Structural Change, Employment, and Inequality in Europe: An Economic Complexity Approach

Caldarola, Bernardo; Mazzilli, Dario; Patelli, Aurelio; Sbardella, Angelica Working Paper Structural Change, Employment, and Inequality in Europe: An Economic Complexity Approach UNU-MERIT Working Papers, No. 2024-033 Provided in Cooperation with: Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), United Nations University (UNU) Suggested Citation: Caldarola, Bernardo; Mazzilli, Dario; Patelli, Aurelio; Sbardella, Angelica (2024) : Structural Change, Employment, and Inequality in Europe: An Economic Complexity Approach, UNU-MERIT Working Papers, No. 2024-033, United Nations University (UNU), Maastricht Economic and Social Research Institute on Innovation and Technology (UNU-MERIT), Maastricht This Version is available at: https://hdl.handle.net/10419/326929 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-sa/4.0/ #2024-033 Structural Change, Employment, and Inequality in Europe: an Economic Complexity Approach Bernardo Caldarola, Dario Mazzilli, Aurelio Patelli and Angelica Sbardella Published 25 November 2024 Maastricht Economic and social Research institute on Innovation and Technology (UNU-MERIT) email: [email protected] | website: http://www.merit.unu.edu Boschstraat 24, 6211 AX Maastricht, The Netherlands Tel: (31) (43) 388 44 00 UNU-MERIT Working Papers ISSN 1871-9872 Maastricht Economic and social Research Institute on Innovation and Technology UNU-MERIT UNU-MERIT Working Papers intend to disseminate preliminary results of research carried out at UNU-MERIT to stimulate discussion on the issues raised. Structural Change, Employment, and Inequality in Europe: an Economic Complexity Approach Bernardo Caldarola∗1,2,3, Dario Mazzilli1, Aurelio Patelli1, and Angelica Sbardella1 1Enrico Fermi Research Center, Rome, Italy 2UNU-MERIT, Maastricht, the Netherlands 3School of Business and Economics, Maastricht University, Maastricht, the Netherlands November 25, 2024 Abstract Structural change consists of industrial diversification towards more productive, knowledge-intensive activities. However, changes in the productive structure bear inherent links with job creation and income distribution. In this paper, by taking an economic complexity approach, we investigate the consequences of structural change defined in terms of labour shifts towards more complex industries on employment growth, wage inequality and functional distribution of income. The analysis is conducted for European countries using data on disaggregated industrial employment shares over the period 2010 – 2018. First, we identify patterns of industrial specialisation by validating a country-industry industrial employment matrix using a bipartite weighted configuration model (BiWCM). Secondly, we introduce a country-level measure of labour-weighted Economic Fitness, which can be decomposed in such a way as to isolate a component that identifies the movement of labour towards more complex industries the structural change component. Thirdly, we link structural change to i) employment growth, ii) wage inequality, and iii) the labour share of the economy. Our findings indicate that the structural change measure we propose is associated negatively with employment growth. However, it is also associated with lower income inequality: as countries move to more complex industries, they drop the least complex ones, so the (low-paid) jobs in the least complex sectors disappear. Finally, structural change predicts a higher labour ratio of the economy; however, this is likely to be due to the increase in wages rather than to job creation. ∗Corresponding author. Email: [email protected] 1 Keywords: Structural Change, Economic Complexity, Inequality, Employment JEL Codes: D63, E24, J31, O11, O52 2 1 Introduction Structural change can be defined by the reallocation of economic activity across broad sectors of the economy (Lewis 1954,Syrquin 1988). The shift of the relative importance of industrial employment or output towards increasingly more productive activities that defines the process of structural change is also an engine of long-term, sustained economic growth (Herrendorf et al. 2014,2022). However, changes in the productive structure are also inherently associated with the distribution of income, as highlighted in the seminal work by Kuznets (1955). For instance, when countries upgrade their productive structure, they also move to sectors with higher knowledge intensity, and lower labour requirements. This may have consequences on the wage and functional distribution of income, as well on the creation of new jobs and the destruction of old ones. In this paper, we use the case of Europe to study the relationship between the movement of labour to increasingly knowledge intensive activities – defined here as structural change – with employment growth, wage inequalities, and the functional distribution of income between labour and capital. As a matter of fact, in the last decades within country inequality has been growing in Europe and in high-income countries. At the same time, these countries have de-industrialised during the last wave of ICT-led globalisation, leading to the substitution of manufacturing with jobs in complex, knowledge intensive industries – such as business services – contributing to increasing inequalities (Antonelli & Tubiana 2020). The mechanism underlying this trend has been explored further in the empirical literature. The Skill-Biased Technological Change (SBTC) literature (Autor et al. 2003,Autor 2022,Acemoglu & Restrepo 2022) has examined the impact of the main driver of structural change – technological change – on employment and on the distribution of earnings, suggesting that waves of technological change tend to complement only high-skilled workers in knowledge intensive industries, leaving low-skilled workers behind. Moreover, structural change bears a strong linkage with employment creation. As economies transform in their productive structures, they create winners and losers (Schumpeter 1934). The current wave of technological change may not reduce total employment (Dauth et al. 2021), but may induce labour market adjustments at the disadvantage of low-skilled workers, whose share over total workers is reduced (Graetz & Michaels 2018). In the case of other technologies, such as ICTs, these may have been the culprits of jobless growth (Brynjolfsson & McAfee 2011,Frey & Osborne 2017) and contributed to the shrinking of the number of middle-skill jobs (Goos et al. 2009,2014). In order to identify the degree of knowledge intensity of European industries we employ 1 an Economic Complexity (EC) perspective. Complexity metrics aim at quantifying the capability structure of countries by looking at their productive basket, preserving information on their specialisation and diversification patterns rather than by simply aggregating their economic output (Hausmann & Hidalgo 2009,Tacchella et al. 2012). By the same token, EC measures allow to identify the capability (knowledge) intensity (Lo Turco & Maggioni 2022) of individual products, industries, or technologies by considering their ubiquity – a proxy for the diffusion of the capabilities that such activities require – weighted by the capability endowment of the countries specialised in such activities – i.e., the countries’ Fitness. Given the strong link of EC metrics with economic diversification, and in particular of the Economic Fitness and Complexity (EFC) approach (Tacchella et al. 2012), they have proved to be a useful tool to describe a long-term dynamic such a structural change (Freire 2021,Castañeda et al. 2022,McNerney et al. 2023), being also highly predictive of GDP growth (Tacchella et al. 2018). Following the analogy between structural change and complexity, the latter can be used to explain the inequality trends that result of changes in the sectoral composition of the economy. The literature on complexity and inequalities (Sbardella et al. 2017,Hartmann et al. 2017,Hartmann & Pinheiro 2022) indicates that higher complexity of the productive structure is associated to lower inequalities at the country level. It is also well known that knowledge concentrates in space (Balland et al. 2020,2022), turning geography into a useful lens through which the income distribution can be analysed. In fact, at the regional level the sign of this relationship flips (Sbardella et al. 2017), as only regions with the strongest knowledge basis are able to diversify into highly complex industries, leaving laggards behind (Pinheiro et al. 2022). Looking at the relationship between complexity and employment, Adam et al. (2023) finds that higher complexity is linked to job creation. Where existing jobs are made redundant, it has been shown that re-employment is easier in presence of related industrial variety at the regional level (Hane-Weijman et al. 2018). One important feature of EC methods is that these allow to quantify the capability requirements of finely disaggregated industries, offering a granular description of the process of structural change both across and within broad sectors of the economy. In order to offer an encompassing assessment of the complexity of European industries and of the labour flows across them, we depart by traditional approach of the EC literature – which infers the productive structure of countries using export data – by using data on industrial employment in European countries between 2010 and 2019. First, we consider the use of employment data to measure industrial Complexity and countries’ employment Fitness as this allows to include also non-tradeable activities, most of which are in the service sector. This is particularly relevant considering that European economies have heavily shifted their productive structure towards the tertiary industries along with the process of globalisation. Secondly, the focus 2 on employment represents a crucial element in the conceptualisation of structural change as the change in relative shares of the economy. Using employment data to measure Economic Fitness and Complexity respectively at the country and industry level resonates with this conceptual framework, and allows to outline the structure of capabilities of labour as a productive input. In the present study, the analysis of the relationship between structural change, inequality, and employment creation is conducted for European countries over the period 2010 - 2018. The empirical approach relies on information on European industrial employment provided by Eurostats Structural Business Survey data, measured at a fine level of disaggregation (NACE 4 digits) across manufacturing and service industries. This information is matched with data on the distribution of wages at the country level, provided by the ILO, and with data on the labour share of the economy and employment growth, obtained by ARDECO. Using this newly constructed dataset, we start by identifying patterns of industrial specialisation by validating the country-industry industrial employment matrix using a bipartite weighted configuration model (BiWCM) (?) , overcoming some limitations imposed by the Balassa (1965) method. Secondly, we introduce a measure of labour-weighted Fitness, which sums up the complexity of industries weighted by their employment share. This measure can be decomposed in such a way as to identify the contribution to changes in labour-weighted Fitness coming from the movement of labour towards more complex industries. We identify such component as the one linked to structural change from a labour viewpoint the structural change component. Thirdly, we link the structural change component to a number of economic outcomes at the country level: i) employment growth, ii) wage inequality, and iii) the functional distribution of income (labour share of the economy). In this way, we analyse the distributional consequences of structural transformation – framed as the movement of labour to more complex industries – and its job-creating potential. An analysis of the specialisation patterns in European countries, conducted by inspecting the significant links of the country-industry specialisation matrix generated by the BiWCM, reveals that the most diversified countries have progressively moved away from specialisation in low-complexity, labour-intensive industries. Moreover, our empirical analysis explores the association between the structural change component of our labour-Fitness measure - derived from the movement of labour towards more complex industries - to employment growth, wage inequality, and the labour share of the economy). The preliminary results of OLS panel regressions with country and time fixed effects indicate that our structural change measure is associated negatively with employment growth, corroborating the evidence that highly complex industries have lower labour requirements. However, it is also associated with lower income inequality measured in terms of the ratio of average wages in the ninth and 3 first deciles of the wage distribution. As countries move to more complex industries, they drop the least complex ones, so the (low-paid) jobs in the least complex sectors disappear, making the first decile of the salary distribution go up. Finally, structural change predicts a higher labour ratio of the economy; however, this is likely to be due to the increase in salaries in highly complex industries rather than by job creation. Very interestingly, the structural change component is the only one to be significant in the regressions, while the increase in within-sector complexity, and the labour-weighted Fitness of countries do not explain our outcome variables. Results are robust to a battery of robustness checks, including leaveone-out regressions and a specification where the structural change component of the labourweighted Fitness is replaced by a similar component obtained decomposing a labour-weighted entropy measure. The remainder of this paper is organised as follows. In Section 2we delve into the Economic Fitness and Complexity framework, and we introduce a new measure – the LabourWeighted Fitness (LWF) the variation of which can be decomposed to extract a structural change component. In Section 3we outlined the data sources used in the analysis, which in turn is described in the following Section (4). Section 5illustrates the preliminary results of the analysis, and Section 6concludes. 2 Analytical Framework In this section, we will outline the analytical framework used to derive a measure of structural change based on the reallocation of labour to complex industries. First, we will define countries’ Fitness and industrial complexity according to the EFC framework (Tacchella et al. 2012). Second, we introduce a new measure – the Labour Weighted Fitness – from which we derive a structural change component. 2.1 Structural change, Fitness and Complexity Economic Complexity is a framework that builds on earlier evolutionary and institutional literature (Cimoli & Dosi 1995,Hirschman 1958,Teece et al. 1994) to tackle the complexity of economic systems. It describes the economy as an evolutionary process of globally interconnected ecosystems. The main recent advance with respect to the earlier literature is the use of newly developed network science and other methods to investigate complex and dynamical systems (Hausmann & Klinger 2006,Hausmann & Hidalgo 2009,Tacchella et al. 2012) to separate the random noise from the underlying signal. The Economic Complexity framework shifts the focus from aggregate quantities (What is the GDP of the country?) to a 4 Where betweenc,t,withinc,t and ∆LWFc,t are respectively the between term, within term, and sum of the previous two, as described by equation 6; in all cases, these quantities are computed for k= 1.Yc,t is a placeholder for measures of employment growth, wage inequality and labour share of the economy. Employment growth is computed as the share of employed, working age population in each country at time tover the same share of employment in the previous time period, as obtained from the ARDECO dataset: gc,t =empc,t empc,t−1 (9) Wage inequality is constructed using the ILO Statistics on Labour Income and Inequality described in Section 3.3. In order to capture the effect of structural change on the distribution of income, we take three different measures of wage inequality: 9th to 1st percentile, 9th to 5th percentile, and 5th to 1st percentile: ineqpth/qth c,t =wagepth c,t wageqth c,t (10) With the superscripts pand qindicating either the 9th, 5th or 1st percentile. Finally, we use the ARDECO data to compute a measure of the labour share of the economy in each country and year, taking the ratio of the total wages paid to workers over the total value added, all measured at constant prices (2015): labsharec,t =wagec,t V Ac,t (11) Finally, Xc,t is a vector of time varying controls at the country level, including GDP per capita, population (from ARDECO), share of R&D investments over GDP, share of exports over GDP (from the World Development Indicators). All equations are estimated using OLS country (σc) and time (τt) fixed effects, with errors clustered by country and year. Preliminary results of the descriptive econometric model are provided in the next section, along with an illustration of the results produced by the construction of the specialisation matrix and structural decompositions. 5 Results 5.1 Descriptive results We begin by commenting on the patterns exhibited by the specialisation matrix, which has been obtained by filtering the country-industry employment matrix for the year 2010 as 11 described in Section 4.1. Figure 1contains information on the statistical significance of individual country-industry linkages, with values going from more significant (red) to less significant (blue). The rows of the matrix represent countries, while the columns identify industries. Countries have been ordered from higher Fitness (top) to lower Fitness (bottom), and industries are sorted from the least (left) to the most (right) complex. 0 25 50 75 100 125 150 175 200 0 5 10 15 20 25 0.0 0.2 0.4 0.6 0.8 1.0 Figure 1: BIWCM Specialisation Matrix At a first visual inspection, the matrix displays a pattern ascribable to the nested pattern (Bustos et al. 2012) that typically characterises country-product RCA matrices, filtered using the Balassa index (Patelli et al. 2023). Precisely, the distribution of the statistically significant linkages between countries and industries that describes countries’ diversification patterns has a triangular shape, which can be used as a broad approximation of the nested nature of the system being observed. While the concept of nestedness has originated in evolutionary biology, it has found vast application in the economic complexity literature: productive systems that exhibit a nested pattern reveal the presence of correlations within the system that cannot be accounted for using linear models or average approximations. In economic terms, nestedness indicates that industrial activities that are pursued by low-Fitness countries are also pursued by high-Fitness countries, which are however also more diversified. In the specific case described in figure 1, the economic interpretation of nestedness holds only in part; as a matter of fact, the top left corner – where high-Fitness countries and lowComplexity industries are located – appears to be less populated that one would expected in a fully nested matrix. While a similar pattern emerges also in other systems – such as the 12 one of scientific production (Patelli et al. 2023) – in this case we find the observed result consistent with the process of structural change. When it comes to the employment dimension of the industrial production, high-Fitness countries tend to have higher diversification and to develop a comparative advantage in capitalor knowledge-intensive industries, at the expensive of more labour-intensive (and low-complexity) ones. As a result, industrialising countries with a comparative advantage in labour intensive activities (often due to lower wages) tend to specialise in the latter industries, and to have lower diversification. In substance, the sample of European countries seems to be divided in two subsamples of countries at different stages of structural transformation, with a minority of more mature countries and a majority of industrialising countries. The dynamic of Fitness rankings shows clearly which countries belong to each group, with the more mature economies being Germany, France, Italy, United Kingdom and Spain, followed by emerging countries such as Poland and Czechia. AT BE BG CH CY CZ DE DK EE EL ES FI FR HR HU IE IT LT LU LV MT NL NO PL PT RO SE SI SK UK AT BE BG CH CY CZ DE DK EE EL ES FI FR HR HU IE IT LT LU LV MT NL NO PL PT RO SE SI SK UK Fitness with BiWCM: ranking dynamics Figure 2: Fitness rankings: 2010-2018 We now move towards an analysis of the variation in LWF over the time period 2010–2018. 13 Figure 3shows the difference in LWF between 2010 and 2018, computed as per equation 6, with its respective between and within component. We shall recall that the former measures the contribution of structural change – defined here as the movement of labour to more complex industries – to the variation in LWF over the time period under consideration; while the latter picks up the change in LWF due to increasing complexity of the industries in which countries develop an (inferred) comparative advantage. Overall, figure 3indicates that the between and within component are often anticorrelated. This can be read in economic terms by considering the reasonable assumption that as industries become more complex, they also become less labour intensive. The within component captures if the industries in which a country is specialised become more complex – for instance, as a result of technical change within the industry. By doing so, labour is shed by such complex industries, flowing into industries with lower complexity – a pattern associated with a between component of the opposite sign to the within component. To have a clearer idea of what happened in European countries over the 2010–2018, figure 4describes the same quantities as figure 3, but for two different time periods: 2010–2014 (left panel) and 2014–2018 (right panel). The figure indicates that the first time period the variation in LWF fitness was negative for most countries, mostly driven by the decreasing within-industry complexity; it must be noted that this period coincides with the aftermaths of the 2008 financial crises, which has severely compromised economic activity in Europe across the board. In the following period, however, European countries show a much better performance, with most countries increasing their LWF, in many case due to within industry complexity and between industry reallocation of labour. One interesting case is the one of Poland, which is an outlier in both period: the country shows a huge within industry growth of complexity in 2010–2014, which is however followed by a negative between industry reallocation of labour in the following time period – hypothetically because of the lower labour intensity of its productive structure that has shifted towards increasingly knowledge-intensive activities. 14 UK BE SE EE LU FI CH NL CZ DE IT LV HU AT SK FR IE DK LT HR MT NO CY ES EL RO BG SI PL PT −2e−04 0e+00 2e−04 4e−04 Labour−weighted Fitness decomposition Country Decomposition Within term Between term Labour−weighted Fitness Figure 3: Labour-weighted Fitness decomposition: 2010-2018 15 CZ UK DE BE FR FI EE HU CH NL SE AT IT SK SI LV NO RO LT IE HR DK LU BG EL ES PT MT CY PL −2e−04 0e+00 2e−04 4e−04 Labour−weighted Fitness decomposition Country PL LU UK CY MT SE BE EE DK IE HR LV NL CH LT ES IT FI EL NO SK BG AT HU RO DE PT CZ FR SI −2e−04 −1e−04 0e+00 1e−04 2e−04 Labour−weighted Fitness decomposition Country Decomposition Within term Between term Labour−weighted Fitness Figure 4: Labour-weighted Fitness decomposition: 2010-2014 and 2014-2018 16 5.2 Structural change and employment growth, wage inequality, and functional distribution of income In this section we report the results of the estimation of equation 8using the different outcome variables for employment growth, wage inequality and functional distribution of income. It should borne in mind that, at the present stage, the results presented below aim exclusively at describe the correlation between structural change and the outcome variables, without any implication for their causal relationships. Table 1summarises the relationship between structural change (between) and employment growth. Looking at columns (1) and (4), the between component of the LWF structural decomposition shows a negative correlation with employment growth at the country level. This indicates that when countries experience a labour flow from lowto high-complexity industries, they also experience slower rates of employment growth. This finding seems to corroborate the hypothesis that high-complexity industries contribute less to aggregate employment, due to their higher levels of knowledge intensity, and lower labour-intensity, as they require fewer (and likely more skilled) workers. It is worth mentioning that neither the within component of the variation in LWF (column 2), nor the sum of both components (column 3) do not show any significant correlation with employment growth, highlighting the importance of structural change in explaining outcomes related to the creation of employment. The coefficient on the between component remains significant also when the within component is also included in the regression. We now move to the relationship between structural change and wage inequality. First, we look at the link between structural change and the wage gap between top (9th decile) and bottom (1st decile) wages. The wage gaps appears to contract as a result of labour reallocation towards knowledge-intensive industries, as indicated by columns (1) and (4) in Table 2. However, it is not clear whether this result is driven by lower wages in the top of the distribution, or higher wages at the bottom. In order to further explore this, Tables 3and 4show respectively the results for the wage gap between 5th/1st and 9th/5th deciles. Interestingly, the wage gap between median and bottom salaries narrows (columns 1 and 4 in Table 3) as a result of structural change, while the ratio between salaries at the 9th and 1st decile doesn’t show any association with structural change (Table 4, columns 1 and 4). We conclude that the reduction in the wage gap between top and bottom wages results from bottom wages growing at a faster rate than top wages. In order to rationalise this result, it is worth going back to the specialisation matrix described in figure 1. As countries move to more complex industries, they drop the least complex ones, as indicated by the "emptier" topleft corner of the matrix. Assuming a positive correlation between complexity and average 17 wages, it can be inferred that the reduction in wage inequality is driven by the fact htat the low-paid jobs in the least complex sectors disappear, making the 1st decile of the salary distribution go up. Although this conjecture is yet to be tested at the present stage of the research, it represents one of the next steps in the analysis conducted in this paper. Finally, Table 5displays the results of regressing labour share (the ratio between aggregate wages and value added in the economy) on the between component of LWF. Columns (1) and (4) show a positive relationship between the two, indicating that as workers move to more complex industries, that contributes positively to the increase of the labour component of production, which appears to become more intensive. In order to understand this results, we should rely once more on the assumption that high-complexity industries pay, on average, higher wages. Bearing in mind the diversification patterns displayed by figure 1, coupled with the results on employment growth in Table 1. If employment growth slows down as a result of structural change towards knowledge-intensive industries, the increase in the labour share of the economy is due to be driven by higher average wages rather than higher participation of the population to the workforce. This hypothesis will be tested in the future steps that we will take to carry on with the analysis. As a final note on the preliminary results presented in this section, it is important to stress that the structural change component is the only one to be significant across the regression specifications summarised by equation 8. As a matter of fact, neither the within component nor the change in labour-weighted fitness explain any of the outcomes observed, indicating that framing structural change as the movement of labour across industries of different complexity – that is, knowledge intensity – can contribute significantly to explain the current trends in employment growth, wage inequality and functional distribution of income in European countries. 18 Table 1: Employment growth regressions Dependent Variable: Yearly employment rate growth (%) FE OLS Model: (1) (2) (3) (4) Variables Between -7,313.359∗∗ -7,527.217∗∗ (3,229.347) (3,248.861) Within 1,266.110 1,412.623 (1,121.540) (1,123.097) ∆C279.081 (993.907) Population (log) -19.942 -19.817 -19.927 -19.785 (12.878) (13.098) (13.057) (12.968) GDPpc (log) -6.179∗-5.824 -6.046 -5.859 (3.563) (3.572) (3.556) (3.570) R&D (%GDP) 0.169∗∗ 0.175∗∗∗ 0.174∗∗ 0.170∗∗ (0.064) (0.063) (0.063) (0.063) Exports (%GDP) 0.142 0.232 0.177 0.222 (0.761) (0.736) (0.756) (0.718) Fixed-effects Country Yes Yes Yes Yes Year Yes Yes Yes Yes Fit statistics Observations 224 224 224 224 R20.526 0.522 0.520 0.528 Within R20.233 0.225 0.223 0.236 Clustered (Country-Year) standard-errors in parentheses. Signif. Codes: ***: 0.01, **: 0.05, *: 0.1 19 Table 2: Wage inequality regressions: ratio 9th/1st dec. Dependent Variable: Ratio 9th/1st dec. FE OLS Model: (1) (2) (3) (4) Variables Between -8,201.860∗∗∗ -8,142.939∗∗∗ (2,884.048) (2,893.616) Within -547.693 -389.195 (671.490) (744.401) ∆C-1,372.345 (956.596) GDPpc (log) -3.781 -3.831 -4.031 -3.870 (2.770) (2.863) (2.861) (2.778) Population (log) -5.596 -5.675 -5.763 -5.640 (7.401) (7.552) (7.514) (7.424) R&D (%GDP) -0.013 -0.008 -0.010 -0.014 (0.024) (0.025) (0.024) (0.024) Exports (%GDP) -1.080 -1.091 -1.141 -1.102 (0.995) (1.000) (1.000) (1.002) Fixed-effects Country Yes Yes Yes Yes Year Yes Yes Yes Yes Fit statistics Observations 224 224 224 224 R20.906 0.903 0.904 0.906 Within R20.073 0.044 0.051 0.074 Clustered (Country-Year) standard-errors in parentheses. Signif. Codes: ***: 0.01, **: 0.05, *: 0.1 20 results, we are planning to devise an identification strategy based on a shift-share instrument, which appears at the moment the most suitable identification strategy given the nature of our (potentially) endogenous explanatory variable of interest. Our results bear important consequences for policy making, as they highlight a trade-off between industrial upgrading and labour creation, while also dissecting the nature of the reduction of wage and functional inequality resulting from the movement of labour towards complex industries. Based on these results, policy-makers will have to device policy tools that steer structural change towards those industries that, at the same time, conflate job creation with higher technological sophistication, in order to sustain a process of inclusive economic growth. Such industries can be identified using the EFC framework with a focus on labour, rather than exports alone. Coupling the geography of productive capabilities with labour and inequality, enables us to advance with respect to the extant literature not only in providing a policy-relevant and data-driven toolkit to embrace socio-economic complexity, but also in offering a novel analytical and empirical understanding of a pathway not only geared towards boosting growth, but also to steering inclusive and sustainable development. 27 References Acemoglu, D. & Restrepo, P. (2022), ‘Tasks, Automation, and the Rise in U.S. Wage Inequality’, Econometrica 90(5), 1973–2016. URL: https://www.econometricsociety.org/doi/10.3982/ECTA19815 Adam, A., Garas, A., Katsaiti, M.-S. & Lapatinas, A. (2023), ‘Economic complexity and jobs: an empirical analysis’, Economics of Innovation and New Technology 32(1), 25–52. URL: https://doi.org/10.1080/10438599.2020.1859751 Antonelli, C. & Tubiana, M. (2020), ‘Income inequality in the knowledge economy’, Structural Change and Economic Dynamics 55, 153–164. URL: https://www.sciencedirect.com/science/article/pii/S0954349X2030374X Autor, D. (2022), The labor market impacts of technological change: From unbridled enthusiasm to qualified optimism to vast uncertainty, Working Paper 30074, National Bureau of Economic Research. URL: http://www.nber.org/papers/w30074 Autor, D. H., Levy, F. & Murnane, R. J. (2003), ‘The skill content of recent technological change: An empirical exploration’, The Quarterly Journal of Economics 118(4), 1279–1333. Balassa, B. (1965), ‘Trade Liberalisation and Revealed Comparative Advantage’, The Manchester School 33(2), 99–123. Balland, P.-A., Broekel, T., Diodato, D., Giuliani, E., Hausmann, R., O’Clery, N. & Rigby, D. (2022), ‘The new paradigm of economic complexity’, Research Policy 51(3), 104450. URL: https://doi.org/10.1016/j.respol.2021.104450 Balland, P.-A., Jara-Figueroa, C., Petralia, S. G., Steijn, M. P. A., Rigby, D. L. & Hidalgo, C. A. (2020), ‘Complex economic activities concentratein large cities’, Nature Human Behaviour 4, 248–254. Bruno, M., Mazzilli, D., Patelli, A., Squartini, T. & Saracco, F. (2023), ‘Inferring comparative advantage via entropy maximization’. _eprint: 2304.12245. Brynjolfsson, E. & McAfee, A. (2011), Race against the machine: How the digital revolution is accelerating innovation, driving productivity, and irreversibly transforming employment and the economy, Brynjolfsson and McAfee. 28 Bustos, S., Gomez, C., Hausmann, R. & Hidalgo, C. A. (2012), ‘The Dynamics of Nestedness Predicts the Evolution of Industrial Ecosystems’, PLoS ONE 7(11), 1–8. Caldarola, B., Mazzilli, D., Napolitano, L., Patelli, A. & Sbardella, A. (2023), ‘Economic complexity and the sustainability transition: A review of data, methods, and literature’. Castañeda, G., Pietronero, L., Romero-Padilla, J. & Zaccaria, A. (2022), ‘The complex dynamic of growth: Fitness and the different patterns of economic activity in the medium and long terms’, Structural Change and Economic Dynamics 62, 231–246. URL: https://www.sciencedirect.com/science/article/pii/S0954349X22000601 Cimoli, M. & Dosi, G. (1995), ‘Technological paradigms, patterns of learning and development: An introductory roadmap’, Journal of Evolutionary Economics 5(3), 243–268. URL: https://doi.org/10.1007/BF01198306 Cristelli, M., Gabrielli, A., Tacchella, A., Caldarelli, G. & Pietronero, L. (2013), ‘Measuring the intangibles: A metrics for the economic complexity of countries and products’, PLoS ONE 8(8). Dauth, W., Findeisen, S., Suedekum, J. & Woessner, N. (2021), ‘The Adjustment of Labor Markets to Robots’, Journal of the European Economic Association 19(6), 3104–3153. URL: https://doi.org/10.1093/jeea/jvab012 de Vries, G., Timmer, M. & de Vries, K. (2015), ‘Structural Transformation in Africa: Static Gains, Dynamic Losses’, The Journal of Development Studies 51(6), 674–688. Fabricant, S. (1942), Employment in manufacturing, 1899-1939: An analysis of its relation to the volume of production, National Bureau of Economic Research, Inc. Freire, C. (2021), Economic Complexity Perspectives on Structural Change, in ‘New Perspectives on Structural Change’, Oxford University Press, pp. 188–214. Frey, C. B. & Osborne, M. A. (2017), ‘The future of employment: How susceptible are jobs to computerisation?’, Technological Forecasting and Social Change 114, 254–280. Garlaschelli, D. & Loffredo, M. I. (2008), ‘Maximum likelihood: Extracting unbiased information from complex networks’, Physical Review E - Statistical, Nonlinear, and Soft Matter Physics 78, 1–5. Goos, M., Manning, A. & Salomons, A. (2009), ‘Job polarization in europe’, American Economic Review 99(2), 58–63. URL: https://www.aeaweb.org/articles?id=10.1257/aer.99.2.58 29 Goos, M., Manning, A. & Salomons, A. (2014), ‘Explaining job polarization: Routine-biased technological change and offshoring’, American Economic Review 104(8), 2509–26. URL: https://www.aeaweb.org/articles?id=10.1257/aer.104.8.2509 Graetz, G. & Michaels, G. (2018), ‘Robots at Work’, 100(5), 753–768. URL: https://doi.org/10.1162/rest_a_00754 Hane-Weijman, E., Eriksson, R. H. & Henning, M. (2018), ‘Returning to work: regional determinants of re-employment after major redundancies’, Regional Studies 52, 768 – 780. URL: https://api.semanticscholar.org/CorpusID:53615559 Hartmann, D., Guevara, M. R., Jara-Figueroa, C., Aristarán, M. & Hidalgo, C. A. (2017), ‘Linking economic complexity, institutions, and income inequality’, World Development 93, 75–93. URL: https://www.sciencedirect.com/science/article/pii/S0305750X15309876 Hartmann, D. & Pinheiro, F. L. (2022), ‘Economic complexity and inequality at the national and regional level’. Hausmann, R. & Hidalgo, C. (2009), ‘The building blocks of economic complexity’, Proceedings of the National Academy of Sciences of the United States of America 106(26), 10570–10575. Hausmann, R. & Klinger, B. (2006), Structural transformation and patterns of comparative advantage in the product space, CID Working Paper RWP06-041, Centre for International Development, Harvard University. Herrendorf, B., Rogerson, R. & Valentinyi, A. (2014), Growth and structural transformation, in ‘Handbook of economic growth’, Vol. 2, Elsevier, pp. 855–941. Herrendorf, B., Rogerson, R. & Ákos Valentinyi (2022), New evidence on sectoral labour productivity: implications for industrialization and development, Working Paper 29834, NBER. Hirschman, A. O. (1958), The Strategy of Economic Development, Yale University Press, New Haven. Krantz, R., Gemmetto, V. & Garlaschelli, D. (2018), ‘Maximum-entropy tools for economic fitness and complexity’, Entropy 20(10). URL: https://www.mdpi.com/1099-4300/20/10/743 30 Kuznets, S. (1955), ‘Economic growth and income inequality’, The American Economic Review 45(1), 1–28. URL: http://www.jstor.org/stable/1811581 Lewis, A. (1954), ‘Economic Development with Unlimited Supplies of Labour’, The Manchester School 22(8), 139–191. Lo Turco, A. & Maggioni, D. (2022), ‘The knowledge and skill content of production complexity’, Research Policy 51(8), 104059. Special Issue on Economic Complexity. URL: https://www.sciencedirect.com/science/article/pii/S0048733320301372 Mazzilli, D., Mariani, M. S., Morone, F. & Patelli, A. (2024), ‘Equivalence between the fitness-complexity and the sinkhorn-knopp algorithms’, Journal of Physics: Complexity . McMillan, M., Rodrik, D. & Sepúlveda, C., eds (2017), Structural change, fundamentals, and growth: A framework and case studies, International Food Policy Research Institute (IFPRI), Washington, D.C. McMillan, M., Rodrik, D. & Íñigo Verduzco-Gallo (2014), ‘Globalization, Structural Change, and Productivity Growth, with an Update on Africa’, World Development 63, 11–32. McNerney, J., Li, Y., Gomez-Lievano, A. & Neffke, F. (2023), ‘Bridging the short-term and long-term dynamics of economic structural change’. Patelli, A., Napolitano, L., Cimini, G., Pugliese, E. & Gabrielli, A. (2023), ‘Capability accumulation patterns across economic, innovation, and knowledge-production activities’, Scientific Reports 13(1), 12988. URL: https://doi.org/10.1038/s41598-023-29979-x Pinheiro, F. L., Balland, P.-A., Boschma, R. & Hartmann, D. (2022), ‘The dark side of the geography of innovation: relatedness, complexity and regional inequality in europe’, Regional Studies Forthcoming special issue: The dark side of innovation and its geography(0), 1–16. URL: https://doi.org/10.1080/00343404.2022.2106362 Sbardella, A., Pugliese, E. & Pietronero, L. (2017), ‘Economic development and wage inequality: A complex system analysis’, PLOS ONE 12(9), 1–26. Schumpeter, J. A. (1934), The Theory of Economic Development: An Inquiry into Profits, Capital, Credit, Interest, and the Business Cycle, Harvard University Press, Cambridge, MA. 31 Syrquin, M. (1988), Patterns of structural change, in H. Chenery & T. Srinivasan, eds, ‘Handbook of Development Economics’, 1 edn, Vol. 1, Elsevier, chapter 07, pp. 203–273. URL: https://EconPapers.repec.org/RePEc:eee:devchp:1-07 Tacchella, A., Cristelli, M., Caldarelli, G., Gabrielli, A. & Pietronero, L. (2012), ‘A New Metrics for Countries Fitness and Products Complexity’, Scientific Reports 2(723), 1–7. Tacchella, A., Mazzilli, D. & Pietronero, L. (2018), ‘A dynamical systems approach to gross domestic product forecasting’, Nature Physics 14(8), 861–865. Teece, D. J., Rumelt, R., Dosi, G. & Winter, S. (1994), ‘Understanding corporate coherence: Theory and evidence’, Journal of Economic Behavior & Organization 23(1), 1–30. Vollrath, T. L. (1991), ‘A theoretical evaluation of alternative trade intensity measures of revealed comparative advantage’, Weltwirtschaftliches Archiv 127(2), 265–280. URL: http://www.jstor.org/stable/40439943 32 A Data reconstruction The Eurostat’s Structural Business Survey data used to compute the employment based EFC measures used in this paper are one of the most comprehensive and detailed data sources of employment in Eurpean industries, classified at a fine level of disaggregation (NACE 4 digit). However, the dataset presents a large number of missing values, which hamper the construction of a specialisation matrix as explained in Section 4.1. In fact, the BiWCM and EFC algorithm require both a matrix with complete information. In order to address this shortcoming in the SBS data, we have pursued, compared and validated seven different interpolation strategies, in order to retrieve a good approximation of the missing data. The first step has been to attempt the reconstruction of the SBS data following 7 alternative strategies, applied to the country-industry time series having a maximum 7 missing data points5: 1. Internal missing values (those after the first available data point, and before the last) are interpolated using linear interpolation, while external missing values (before and after, respectively, the first and last data point available) have been extrapolated backwards taking the first available value as constant; 2. Internal missing values are interpolated using linear interpolation, while external missing values are extrapolated using the closest growth rate available, obtained by after interpolating internal missing values; 3. Internal missing values are interpolated using linear interpolation, while external missing values are extrapolated using average growth rate applied to the fist and last data points available after interpolating. This strategy produced negative values, which were constrained to 0; 4. Internal missing values are interpolated using linear interpolation, while external missing values are extrapolated using the first available growth rate by filling up/downwards available growth rates; 5. Internal missing values are interpolated using linear interpolation, while external missing values are extrapolated using moving average of growth rates, constructed using a rolling window of 3 years. The first two years are always ‘NA‘; second year is filled with moving average for t=2, and first year is just the growth with respect to the previous year; 5In the case of fully missing series, the country-industry couple has been excluded. In case of 8 out of 9 missing data points, the value has been assumed to be constant across the time series, extrapolating at constant values from the only available data point in the series. 33 6. Interpolate and extrapolate with linear fit computed using the available data points in each country-industry series; 7. Internal missing values are interpolated using linear interpolation, while external missing values are extrapolated using linear fit. Each strategy is compared and validated using information from the complete countryindustry time series. After filtering only the complete series from the dataset, we have created five different validation subsamples by randomising missing values in the complete dataset, imposing the following conditions: •Reflect NA frequency in original data (14% of data missing) •Reflect frequency in original data ś 5%, •Reflect frequency in original data + 10% •Impose 50% of missing values. In order to pick the data reconstruction strategy that best approximates the observed data, we compare the Mean Absolute Error of each strategy across the 5 validation samples. As shown by table 7, the strategy that minimises the prediction error across the different validation subsamples is the first – i.e. the one that interpolates internal values linearly, and extrapolates external values taking the first and last available data point as constant, both backward and forward. Table 7: Mean Absolute Errors Reconstruction strategy Subsample 1 2 3 4 5 6 7 NA 9% 174.96 189.05 179.47 192.23 189.44 222.55 197.73 NA 14% 214.95 227.60 215.10 230.20 227.45 273.22 240.86 NA 19% 254.61 294.14 267.06 297.80 292.47 323.64 286.27 NA 24% 311.79 329.90 315.13 331.72 325.46 395.41 355.26 NA 50% 692.86 788.71 749.16 796.20 784.38 819.18 777.39 34 The UNU-MERIT WORKING Paper Series 2024-01 The green transformation as a new direction for techno-economic development by Rasmus Lema and Carlota Perez 2024-02 Evidence on an endogenous growth model with public R&D by Thomas H.W. Ziesemer 2024-03 Higher Educational Systems and E-resilience by Mindel van de Laar, Julieta Marotta and Lisa de Graaf 2024-04 Estimating the wage premia of refugee immigrants: Lessons from Sweden by Christopher F. Baum, Hans Lööf, Andreas Stephan and Klaus F. Zimmermann 2024-05 Green window of opportunity through global value chains of critical minerals: An empirical test for refining copper and lithium industries By Jorge ValverdeCarbonell and Alejandro Micco A. 2024-06 Migration and Consumption by Roberta Misuraca and Klaus F. Zimmermann 2024-07 The power of narratives: How framing refugee migration impacts attitudes and political support towards refugees by Tobias Hillenbrand, Bruno Martorano, Laura Metzger and Melissa Siegel 2024-08 Strategies of search and patenting under different IPR regimes by Robin Cowan, Nicolas Jonard and Ruth Samson 2024-09 From drains to bridges: The role of internationally mobile PhD students in linking non-mobile with foreign scientists by Rodrigo Ito, Diego Chavarro, Tommaso Ciarli, Robin Cowan and Fabiana Visentin 2024-10 Expanding horizons: The impact of post-doctoral mobility grants on research diversification by Yağmur Yıldız, Diego Chavarro, Fabiana Visentin and Tommaso Ciarli 2024-11 Scorching heat and shrinking horizons: The impact of rising temperatures on marriages and migration in rural India by Manisha Mukherjee, Bruno Martorano and Melissa Siegel 2024-12 Labour market integration for refugees: A social justice framework by Tamara Kool and Zina Nimeh 2024-13 The evolution of development with trade in global value chains by Önder Nomaler and Bart Verspagen 2024-14 The impacts of the Microfinance Multiplied approach on seasonal food insecurity: Evidence from a high-frequency panel survey in Uganda by Ricardo Morel, Franziska Gassmann, Bruno Martorano, Nyasha Tirivayi and John Kamau 2024-15 Building CoPS Capability for Catching Up During Transitions: The Case of ACWA Power in Saudi Arabia by Kaori Mita and Michiko Iizuka 2024-16 The KSTE+I approach and the AI technologies: evidence from the European regions by Francesco D’Allesandro, Enrico Santarelli and Marco Vivarelli 2024-17 Theory and measurement in SFC models: the role of the financial sector by Huub Meijers and Joan Muysken 2024-18 Is artificial intelligence generating a new paradigm? Evidence from the emerging phase by Giacomo Damioli, Vincent Van Roy, Daniel Vertesy and Marco Vivarelli 2024-19 Industrializing Africa - a science, technology, and innovation policy perspective by Bertha Vallejo 2024-20 Innovation in the wind energy sector by Rasmus Lema, Davide Bonaglia and Ulrich Elmer Hansen 2024-21 The Investment Landscape in the Lithium Triangle: Implications for the EU’s Critical Raw Materials Act Agenda by Beatriz Calzada Olvera and Melissa VergaraFernández 2024-22 3G internet diffusion and secondary education attainment: Evidence of opportunity cost in Viet Nam by Trang Thi Pham and Bernardo Caldarola 2024-23 Influence of the Sectoral Composition of Production on Occupational and Sectoral Wage Structures: Evidence from Multiple Economies by Nelson Marconi, Danilo Spinola, Tiago Porto and Eliane Araújo 2024-24 Transition from Information and Communications Technology to Educational Technologies and Game-Based Learning in the Republic of Moldova: The Case of The Mobility Era Game by Irina Burlacu de Paoli 2024-25 When one teaches, two learn: The bidirectional learning process supervisor-PhD student by Pierre Boutros, Michele Pezzoni, Sotaro Shibayama and Fabiana Visentin 2024-26 Cloud technologies, firm growth and industry concentration: Evidence from France by Bernardo Caldarola and Luca Fontanelli 2024-27 The Innovation and Structural Transformation Database: A Guide by Neil FosterMcGregor, Önder Nomaler & Bart Verspagen 2024-28 The Political Economy of Targeting: An Empirical Assessment of Poland’s Family 500+ Programme by Stefanie Roost and Franziska Gassmann 2024-29 Melkert Jobs, a unique experiment by Kees Mosselman and Joan Muysken 2024-30 A post-mortem of interest rate policy: beware of financial fragility by Huub Meijers and Joan Muysken 2024-31 Rankings and Job Market Dynamics: A Model of Academic Science by Robin Cowan and Nicolas Jonard 2024-32 Twin transition trade based on multi-dimensional economic complexity by María de las Mercedes Menéndez 2024-33 Structural Change, Employment, and Inequality in Europe: an Economic Complexity Approach by Bernardo Caldarola, Dario Mazzilli, Aurelio Patelli and Angelica Sbardella