productivity at a granular micro-level can build on three categories of complementary and intertwined indicators:
Firm productivity: indicators related to productivity dispersion, the heterogeneity of productivity across different business types, as well as within-business productivity growth;
Resource (re-)allocation: indicators related to the (re-)allocation of resources, the state of competition and zombie firms;
Business dynamics: indicators related to creative destruction (firm entry and exit), the contribution of entry/exit to productivity and the firm-life cycle.
Figure 2.4 outlines the main indicators for each pillar to consider in conducting a comprehensive analysis of productivity at the micro level.
Source: OECD elaborations.
This section delves into the relevance of each of the three pillars and related indicators and provides an assessment of the current analyses conducted in Thailand through the lens of this framework. It also suggests potential opportunities to enhance the analysis based on the implementation of relevant indicators. An overview and summary of the set of indicators discussed below is also provided in Table 2.6.
Thai institutions – such as the Office of Industrial Economics (OIE) and the Puey Ungphakorn Institute for Economic Research (PIER) – conduct extensive analysis on productivity using the available micro-data. These current analyses serve as a foundation for further development and enhancements. While data constraints limit certain aspects of the analysis, there is significant potential for actionable insights and improvements to support policymaking.
Strengthening firm productivity analysis
Measuring the heterogeneity of productivity across businesses, understanding how this relates to their characteristics and assessing the extent to which businesses improve their productivity are the first steps of the analysis of firm productivity. Some Thai institutions already perform analysis along these dimensions, leveraging mainly the Industrial Census and OIE’s survey of the manufacturing sector. However, there is still scope for broadening the analysis and conducting it more systematically across Thai institutions. This would nonetheless require overcoming data limitations. This report reviews three categories of indicators that are key to conduct this analysis: indicators related to (i) productivity dispersion and its evolution, (ii) firms’ heterogeneity according to their characteristics, (iii) within-firm productivity growth.
First, Thai analyses insufficiently focus on productivity dispersion, i.e. the gaps between leading firms at the frontier and firms lagging (“laggards”) both cross sectionally and over time. Productivity dispersion is crucial for policymakers. Large disparities in productivity within industries can arise from issues such as policy distortions, financial frictions that reduce investment, a scarcity of demanded skills, or many other issues that may require the attention of policymakers. Conversely, a significant productivity advantage among leading firms may indicate successful innovation and other positive factors. In addition, productivity dispersion is tightly linked to inclusiveness, as evident from the connection between divergence in productivity and wages (Berlingieri, Blanchenay and Criscuolo, 2017[51]). It is therefore essential to analyse productivity dispersion comprehensively. This includes segmenting the dispersion into gaps between top-performing firms and those with medium performance, as well as the distance between the least productive firms and the rest. Such segmentation provides nuanced insights, for instance, helping to determine whether productivity gaps are driven by innovative leaders who can uplift the industry and contribute to aggregate productivity, or by a prevalence of inefficient firms, signalling weak market selection and slow diffusion of technology and knowledge. Additionally, understanding productivity dispersion is relevant for addressing inclusiveness, as differences in firm productivity contribute to wage inequalities.
Beyond simply observing productivity dispersion at any given point in time, it is essential for policymakers to monitor the evolution of productivity levels and gaps between firms over time. Tracking these changes can reveal important insights. Rapid growth at the productivity frontier (the best performers in an industry) may indicate strong innovation and knowledge development, while a stagnating frontier could signal the opposite (Andrews, Criscuolo and Gal, 2016[52]). At the lower end, observing the least productive businesses can inform about the diffusion of technology as well as market efficiency in selecting the most productive firms and facilitating the exit of less productive ones. OECD analyses highlight the importance of these indicators (see for instance Berlingieri, Blanchenay and Criscuolo (2017[51])). They show that the aggregate productivity slowdown in OECD countries is characterised by increasing gaps, especially among the least productive firms. This trend indicates difficulties for these firms to keep pace with the best performers, particularly in digital and intangible-intensive sectors. Based on these findings, the OECD recommends policies to enhance technology diffusion and knowledge transfer, alongside initiatives to foster innovation and advance the productivity frontier.
Monitoring productivity dispersion and its evolution is becoming more common across countries and would be highly relevant for Thailand. However, the existing indicators fall short in capturing the dynamic evolution of productivity differences over time. The OIE report (2022[53]) provides some insights into the differences between the best and worst performing firms and the characteristics of firms with different productivity levels. However, the analysis measures neither productivity dispersion directly nor its evolution. Importantly, the data used in the report covers only manufacturing. It may not be fully suitable to provide a comprehensive assessment of the heterogeneity across businesses due to the limited coverage of small and young firms previously discussed. A better monitoring of the evolution of dispersion, the growth at the frontier and of the least productive firms would align with the 13th NESDP's strategies to promote innovation as well as technology and knowledge adoption. National statistical offices and pro-productivity institutions across the world increasingly engage in this endeavour. Recognising the importance of such data, the U.S. Bureau of Labor Statistics and the U.S. Census Bureau introduced an experimental data product, the Dispersion Statistics on Productivity, which measure productivity difference across establishments (see Box 2.11). This offers a model for the implementation of dispersion statistics.
A second approach to enhancing the analysis of firm productivity in Thailand would be to better connect the heterogeneity of productivity across businesses to their specific characteristics, leveraging existing evidence and extensive data available in Thailand. This approach achieves several objectives. It offers a comprehensive overview of the business landscape based on productivity, revealing systematic patterns and identifying leaders and laggards. For instance, studies uncover consistent productivity variations within sectors across different firm sizes, notably in manufacturing, and across age groups or other dimensions such as ownership structures. Furthermore, delving into the attributes of firms with varying productivity levels contribute to the understanding of their productivity determinants and illuminates the potential of firms to undergo productivity improvements. The Industrial Census offers rich data to examine heterogeneity in productivity performance across firms, mainly in a cross-sectional approach.50 Analyses have partially examined these differences (Paweenawat, Chucherd and Amarase, 2017[54]), and the World Bank and Bank of Thailand (2020[14])). However, they lack comprehensive examination of the most updated census waves to help discerning the persistence of these disparities.51 To date, analyses also lack investigations of productivity gaps across businesses in the service sector.
Since 2019, the U.S. Bureau of Labor Statistics (BLS) and the U.S. Census Bureau publish statistics on establishment-level distributions of real gross output per hour worked and total factor productivity. The DiSP database covers all 86 4-digit North American Industry Classification System (NAICS) manufacturing industries and provide annual data for the period 1987-2020. These statistics are based on the combination of primary data sources including micro-data from the Census Bureau's Annual Survey of Manufactures (ASM), Census of Manufactures (CM), and Longitudinal Business Database (LBD).
DiSP provides a range of complementary statistics that picture the productivity distribution of each industry. This includes standard deviations, inter-quartile ranges (comparing productivity at the 75th and 25th percentiles), and inter-decile ranges (comparing productivity at the 90th and 10th percentiles) of the within-industry distributions of establishment-level productivity levels. DiSP also provides comparisons between the 10th and 1st percentiles and between the 90th and 99th percentiles of the distribution. The methodology and data sources are extensively discussed in Cunningham et. al., (2022[55]), and readers can find further information on the dedicated webpage.
The U.S. BLS summarises relevant findings using these data:
Productivity differences across establishments are large within detailed industries. On average, an establishment at the 75th percentile of the labour productivity distribution is more than twice as productive as an establishment at the 25th percentile of the distribution.
Dispersion is also significant among the most and least productive establishments in an industry. For instance, the 1% most productive establishments are not only significantly more productive than average-productivity firms but also stand out within the group of the 10% most productive establishments, revealing the presence of “superstar" firms.
Productivity dispersion has increased between 1997 and 2020.
Most of the within-industry dispersion in productivity is not accounted for by standard establishment characteristics like business size, age, or location suggesting that other factors are at play.
Researchers have used these data in various ways. Cunningham et al., (2021[56]) show that changes in dispersion are related to periods of innovations and surging entry rates with rising dispersion and potentially declining aggregate productivity growth, followed by a shakeout process characterised by declining dispersion and higher aggregate productivity growth. Blackwood et al. (2022[57]) explore the role of tasks and skills mix in contributing to differences in productivity and find that within-industry productivity dispersion is strongly positively related to within-industry task/skill dispersion.
Analyses could also include a broader set of factors, following a rich literature from academia and public institutions. Syverson (2011[58]) gives an overview of the determinants of differences in productivity performance which provides a relevant starting point to structure the analysis.52
Finally, the analysis of within-business productivity growth remains underexplored. Improvements in individual firms' productivity significantly enhance overall economic performance, driven by their capacity to innovate, adopt new technologies, implement organisational changes and upgrade workforce skills. Examining within-firm performance can uncover obstacles to these improvements, such as barriers to technology and knowledge diffusion across sectors, or, conversely, identify factors that boost productivity growth. Despite its limitation, the OIE Manufacturing Survey seems to be the best source available for measuring within-business productivity growth, as well as analysing correlated factors and potential determinants such as R&D, skills and technology adoption. For instance, Sangsubhan, Pornpattanapaisankul and Kambuya (2023[38]) use the OIE Manufacturing Survey to demonstrate the positive impact of automation adoption on firm multifactor productivity. Thai institutions or researchers could further leverage this dataset to investigate other determinants of firm growth. The information from the OIE report may offer relevant insights into the role of input quality, management practices, innovation and finances.
Strengthening the analysis of allocative efficiency and the state of competition
The lack of allocative efficiency appears to be an issue for the Thai economy. Thai institutions responsible for productivity analysis could monitor this more closely by replicating existing studies of misallocation on more recent data, extending the evidence to the service sector, and integrating to productivity analyses additional and readily available indicators measuring the extent of reallocation. This set of indicators could also more comprehensively integrate measures of the strength of competition. These indicators are essential to identify and eliminate policy distortions and assess the productivity potential of policies that favour reallocation and competition. Four categories of indicators are particularly relevant for this analysis: indicators on (i) allocative efficiency, (ii) re-allocation of resources across business, (iii) state of competition and (iv) zombie firms.
Existing evidence indicates that the misallocation of resources is a pressing concern for policymakers. The evolution of this issue could be investigated further to inform policy action. Paweenawat et.al. (2017[54]) uncover a large effect of the misallocation of resources on productivity in Thailand between 1996 and 2011. Using the methodology developed by Hsieh and Klenow (2009[59]; 2014[60]), they estimate that total output could increase by 75–132% if the Thai manufacturing sector had the same allocative efficiency as the U.S. Their analysis identifies that this issue may arise from distortions linked to size-contingent policies and the presence of state-owned enterprises.53 Thai institutions could conduct further analysis to update this evidence and extend it to services, but also to identify other sources of distortions or inefficiencies. Indeed, various factors can also affect allocative efficiency, beyond those already analysed. For instance, employment legislation, incomplete tax enforcement, product market regulations, or imperfect financial markets can also influence the allocative efficiency (see Berlingieri et al. (2017[61]), Andrews and Cingano (2012[62])). Further to the Hsieh and Klenow (2009[59]; 2014[60]) methodology, Thai institutions may also rely on other measures (easier to implement and interpret) to assess allocative efficiency and understand the influence of different factors. For example, the Olley and Pakes (1996[63]) measure evaluates the extent to which more productive firms receive more resources by examining the strength of the link between productivity and business size.54 Thai agencies and researchers can also use these indicators to explore sources of inefficiencies but also to evaluate the effect of potential reforms (see Gnocato, Modena and Tomasi (2020[64]) for an example).
In addition to allocative efficiency at a given point in time, Thai agencies should also assess the strength of the reallocation of resources across businesses. Productivity-enhancing reallocation occurs when more productive businesses expand while less productive businesses contract, as well as through firm entry and exit (discussed later). The contribution of reallocation to productivity can be directly assessed through decompositions of aggregate productivity growth. For instance, Paweenawat et.al. (2017[54]) have applied the widely used Melitz and Polanec (2015[65]) decomposition to the Thai manufacturing sector and found that dynamic reallocation positively contributes to aggregate productivity growth. Productivity analyses by Thai agencies could also replicate the exercise on more recent editions of the Industrial Census, and possibly extend this to the service sector as well. Beyond this decomposition, job reallocation rates are a relevant indicator of the extent of strength of this process, capturing the simultaneous creation and destruction of jobs within sectors. Changes in market shares and employment shares of firms with varying productivity levels and growth rates can assess whether shifts in resources are enhancing productivity. However, data limitations, previously discussed in the report, are hampering the computation and analysis of these indicators, which are currently missing for a broad diagnostic and understanding of Thailand’s productivity.
A competitive environment is key for productivity but measures of competition in Thailand display concerning trends that warrant a closer monitoring. Apaitan et al. (2020[39]) reveal that economic dynamism has decreased over time in Thailand, while the market power of Thai firms has been increasing. The study also finds that this evolution is negatively associated with firms' investment levels, raising concerns for future productivity growth. Thai institutions could contribute to update evidence of these trends, as the information is currently available until 2016 only. In addition, they could broaden the range of competition measures analysed. Indeed, measuring competition is challenging and requires combining approaches. A recent OECD report offers an in-depth analysis of the evolution of competition in Europe, providing a valuable guideline on how to measure and combine relevant indicators (OECD, 2024[66]). This includes indicators of markups and sales concentration (also discussed in Apaitan et al. (2020[39])) but also complementary measures of entrenchment, which reflect the turnover in market leaders over time and could easily be implemented using the CPFS data.
Finally, Thai institutions could also pursue the analysis of zombie firms and further investigate their implications for productivity. Despite evidence for an overall positive contribution of reallocation, information on the profitability of businesses suggests that zombie firms may hinder efficient resource allocation and productivity growth in Thailand. Zombie firms, characterised by persistent difficulties in servicing debt due to low profitability and productivity, represented 9.1% of firms in Thailand in 2019, according to SCB (2020[67]), with an upward trend since 2009. Recent evidence indicates that the rise of zombie firms may continue to pose challenges for the Thai economy (O’Connor, 2023[68]). This may require policy action, discussed in SCB (2020[67]), such as improving business and labour market dynamism and removing barriers to competition, addressing debt overhang along with continuous debt restructuring and supporting innovation. Thai institutions could rely on the methodology established by the Economic Intelligence Center of the Siam Commercial Bank in its report on zombie firms to report the most recent trends and projected evolution of the share of zombie firms more frequently. However, the current analysis by (SCB, 2020[67]) does not allow to gauge the productivity benefits of curbing the share of zombie firms. This understanding would help raise attention on the extent to which this is a pressing issue for the Thai economy, thereby contributing to prioritisation of policy intervention. This evaluation requires an estimate of the share of total capital and employment trapped in zombie firms (McGowan, Andrews and Millot, 2017[69]) and an understanding of how this affects investment of more productive firms (Banerjee and Hofmann, 2018[70]).55
Strengthening the analysis of business dynamics and creative destruction
Creative destruction plays a crucial role in driving productivity. This involves a continuous process of old, less productive firms being replaced by newer, more innovative ones. New technologies and business models can disrupt existing markets, forcing firms to adapt, innovate, or exit the market. These dynamics contribute to bringing innovations to the market, to the reallocation of resources to more productive uses, and contribute to maintaining a competitive pressure on incumbent firms and market leaders, which incentivise their innovation and investment.
Despite its relevance for productivity, creative destruction remains understudied in Thailand. It features prominently in some research conducted by the PIER, but this remains insufficient to inform policymaking. This evidence needs to be updated and extended. Three categories of indicators are relevant for the analysis: (i) indicators of the evolution of entry and exit, (ii) indicators of the contribution of entry and exit to productivity and (iii) indicators of firms’ life cycle, including post-entry growth and market selection. Data limitations are still impairing a broad implementation of these indicators and the integrated data framework previously discussed in section would generally address these issues. The relevance and implementation of these indicators for Thailand is discussed in more detail below.
The Thai economy has experienced a decline in firm entry and exit rates that requires further attention and analysis by Thai institutions. Entry and exit rates are crucial metrics of the business environment's dynamism. High entry rates suggest a dynamic entrepreneurial environment and potential for innovation, while exit rates illuminate the efficiency of market selection processes. The sum of entry rates and exit rates provides a measure of firm churning, helping to understand the extent of resource reallocation that occurs through creative destruction. Apaitan et al. (2020[39]) show that entry rates and exit rates have declined over the period 2004-2016, in line with long-term trends observed in the US (Decker et al., 2016[71]) and OECD countries (Calvino, Criscuolo and Verlhac, 2020[72]). Further analysis may provide updated data for the most recent period and a better understanding of the drivers of these trends, focusing on structural and policy determinants, following for instance Calvino, Criscuolo and Verlhac (2020[72]). Thai institutions should provide entry and exit indicators on a regular basis, also by detailed industries, to spur such analyses. These indicators are widely monitored across countries and international organisations. The OECD/Eurostat Manual (2008[45]) provides guidelines for the implementation of business demography statistics.56 Currently the CPFS data would be a primary source for computing these indicators, but setting up a Statistical Business Register (SBS) would be the preferred option to establish official business demography statistics.57
Further to establishing business demography statistics, Thai institutions may assess the contribution of creative destruction to productivity growth, requiring some data enhancements. Assessing this contribution requires comparing the productivity of new entrants and exiting firms with that of incumbent firms. Such contributions have been calculated for Thailand (World Bank and Bank of Thailand, 2020[14]) using different waves of the Industrial Census and point to a positive contribution of creative destruction to overall productivity growth. However, the data used and available may not allow to fully evaluate the contribution of creative destruction. In particular, the Census does not cover the whole population of micro-establishments and is therefore not suitable to fully capture total entry and exit of establishments.58 On the contrary, the CPFS data are suitable to analyse entry and exit (of incorporated businesses) but do not contain all the information for productivity analysis.
To complete the picture of business dynamics and creative destruction in Thailand, analyses could further investigate the firm’s life cycle, i.e. the extent to which businesses improve their productivity and grow after entry and as they age or exit the market otherwise. Creative destruction is intricately intertwined with the process of experimentation and market selection. Typically, businesses enter with relatively low productivity levels and many firms exit the market in the first three years of existence (Calvino, Criscuolo and Menon, 2015[73]). However, successful entrants are expected to undergo rapid improvements in productivity, accompanied by significant growth in size (employment and sales). Paweenawat et.al. (2017[54]) provide evidence on this by investigating the life cycle of firms using the Industrial Census and show that barriers to firm growth contribute to the resource misallocation issue previously discussed. The indicators they use are well-established in the literature and future productivity analyses should focus on updating them using more recent data and extending the methodology. They could in particular provide updated evidence on the productivity-age profile and could further adopt the synthetic cohort approach presented in Hsieh and Klenow (2014[60]) to refine the analysis.59 However, further desirable extensions for an in-depth understanding of the firm life cycle and market selection in Thailand remain limited by the data available for analysis. For instance, a more direct approach necessitates tracking firms across different time periods and measuring post-entry employment and productivity growth to gauge whether young firms exhibit "up or out" dynamics (Criscuolo, Gal and Menon, 2014[74]). Linking firms’ likelihood of ceasing operations with their productivity levels and ascertain whether surviving low-productivity firms manage to enhance their productivity would also provide relevant insights.
A detailed understanding of productivity requires suitable micro-data, that also serve more generally to establish sound business statistics. In Thailand, there is a rich set of data that can directly, or indirectly, contribute to productivity analysis at a granular level. This includes censuses, survey data and administrative sources collected by several Thai institutions, including the National Statistical Office, the Office of Industrial Economics (from the Ministry of Industry), the Department of Business Development (Ministry of Commerce), the Revenue Department (Ministry of Finance) and several other institutions. However, Thailand does not yet fully leverage the potential of these micro-data, especially from administrative sources.
Despite the availability of relevant micro-data, Thailand needs to address important limitations of these sources for productivity analysis and more broadly for business statistics. Important limitations remain in terms of the frequency of the available data, business coverage and information provided for productivity analysis. Recommendations 7 to 9 discuss how Thailand can enhance business statistics, through better data linkages and business identification, the further integration of administrative data, and the creation of a Statistical Business Register.
Despite their limitations, existing micro-data already enable the computation of some key indicators for the three pillars of productivity analysis at a granular level: firm productivity, allocative efficiency and business dynamics. However, existing analyses do not systematically compute and analyse these indicators which are mostly limited to ad-hoc research. Extending and regularly updating these analyses would provide key insights to inform policies in favour of productivity. Furthermore, advancements in data capacity are necessary (discussed in recommendations 7-9) to expand the scope of certain indicators, such as gaining a deeper understanding of business demographics and growth. Recommendation 10 presents how Thailand can tap into the existing potential of micro-data to provide more comprehensive productivity analysis relying on an analytical framework.
Micro-data currently supports economic research on productivity, but Thai pro-productivity institutions should encourage a more systematic use of such data to support evidence-based policy. The lack of a more systematic, and ideally centralised, solution for sharing micro-data may hamper its broader adoption in economic analyses of productivity in Thailand, both among Thai institutions, external researchers and international organisations. This shortfall deprives policymakers of a valuable source of evidence and limits researchers' ability to contribute to improvements in data quality and analysis. Recommendation 11 presents how Thailand can boost the use of micro-data through appropriate dissemination solutions.
Enhancing data linkages and leveraging more systematically administrative sources to improve business statistics
This report identifies three main data sources for granular productivity analysis in Thailand. These are the Economic Censuses from the National Statistical Office (NSO), the Corporate Profile and Financial Statements from the Ministry of Commerce and the Manufacturing Survey (Annual Factory Operation Information Form) of the Office of Industrial Economics. However, each data source presents some limitations for productivity analysis. Relevant limitations are the low frequency of data collection, limited possibility to follow individual businesses over time, limited coverage of the business population and the lack of key information for calculating productivity, such as employment in the Corporate Profile and Financial Statements data.
Linking currently available data sources can help address limitations of the existing data for micro-level analysis (on productivity and beyond). However, this is currently hampered by the use of different business identifiers across several sources, as these identifiers may be specific to each institution. While Thailand can work on improving data linkages within the current system of identification, relying on the expertise of the NSO and administrative data owners, a key improvement for business statistics would be to set up a unique business identifier across administrations (recommendation 7).
A complementary target to improve business statistics would be to further repurpose rich administrative data for their integration into statistical information systems (recommendation 8) and to pursue the development of a Statistical Business Register (SBR), as discussed in recommendation 9.
Recommendations 7 to 9 collectively establish a comprehensive strategy for building an integrated business statistics system in Thailand based on micro-data. Before presenting each recommendation individually in more detail, the summary below discusses how they can be combined in a unified approach.
Recommendation 7 focuses on establishing a consistent business identification system across public entities to enable data linkages. This unique identifier serves as a foundation for integrating data from various sources, such as tax records and social security data, but also ensuring the consistent longitudinal identification of businesses over time.
Recommendation 8 complements this by advocating for a more frequent and consolidated use of administrative data for statistical purposes. By improving data access and developing standardised methodologies for repurposing these data, agencies can better link and harmonise information using the identifiers from recommendation 7. This process enables more accurate and comprehensive productivity analysis and supports the broader use of micro-data in business statistics.
Recommendation 9 also builds on the previous two by proposing the development of an SBR. The SBR would leverage the standardised business identifiers and enhanced administrative data access to compile a comprehensive register that consolidates business information across multiple sources. This register would facilitate consistent data usage, improve the quality of national statistics, and contribute to more accurate productivity measurement but also measurement of related indicators, for instance on business demography.
Together, these recommendations form an interconnected roadmap: the consistent business identification from recommendation 7 enables data linkages for enhanced administrative data usage outlined in recommendation 8, which, in turn, lays the groundwork for creating a comprehensive SBR as described in recommendation 9. At the same time, the SBR can in turn facilitate the integration and use of new sources, further facilitating the implementation of recommendation 8. Furthermore, reinforcing inter-agencies collaboration and data exchange can help identify opportunities and challenges to the system of unique business identifiers. Implementing these recommendations jointly ensures that each step supports and strengthens the next, fostering a cohesive and efficient business statistics system in Thailand.
For each recommendation, the implementation is organised in two sequential phases. The implementation of phase 1 of recommendations 7 to 9 can, to some extent, be carried out simultaneously. This approach allows multiple recommendations to progress in parallel, speeding up the overall implementation process. However, the full implementation of the SBR and integration of data for business statistics (recommendation 9) requires the effective achievement of previous recommendations 7 and 8.
The sequencing described below proposes how each phase of recommendations 7 to 9 can be further decomposed into different steps articulating the different recommendations.
Phase 1: Identifying challenges and roadmaps for data integration in business statistics, based on the following steps:
Step 1: Assessment of current barriers to data integration and strengthening of inter-agency collaboration. This initial step integrates elements from phase 1 of recommendations 7 to 9, aiming to provide a comprehensive evaluation of existing barriers to data integration. The cornerstone of this process is the assessment of the maturity of the Statistical Business Register (SBR), as outlined in phase 1 of recommendation 9. This assessment will help guide and structure the audits proposed in phase 1 of recommendations 7 and 8. The focus should be on examining the legal and institutional frameworks related to data sharing, as well as identifying challenges with data inter-operability and identifying potential obstacles to establishing a unique business identification system. In addition to identifying these barriers, this step should enhance inter-agency dialogue and collaboration, which is necessary for the successful implementation of subsequent steps.
Step 2: Develop a roadmap to address barriers to data sharing and a roadmap for the transition to a unified business identification system.
Based on the findings from the initial assessments (step 1), the next step involves creating a detailed roadmap for overcoming data-sharing barriers (as recommended in phase 1 of recommendation 8) and a roadmap for transitioning to a unique business identifier (as recommended in phase 1 of recommendation 7). The roadmaps should outline necessary amendments to the legal framework for data sharing, and potential amendments to the business registration system. Additionally, the roadmaps should propose technical solutions to enhance data inter-operability.
Step 3: Develop a roadmap for the creation and development of the SBR. Mandate the NSO with the responsibility of developing the SBR and outline a roadmap for its creation (phase 1 of recommendation 9), building on the previous SBR maturity assessment (step 1). This plan should align with the roadmaps defined in Step 2 which determine the feasibility of SBR development. Beyond defining the implementation roadmap, Thailand should start the development of standardised methodologies for using individual administrative data for statistical purposes.
Phase 2: Implementation of the SBR and annual business statistics:
Step 1: Initiate the SBR. Develop an initial SBR leveraging available data sources that the National Statistical Office can currently access and relying on methodological development discussed in step 3 of phase 1. To this aim, the NSO should establish and standardise methodologies for harmonising and reconciling information across diverse data sources to ensure data consistency and accuracy (see also phase 2 of recommendation 9).
Step 2: Address barriers to administrative data-sharing (phase 2 of recommendation 8). Implement necessary updates and modifications to the existing legal framework to facilitate data sharing between relevant entities.
Step 3: Phase-in the unique business identification system (phase 2 of recommendation 7). Gradually phase in the unique business identification system to support comprehensive integration of data and support the longitudinal identification of businesses over time to support the analysis.
Step 4: Fully implement the SBR and annual business statistics and complement this with regular surveys on informal sector. While step 1, 2 and 3 of phase 2 could to some extent proceed in parallel, step 4 requires the effective implementation of step 1 and step 2 and could be facilitated if step 3 is implemented first.
Recommendation 7: Facilitate analysis and data linkages through consistent business identification.
To achieve consistent identification of businesses across data sources and over time, all public administrations in Thailand should use common identifiers when collecting data on businesses. This common identifier is generally “structured as a set of numeric or alphanumeric characters, unique to the business to which it has been allocated, and remain unchanged and is not reallocated following any deregistration of the business” (UNCITRAL, 2019[75]). The basis of a unique business identifier might be the tax authority’s identification number, the business registration number, or another existing identifier, but can also be newly created.
The availability of unique longitudinal identifier is key to enhance the measurement and analysis of productivity by Thai institutions and researchers more broadly. Ensuring that data users can follow individual businesses over time will provide a more comprehensive assessment of within-business productivity growth, as discussed in recommendation 10 (this is particularly relevant, for instance, in the Economic Censuses of the NSO). Moreover, the use of common identifiers across various data sources would ensure that users can match businesses across different sources. A common business identifier would thus significantly improve the feasibility of measuring firm-level productivity at an annual frequency and for a broad coverage of the business population, for instance linking balance sheet information from the CPFS data, with information on employment from social security data, but also information from corporate income tax data. It would also improve analysis of factors related to productivity determinants available across data sources, for instance allowing to link administrative data also with surveys on R&D expenditures in businesses and surveys on the use of ICT.
A common business identification number across all public administrations also provides several advantages that go beyond statistical purposes. For instance, it facilitates the inter-administrative exchange of micro-data, improves data quality (for instance reducing duplications and omissions of units), but also simplifies administrative procedures for enterprises (UNCITRAL, 2019[75]). As an example, it can contribute to integrated registration systems that improve the overall business environment through reduced administrative burden for business creation (World Bank, 2016[76]). Evidence from other countries shows that it can also contribute to increase tax compliance and the number of registered firms, contributing to addressing challenges related to informality (World Bank, 2016[76]). Thus, further to improvements of measurement, this recommendation aligns well with a broader pro-productivity policy agenda that should aim at providing appropriate framework conditions for businesses (see also the analytical framework for pro-productivity public policies presented in Chapter 3).
Setting up a system for a consistent business identification involves several considerations (see also details in Box 2.12 together with examples of practices in other countries) and coordination among relevant entities. To implement this recommendation, Thailand can rely on international guidelines and international cooperation. For instance, Thailand can follow the World Bank’s guidance note and case studies on implementing a unique business identifier in government (World Bank, 2016[76]). Broadly, the implementation can focus on short-term objectives with immediate actions in phase 1, and medium-term objectives in phase 2 (see for instance (World Bank, 2016[76]) for a practical and detailed roadmap):
Phase 1: Make existing longitudinal business identifiers available for analysis and develop a roadmap to set up a unified system of business identification. First, in the context of productivity analyses, all Thai institutions collecting relevant micro-data should ensure that existing business identifiers are made available to conduct analyses on these data. This identifier can be appropriately anonymised if needed, when sharing data with external researchers (see recommendation 11 on micro-data dissemination).
Then, the first stage of implementing a unique business identifier corresponds to the design of the system and the definition of a roadmap. This phase should address key questions, such as assessing the need for a new identifier or the possibility to use an existing one, the need for changes in relevant legislation, but also identifying challenges in the interoperability across agencies. This phase should lead to the development of a clear roadmap. Achieving this objective requires a coordination across all relevant entities collecting business information (including – but not limited to – the NSO, the Ministry of Commerce, the Ministry of Finance, the OIE and the Ministry of Labour). Given its role as a central state agency in charge of the technical statistics work, the NSO may play a key role in coordinating the (re-)designing of administrative data in a way that can facilitate data linkages and business identification. Additionally, the productivity council can promote this agenda at the highest levels of the government and can to some extent further support the coordination of relevant agencies by contributing to the organisation of inter-institutions consultations, workshops and meetings with national and international experts.
Phase 2: Implement unique business identifiers across all public entities. After the initial phase and the adoption of a roadmap, Thailand should pursue the efforts towards the implementation of the unique business identifier by designing adequate solutions, ensuring their development and rollout. This includes, for instance, defining the structure of the identifier for legal entities and their branches, the solutions of interoperability across agencies, implementing the necessary legislative changes previously identified, and designing solutions to convert existing data to the new system. The responsibilities of relevant agencies will be defined in the initial planning phase, but the NSO and the productivity council could continue to play a key role in promoting this agenda and facilitating inter-agencies cooperation.
The development of a national business identification system generally considers a harmonised system for legal entities and their branches. The identification system typically includes a unique number for legal entities and another one for their branches (this number is often based on a company number with an additional code for the local units, i.e. branches). The identification number should be generated immediately after registration with the relevant authorities (UNCITRAL, 2019[75]). The same unique identifier should then be used by all public authorities to share information about that particular registered business. In South Korea, a 10-digit Business Registration Number (BRN) is assigned by the tax office to entities starting a business (UNECE, 2018[47]). The BRN is also used for taxation purposes. A 13-digit Corporation Registration Number (CRN) is, instead, assigned to legal persons registered at Courts. Statistics Korea currently uses both business identifiers for the identification of statistical units (enterprises) and data linkages.
Relevant legislation can establish such a business identification system. The legislation can specify the national identification system and should include a legal mandate of the public authorities to use the unique identifier (UNCITRAL, 2019[75]). For example, the articles R123-220 to R123-234 of the Commercial Code in France establish a national identification system for natural and legal persons and their establishments based on the business and establishment register, known as Sirene.
Developing this system requires coordination of several authorities. The development of a national identification system is not a simple process and involves coordination of several stakeholders. In Denmark, for instance, the establishment of a group of relevant ministries and other organisations proved instrumental in overcoming challenges related to developing a unique national identification number (UN, 2024[34]).
Source : INSEE
Recommendation 8: Bolster the use of administrative data for statistical purposes.
Administrative data have emerged as a key source of information for statistical and economic analysis. Thailand could follow international best-practice by leveraging this information more systematically for productivity analysis and broadly for business statistics. Additionally, the NSO may target an integration of these sources together, also with information from statistical surveys, to produce structural business statistics. This would further strengthen National Accounts data and other official data at the macro-level (that builds, across countries, on a variety of primary and secondary sources, including firm-level data). This would also improve productivity measurement at aggregate levels (previously discussed in recommendation 6). Additionally, promoting the use of administrative data is key for the creation of a Statistical Business Register (recommendation 9).
Thailand should thus ensure that existing valuable administrative data are used for statistical and economic analysis across agencies and especially to the National Statistical Office. This relies on two main dimensions:
Dimension 1: Data access. Ensuring access to relevant administrative data, for the purpose of improving productivity statistics, but also as part of the creation of the Statistical Business Register discussed next.
Dimension 2: Statistical use of administrative data. Ensuring that the administrative data are suitable for statistical purpose. The NSO should thus focus on evaluating the quality of the data and repurposing the data for statistical usage as necessary. This includes resolving problems related to the definition of units, removing duplicated observations, treating anormal values or partial non-response within individual sources and resolving inconsistencies across different sources.
Increasing the use of administrative data based on these two dimensions may require developing relevant legal, technical and organisational frameworks as well as investments in human capital. Again, international guidelines provide detailed discussions of principles and practices on which Thailand should rely (see for instance UNECE (2011[77]) and the Collaborative on Administrative Data from the United Nations Statistics Division and the Global Partnership for Sustainable Development Data)60. Further to these guidelines, this report identifies two phases relevant for productivity analysis.
Phase 1: Scope the need for further inter-agencies cooperation, assess current barriers to data-sharing and develop methodologies for the statistical use of administrative data. Thailand should ensure that the NSO (but also possibly other relevant stakeholders) can routinely access administrative data for statistical purpose. Regarding data access, Thailand can rely on the effective implementation of the 2007 Statistical law and the Statistical Master Plans (including its current revision) that empower the National Statistical Office to access administrative data and on the Government Data Exchange that already provide technical solutions for exchanging confidential data across agencies. The NSO already accesses data from the Department of Business Development (Ministry of Commerce) and the Social Security Office (Ministry of Labour) based on MoU. However, it is also crucial for the NSO to secure a long-term access to corporate and personal income tax data from the Revenue Department (Ministry of Finance). To consolidate the access of the NSO but also more broadly of other agencies to micro-data, Thailand could (for instance under the joint initiative of the NSO and the productivity council) organise a consultation of stakeholders to assess the potential need, use and legal, administrative and technical barriers to sharing micro-data. Generalising existing practices (for instance of MoU between agencies) may help address these barriers. More structural barriers should be addressed in the second phase of implementation.
Beyond the consolidation of data access, the NSO should develop standardised procedures for repurposing these administrative data for analysis. While the NSO could, in a second phase, centralise the statistical treatment of administrative data, it could initially provide guidelines and share expertise with other agencies on how administrative data can be used in analysis. This may require developing further expertise in the NSO (for instance by expanding dedicated human resources).
Phase 2: Address barriers to data-sharing previously identified and consolidate methodologies to integrate administrative data into a structural business statistics system. In the medium term, the Thai National Statistical Office could act as a central actor to create a broad statistical network and obtain timely data from a broader range of public agencies. To this aim, Thailand may have to overcome legislative barriers and modernise its statistical legislation to ensure that some key confidential data, such as tax records from the Revenue Department, are shared with the NSO and other relevant agencies. Indeed, barriers to data sharing across Thai agencies remain. In particular, legal barriers may relate to insufficiently explicit statistical law or conflicting legal acts. To ensure that relevant data, such as tax data, are shared for statistical purposes, Thailand can follow international practices and further legitimise the use of administrative data by NSO for statistical purpose by modernising its statistical law (see Box 2.13 discussing how other countries overcome issues in sharing confidential administrative data). To ensure the full effectiveness of the legislation and its modernisation, it is important that this agenda is supported at a high governmental level. Additional specific legislation or other practices to avoid disclosure (e.g. restricting data access to particular teams within agencies) may also reinforce the use of sensitive data.
In addition to data access consideration, the NSO should continue to develop sound methodologies to repurpose administrative data for statistical use and to integrate these sources together. This implies for instance linking data (see recommendation 8 on the system of business identifiers), but also ensuring consistency and complementary across sources. An ambitious target would be to organise these data in a single, harmonised and comprehensive system of business statistics with a single output file (see discussion of the French ESANE system in Box 2.9).
In Thailand the exchange and use of personal data is restricted by the Personal Data Protection Act, preventing agencies from using and connecting databases for statistical purposes on a broad scale. Thus, legal constraints may limit access to taxpayer information from the Revenue Department and to the in-depth population data from the Ministry of Interior. To address those challenges, Thailand could follow international recommendations on statistical legislation provided by UNECE (2019[78]) and other countries’ practices in modernising statistical legislation.
The modernisation of the statistical law can legitimise access to administrative data. Several countries now include in their statistics law a mandate for the National Statistical Office to access – or an obligation to use – administrative data sources for statistical purposes (UN, 2024[34]). The “use for statistical purposes” is a key justification for engaging in data exchange. The “use for statistical purposes” is defined as “the exclusive use of data for the development, production, dissemination and communication of official statistics, quality improvement, statistical analyses and statistical services, including all activities regulated by the statistical law” (UNECE, 2019[78]). Relevant examples of NSO successfully gaining access to administrative data can be found in Box 6.5 of UN (2024[34]).
Specific legislation can reinforce data-sharing for more sensitive data. Even when the statistical law provides a mandate to the NSO to access and use administrative data, a specific legislation may reinforce data sharing of some sensitive information. For example, the NSO of the Republic of Slovenia (SURS) has a clear mandate to use administrative data for statistical purposes in the National Statistics Act (NSA, article 32). In addition, for some specific data – including tax data – there are other specific-sector legislation that reinforces the mandate of the SURS to organise the transmission of such information (UNECE, 2019[78]). Some countries also strengthen their data sharing frameworks with memoranda of understanding (MoUs). Ideally the legislation should detail the requirements and restrictions for data sharing, to maintain trust and ensure continuous access to the data (IMF, 2018[44])
Other solutions can help preserve the confidentiality of information. A practical solution to overcome confidentiality when sharing sensitive data, such as tax data, can be to restrict the data access to a limited group within the NSO (or within other relevant statistical agencies) (IMF, 2018[44]), and add dedicated measures in case of breaches of security and confidentiality (OECD, 2015[79]).
Recommendation 9: Continue efforts to establish a Statistical Business Register using the existing common framework and integrating additional sources.
To support business statistics, the National Statistical Office could create a Statistical Business Register leveraging the existing common frame already used for the Industrial and Business Service Census. The NSO can expand its coverage with other key administrative data sources, such as tax registers (i.e. value added tax, corporate as well as personal income tax data) from the Revenue Department (Ministry of Finance) and other relevant administrative sources (e.g. Social Security data). The integration of several administrative sources into a Statistical Business Register could serve to increase business coverage and to retrieve information missing in individual data sources (i.e. employment), as practices from other countries illustrate (see Box 2.14). The Economic Censuses could serve as valuable sources to complement and update the SBR.
Following international practice, Thailand could then use the Statistical Business Register to consolidate the sampling frame for business surveys, to link various micro-data, and to establish official business demography statistics that are relevant to productivity analyses. The creation and maintenance of a Statistical Business Register would not only support micro-level analysis but could also enhance National Accounts data through high-quality and consistent business statistics. As such, it would also contribute to better measurement of productivity at aggregate, sectoral and industry levels (discussed in recommendation 6).
Importantly, the SBR purposes is to collect information on registered firms. In economies with a high level of informality like Thailand, regular surveys of the informal sectors should complement administrative data and the SBR. This can help measure the contribution of the informal economy to economic output and further identify investigate unregistered firms and/or those that hire workers informally.
To implement this recommendation, Thailand can follow international practice and examples, and draw on the expertise of international organisations, which publish guidelines providing the key characteristics of business registers (UN, 2024[34]; OECD/Eurostat, 2008[45]; Eurostat, 2021[46]; UNECE, 2018[47]). The practical steps to set up and improve the SBR depend on the stage at which Thailand stands along different dimensions. To evaluate this, the UNSD defines a Maturity Model for Statistical Business Registers (UNSD, 2023[80]). The maturity model identifies seven dimensions that characterise SBRs and propose criteria for countries to provide a self-assessment on the degree of maturity (preliminary, early, mature, advanced). The dimensions refer to the (1) legal and institutional framework relevant for establishing and maintaining the SBR, (2) data sources for the SBR, (3) maintenance and update of the SBR, (4) coverage of the SBR, (5) use of SBR, (6) IT environment and (7) interoperability.
Developing a comprehensive business register may take time, but Thailand can already support the NSO's efforts by incorporating the creation of a business register into its pro-productivity agenda and backing it with short-term actions as well as longer-run investments articulated in two phases.
Phase 1: Develop a roadmap for the SBR informed by an assessment of Thailand's SBR maturity. A first step in the implementation process would be for the NSO to develop a roadmap, informed by an assessment of Thailand's maturity across the various dimensions of the SBR. International organisations offer key resources to this aim. For instance, the UN “maturity model toolkit” provides a questionnaire for countries’ self-assessments of the maturity of the SBR together with training material. Simultaneously, Thailand should continue investing in its statistical system by further developing the NSO's expertise and human capital and ensuring access to technical resources. This can be achieved through participation in relevant international groups and events, investing in staff training and ensuring the NSO has access to essential technical resources. In parallel, the NSO can, possibly with the help of the productivity council, raise awareness about the need for such data and how this endeavour can support business statistics. This may facilitate collaboration and data exchanges across Thai agencies (see recommendation 8).
Phase 2: Continue efforts to develop a SBR by combining several administrative data. The National Statistical Office could establish the SBR, by expanding the common frame with other key administrative data sources, such as tax registers (i.e. value added tax, corporate as well as personal income tax data) from the Revenue Department (Ministry of Finance) and other relevant administrative sources (e.g. Social Security data). To strengthen the implementation of recommendation 9, Thailand should ensure that the NSO can rely on a strong mandate to collaborate with administrative data holders. As discussed in recommendation 8, this may require adapting the legal framework governing data exchanges. Furthermore, this will require the NSO to integrate these sources according to international guidelines, for instance addressing challenges related to the harmonisation of sources, definition of statistical units and update of the SBR.
Most countries combine several data sources to construct a Statistical Business Register (SBR), which is usually primarily based on administrative data (UN, 2024[34]).Those data often include compulsory registration systems, tax records, social security data, and can also integrate information from statistical surveys.
The sources of the SBR should ensure high coverage and quality standards. The UN details six criteria to select primary and relevant administrative sources (see Figure 6.1 in (UN, 2024[34])): (i) coverage (it should cover a broad range of the intended SBR population), (ii) relevant variables (the administrative source covers basic identification information on businesses, but further information on employment and turnover is usually required), (iii) register unit and identifiers (the administrative unit may not need to correspond with the statistical unit, but identifiers of legal units can be used as the enterprise unit for most of them), (iv) frequency (monthly or quarterly data is preferred to continuously update the BR), (v) timeliness (register information should be updated in a timely way), (vi) quality (the quality of the information provided by the data owner should be continuously monitored, discussed and if necessary improved together with the data owner). Different countries are using different primary data sources for their business register. Statistics Canada, for example, uses as its primary source for the SBR the register of taxpayers with business income, which is maintained by Revenue Canada. Statistics Netherlands uses the Trade Register of the Chamber of Commerce (see additional details on the frequent data sources and additional examples in UN (2024[34]) Box 6.5). The Korean SBR, for example, merges several administrative data – including data from the National Tax Service – and survey data, including the Census on Establishment.
Linking different data can enhance the quality and coverage of the SBR. Combining and linking different sources is also key to meet the above criteria (for instance a single data source may not cover the whole targeted population, and complementary sources can also enhance the quality and timeliness). Costa Rica, for example, is using the Social Security Registry (SSR), Tax records and the Register of exporting enterprises to improve the coverage of the business register (REE) through data integration (UN, 2024[34]). Merging different sources is also key to accurately derive statistical units (i.e. enterprises) from administrative units (legal units). Statistics Netherlands, for instance, supplements its primary Trade Register data with tax and social security data when they provide more accurate information, ensuring the creation of robust statistical units that reflect economic reality. Finally, merging several datasets is key to retrieve missing (or better measured) information, such as employment, which is often obtained from social security records (e.g. Italy and other countries).
Providing a comprehensive assessment of productivity at a granular level using micro-data
Thai institutions, such as the Office of Industrial Economics (OIE) and the Puey Ungphakorn Institute for Economic Research, already conduct rich productivity analyses using available micro-data. These current analyses serve as a foundation for further development and enhancements of firm-level analyses, as these do not predominantly feature in policy discussion (see Chapter 3). While data constraints limit certain aspects of the analysis, there is significant potential for actionable insights and improvements to support policymaking. For instance, the administrative business register maintained by the Department of Business Development (Ministry of Commerce) could be used further for analysis and statistical purposes, in particular for business demography, firm dynamics and productivity analysis if complemented with additional information on firm-level employment.
Recommendation 10: Strengthen the analysis of productivity at the firm-level.
Analyses linking firm-level dynamics to industry and aggregate productivity are crucial to assess strengths and weaknesses of the Thai economy and identify relevant policy actions according to the policy framework discussed in Chapter 3. Such granular analyses, grounded in cutting-edge economic research, have become an integral part of policy analyses conducted by national pro-productivity institutions and international organisations.
To support evidence-based pro-productivity policies, Thailand should target comprehensive analyses covering three key aspects of firm-level analyses.
Dimension 1: Firm productivity. Enhancing Thailand's productivity analysis at a granular level requires incorporating an examination of the dispersion (i.e. differences) in productivity between businesses that are leading (the productivity frontier) and those that are lagging behind (laggards). Monitoring the evolution of these gaps over time and across industries is also necessary. Moreover, a more systematic and thorough investigation of the broad range of factors related to productivity – size, age, ownership structure, use of ICT and automation technologies, international trade activities, workforce composition, etc. – would provide a snapshot of the business population in terms of productivity and strengthen the understanding of business performance. Understanding productivity heterogeneity across firms is also key to promote inclusiveness and ensure that productivity gains are widely shared across individuals. Enhancing the analysis of within-business productivity growth is also essential. However, it may necessitate further investment in micro-data, leveraging previous recommendations to link and use administrative data and surveys in integrated statistical information systems. Moreover, the role of firms’ informality remains largely understudied in current productivity analysis. This is also due to the lack of available micro-data on unregistered firms and their performance. Given the large share of informal establishments in Thailand, a better understanding of un-registered firms, including their characteristics and performance, would be valuable. Such analysis could help identify targeted policies to encourage formal registration, thereby enhancing overall economic performance.Table 2.6As shown in Table 2.6, the PIER institute and the OIE of the Ministry of Industry are the most active institutions already conducting firm productivity analysis. Their analyses can then be extended to include a scrutiny of new indicators, such as productivity dispersion for all sectors.
Dimension 2: Allocative efficiency. In previous analyses, Thai institutions already investigated the efficiency of resources allocations across businesses in the manufacturing sector. A more detailed and systematic examination of four categories of indicators could enhance the analysis. These indicators measure (i) the allocative efficiency, (ii) the re-allocation of resources across business, (iii) the state of competition, (iv) the extent to which zombie firms reduce allocative efficiency. These indicators should be documented more regularly, replicating existing studies (notably those from the Puey Ungphakorn Institute for Economic Research) using the most recent editions of the Censuses for the (re-)allocation analysis and updating the series of the CPFS for the analysis of the competitive environment and zombie firms. Additionally, performing similar analyses for the service sector – which as of now remains under-studied – would provide valuable insights.
Dimension 3: Business dynamics. The analysis of business dynamics in Thailand remains limited and should be fostered. Thailand could regularly monitor three categories of indicators for such analysis: (i) indicators of the evolution of firms’ entry and exit, (ii) indicators of the contribution of entry and exit to productivity and (iii) indicators of firms’ life cycle, notably post-entry growth and market selection. Data limitations are still impairing a broad implementation of these indicators, but the construction of SBR (discussed above) could be key to address some of these issues and establish business demography statistics.
Thailand can focus on short-term and medium-term objectives for the implementation of the recommendation.
Phase 1: Integrate the firm-level perspective into the council’s bulletins. To comprehensively cover these three dimensions, Thailand can integrate this framework) into the council's work, as outlined in recommendation 3, to cover these relevant dimensions of micro-level productivity analyses. In the first phase of implementation, the council can draw on existing evidence, notably based on the research from the Puey Ungphakorn Institute for Economic Research and the annual reports of the Office of Industrial Economics. The second phase will aim at updating and extending the set of evidence based on micro-data.
Phase 2:
Integrate the firm-level perspective into the council’s annual reports. To comprehensively cover these three dimensions, Thailand can integrate this framework (Box 2.4) into the council's work, as outlined in recommendation 3. The council should ensure that these dimensions are regularly reflected in its annual reports, drawing on existing research by Thai pro-productivity institutions, external researchers and international organisations, but also on new evidence generated by the secretariat as it builds its analytical expertise and capacity. Given their expertise in conducting the recommended analyses, the Puey Ungphakorn Institute for Economic Research and the Office of Industrial Economics are well-positioned to update existing indicators based on previous research and to develop new ones that maximise the use of micro-data Box 2.15.illustrates how national productivity boards incorporate firm-level analysis into their annual reports, while Table 2.6 outlines relevant indicators and their applications to the Thai economy. Table 2.6 also proposes a potential schedule for the council to integrate these dimensions into its work. Not all indicators need to be monitored annually. Instead, their frequency should align with policy priorities and key challenges previously identified, particularly in phase 1 of recommendation 3.
Produce official statistics based on firm-level data. Thailand can build on these developments, as well as data improvements recommended in this chapter (recommendations 8-10), to publish relevant official statistics based on firm-level data. The NSO could establish business demography statistics based on the Statistical Business Register following international guidelines (OECD/Eurostat, 2008[45]) but also establish other indicators, such as dispersion statistics (see for example the US dispersion statistics presented in Box 2.11). Eventually, further integration of data (administrative data, survey data) would more generally allow the NSO to publish structural business statistics (see Box 2.9).
Firm-level data is widely used in annual reports of National Productivity Boards. The 2024 European Commission report provides examples on how annual reports from European National Productivity Boards are using firm-level data (European Commission, 2024[81]). While, in most cases, firm-level analyses are performed using national data sources (for instance from the National Statistical Office), in some cases they rely on international databases. For instance, Finland or France used the OECD MultiProd data to analyse productivity dispersion, Luxembourg used the Eurostat’s community and innovation survey on structural business statistics to assess the relationship between management practices and labour productivity.
Insights from granular, firm-level, data are compiled both through analysis carried out internally or drawing from existing evidence. Further to their own analysis, productivity boards’ reports summarise the available evidence produced by external researchers from academia or other institutions. For instance, the Finnish Productivity Board report of 2023, included a discussion of the impacts of AI on productivity, based on a summary of evidence from existing literature. Similarly, the 2023 report of the French National Productivity Board provides a literature review on the potential effect of the green transition on firm productivity. Leveraging the expertise of researchers at Statistics Netherlands, the 2023 report from the Dutch National Productivity Board includes a chapter dedicated to a firm-level perspective on productivity, including firm-level heterogeneity and productivity dispersion, and the role of business dynamics (firm entry and exit), reallocation and within firm growth (CPB, 2024[82]).
Source: CPB Netherlands Bureau of Economic Policy Analysis: 2023 annual report (CPB, 2024[82]); Conseil National de Productivité: statistics of May 2022 (Conseil National de Productivité, 2022[83]) and 2023 annual report (Conseil National de la Productivité, 2023[84]); Finnish Productivity Board: 2023 annual report (Finnish Productivity Board, 2023[29]).
Note: This table provides a summary of indicators that can be considered in recurrent analysis of productivity in Thailand, using micro-economic data (establishment or firm-level). The list of indicators is not exhaustive. 1 The list of references provided is not exhaustive, it focuses on references discussed in the main text of the report, focusing, when possible, on applications to Thailand, as well as other complementary examples (readers should also refer to discussions and references herein).2 The suggested frequencies are “very frequent” (e.g. annual), “frequent”, “regular” (e.g. analysis conducted at least for each sub-period of the National Economic and Social Development Plan). The frequency is purely indicative and can be adapted depending on the feasibility, the analytical needs, and topics of particular interest in the context of Thailand for the period under consideration.
Source: OECD elaborations.
Enhancing the dissemination of micro-data through appropriate solutions for data-sharing.
Micro-data are crucial for understanding a country’s productivity performance. However, the current conditions for sharing and accessing these data, often through ad-hoc agreements, may still limit their use for productivity analysis in Thailand. Implementing more systematic, ideally centralised, and secure data repositories, in line with international practices, can standardise data-sharing and access procedures, to help maximise the value derived from these data for policymaking. Allowing relevant Thai institutions, but also external researchers (from academia, international organisations and to some extent the private sector) to access and use the data more easily offers several advantages. First, the research output is directly beneficial to policymakers and advances knowledge on specific issues. Additionally, it enables technical experts to contribute to data improvements and documentation.
Recommendation 11: Develop a centralised solution to access and use micro-data for analysis.
Thailand could establish a centralised data-sharing solution, such as a research data centre, that allow users to access micro-data easily and securely in one location. Thailand can choose among different (non-mutually exclusive) solutions, including physical locations for data access (e.g. safe rooms equipped with computers) or solutions that eliminate the need for researchers to be physically present, such as remote access using VPN and/or remote execution solutions. Clear information on how to access the data, including the authorisation process, and comprehensive metadata should be available. Thailand can build on international standards that guide safe micro-data sharing (e.g. the Five Safes Framework) and practices that ensure the preservation of confidential information. This includes for instance the use of safe rooms with relevant technical specifications (e.g. no internet connection, video-surveillance) and appropriate anonymisation/semi-anonymisation of data, as well as clear rules and procedures to implement statistical disclosure control. See Box 2.16 for examples of practices from different countries. UN (2022[32]) provides additional guidelines on data dissemination, and UNECE (2007[85]) also provides guidelines to provide access to micro-data preserving confidentiality, discussing case studies from selected countries. 61
The initiative could be coordinated by the National Statistical Office, which currently maintains the Censuses and has experience in handling large micro-data sets and granting access to both data and metadata. Thailand can also build on the experience of the Government Data Catalogue and Government Data Exchange systems that improved technical solutions for data dissemination and data sharing of sensitive solutions.
Phase 1: Establishing physical locations to provide access to NSO micro-data. In the short run, a first building block could be to establish physical locations with computers at NSO offices to provide researchers with a possibility to access NSO confidential micro-data. This is the most common and simpler practice used worldwide (see Box 2.16). The NSO could establish safe rooms with relevant technical settings to enforce data protection (e.g. no internet connection, use of video-surveillance). Ideally the micro-data accessible in such locations could be merged by the presence of a unique identifier for businesses, if necessary, adequately anonymised.
Phase 2: Complement physical locations with remote access and expand the access to other data. In the medium run, the goal would be to complement physical locations with remote access solutions that overcome the need for researchers to travel for accessing the data (see for example the case of NSO in Italy that establishes a remote access at the Bank of Italy – Box 2.16). In addition, a single repository should be dedicated where to access together several data collected by different agencies (see for example the case of Korea in Box 2.16) ideally combining administrative and survey data. In addition to the NSO data, this repository could integrate the Corporate Profile and Financial Statement (CPFS) data from the Department of Business Development in the Ministry of Commerce, and micro-data from other owners, such as the Office of Industrial Economics. This data dissemination will benefit from, and support, the integration of administrative data to statistical systems and broader use of these sources for productivity analysis (recommendation 8).
Countries worldwide are facilitating micro-data sharing with external researchers, ensuring preservation of confidential information. As a first solution, most countries have established physical locations (e.g. safe rooms which prevent data leaks) at national offices or central banks where researchers can come to access and work on the micro-data.
The simpler solution (at least in the first instance) is to provide access to the data that are collected by the institution setting the data-sharing facility (as the RDC of the Bank of Italy presented in Box 2.10). More advanced solutions also offer access to a broad range of data collected by different public institutions in a single location (as CASD in France presented in Box 2.10 or the micro-data integrated service of Korea presented below). Such a solution requires a longer term to be settled and the establishment of data-sharing agreements between data providers.
Below are provided two examples of countries setting physical and remote access solutions.
Italian National Statistical Office – Physical locations and remote access at Italian Central Bank
The Laboratory for the Analysis of ELEmentari Data (ADELE Laboratory) is the data research centre where the Italian National Statistical office, ISTAT, provides access for scientific purposes to secure use files, which are micro-data files coming from surveys free of direct identifiers or special categories of personal data and/or data related to criminal convictions and offenses. Access to the ADELE Laboratory is granted at ISTAT headquarters and at all territorial offices of the Institute. Elementary data files are provided free of charge. Access to ADELE Laboratory is also free of charge. 62 Micro-data can also be accessed under special research agreements within research projects conducted jointly by ISTAT and recognised research organizations. ISTAT completed an experimental project for remote access, led to the opening of a remote elementary data access laboratory at the Bank of Italy headquarters, via VPN connection.
Micro-data integrated service, Korea – Physical access and remote access
Statistics Korea collects micro-data of self-produced statistics but also collects micro-data of other statistical agencies such as government departments, local governments, and research institutes in one place, so that people can conveniently use various statistical data through the micro-data integrated service (MDIS). Statistics Korea offers two types of services to access its data. A free of charge service where users can download materials to their pc and access online statistics (excluding individual identification information and sensitive variables etc.) and a surcharge service. The latter includes remote access and a research data centre (RDC). The remote access allows to access more detailed information, while RDC is the most detailed data which also include linkage identification information (using encryption replacement key) rather than remote access data.
Source: Italy’s Laboratory for Elementary Data Analysis – Istat and Banca d’Italia; Korea’s Microdata Integrated Service
References
[62] Andrews, D. and F. Cingano (2012), “Public Policy and Resource Allocation: Evidence from Firms in OECD Countries”, OECD Economics Department Working Papers, No. 996, OECD Publishing, Paris,.
[52] Andrews, D., C. Criscuolo and P. Gal (2016), “The Best versus the Rest: The Global Productivity Slowdown, Divergence across Firms and the Role of Public Policy”, OECD Productivity Working Papers, No. 5, OECD Publishing, Paris, https://doi.org/10.1787/63629cc9-en.
[39] Apaitan, T. et al. (2020), Common Ownership, Domestic Competition, and Export: Evidence from Thailand, Puey Ungphakorn Institute for Economic Research, Bangkok, https://www.pier.or.th/files/dp/pier_dp_140.pdf.
[15] APO (2023), APO Productivity Databook 2023, Asian Productivity Organization, https://doi.org/10.61145/TRKP9496.
[3] APO (2023), Productivity Outlook 2023. Manufacturing labour productivity: Trends and Linkages, Asian Productivity Organization (APO).
[2] APO/OECD (2021), Towards Improved and Comparable Productivity Statistics: A Set of Recommendations for Statistical Policy, OECD Publishing, Paris, https://doi.org/10.1787/1ae0ec74-en.
[70] Banerjee, R. and B. Hofmann (2018), “The rise of zombie firms: causes and consequences”, BIS Quarterly Review, https://www.bis.org/publ/qtrpdf/r_qt1809g.pdf.
[27] Belgium National Productivity Board (2023), “Annual Productivity Report 2023”, https://cnp-nrp.belgium.be/uploaded/files/202402091439280.CNP_Rapport_annuel_2023_24_EN.pdf.
[28] Belgium National Productivity Board (2022), “Annual Productivity Report 2022”, https://www.cnp-nrp.belgium.be/uploaded/files/202301200846380.CNP_rapport_2022_EN.pdf.
[26] Belgium National Productivity Board (2021), “Annual Productivity Report 2021”, https://cnp-nrp.belgium.be/uploaded/files/202110251000540.CNP_rapport_2021_EN.pdf.
[61] Berlingieri, G. et al. (2017), “The Multiprod project: A comprehensive overview”, OECD Science, Technology and Industry Working Papers, No. 2017/04, OECD Publishing, Paris.
[51] Berlingieri, G., P. Blanchenay and C. Criscuolo (2017), “The great divergence(s)”, OECD Science, Technology and Industry Policy Papers, No. 39, OECD Publishing, Paris, https://doi.org/10.1787/953f3853-en.
[57] Blackwood et. al. (2022), Opening the Black Box: Task and Skill Mix and Productivity Dispersion, Presented at NBER CRIW.
[36] Calligaris, S. et al. (2023), “Is there a trade-off between productivity and employment?: A cross-country micro-to-macro study”, OECD Science, Technology and Industry Policy Papers, No. 157, OECD Publishing, Paris, https://doi.org/10.1787/99bede51-en.
[42] Calligaris, S. et al. (2023), “Employment dynamics across firms during COVID-19: The role of job retention schemes”, OECD Economics Department Working Papers, No. 1788, OECD Publishing, Paris, https://doi.org/10.1787/33388537-en.
[73] Calvino, F., C. Criscuolo and C. Menon (2015), “Cross-Country Evidence on Startup Dynamics”, OECD Science, Technology and Industry Working Papers, No. 2015/06, OECD Publishing, Paris.
[72] Calvino, F., C. Criscuolo and R. Verlhac (2020), “Declining business dynamism: Structural and policy determinants”, OECD Science, Technology and Innovation Policy Papers, No. 94, OECD Publishing, Paris.
[43] Charoenporn, P. and S. Choksawatpaisan (2024), “Innovation and Productivity in the Service Sector: The Case of Thailand”, Thammasat Review of Economic and Social Policy, Vol. 9/1, https://so04.tci-thaijo.org/index.php/TRESP/issue/view/17911.
[84] Conseil National de la Productivité (2023), “Bilan des crises Compétitivité, productivité et transition climatique ; Quatrième rapport du Conseil national de productivité [Review of crises in competitiveness, productivity and climate transition; Fourth report of the National Productivity Council]”.
[83] Conseil National de Productivité (2022), “Productivité et compétitivité : analyses conjoncturelles et structurelles post-Covid [Productivity and competitiveness: post-Covid economic and structural analyses]”, https://www.strategie.gouv.fr/sites/strategie.gouv.fr/files/atoms/files/cnp2022-complement-tableau_de_bord_statistiques-mai.pdf.
[18] Coyle, D. and J. Mei (2022), “Diagnosing the UK Productivity Slowdown: Which Sectors Matter and Why?”, The Productivity Institute, Vol. Working Paper No. 018 version 1.
[82] CPB (2024), “National Productivity Board 2023 Annual Report”, CBP Netherlands Bureau for Economic Policy Analysis, https://www.cpb.nl/sites/default/files/omnidownload/CPB-Publicatie-national-productivity-board-2023-annual-report_1.pdf.
[35] Criscuolo, C. et al. (2021), “The human side of productivity: Uncovering the role of skills and diversity for firm productivity”, OECD Productivity Working Papers, No. 29, OECD Publishing, Paris.
[74] Criscuolo, C., P. Gal and C. Menon (2014), “The Dynamics of Employment Growth: New Evidence from 18 Countries”, OECD Science, Technology and Industry Policy Papers, No. 14, OECD Publishing, Paris.
[55] Cunningham et. al (2022), “Dispersion in Dispersion: Measuring Establishment-Level Differences in Productivity”, BLS WORKING PAPERS, https://www.bls.gov/osmr/research-papers/2020/pdf/ec200120.pdf.
[56] Cunningham et. al. (2021), “Chaos Before Order: Productivity Patterns in U.S. Manufacturing”, International productivity monitor, 41 (fall), https://www.productivity.ac.uk/wp-content/uploads/2021/12/IPM_41_Cunningham.pdf.
[71] Decker, R. et al. (2016), “Declining Business Dynamism: What We Know and the Way Forward”, American Economic Review, Vol. 106/05, pp. 203-07, https://doi.org/10.1257/aer.p20161050.
[23] EU Council (2016), “Council Recommendation of 20 September 2016 on the establishment of National Productivity”, Official Journal of the European Union, 2016/C 349/01.
[81] European Commission (2024), “National Productivity Boards after Seven Years: An assessment”, European Economy Discussion Papers 203, https://economy-finance.ec.europa.eu/document/download/75335140-1d65-4fe7-8342-b71da6272537_en?filename=dp203_en.pdf.
[24] European Commission (2021), “Follow-up of the progress report on the implementation of the Council Recommendation of 20 September 2016 on the establishment of National Productivity Boards”, COMMISSION STAFF WORKING DOCUMENT, https://economy-finance.ec.europa.eu/document/download/4fe2a633-4379-4691-9def-c3ca00b19a41_en?filename=swd2016216.final_.pdf.
[46] Eurostat (2021), “European business statistics methodological manual for statistical business register”, https://ec.europa.eu/eurostat/documents/3859598/12433023/KS-GQ-20-006-EN-N.pdf/0c31c77a-5d20-9954-9223-2b856fdb93c9?t=1613481618625.
[87] FAO (2018), Guidelines for the measurement of productivity and efficiency in agriculture, https://openknowledge.fao.org/items/91366037-12f6-4b6c-a0d7-01fe111e522c.
[29] Finnish Productivity Board (2023), “Skilled people create productivity”, PUBLICATIONS OF THE MINISTRY OF FINANCE – 2023:88, https://julkaisut.valtioneuvosto.fi/bitstream/handle/10024/165291/VM_2023_88.pdf?sequence=1&isAllowed=y.
[25] German Council of Economic Experts (2023), “Strengthening potential growth through capital formation”, GCEE Annual Report 2023/24 - Council of Economic Experts, https://economy-finance.ec.europa.eu/document/download/1d580474-9093-45f9-9cf2-cb4bc0e52411_en?filename=JG202324_Chapter_2%20%281%29.pdf.
[64] Gnocato, N., F. Modena and C. Tomasi (2020), “Labor market reforms and allocative efficiency in Italy”, Labour Economics, Vol. 67.
[60] Hsieh, C. and P. Klenow (2014), “The life cycle of plants in India and Mexico”, The Quarterly Journal of Economics, Vol. 129/3, pp. 1035-1084.
[59] Hsieh, C. and P. Klenow (2009), “Misallocation and manufacturing TFP in China and India”, The Quarterly Journal of Economics, Vol. 124/4, pp. 1403-1448.
[91] ILO (2013), “Measuring informality: A statistical manual on the informal sector and informal employment”, International Labour Organization, https://www.ilo.org/publications/measuring-informality-statistical-manual-informal-sector-and-informal.
[44] IMF (2018), “Using Administrative Data to Enhance Policymaking in Developing Countries: Tax Data and the National Accounts”, IMF Working Paper No. 2018/175.
[9] Jorgenson, D. (1963), “Capital theory and investment behaviour”, American Economic Review, Vol. 53.
[10] Jorgenson, D. and Z. Griliches (1967), “The Explanation of Productivity Change”, Review of Economic.
[20] Klyuev, V. (2015), Structural Transformation — How Does Thailand Compare?, IMF Working Paper No. 2015/051.
[48] Korwatanasakul, U. (2021), Determinants, wage inequality, and occupational risk exposure of informal workers: A comprehensive analysis with the case study of Thailand, Puey Ungphakorn Institute for Economic Research, Bangkok.
[40] Limjaroenrat, V. (2016), Firm-level Perspective of Thailand’s Low Investment Puzzle, Puey Ungphakorn Institute for Economic Research, Bangkok, https://www.pier.or.th/en/dp/042/.
[33] London Economics and DIW Econ (2017), “Review of international best practice in the production of productivity statistics”, https://www.ons.gov.uk/economy/economicoutputandproductivity/productivitymeasures/articles/reviewofinternationalbestpracticeintheproductionofproductivitystatistics/2018-02-07.
[69] McGowan, A., M. Andrews and V. Millot (2017), “The Walking Dead?: Zombie Firms and Productivity Performance in OECD Countries”, OECD Economics Department Working Papers, No. 1372,, https://doi.org/10.1787/180d80ad-en.
[65] Melitz, M. and S. Polanec (2015), “Dynamic Olley-Pakes productivity decomposition with entry and exit”, RAND Journal of Economics, Vol. 46/2, pp. 362-375.
[11] MOL (2022), ผลิตภาพแรงงานไทย: ขับเคลื่อนอย่างไรในยุคโลกเปลี่ยน [Thai labour productivity: how to drive it in an era of global change?].
[41] Muthitacharoen, A., A. Paweenawat and K. Samphantharak (2024), The Double-Edged Sword: Unintended Consequences of SME Promotion Policy, Puey Ungphakorn Institute for Economic Research, Bangkok, https://www.pier.or.th/dp/148/.
[12] NESDC (2024), Revisiting Thailand’s Productivity: Concepts and Measurement, NESDC, Bangkok, https://www.nesdc.go.th/download/document/Macroseminar/2024/Report-01.pdf.
[7] NESDC (2022), Capital Stock of Thailand, 2022 Edition, Office of the National Economic and Social Development Council, https://www.nesdc.go.th/nesdb_en/ewt_news.php?nid=4503&filename=index.
[92] OAE (2022), การศึกษาผลิตภาพการผลิตโดยรวมในภาคเกษตร [A Study on Total Factor Productivity in the Agricultural Sector].
[68] O’Connor, J. (2023), “Zombie Thai firms holding back economic growth as they struggle just to pay interest on bank debt”, Economy, Foreigners, Living , More News from Thailand, https://www.thaiexaminer.com/thai-news-foreigners/2023/11/23/zombie-firms-holding-back-economic-growth-banks-cut-back-loans/.
[66] OECD (2024), “Exploring the state of Competition in the EU”.
[17] OECD (2024), OECD Compendium of Productivity Indicators 2024, OECD Publishing, Paris, https://doi.org/10.1787/b96cd88a-en.
[49] OECD (2023), “Boosting evidence-based policy making for economic development policies in Italy”, https://www.oecd.org/economy/surveys/Italy-boosting-evidence-based-policy-making-for-economic-development-report-2023.pdf.
[30] OECD (2023), OECD Compendium of Productivity Indicators 2023, OECD Publishing, Paris, https://doi.org/10.1787/74623e5b-en.
[16] OECD (2023), OECD Economic Surveys: Thailand 2023, OECD Publishing, Paris, https://doi.org/10.1787/4815cb4b-en.
[89] OECD (2022), “Frascati Manual 2002: Proposed Standard Practice for Surveys on Research and Experimental Development, The Measurement of Scientific and Technological Activities”, OECD Publishing, Paris, https://doi.org/10.1787/9789264199040-en.
[8] OECD (2019), “OECD Compendium of productivity indicators 2019”, OECD Publishing Paris, https://doi.org/10.1787/b2774f97-en.
[37] OECD (2018), The Productivity-Inclusiveness Nexus, OECD Publishing, Paris, https://doi.org/10.1787/9789264292932-en.
[79] OECD (2015), Improving Co-operation Between Tax and Anti-Money Laundering Authorities, OECD Publishing, Paris, https://doi.org/10.1787/215d3015-en.
[22] OECD (2015), The Future of Productivity, OECD Publishing, Paris, https://doi.org/10.1787/9789264248533-en.
[6] OECD (2009), Measuring Capital - OECD Manual 2009: Second edition, OECD Publishing, Paris, https://doi.org/10.1787/9789264068476-en.
[4] OECD (2001), Measuring Productivity - OECD Manual: Measurement of Aggregate and Industry-level Productivity Growth, OECD Publishing, Paris, https://doi.org/10.1787/9789264194519-en.
[5] OECD/APO (2022), Identifying the Main Drivers of Productivity Growth: A Literature Review, OECD Publishing, Paris, https://doi.org/10.1787/00435b80-en.
[45] OECD/Eurostat (2008), Eurostat-OECD Manual on Business Demography Statistics, OECD Publishing, Paris, https://doi.org/10.1787/9789264041882-en.
[90] OECD/Eurostat/European Union (1997), “Proposed Guidelines for Collecting and Interpreting Technological Innovation Data: Oslo Manual, The Measurement of Scientific and Technological Activities”, OECD Publishing, Paris, https://doi.org/10.1787/9789264192263-.
[53] OIE (2022), Productivity and Performance of Thai Industry Report 2022, Ministry of Industry, Office of Industrial Economics, Bangkok.
[31] OIE (2019), Productivity and Performance of Thai Industry Report 2019, Ministry of Industry, Office of Industrial Economics, Bangkok.
[63] Olley, S. and A. Pakes (1996), “The Dynamics of Productivity in the Telecommunications Equipment Industry”, Econometrica, Vol. 64/6, pp. 1623-1297, https://doi.org/10.2307/2171831.
[54] Paweenawat, A., T. Chucherd and N. Amarase (2017), Uncovering Productivity Puzzles in Thailand: Lesson from Microdata, Puey Ungphakorn Institute for Economic Research, Bangkok, https://www.pier.or.th/en/dp/073/.
[38] Sangsubhan, K., K. Pornpattanapaisankul and P. Kambuya (2023), Automation and Productivity: Evidence from Thai Manufacturing Firms, Puey Ungphakorn Institute for Economic Research, Bangkok, https://www.pier.or.th/files/dp/pier_dp_199.pdf.
[67] SCB (2020), “The Impact of COVID-19 on Zombie Firms and Implications for the Thai Corporate Sector”, Economic Intelligence Centre, https://www.scbeic.com/en/detail/file/product/7271/fu7egpcdcc/EIC_Note-Zombie-Firm_ENG_20201216.pdf.
[1] SNA, 2. (2009), System of National Accounts 2008, UN, EU, IMF, OECD, WB.
[58] Syverson, C. (2011), “What Determines Productivity?”, Journal of Eocnomic Literature, Vol. 49/2, pp. 326-65, https://doi.org/10.1257/jel.49.2.326.
[86] Thailand, N. (21-25 May 2018), Regional course on statistical business register: Thailand’s experience.
[50] Tumpane, S. (2023), 64th ISI World Statistics Congress - Ottawa, Canada.
[34] UN (2024), “Guidelines on Statistical Business Registers”, UN Department of Economic and Social Affairs, Statistics Division.
[32] UN (2022), “Chapter 10: Dissemination of Official Statistics”, Handbook on Management and Organization of National Statistical Systems, 4th Edition of the Handbook of Statistical Organization.
[75] UNCITRAL (2019), “Legislative Guide on Key Principles of a Business Registry”, https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwiO_-ya5OuEAxVhVKQEHR16AxkQFnoECBQQAQ&url=https%3A%2F%2Funcitral.un.org%2Fsites%2Funcitral.un.org%2Ffiles%2Fmedia-documents%2Funcitral%2Fen%2Flg_business_registry-e.pdf&usg=AOvVaw1J8.
[78] UNECE (2019), “Guidance on Modernizing Statistical Legislation”, https://unece.org/statistics/publications/guidance-modernizing-statistical-legislation.
[47] UNECE (2018), Guidelines on the use of statistical business registers for business demography and entrepreneurship statistics, United Nations.
[77] UNECE (2011), Using Administrative and Secondary Sources for Official Statistics - A Handbook of Principles and Practices, United Nations Economic Commission for Europe.
[85] UNECE (2007), “Managing Statistical Confidentiality & Microdata access”, https://unece.org/fileadmin/DAM/stats/publications/Managing.statistical.confidentiality.and.microdata.access.pdf.
[80] UNSD (2023), Manual on the Maturity Model for Statistical Business Registers, United Nations Statistics Division.
[88] UNSD (2023), Microdata Dissemination - Summary report prepared for the Intersecretariat Working on Household Surveys, United Nations Statistics Division, Department of Economic and Social Affairs, Inter-secretariat Working Group on Household Surveys (ISWGHS), https://unstats.un.org/iswghs/documents/Microdata_Dissemination_Taskforce_Summary_Report.pdf.
[13] Ward, A. and B. Zinni (2024), “The composition of labour input: Sensitivity testing and results for productivity analysis”, OECD Statistics Working Papers, No. 2024/06, OECD Publishing, Paris, https://doi.org/10.1787/5d9b866a-en.
[19] Warr, P. and W. Suphannachart (2022), Benign Growth: Structural Transformation and Inclusive Growth in Thailand, Oxford Academic, https://doi.org/10.1093/oso/9780192855299.003.0005.
[21] World Bank (2021), Sectoral Sources of Productivity Growth, https://doi.org/10.1596/978-1-4648-1608-6_ch7.
[76] World Bank (2016), Implementing a Unique Business Identifier in Government: Guidance Note for Practitioners and Nine Country Case Studies, World Bank, Washington, DC, License: Creative Commons Attribution CC BY 3.0.
[14] World Bank and Bank of Thailand (2020), Thailand Manufacturing Firm Productivity Report, World Bank, Washington, DC, https://doi.org/10.1596/34882.
← 1. The quality of productivity indicators therefore primarily depends on the application of international standards in measuring economic activity. Thailand has adopted the 2008 SNA standards in 2016 (prior to that the compilation of National Accounts data relied on the previous 1993 SNA framework), with backward estimates implemented back to 1998 (APO, 2023[3]). Furthermore, economic activities are classified according to TSIC-2009 (the Thailand Standard Industrial Classification) which follows the International Standard Industrial Classification of All Economic Activities (ISIC) revision 4. The adoption of these frameworks allows for the computation and analysis of aggregate productivity following international recommendations and largely contributes to international comparability, although challenges remain (OECD/APO, 2022[5]).
← 2. National Accounts data in Thailand are primarily collected by the Office of the National Economic and Social Development Council (NESDC), which provides long time series data of GDP and the capital stock disaggregated by 21 industries. GDP is also collected by industry at the local level for both provinces and regions of Thailand. Data on employment and hours worked, necessary for productivity analysis, are collected by the National Statistical Office, based on the Labour Force Survey by industries and regions.
← 3. The Office of Industrial Economics, part of the Ministry of Industry, also provides productivity statistics at the industry level. These statistics are based on survey data for a sample of firms in the manufacturing sector. The OIE data will be discussed in the section on micro-data.
← 5. FTPI’s productivity dashboard is provided on its website in the “measurement assessment” section. See https://piu.ftpi.or.th/category/measurement/
← 6. Data on productivity growth are available at https://www.apo-tokyo.org/productivitydatabook/ over a long time period (1970-2021, for the APO Productivity Database 2023, Ver.1, November 07, 2023).
← 7. For instance, Thailand subscribed to the IMF Special Data Dissemination Standard (SDDS, one of the most common international standards for dissemination of statistical indicators and publications) in 1996, and met all the SDDS requirements in 2000 (see https://dsbb.imf.org/sdds/country/THA/category).
← 8. The availability, timeliness and dissemination of indicators could be improved beyond productivity statistics for other relevant data. For instance, the NESDC could ensure that Input/Output tables are regularly produced (ideally annually), as they are currently only publicly available up to 2015.
← 9. MFP is computed as a residual after accounting for the contribution of a combined measure of labour and capital. An index of combined inputs can be obtained by appropriately weighting the different inputs, usually using factor income shares. Income factor shares refer, for instance, to the share of employee compensation in total cost. The OECD manual on measuring productivity (OECD, 2001[4]) provides detailed guidelines on how to properly compute these shares.
← 10. An analysis of the within- and between-sector contributions to aggregate productivity can be extended and applied to any stratifying variable, such as industries, regions and industry-regions. An analysis of Thailand’s productivity growth can also benefit from examining sub-national data jointly with industry data. Jointly analysing detailed information by regions and industries would, for instance, enable a policy-relevant understanding of productivity disparities. This analysis would clarify whether regional disparities stem from sectoral compositions within regions, where regions specialise in certain activities, or from variations in firm performance within industries across different regions. Such analysis has important policy implications for regional development goals.
← 11. Such decompositions, at broad and more detailed industry levels, are implemented in the OECD Compendium of Productivity Indicators (APO, 2023[15]), showing that in OECD countries manufacturing was the main contributor to the slowdown in productivity after the 2008-09 crisis, while trade, hotels and transport services were the main contributors to the slowdown in the labour productivity growth of business services in about half of OECD countries. These decompositions are also routinely implemented by the Asian Productivity Organization (APO, 2023[15]). The decompositions of aggregate productivity growth into within- and between-effects are frequently implemented using labour productivity, as this indicator is more frequently available at industry levels.
← 12. This is relevant for Thailand given its persistently high share of employment in agriculture (World Bank and Bank of Thailand, 2020[14]). However, the increasing importance of the service sector may also raise new challenges since it usually features lower levels of productivity than manufacturing across countries. Additionally, the digitalisation of economies and the green transition may lead to significant reallocation effects across industries, strengthening the relevance of monitoring their effects on productivity.
← 13. To decompose aggregate growth into sectoral contributions, each sector’s growth is weighted according to its economic weight (for instance, in terms of employment).
← 14. Annual per-worker labour productivity growth over the period 2001-2021 is decomposed into the contribution of the following nine sectors: 1. Agriculture, 2. Mining, 3. Manufacturing, 4. Electricity, gas, and water supply, 5. Construction, 6. Wholesale and retail trade, hotels, and restaurants, 7. Transport, storage, and communications, 8. Finance, real estate, and business activities, 9. Community, social, and personal services. The annual productivity growth of each sector is weighted by its employment share.
← 15. A recent study from the Office of Agricultural economics (OAE, 2022[92]) assesses factors affecting overall productivity in the agricultural sector of Thailand, such as the rate of change in export and import, and in rainfall.
← 16. For instance, the French National Productivity Board complemented its 4th annual report with a statistical dashboard containing 68 charts covering the following seven areas: (i) productivity, (ii) competitiveness, (iii) business demographics, (iv) business dynamism, (v) business investments in ICT, (vi) constraints on businesses and (vii) businesses facing environmental challenges. See https://www.strategie.gouv.fr/sites/strategie.gouv.fr/files/atoms/files/cnp2023-tableau_de_bord-octobre.pdf
← 17. This report refers to the APO/OECD (2021[2]) for a thorough discussion of the challenges and more detailed recommendations on how to address them.
← 18. This is for instance required for the implementation of the state-of-the-art methodology to estimate MFP (the “control function approach”) but also for the analysis when investigating within-unit productivity growth across groups of firms.
← 19. Legal units are recognised by law or society and form the basis to define statistical units. Most administrative registers are based on legal units. An administrative unit is a unit created for administrative purposes, for instance to comply with an administrative regulation, for example VAT or Social Security.
← 20. Other internationally accepted classifications are kind-of-activity units (KAUs) and local units.
← 21. Productivity analyses can also leverage other relevant data available in Thailand which are not the main focus of this report. This includes, for instance, the R&D survey from the Office of National Higher Education Science Research and Innovation Policy Council (NXPO) and the business ICT survey from the NSO. Such data would best serve productivity analyses when used in an integrated data framework, as discussed in a following sub-section.
← 22. The NSO also conducts a census for the agricultural sector. The assessments and recommendations for measuring and analysing productivity included in this report also apply to this sector. However, measuring and analysing productivity in the agricultural sector involves specific characteristics that need attention. FAO (2018[87]) provides guidelines for the measurement of productivity and efficiency in agriculture, which account for the specificities of the sector as well as of developing countries. The OECD launched a Network on Agricultural TFP and the Environment in 2017 to help implement measures of agricultural performance that account for the environment.
← 23. The NSO conducted the Industrial Census in 1964, 1997, 2007, 2012, 2017, 2022 with each wave collecting information pertaining to the previous calendar year. The NSO planned a data collection every ten years starting with the 2nd Industrial Census, and revised the frequency to conduct the survey every five years, starting with the 3rd Industrial Census of 2007. See https://www.nso.go.th/nsoweb/main/summano/Pf?set_lang=en.
← 24. The NSO conducted the first wave in 1966, and the second in 1988. From 2002 onwards, the NSO conducted the Business Trade and Service Census every 10 years (in 2012 and 2022) jointly with the Industrial Census.
← 25. The data integrated by the NSO includes the database from the Department of Business Development (Ministry of Commerce), the Department of Industrial Works (Ministry of Industry), the Department of Agricultural Extension (Ministry of Agriculture) and the Social Security Office of Thailand (Ministry of Labour).
← 26. Specifically, in the stratification, Bangkok and each province constituted a stratum. There were altogether 77 strata. Each stratum was classified by TSIC-2009 into 818 sub-strata at 409 activity levels, and each sub-stratum was divided into two sub-sizes according to number of persons engaged: one to five and six to ten. In 2022 the sample includes 49,952 units from a total of 398,645 establishments, representing around 12% of the total population of establishments with one to ten persons engaged.
← 27. The 2022 edition provides a complete enumeration of 60,558 establishments.
← 28. Indeed, prior to this edition (for the 2017 and previous waves), the NSO constructed the sampling frame from past censuses and surveys, which was costly to compile and maintain. The information from all establishments had to be checked by the enumerators, who visited all establishments on the list prepared by the NSO and newly formed establishment (Thailand, 21-25 May 2018[86]).
← 29. Although the frequency of the Business Trade census could be aligned with those of the Industrial Census, the NSO conducts censuses with a periodicity consistent with international practices. In this respect, the limitation of the censuses for productivity analyses is not specific to Thailand. A pathway for improvement would be an integrated statistical information system leveraging administrative sources and surveys, along with the census.
← 30. For instance, the World Bank and Bank of Thailand report (2020[14]) indicates that the panel component reduces the data to around half of the original sample when combining the 2006, 2011 and 2016 waves of the Census. It is however difficult to assess the extent to which this results from the sampling strategy – with a small portion of micro-units being sampled across waves– or from firm exiting the market.
← 31. This refers to the survey data from the Annual Factory Operation Information Form (R.N.9).
← 32. The unit-identifier relies on the Factory Registration Number, which may be common with other data. It may therefore allow to merge this survey with other data to complement the set of information, but the OIE may further assess the feasibility of linkages with datasets that contain complementary information (notably on employment).
← 33. It is mandatory to complete this survey, but in practice firms compile the questionnaire on a voluntary basis. This reduces the representativeness of the data at the “bottom of the productivity distribution” – see discussion in Berlingieri et al. (2017[51]). The OIE nonetheless conducts relevant analyses of the distinct characteristics of low-efficiency sample groups (“worst practices”) using statistical testing methods, as described OIE (2022[53]).
← 34. Around 15% of the sample of registered firms does not provide financial statements. Those firms are generally two years old or less, and they might not report financial statements because they do not generate revenue yet.
← 35. The businesses that register at the Ministry of Commerce represent more than 1.2 million juristic persons.
← 36. From 1998 the survey has been conducted quarterly, and since 1994 the NSO has expanded the sample size. Since September 2001, data have been presented monthly.
← 37. The January-March 2024 survey selected a total of more than 80k private households for enumeration. The primary and secondary sampling units were enumeration areas (EAs) for municipal areas and non-municipal areas, and private households and persons in the collective households respectively. At the first stage, the EAs based on the 2010 census frame were updated from other sample surveys and selected separately and independently in each stratum by using probability proportional to size, giving the total number of households. The total number of sampled EAs was 5430 from 139325 EAs. At the second stage, private households and persons in the collective households were the ultimate sampling units. A new listing of private households was made for every sampled EA to serve as the sampling frame. In each sampled EAs, a systematic sample of private households were selected with the following sample size: Municipal areas: 16 sample households per EAs and non-municipal areas: 16 sample households per EAs. https://www.nso.go.th/nsoweb/storage/survey_detail/2024/20240521125659_79859.pdf
← 38. More specifically, since 2002 the NSO based the survey on a rotating sample design to improve the quality of estimator, with four rotation groups and a “2-2-2 pattern”. This means that, for each household sampled, information is collected for two consecutive quarters, after which the household is omitted from the survey for two consecutive quarters and then surveyed again for the next two quarters. This rotating sample design implies that for every year, around 50% of the sample in the labour force survey at time t-1 is also surveyed at time t.
← 39. The Department of Skill Development from the Ministry of Labour also collects useful information on workers that receive training, to assess the impact of such training on workers’ productivity (proxied with wages).
← 40. The survey follows the international definitions. It follows the OECD Frascati Manual (OECD, 2022[89]) for R&D, the Oslo Manual (OECD/Eurostat/European Union, 1997[90]) for innovation activity, and the FoS 2007 for the classification of research fields.
← 41. S-curve industries include: Next Generation Automotive, Smart Electronics, Affluent, Medical and Wellness, Tourism, Agriculture and Biotechnology, Food for the Future, Robotics, Aviation and Logistics, Biofuels and Biochemicals, Digital, and Medical Hub. The conventional classifications of industry followed by the RDI survey are not always suitable for the new agenda focusing on forward-looking, innovative target industries. New definitions are often made on a case-by-case basis on ad-hoc projects.
← 42. The project is under the cooperative effort of the Office of National Higher Education Science Research and Innovation Policy Council (NXPO), the Thailand Science Research and Innovation (TSRI), the Office of MHESI Permanent Secretary and the National Research Council of Thailand (NRCT).
← 44. See more information on the following webpage of the Digital Government Development Agency: https://www.dga.or.th/our-services/digital-platform-services/gdx/
← 45. For more information on guidelines to measure the informal sector, see the statistical manual produced by the United Nations Expert Group on Informal Sector Statistics (Delhi Group) and ILO (2013[91]).
← 46. From 2006 to 2016 the data collected comprised approximately 80 000 randomly selected households, for a total of around 200 000 individuals, representing around 0.1-0.5 percent of the total Thai population (Korwatanasakul, 2021[48]).
← 47. Some systems (for instance the United States Census Bureau) even integrate this into their requirements for data use.
← 48. This framework is widely used by national statistical agencies (such as the Australian Bureau of Statistics, Statistics Canada, Stats NZ), research institutions (such as the IAB in Germany), provinces and individual agencies (e.g. Province of British Columbia, BC Ministry of Citizens’ Services).
← 50. Variables include age, size, and location, as well as financials, digital readiness, workforce composition, innovative activities like R&D spending, and operational obstacles.
← 51. While Paweenawat et al., (2017[54]) delve into these characteristics through an econometric lens, the World Bank and Bank of Thailand report (2020[14]) provides descriptive statistics on productivity for only a limited set of factors. Enhancing the joint report by the World Bank and the Bank of Thailand to include a more comprehensive characterisation or replicating Paweenawat et.al., (2017[54]) analysis across multiple waves would be highly beneficial to Thai policymakers, since the analyses mentioned only cover Census’ editions up to the 2012 wave.
← 52. This includes factors internal to the firms (like workers’ skills, use of digital technologies) or external factors (like the quality of infrastructure, competition and regulations).
← 53. Additionally, Muthitacharoen et al. (2024[41]) discuss how policies intended to promote SMEs, such as a revenue cap in the corporate income tax exemption inadvertently led to lower revenue growth and investment, particularly among initially low-profitable firms. This reduced investment can further translate into lower productivity and higher market shares for large firms, potentially undermining the competitive environment. This underscores the need for a careful policy design to avoid unintended economic distortions and warrants further analysis, but also the need for in-depth analyses of the reallocation process and the competitive environment.
← 54. This indicator is particularly suited for cross-sectional data, such as the Industrial Census, and can be applied to the services sector as well. Descriptive statistics of the size-productivity relationship can also offer valuable insights into the efficiency of resource allocation.
← 55. The set of information available in the CPFS data, used by (SCB), allows the computation of the share of capital in zombie firms but not the share of employment which would require linking CPFS data with other sources.
← 56. See for instance the OECD DynEmp project, the OECD Structural and Demographic Business Statistics or the Eurostat Business Demography Statistics.
← 57. The CPFS data do not cover unincorporated businesses, a common limitation in some administrative datasets. Understanding the entry dynamics of these businesses is nonetheless important for a comprehensive view of business dynamics and its contribution to employment, especially in countries with a large share of self-employed individuals.
← 58. The census collects information for only around 10% of micro firms. However, most firms enter the market with fewer than 10 employees (Calvino, Criscuolo and Menon, 2015[73]), implying that entry may be understated in the Census data.
← 59. Hsieh and Klenow (2014[60]) propose a method to disentangle the effects of firm growth from cohort-specific influences related to economic environment or business cycle at the time of their inception.
← 60. The UNSD describes the collaborative as follows: “The collaborative is a platform to share resources, tools, best practices and experiences, and contributes to raising awareness among all members of national statistical systems about the benefits of sharing and combining administrative sources to enhance the quality, timeliness, coverage and level of disaggregation of statistical data”. https://unstats.un.org/UNSDWebsite/capacity-development/admin-data/
← 61. See also relevant resources from Eurostat: https://ec.europa.eu/eurostat/web/microdata. SDC2023_S1_2_Eurostat_Espelage_D_0.pdf (unece.org); UNSD (2023[88]) discusses principles of microdata dissemination and examples from countries that are investing in their data sharing.
← 62. Elementary data files appropriately processed through methods that, by limiting the original information content, make it possible to restrain the risk of breaching the confidentiality of data subjects.