Adult skills and productivity: New evidence from PIAAC 2023

After two decades of low productivity growth, policymakers maintain a keen interest in understanding the drivers of productivity and what policy can do to restore productivity prospects. Economic research has identified a range of structural headwinds to productivity, including slowing innovation (Brynjolfsson et al. 2021), stalling technological diffusion and “winner-take-all” dynamics (Andrews, Criscuolo and Gal, 2016), declining reallocation (Decker et al. 2020) and rising market power (De Loecker et al. 2020). In contrast, the role of human capital accumulation has received less attention. One exception is a recent OECD study, which shows the relevance of human capital accumulation for productivity performance by exploiting changes in student-level educational performance, measured by student test scores (Andrews et al. 2024).

Against this backdrop, the 2023 Survey of Adult Skills (PIAAC – Programme for the International Assessment of Adult Competencies) – a comprehensive survey of adults aged between 16 and 65 in 31 OECD countries and 6 non-OECD countries (Box 1) – provides another opportunity to analyse the economic consequences of human capital accumulation. In this regard, we study the productivity implications of cross-country differences in the level of adult skills (Figure 1), which, according to PIAAC, are highest in Finland, Japan and Norway and lowest in Chile, Poland and Portugal. And the scale of these differences are material: average PIAAC scores in the top three performing countries are around 10 percent higher than the OECD average and 25 percent higher than in the bottom three performing countries.

We exploit cross-country variation derived from industry-level data to better understand the key channels that link human capital to productivity. This concerns both the direct effect of skills and how skills can be allocated to generate productivity gains. The latter channel relates to the capacity of economies to more efficiently allocate the existing stock of skills at any point in time – by minimising labour market mismatches (i.e. static allocative efficiency) – as well as to redeploy scarce high-skilled labour over time to underwrite the expansion of dynamic firms (i.e. dynamic allocative efficiency). This note sheds new empirical light on these mechanisms, quantifies their economic magnitude and offers some preliminary reflections for policy.

Human capital is widely recognised as a cornerstone to economic growth in the theoretical literature (Jones, 2016). Statistical decomposition exercises indicate that human capital contributes significantly to economic growth and explains differences in per capita income across countries (Hall and Jones, 1999). Consistent with this evidence, the OECD has long identified enhancing human capital as a key structural policy priority for many countries (OECD, 2018).

Recent OECD work (Andrews et al., 2024; and Égert et al., 2024) that uses a novel measure of human capital – constructed by combining cohort-level data on PIAAC, student test scores (PISA) and mean years of schooling – identifies a robust positive link between human capital accumulation and aggregate productivity in OECD countries (Égert et al., 2024). At the aggregate level, however, the productivity impacts of human capital vary significant across countries. Some countries are more efficient at allocating the existing stock of human capital – as illustrated by a stronger pass-through of human capital to productivity (Appendix A) – and in ways that are connected to worker-firm matching efficiency and the policy environment (Andrews et al., 2024).

In fact, a nuanced understanding of the latest PIAAC results requires an analysis of granular – industry-level – data to pin down the key channels linking adult skills and productivity. A key observation is that within any given industry, a higher average level of adult skills will directly support aggregate productivity performance (Figure 2). Indeed, a higher share of high skilled workers – relative to lower skilled workers – will enable both the generation of new ideas and the broader diffusion of existing ideas. But adult skills will also impact productivity via indirect channels.

Since the stock of adult skills is relatively fixed in the short to medium term, the allocation of skilled workers across firms will also matter for industry-level productivity (Figure 2). Earlier OECD research found that higher rates of skill mismatch within industries go hand in hand with lower labour productivity due to inefficient resource allocation (Adalet McGowan and Andrews, 2017). From the perspective of a single firm, hiring an over-skilled worker may be beneficial for productivity.1 But over-skilling in any given firm could be harmful for aggregate productivity if there exist more productive firms that could better utilise these skills but find it difficult to expand due to a lack of suitable labour.

In an economy where firms are relatively homogenous, the potential gains to aggregate productivity from such a reallocation of mismatched workers would be relatively small. In practice, however, the degree of firm heterogeneity is striking:

• First, highly productive firms coexist with low productivity firms: even within narrowly defined industries in the United States, firms at the 90^th percentile of the total factor productivity (TFP) distribution are twice as productive as firms at the 10th percentile (Syverson, 2004).

• Second, the same is true with respect to the firm size distribution, with many small firms co-existing with a smaller number of very large firms, which are typically more productive (Bartelsman et al., 2013).

• Finally, firms vary greatly in their growth potential: many firms do not grow at all, a small cadre of young firms tend to disproportionately drive net job creation, while small and old firms tend to destroy jobs on net (Haltiwanger et al 2013; Criscuolo et al 2014).

This widespread firm heterogeneity implies that aggregate productivity will also depend upon how skilled workers – which are currently in short supply (OECD, 2024a) – are allocated across firms and matched to various job roles. At any point in time, aggregate productivity is an increasing function of static allocative efficiency, which measures the extent to which scarce resources are allocated to their highest valued use in the form of higher quality (i.e. more productive) firms (Haltiwanger, 2011; Andrews and Hansell, 2021). Dynamic allocative efficiency captures the extent to which resources are moving towards higher quality firms over time. Achieving static allocative efficiency in one period requires sufficient dynamic allocative efficiency in preceding periods, and differences in this process is now a leading explanation for why some countries are more productive than others (Bartelsman et al., 2013; Hsieh and Klenow, 2009).

This note exploits a novel cross-country industry-level database created through the merger of sector-level labour productivity data from the OECD National Accounts database and a range of skills indicators drawn from the 2023 PIAAC dataset. This dataset covers twelve one-digit industries in almost 30 OECD countries. Only non-farm business sectors were considered; non-market-based sectors such as agriculture, public administration, arts, health, education and the mining sector are excluded from the analysis.

The different channels through which adult skills shape industry-level productivity are examined in regression analysis (summarised in Table 1). In the baseline regression, the log-level of industry-specific labour productivity is regressed on the log-level of average PIAAC scores in corresponding industries (i.e. the direct effect) as well as on variables to capture the static and dynamic allocative efficiency processes (defined below).2 Both country and industry fixed effects are also included to control for country-specific factors (e.g. country-specific policies) and common industry-specific factors across countries (e.g. technological factors) that shape productivity and skills.

$\log \left({LP}_{c,s}\right)=\mathrm{ }\alpha \bullet \log \left({PIAAC}_{c,s}\right)+\beta \bullet {static\mathrm{ }allocation}_{c,s}+\gamma \bullet {dynamic\mathrm{ }allocation}_{c,s}+{CFE}_c{+SFE}_s+{\epsilon }_{c,s}$ (1)

where c and s denote countries and sectors and CFE and SFE stand for country and sector fixed effects, respectively. Static allocative efficiency is measured by labour market mismatch, measured by the share of workers that are mismatched both in terms of qualification and the field of study as well as the distribution of worker skills by firm size. Dynamic allocative efficiency is measured by the difference in the average skills (i.e. PIAAC score) of workers in growing and declining firms (in terms of employment) and the average skills of workers sunk in declining firms. The estimated regression coefficients are used to inform back-of-the envelope calculations to shed light on the relative importance of these channels and their aggregate relevance, as discussed below.

Adult skills of the working-age population, captured by the PIAAC test results, and aggregated at the country level are strongly correlated with aggregate productivity (Égert et al., 2024). PIAAC scores of workers also show a positive correlation with sector-level labour productivity (Figure 3).3 This positive relationship upholds in the regression (Table 1).4 Taken literally, our baseline estimates imply that closing the gap between the OECD average and the top 3 best PIAAC performing countries would be associated with an 18 percent increase in average OECD labour productivity. This overall estimate is based on the average gap in PIAAC results across industries in each country vis-à-vis the three countries with best average PIAAC results. More granular estimates can be obtained by considering the impact of closing gaps industry by industry, taking into account that the top performing countries may vary across industries. This is explored next.

How important are adult skills for understanding labour productivity gaps across countries? To investigate this question, we conduct a simulation exercise. First, sectoral productivity gaps are calculated as the difference between a country-sector’s PIAAC level and the PIAAC-level in the corresponding sector of the top 3 performing OECD countries. Second, the estimated coefficient linking PIAAC to productivity (first column from Table 1) is applied to derive the country-sector productivity gap implied by the PIAAC gap. This is then compared to the observed sector-level productivity gaps relative to top performing countries. Finally, sectoral shares in value-added are used to calculate the contributions to the productivity gap at the country level.

The results of this exercise suggest that on average across the OECD, differences in adult skills can potentially account for one-quarter of the cross-country gaps in average industry-level labour productivity (Figure 4).5 But it is notable that in some countries – particularly those in Southern and Eastern European – differences in adult skills can potentially account for one-third of productivity gaps. These estimates leave plenty of scope for other factors to explain the productivity gap – including the efficient allocation of skills – which is explored below. Figure 4. Country-level productivity gap explained by sector-level PIAAC differences.

The results reported in Figure 4 are based on the average positive relationship between labour productivity and PIAAC scores at the industry level (Table 1, column 1). This average relationship masks significant differences across group of countries in the pass-through of adult skills to productivity: it is in fact much stronger in Nordic countries than elsewhere in the OECD (Table 1, Column 2). This could reflect the idea that Nordic economies are more efficient at allocating human capital, possibly due to structural policy frameworks that support reallocation and adaptability (see Andrews et al., 2024). In any case, this heterogeneity across countries motivates a deep dive into the link between human capital allocation and productivity in the next section.

While labour productivity is clearly connected to the average level of adult skills, so far we have been silent on how those skilled workers are allocated across firms and jobs within a given sector. To address this question, we consider how efficiently skills are allocated at any point in time (i.e. static allocative efficiency) and whether skills are being allocated to better firms over time (i.e. dynamic allocative efficiency). While PIAAC does not contain data on firm productivity, it does contain information on both the firms’ size (i.e. level of headcount) as well as its growth status (i.e. headcount is growing, static or declining), from which we can draw inferences about firm performance. Moreover, PIAAC contains various measures of labour market mismatch, which previous OECD research showed has a close (theoretical and empirical) link with static allocative efficiency (Adalet McGowan and Andrews, 2017).

Static allocative efficiency can be captured by the concept of labour market mismatch and the distribution of worker skills by firm size. Mismatches arise when workers are employed in jobs that are either too demanding or not challenging enough. This can occur either because a worker’s qualification does not match the required qualification or because the worker’s field of study is misaligned with the type of job he/she is employed in (Box 2). A particularly compelling definition of labour market mismatch combines both qualification and field of study mismatches: workers are deemed mismatched if they are misaligned in terms of both their specialisation (field of study mismatch) and their qualification (qualification mismatch).

Qualification and field of study mismatches are prevalent in OECD countries. On average, nearly 35 percent of workers in the OECD are employed in jobs that require a lower or higher qualification than their highest level of qualification (Panel A in Figure 5). Similarly, over 35 percent of workers hold jobs that do not align with their field of study (Panel B in Figure 5). These mismatches differ across countries. For instance, Korea has the highest level of field of study mismatch but performs much better in terms of qualification mismatch. Conversely, Switzerland excels in minimising qualification mismatch but faces challenges in terms of field of study mismatch.

Labour market mismatch is potentially most acute when measured by the combination of qualification and field of study mismatch – that is, when workers are mismatched both in terms of qualification and field of study. According to this metric, about 11 percent of workers on average across the OECD experience mismatch. The incidence of mismatch tends to be higher in industries such as transport, hospitality and administrative services and lower than average in ICT, finance and professional services. (Panel C in Figure 5).

There is consensus in the literature that larger firms are often more productive than smaller firms and can better deploy skilled workers and that better-skilled workers can perform jobs more productively in larger firms (Haltiwanger, 2011). The new 2023 PIAAC data confirms that average worker skills are higher in larger firms (Figure 6). However, it also shows that this is not always the case in every country. While workers’ skill increases monotonically with firm size in New Zealand, workers exhibit better skills only in very large firms in Italy. By contrast, in Slovakia, PIAAC scores in large firms are inferior to those in smaller firms (see Appendix B).

Dynamic allocative efficiency can be best measured through the skills of workers employed in growing, static and declining firms in terms of employment. Improving allocative efficiency over time implies that growing firms would attract and employ more skilled workers while declining firms would be left with workers with lower skills. From a dynamic perspective, on average across the OECD, skills are moving in the right direction over time: growing firms employ more skilled workers than declining firms. However, this is not always the case: in Italy, declining firms employ workers with better skills than growing and static firms (Figure 7).

The industry-level analysis utilised in this note can shed light on the relative importance and aggregate significance of these two channels. The cross-country industry-level estimation results, displayed in Table 1 (columns 3 to 6), indicate that the share of mismatched workers tends to be negatively related to sector-level productivity, particularly when field of study and qualification mismatches are combined. Put differently, countries and industries that achieve a more efficient matching of workers in terms of qualification and specialisation are found to exhibit higher labour productivity.

The way in which skilled workers are allocated across firms of different size also matters (Table 1, columns 3 and 4). First, productivity is greater in industries where larger firms employ proportionately more skilled workers compared to smaller firms. Consistent with this is the finding that more skilled workers in smaller firms tends to act as a drag on sectoral productivity. On the dynamic side, allocating higher-skilled workers to expanding firms at the expense of contracting firms is associated with higher productivity (Table 1, columns 5 and 6). The corollary is that productivity is lower when higher-skilled workers are trapped in declining firms.

How much of the cross-country labour productivity gaps can be explained by allocative efficiency? To investigate this question, we conduct a “closing the gap” simulation exercise, focusing on recovering the contributions of labour market mismatch to the overall productivity gap. First, we calculate the difference in labour market mismatch of all countries in a specific industry relative to the average of the top 3 countries with the least mismatch. Coefficient estimates from column 3 of Table 1 are used to derive the implied labour productivity gap, which is then compared to the observed productivity gap. Sectoral value-added shares are used to calculate the contribution of mismatches to the country level productivity gap.

Back-of-the-envelope calculations suggests that for an average OECD country, static allocative efficiency, that is labour market mismatch, accounts for about 12 percent of cross-country labour productivity gaps (Figure 8). This is a conservative estimate because dynamic allocative efficiency effects will eventually feed into static allocative efficiency. For instance, if Italy’s PIAAC allocation across growing and declining firms were to move to that of the OECD average (Figure 8), market sector labour productivity could be more than 4 percent higher. While this suggests that the aggregate gains from redeploying high skilled workers to dynamic firms are likely to be material, a more systematic accounting exercise – of the contribution of dynamic allocative to cross-country productive performance – is left to future research.

The results – which illustrate the importance of adult skills for productivity performance – raise at least two questions for policy, which should be the focus of future research. First, given the strong positive relationship between industry labour productivity and the average level of workers skills, what can policy do to increase the skillset of workers in the economy? Second, how can countries more productively allocate the existing stock of adult skills in an economy, noting that industry-level labour productivity tends to be higher when labour market mismatch is lower and skilled workers are more likely to be allocated to better performing firms?

On the first question, work-related training is crucial for improving adult skills as it significantly enhances adaptability and productivity in the workforce. Participation in training programmes is strongly correlated with higher PIAAC scores. But in many OECD countries, the incidence of training is very low. This is particularly the case for low-educated individuals who potentially have the most to gain from participating in work-related training programmes (Figure 9). This underscores the importance of education policies at younger ages to build foundational skills. Recent OECD research emphasises that participation in good-quality early-childhood education, exposure to high-quality teachers, school assistance to homework and a regulated use of personal digital devices in school all help improve student test performance (Andrews et al. 2024). These fundamental skills will make children and workers more adaptable to changing job requirements and technological advancements.

On the second question, while the economic magnitude of the allocation of skills is more modest than that of the average skills level effects, the scope for policy to leverage this channel is higher in the near term. On this front, it is significant that the share of workers employed in growing businesses is positively associated with the extent to which policy supports reallocation and adaptability in labour markets, as proxied by a composite indicator that combines product market regulations, employment protection legislation, insolvency regimes and ALMP spending (see Andrews et al. 2024; Figure 10). This highlights the continued relevance of structural policies that promote more fluid labour markets and firm dynamism to support efficient matching, at a time when the headwinds to human capital accumulation have never been stronger.

Adalet McGowan, M. and D. Andrews (2017), "Labor Market Mismatch and Labor Productivity: Evidence from PIAAC Data," Research in Labor Economics, in: Skill Mismatch in Labor Markets, Volume 45, pp. 199-241.

Andrews, D. and D. Hansell (2021), Productivity-enhancing labour reallocation in Australia, Economic Record 97(317), pp. 157-169.

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.

Andrews, D., B. Égert and C. de La Maisonneuve (2024), From decline to revival: Policies to unlock human capital and productivity, OECD Economics Department Working Paper No. 1827.

Bartelsman, E., J. Haltiwanger, and S. Scarpetta (2013), Cross Country Differences in Productivity: The Role of Allocation and Selection, American Economic Review, 103(1), 305-334.

Brynjolfsson, E., Rock, D. Syverson, C. (2021), “The productivity j-curve: How intangibles complement general purpose technologies”, American Economic Journal: Macroeconomics 13(1), 333–72.

Criscuolo, C., P.N. Gal and C. Menon (2014), The dynamics of employment growth: New evidence from 18 countries, OECD Science, Technology and Industry Policy Papers 14.

Decker, R., J. Haltiwanger, R. Jarmin, and J. Miranda (2020), Changing Business Dynamism and Productivity: Shocks vs. Responsiveness, American Economic Review, 110 (12): 3952–90.

De Loecker, J., J. Eeckhout and G. Unger (2020), The Rise of Market Power and the Macroeconomic Implications, The Quarterly Journal of Economics, 135(2), 561–644, https://doi.org/10.1093/qje/qjz041

Égert, B., C. de La Maisonneuve and D. Turner (2024), A new macroeconomics measure of human capital exploiting PISA and PIAAC: linking education policies to productivity, Education Economics, 32(6), 745-761.

Haltiwanger, J. (2011), Firm dynamics and productivity growth, European Investment Bank Economic Department Working Paper No 5/2011.

Haltiwanger, J., R. Jarmin and J. Miranda (2013), Who Creates Jobs? Small Versus Large Versus Young The Review of Economics and Statistics, Vol. 95, Issue 2 (May 2013) , pp. 347-361.

Hsieh, C. and P. Klenow (2009), Misallocation and Manufacturing in TFP in China and India, Quarterly Journal of Economics, 124(4), 1403-1448.

Jones, C. I. (2016), The Facts of Economic Growth, Handbook of Macroeconomics, Volume 2A, Chapter 1, 3-69.

Hall, R. and C. Jones (1999), Why Do Some Countries Produce So Much More Per Worker Than Others?, The Quarterly Journal of Economic, 114(1), 83-116.

OECD (2018), Economic Policy Reforms 2018: Going for Growth Interim Report, OECD Publishing, Paris.

OECD (2024a), OECD Economic Outlook, Volume 2024 Issue 2. OECD Publishing, Paris.

OECD (2024b), Do Adults Have the Skills They Need to Thrive in a Changing World?: Survey of Adult Skills 2023, OECD Skills Studies, OECD Publishing, Paris, https://doi.org/10.1787/b263dc5d-en.

Syverson, C. (2004), “Sustainability and Product Dispersion”, Review of Economics and Statistics, Vol.

86(2), pp. 534-550.

Dan ANDREWS (✉ Dan.Andrews@oecd.org)

Balázs ÉGERT (✉ Balazs.Egert@oecd.org)

Christine de LA MAISONNEUVE (✉ Christine.Maisonneuve@oecd.org)