Artificial Intelligence and Its Impact on Unemployment: A Comparative Analysis of Old and New EU Member States

Artificial Intelligence and Its Impact on Unemployment: A Comparative Analysis of Old and New EU Member States | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Artificial Intelligence and Its Impact on Unemployment: A Comparative Analysis of Old and New EU Member States Jordan Kjosevski This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6641180/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This study examines the impact of venture capital (VC) investments in artificial intelligence (AI) on unemployment rates across 27 EU member states, distinguishing between old and new EU countries. Utilizing annual data from 2012 to 2023, we explore whether AI investments significantly influence unemployment and how these effects vary between advanced economies and those still developing their digital infrastructure. Employing the two-step system Generalized Method of Moments (GMM), we effectively address endogeneity and the dynamic nature of unemployment, making this method well-suited for our panel dataset covering 27 countries over 12 years. Our findings reveal that AI investments correlate with higher unemployment in old EU countries while positively impacting job creation in new EU member states. Based on these results, we recommend targeted policies to enhance AI adoption, improve digital infrastructure, and promote workforce training, particularly in new member states, to optimize the benefits of AI investments and mitigate potential job displacement. JEL Codes: O1, O3 1. Introduction Artificial Intelligence (AI) has emerged as one of the most transformative technologies of the 21st century, influencing various aspects of human life, particularly in the workplace. As AI continues to evolve and integrate into different sectors, there is growing concern about its impact on employment and unemployment rates. While AI has the potential to enhance productivity and create new job opportunities, it also poses significant risks to existing jobs, particularly those that involve routine and repetitive tasks. This study explores the multifaceted impact of AI on unemployment, analyzing theoretical perspectives and empirical evidence. It also considers the implications of AI-driven changes in the labor market and discusses potential policy responses to mitigate the adverse effects on employment. AI refers to the simulation of human intelligence in machines programmed to think and learn. These machines can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation (Russell and Norvig, 2016). AI can be broadly categorized into two types: narrow AI, which is designed for specific tasks, and general AI, which possesses the ability to perform any intellectual task that a human can do (Goodfellow, Bengio, and Courville, 2016). The scope of AI extends across various sectors, including manufacturing, healthcare, finance, and customer service. In manufacturing, AI-powered robots and automation systems are used to enhance production efficiency (Brynjolfsson and McAfee, 2014 ). In healthcare, AI is utilized for diagnostic purposes, personalized treatment plans, and robotic surgeries (Topol, 2019). The financial sector leverages AI for algorithmic trading, fraud detection, and customer service chatbots (Baldwin, 2019 ). These applications illustrate the widespread adoption of AI, which is poised to have profound effects on employment. The impact of AI on employment is a subject of considerable debate among economists and policymakers. One of the primary concerns is the potential for AI to displace jobs through automation. The concept of "technological unemployment" was first introduced by Keynes (1933), who suggested that technological advancements could lead to a net loss of jobs. AI, with its ability to perform complex tasks that were once the domain of humans, could exacerbate this phenomenon, particularly in sectors where routine and manual tasks are prevalent (Acemoglu and Restrepo, 2018 ). However, AI also has the potential to create new jobs, particularly in emerging fields such as AI development, data science, and machine learning (Arntz, Gregory, and Zierahn, 2016). The introduction of AI could lead to job creation through increased demand for new products and services, as well as the need for workers to develop, maintain, and supervise AI systems. This dual effect of AI on employment—job displacement and job creation—reflects the broader theory of creative destruction, where technological progress leads to the obsolescence of certain jobs while giving rise to new ones (Schumpeter, 1942). The concept of skill-biased technological change (SBTC) is also relevant in the context of AI. SBTC refers to technological advancements that favor workers with higher skill levels, leading to increased demand for skilled labor and reduced demand for unskilled labor (Levy and Murnane, 2003 ). AI, with its reliance on data and complex algorithms, is likely to exacerbate this trend, widening the gap between skilled and unskilled workers and contributing to rising income inequality (Goos, Manning, and Salomons, 2014). Given the diverse perspectives on the effects of AI, a major concern revolves around its impact on unemployment levels. The World Economic Forum's "Future of Jobs" report (2018) notes that AI is increasingly being adopted across various economic sectors, potentially altering employment opportunities and skill development. Similarly, the European Commission's "Artificial Intelligence for Europe" statement (2018) underscores AI's profound influence on multiple economic sectors, advocating for its responsible use in line with ethical standards. Despite extensive research on AI's effects on employment and income distribution, there remains a gap in understanding its specific implications for the European economy. Much of the existing literature is either global (Mutascu, 2021; Guliyev, 2023) or country-specific (Acemoglu and Restrepo, 2020 ; Carbonero, Ernst, and Weber, 2020 ), leaving a need for focused investigations into AI's impact on the European Union (except Bowles 2017 ; Chiacchio et al. 2018 ). This paper aims to fill this research gap by examining the relationship between AI and unemployment in the 27 EU countries, using worldwide venture capital (VC) investments in AI and AI-related data start-ups as key indicators. This study distinguishes itself by providing a comprehensive analysis of the 27 EU countries and by being among the first to apply a dynamic panel data estimation method to assess AI's impact on unemployment. Furthermore, one of the key contributions of this study is the use of worldwide venture capital (VC) investments in AI and data start-ups as a proxy for AI adoption and innovation within the EU. This approach provides a novel way of measuring AI's influence on the labor market, capturing the intensity and scale of AI-related activities that are likely to affect employment. By focusing on venture capital investments, this study not only considers the current state of AI adoption but also the future potential for AI-driven economic transformation in the region. The subsequent sections of this study are structured as follows: The second section provides a review of the relevant academic literature. In the third section, we delve into detail regarding our empirical approach, including the data sources and variables used in our analysis. Next, the methodology and econometric framework are outlined in the subsequent section. Moving forward to the fifth section, we present the empirical findings. Finally, the sixth section concludes the study and discusses policy implications. 2. Literature review Artificial intelligence (AI) has profound implications for the socioeconomic landscape, influencing various areas such as unemployment, inequality, and productivity. Over the past few decades, the literature on the impact of AI on unemployment has expanded considerably (Ernst et al., 2019; Martens and Tolan, 2018), though many studies treat AI as part of a broader automation process (Wang and Siau, 2019). Empirical evidence on the impact of AI on unemployment is mixed. Broadly speaking, this literature can be divided into two main strands: the "replacement effect" and the "displacement effect.", with outcomes varying across sectors and regions. The first strand supports the "replacement effect," which includes both theoretical and empirical studies. This perspective argues that AI negatively impacts the labor market by replacing jobs and increasing unemployment. Leontief (1983) was one of the first to theorize about AI's potential to destroy jobs, predicting that almost all jobs could be replaced by AI in the coming decades, leading to higher unemployment unless offset by government redistributive policies. Bessen (2018, 2020) introduces the concept of demand elasticity, arguing that automation only reduces employment if demand is inelastic; otherwise, there could be positive effects on job creation. Atkinson (2018) describes AI's destructive role in employment, noting that as automation reduces jobs in one sector, such as agriculture, people move to new sectors, but eventually, there will be no new sectors left to absorb the displaced workers. Acemoglu and Restrepo ( 2018 ) expand on this by modeling the race between humans and machines, showing that automation reduces employment and wages when capital is fixed and technology is exogenous. They later (Acemoglu and Restrepo, 2019 ) argue that increased AI use lowers labor demand, national income, and productivity growth while exacerbating inequality. Empirical studies also support this strand, with Frey and Osborne ( 2017 ) estimating that automation could lead to job losses of 47 percent in the U.S. over the next two decades. Arntz et al. (2017) provide a more conservative estimate of a 9 percent job loss in 21 OECD countries. Acemoglu and Restrepo (2017) find that increased industrial robot usage between 1990 and 2007 negatively impacted U.S. local labor markets, reducing employment and wages. Bowles ( 2017 ) and Chiacchio et al. ( 2018 ) find similar negative effects in the EU. The second strand of literature promotes the "displacement effect," arguing that AI positively impacts the labor market by creating new jobs and reducing unemployment. Albus (1983) was one of the first to argue that technological progress, including AI, increases productivity and generates new markets, thereby creating more jobs and reducing unemployment. Boden (1987) echoes this sentiment, suggesting that AI will eventually lead to more jobs than before, despite transitional challenges. Dauth et al. (2017) find no net job losses in Germany due to automation. Frey and Osborne ( 2017 ) also introduce the concept of sectors with varying risks of automation, noting that unemployment increases only in sectors most exposed to automation. Su (2018) argues that AI reduces unemployment by changing job structures, with policymakers needing to stimulate job creation in new sectors to offset job losses. Gries and Naudé (2018) incorporate AI into an economic growth model and find that AI does not immediately increase unemployment. Acemoglu and Restrepo ( 2018 ) claim that AI improves productivity and creates new jobs through increased capital accumulation. Cautionary perspectives also exist within this strand. Trehan (2003) empirically shows that AI's positive impact on unemployment is only valid under specific conditions and may have a retrograde effect over time. Autor (2015) argues that automation changes the nature of jobs rather than eliminating them, with new jobs replacing old ones. Berriman and Hawksworth (2017) suggest that the net impact of automation on jobs is neutral, as job losses are compensated by newly created jobs. Gordon (2018) finds that automation slowly replaces jobs, mainly in a minority of sectors, while Carbonero et al. (2018) show that robots have a small negative impact on jobs in developing countries. Recent studies such as Mutascu (2021) and Guliyev (2023) contribute to this field by examining the impact of AI on unemployment in high-tech, developed countries. Mutascu (2021) uses a nonlinear approach with panel threshold and GMM-system estimations to analyze data from 23 countries over the period 1998–2016. The results indicate that AI reduces unemployment at low levels of inflation but has a neutral effect otherwise, with no "switch effect" between the "displacement" and "replacement" effects. Guliyev (2023) extends this analysis by examining the relationship between AI-related Google Trends data and unemployment rates in 24 high-tech countries from 2005 to 2021, using dynamic panel data and GMM-system estimation. The study finds that AI decreases unemployment, validating the "displacement effect." 3. Variables and Data Our study dataset consists of a sample of 27 EU member states and uses annual data from 2012 to 2023. 1 The panel is unbalanced given that the study features many “new” EU member states, and not all of the indicators were available for each country over the entire period (Ostrihoň, Siranova, and Workie Tiruneh,2023). On the other hand, there also existed differences among these countries that made them a remarkably heterogeneous group. These differences were mostly reflected in the high disparities observed concerning the level of public debt, GDP growth, unemployment, and so on. The core determinants selected in our model have been previously used in academic literature (Mutascu, 2021; Guliyev, 2023 among others). The internationally comparable and reliable data were obtained from the OECD.AI International Labor Organization (ILOSTAT) and the World Bank’s World Development Indicators. The dependent variable, expressed as a percentage of the total labor force, represents the unemployment rate, which reflects the portion of the labor force that is unemployed, available for work, and actively seeking employment. This variable was used in studies by Mutascu (2021) and Guliyev (2023) to analyze labor market dynamics and the impact of various economic factors on unemployment. Additionally, the unemployment rate serves as a key indicator for policymakers, helping to guide decisions related to job creation and economic growth strategies. The consistent use of this variable in empirical research underscores its importance in understanding and addressing unemployment issues. In this study, we utilize Preqin's data on global venture capital (VC) investments in AI and data startups, as reported by OECD.AI. To provide a more accurate and meaningful comparison across countries, we normalize these investments by calculating the number of investments per million people. Adjusting for population size allows for a standardized metric that better reflects the intensity of AI investment activity across different nations. This approach helps account for variations in market scale and investment capacity, offering a clearer picture of where AI investments are most concentrated and identifying areas of potential growth. Normalizing investments per capita is particularly valuable for enabling researchers and policymakers to make fair comparisons between countries with differing population sizes. It highlights investment concentration and reveals gaps or opportunities in AI development. Additionally, this metric sheds light on the relationship between investment levels and a country’s technological advancement per capita, offering insights into the nation’s capacity for innovation. By focusing on investments per million people, stakeholders can make more informed decisions regarding resource allocation, strategy planning, and policy development, fostering sustainable growth and competitiveness in the AI sector. A set of control variables is considered to isolate the effects of interest variables. They are inspired by both the theoretical model and the literature: economic growth, inflation rate, government spending, school enrollment, trade, and population growth. The first control variable is GDP growth. Economic growth and the unemployment rate are the key indicators that are simultaneously monitored by both policymakers and the public, as they create a clear picture of the economic development of a country. Moreover, the linkage between the unemployment rate and economic growth as a relevant macroeconomic issue covers a wide area of theoretical and empirical research. It is a widely accepted view in economics that a higher growth rate of the GDP of an economy increases employment and reduces unemployment. This theoretical proposition relating output and unemployment is known as “Okun’s Law.” This relationship is among the most prominent in macroeconomic theory and is held for several countries and regions, mainly in developed countries (Lee, 2000; Farsio and Quade, 2003). Inflation can influence unemployment by affecting real wages and purchasing power. According to the Phillips curve, there is a short-term trade-off between inflation and unemployment, where higher inflation can lower unemployment by reducing real wages and making labor more affordable for employers (Phillips, 1958). However, over the long term, this relationship tends to diminish, as empirical studies indicate that the impact of inflation on unemployment depends on the specific country and its economic conditions (Ball, Mankiw, and Romer, 1988 ). Inflation, measured through the consumer price index, reflects the annual rate of price fluctuations in an economy. The theoretical literature often finds a negative correlation between inflation and unemployment (Phillips, 1958). The Phillips curve suggests a stable relationship between inflation and unemployment, arguing that rising inflation can increase employment opportunities and reduce unemployment (Wulandari et al., 2019). Government spending, expressed as a percentage of GDP includes cash payments for operating activities related to providing goods and services. Fiscal actions can have mixed effects on unemployment: while increased spending may stimulate job creation, it can also raise labor supply and unemployment, even if output rises. Conversely, cuts in government spending might lead to higher private-sector output and productivity, potentially causing GDP growth alongside rising unemployment. This ambiguity is reflected in recent research. Some studies, such as Monacelli et al. (2010), Auerbach and Gorodnichenko ( 2012 ), Ramey (2012), and Holden and Sparrman (2018), find that higher government spending reduces unemployment. However, other studies suggest that large-scale government spending may increase unemployment, with Abrams (1999), Christopoulos et al. (2005), Feldmann (2006), and Brückner and Pappa (2012) indicating that increased government purchases can lead to higher unemployment. Education, particularly secondary school enrollment, is a critical determinant of labor market outcomes. Higher enrollment rates can lead to a more skilled workforce, which may reduce unemployment by increasing employability and adaptability to technological changes brought about by AI. The human capital theory posits that investment in education improves individual productivity, thereby lowering unemployment (Becker, 1993 ). Empirical evidence suggests that regions with higher secondary school enrollment rates tend to experience lower unemployment rates due to the increased availability of skilled labor (Hanushek and Woessmann, 2008 ). Population growth affects labor supply, and its impact on unemployment depends on the economy's ability to absorb new entrants into the labor market. High population growth can lead to higher unemployment if job creation does not keep pace with the growing labor force. Malthusian theory suggests that rapid population growth can outstrip economic growth, leading to higher unemployment (Malthus, 1798). However, empirical studies indicate that population growth can have varied effects on unemployment, depending on the structure of the economy and the rate of economic expansion (Bloom, Canning, and Sevilla, 2003 ). Theoretical models on the effect of trade on unemployment offer mixed conclusions, with no consensus on whether trade increases or decreases unemployment. Some argue for a negative association between trade and unemployment, as trade enhances the economy-wide value of the marginal product of labor. Dutt et al. ( 2009 ) suggest that trade openness improves aggregate labor productivity, fostering job creation and job search, thus reducing unemployment. Similarly, Felbermayr et al. ( 2011 ) argue that trade liberalization lowers unemployment by reallocating labor to more productive firms after driving out less efficient ones. Matusz ( 1996 ) also claims that trade increases productivity through greater labor division, reducing unemployment. On the other hand, some models suggest trade may increase unemployment. Helpman and Itskhoki (2010) argue that reduced trade barriers boost the profitability of exporting firms, leading to an expansion of the trading sector, which could raise unemployment if this sector faces greater labor market frictions. Janiak ( 2006 ) finds that greater trade exposure can lead to more job destruction among small, low-productivity firms than job creation by large firms, increasing unemployment. Finally, other models argue the effect of trade on unemployment is ambiguous. Sener ( 2001 ) and Moore and Ranjan ( 2005 ) find that while trade may increase demand for skilled labor, it can also accelerate job destruction for unskilled workers, leading to frictional unemployment. Moore and Ranjan ( 2005 ) further note that the impact depends on a country’s labor composition: skilled-labor-abundant countries may see lower unemployment, while unskilled-labor-abundant countries may experience higher unemployment. In conclusion, the overall impact of trade on unemployment remains unclear, with outcomes varying based on factors such as labor market frictions and a country’s labor composition. Labor productivity, measured as output per worker, is typically derived from the ratio of Gross Domestic Product (GDP) to the total employed population. The relationship between labor productivity and unemployment is complex and can vary depending on the timeframe and context. Namely, Okun's Law traditionally suggests that higher productivity growth should reduce unemployment by boosting economic output, but the relationship is not always straightforward. In this context, some studies suggest a negative correlation between productivity and unemployment. For instance, Basu et al. ( 2006 ) highlight the adverse effect of technological shocks, which can increase productivity but reduce labor demand, thus raising unemployment in the short term. On the other hand, Gallegati et al. ( 2014 ) argue that labor productivity can have a positive impact on unemployment, but this effect is more pronounced in the long term, using a wavelet approach to capture these dynamics. In addition, as pointed out by Gordon ( 2010 ), in periods of rapid technological change, the displacement of labor by technology can create frictional unemployment, even as overall productivity improves. In other words underscores the role of automation in reshaping labor markets, suggesting that while productivity gains benefit the economy, they may contribute to job polarization, leading to varying impacts on unemployment depending on the skill level of workers. Thus, the net effect of labor productivity on unemployment depends on a range of factors, including technological advancements, labor market flexibility, and the adaptability of the workforce. Yet, in the 1990s, Europe was seen to suffer from higher growth rates of productivity that did not show up in the labor market as higher employment Gallegati et al. ( 2014 ) Table 1 presents the descriptive statistics for all the variables used in the regressions Table 1 Definition of variables Variables Symbol Units Source VC investments in AI AI Number of investments per million people OECD.AI Unemployment, total UNP (% of total labor force) (modeled ILO estimate) World Development Indicators GDP growth (annual %) GDPG Classification code of various exchange regimes World Development Indicators General government final consumption expenditure (% of GDP) GGC (annual %) World Development Indicators The sum of exports and imports of goods and services TRADE (% of GDP) World Development Indicators Inflation, consumer prices INF (annual %) World Development Indicators School enrollment, secondary EDU (% gross) World Development Indicators Population growth POP (annual %) World Development Indicators Annual growth rate of output per worker LP (GDP constant 2017 international $ at PPP) (%) International Labor Organization (ILOSTAT) We also present descriptive statistics for all countries. Table 2 Descriptive statistics Variable Obs Mean Std. Dev. Min Max AI 278 1.38 1.89 0.03 14.09 UNP 324 8.07 4.52 2.02 27.69 GDPG 324 2.20 3.77 -11.17 24.48 GGC 324 19.95 3.19 11.38 26.47 TRADE 324 135.81 70.42 54.87 394.22 INF 324 2.66 3.66 -2.10 19.71 EDU 278 111.12 16.38 86.20 164.08 POP 324 0.26 0.98 -6.19 4.08 LP 324 1.12 3.10 -11.39 20.23 Source: Authors’ calculations. Summary statistics for all variables used in the analysis, presented in Table 2 , demonstrate considerable heterogeneity across countries and over time. Table 3 Correlation matrix AI EDU UNP TRADE POP GDPG GGC AI 1 EDU 0.2557 1 UNP -0.1712 0.0932 1 TRADE 0.2359 -0.0564 -0.3191 1 POP 0.3253 0.23 -0.1216 0.5094 1 GDPG 0.0761 0.0037 -0.1629 0.2402 -0.0529 1 GGC 0.0332 0.4398 -0.0476 -0.3661 0.0284 -0.3315 1 INF 0.269 -0.07 -0.2823 0.1284 -0.0792 0.1263 0.034 Source Authors’ calculations. Before analyzing the regression panel model, a correlation matrix was formed between the dependent and independent variables, and an analysis of Pearson's correlation coefficients was carried out. Namely, we estimate the correlation between selected determinants to check possible problems of multicollinearity between them. We have a multicollinearity problem if the correlation between selected determinants is above 0.80 Gujarati and Porter (2009) and simultaneous inclusion of the variable in the model should be avoided. According to the results from Table 3 , there are no multicollinearity problems between selected determinants. 4. Methodology This section outlines the econometric method chosen to investigate the relationship between AI and unemployment. Unemployment is typically measured as the number of people without a job over a specific period, such as a month, quarter, or year. It reflects the current state of the labor market by indicating how many individuals are actively seeking work but unable to find it at that time. The dynamic lag effect is crucial in the context of unemployment because economic conditions and policies often take time to influence the labor market. For instance, a government stimulus program might require several months for implementation and for its impact on job creation or retention to become apparent. Similarly, during a recession, the unemployment rate may not immediately reflect economic downturns, as businesses might initially opt to reduce hours or wages rather than lay off workers. Furthermore, even after an economic recovery, the unemployment rate may take time to return to pre-recession levels due to factors like workers leaving the labor force or becoming discouraged during the downturn, which can delay their re-entry into the job market. Therefore, incorporating the dynamic effect when analyzing labor market changes provides a more accurate and comprehensive understanding of unemployment dynamics and contributing factors. To capture these effects, this study employs a dynamic panel data model, which is particularly suited for examining the dynamic nature of unemployment. Bond (2002) emphasizes that dynamic relations in analyzing the base process can be decisive for proper and consistent estimations of parameters of the observed independent variables. Specific characteristics of the sample covering 27 countries over 12 years, or the situation when T < N, is an important argument for choosing a dynamic panel model (Greene, 2008). The dynamic panel model is also a good method of estimation when potential endogeneity is considered, which is the case in our model (Greene, 2008). The dynamic panel model offers the possibility of generating internal instruments (external ones are typically difficult to find), so the treatment of potential endogeneity is comprehensive and the estimations more consistent (Roodman, 2006, 2007; Baum, 2006). Various econometric methods are available for estimating dynamic panel data models. The dynamic panel data approach is advantageous as it accounts for the endogeneity of explanatory variables and considers unobservable, time-invariant country effects (Baek, 2016; Yerdelen Tatoglu, 2018). To address the issue of endogeneity, Arellano and Bond (1991) introduced the difference generalized method of moments (difference GMM), which uses instrumental variables to derive GMM estimates of the relevant moment conditions. This method involves taking the first difference of the regression equation to eliminate individual fixed effects, using lagged variables as instruments for endogenous variables in the differenced equation. While this approach effectively mitigates endogeneity concerns, it may suffer from "weak instruments" in finite samples, leading to reduced precision (Bond et al., 2001). To overcome this limitation, Arellano and Bover (1995) and Blundell and Bond (1998) proposed the system GMM estimator, which enhances the difference GMM by using lagged variables as instruments in both the differenced and level equations. Where the αj is p parameters to be estimated, it is a 1 × k 1 vector of strictly exogenous covariates, β 1 is a k 1 × 1 vector of parameters to be estimated, it is a 1 × k 2 vector of predetermined or endogenous covariates, β 2 is a k 2 × 1 vector of parameters to be estimated, νi are the panel-level effects (which may be correlated with the covariates), and ε it are i.i. d . over the whole sample with variance σ 2∈. The νi and the ∈ it are assumed to be independent for each i overall t . There are several compelling reasons for favoring system GMM over difference GMM in empirical analysis. First, when dealing with variables close to a random walk, lagged levels of the regressors, as utilized in difference GMM, tend to be poor instrumental variables (IVs) for the first-differenced regressors (Arellano and Bover, 1995). Second, unbalanced panels can exacerbate data gaps when employing first differences, a characteristic of difference GMM. Third, the standard errors associated with different GMM are often biased downwards (Blundell and Bond, 1998). Fourth, difference GMM tends to perform inadequately with small sample sizes, whereas system GMM exhibits better properties in such scenarios (Blundell and Bond, 2000). Fifth, system GMM provides additional IVs by incorporating a second equation, wherein variables in levels are instrumented with their first differences (Arellano and Bover, 1995). Sixth, the use of more IVs in system GMM enhances estimation efficiency compared to difference GMM. Seventh, system GMM is better suited for modeling non-stationary data and accommodating predetermined explanatory variables. Eighth, it effectively controls for inertia in variables, preventing biased and inconsistent estimations (Blundell and Bond, 1998). Most of our System GMM estimations are based on two to three lags of the endogenous variable in our instrument set. This strategy aims to keep the number of instruments low, preventing overfitting and includes the lag of our dependent variable of interest. The latter helps avoid any bias due to the large number of instruments in a relatively small sample (Petkovski et al., 2023). Related to this, we address the downward bias of standard errors in two-step GMM by using the correction proposed by Windmeijer (2005), which is implemented by the xtabond2 syntax. System GMM has two variants: one-step and two-step estimation methods. The distinction lies in whether the weight matrix is homoscedastic or heteroscedastic. The two-step estimators are often considered more efficient because they reduce bias in the standard errors of estimation values, particularly in small samples (Bond et al., 2001). However, as the number of periods increases, system GMM may generate numerous instrumental variables, which can lead to model overfitting and poor model specification (Roodman, 2009). Thus, a one-step system GMM is recommended for models involving a small number of countries over a longer period, while a two-step system GMM is more suitable for models with a larger number of countries over a shorter period (Teixeira and Queirós, 2016). Given that our analysis covers data from 27 EU countries over 11 years, we opt for the two-step system GMM over the one-step approach To ensure the consistency of our GMM estimation results, we subject our instruments to two specification tests. Firstly, the application of the Hansen test of over-identifying restrictions provides no evidence to reject the validity of the instruments. Secondly, we employ a second-order autocorrelation test to assess whether our error terms exhibit serial correlation. M1 and m2 tests demonstrate a first-order serial correlation in the first-difference equation but detect no evidence of second-order serial correlation. 5. Empirical Results The analysis of the system GMM estimations in Table 3 provides insights into how AI investments, along with other key variables, influence unemployment across different groups of European Union (EU) countries. The estimations are divided into three groups: all EU countries, old EU member states, and new EU member states. This approach allows for a nuanced examination of how AI investments and other economic factors affect unemployment in these regions. Table 3 System GMM estimations Variable EU Countries (all) EU Countries (old) EU Countries (new) UNP (-1) 0.935*** (0.018) 0.918*** (0.031) 0.956*** (0.145) AI 0.051*** (0.022) 0.149* (0.089) -0.496* (0.396) EDU -0.001 (0.004) -0.019 (0.016) 0.136 (0.205) POP -0.103 (0.141) 0.117 (0.173) 0.711 (1.156) TRADE 0.008 (0.002) 0.003 (0.002) 0.004 (0.019) GDPG -0.359*** (0.062) -0.304*** (0.105) -0.730*** (0.231) GGC 0.243*** (0.059) 0.003 (0.002) 0.004 (0.019) INF -0.037 (0.028) 0.149*** (0.050) 0.284 (0.184) LP 0.376*** (0.069) -0.109*** (0.045) 0.175 (0.124) Consant -5.741*** (1.692) 0.351*** (0.108) 0.742*** (0.260) Number of observations 219 135 84 Number of countries 27 16 11 Arellano-Bond test for AR(1) in first differences 0.014 0.083 0.015 Arellano-Bond test for AR(2) in first differences 0.543 0.471 0.227 Hansen test of the validity of instruments ( p -value) 0.711 0.316 0.487 The results indicate substantial variation in the impact of AI investments on unemployment across the three groups: All EU Countries: The coefficient for AI investments is 0.051, which is positive and statistically significant. This suggests that an increase in AI investment per million people leads to higher unemployment in the EU as a whole. The positive effect implies that technological advancements, particularly automation driven by AI, are causing short-term disruptions in the labor market, displacing workers in industries susceptible to automation. Old EU Countries: The coefficient for AI investments in old member states is 0.149, also positive but significant only at the 10% level. The larger coefficient compared to the overall sample suggests that in advanced economies, where AI technologies are more integrated into traditional industries, the displacement of workers is more pronounced, contributing to higher unemployment. New EU Countries: In contrast, the AI coefficient for new EU member states is -0.496, which is negative and statistically significant. This finding indicates that AI investments in these countries are associated with a reduction in unemployment. AI may be fostering job creation in new industries, particularly in IT and AI-related sectors, which offsets the displacement effects seen in more technologically mature economies. Across all groups, the lagged unemployment variable is highly significant, with coefficients close to 1, highlighting the persistence of unemployment. The highest persistence is observed in new member states (0.956), indicating slower adjustment in labor markets compared to old member states (0.918), where the slightly lower persistence may reflect more dynamic labor market policies. The effect of school enrollment on unemployment is largely insignificant across the three groups, including in new member states where the coefficient is positive but not statistically significant. This result suggests that secondary school enrollment alone does not have a meaningful impact on unemployment in the short term, possibly due to a mismatch between the education provided and the skills demanded in a rapidly changing labor market. Population growth shows mixed and generally insignificant effects. In the overall EU sample, the coefficient is negative but not significant, while it is positive and insignificant in both old and new member states. These results suggest that population growth does not exert a strong direct influence on unemployment, although it could interact with other economic factors in more complex ways. The results for the trade variable, show positive effect across all three groups of EU countries, but they are not statistically significant, meaning that we cannot confidently conclude that trade has a reliable impact on unemployment in these cases. Nevertheless, the positive coefficients suggest a potential association between increased trade and higher unemployment, although it is important to interpret these results cautiously due to the lack of significance. However, the positive association between trade and unemployment in these results could reflect short-term labor market frictions or structural adjustments as industries adapt to global competition. The fact that the coefficients are not statistically significant suggests that trade's impact on unemployment might be complex and not easily captured in a straightforward positive or negative relationship, especially when considering different dynamics in EU member states. Economic growth is consistently negative and significant in all groups, confirming its expected role in reducing unemployment. The strongest effect is found in new member states (-0.730), where GDP growth has a particularly pronounced impact on lowering unemployment, likely reflecting the ongoing process of economic convergence with older EU members. In old EU countries, the coefficient is smaller (-0.304), possibly due to more stable and developed labor markets where growth translates less directly into job creation. General government consumption has a positive and significant effect on unemployment in the overall EU sample, suggesting that higher government spending may not be effectively targeted at reducing unemployment in the short term. However, the effect is insignificant in both old and new member states, indicating that government consumption may not be a major factor in determining unemployment at the regional level. Inflation has a positive and significant effect on unemployment in old EU countries, suggesting that rising prices may be contributing to unemployment, possibly through mechanisms like cost-push inflation. In new member states, however, the effect is insignificant, pointing to different inflation dynamics in these countries, which are likely still adjusting to the economic integration within the EU. The effect of labor productivity on unemployment is positive and significant in the overall sample, which may seem counterintuitive. One possible explanation is that productivity improvements, often driven by technological advances like AI, lead to job displacement as firms automate production processes. In old EU countries, the coefficient is negative and significant, indicating that productivity gains help reduce unemployment, likely due to better labor market adaptation. In new member states, the positive and insignificant coefficient may reflect the early stages of productivity growth, where the displacement effects of automation are more apparent. The contrasting effects of AI investments on unemployment in old and new EU member states can be explained by several factors: Stage of Technological Adoption : Old EU member states are more advanced in integrating AI technologies, leading to higher displacement of routine jobs in sectors like manufacturing and services. This is reflected in the positive AI coefficient (0.149) for these countries, where unemployment rises as firms adopt labor-saving technologies. Job Creation vs. Displacement : In new EU member states, AI investments appear to be generating more jobs, especially in emerging industries such as IT and software development. The negative AI coefficient (-0.496) suggests that job creation outweighs displacement in these countries, where the AI sector is still in a growth phase. Labor Market Structure : Old EU countries generally have more rigid labor markets, with higher wages and stronger protections for workers, making it harder for labor markets to adjust to AI-driven changes. In contrast, new member states have more flexible labor markets, allowing for easier transitions to new industries created by AI investments. Sectoral Composition : The economic composition of old and new member states differs, with old EU countries having a larger share of high-tech industries where AI adoption is faster and more disruptive. New EU countries, with a larger share of labor-intensive sectors, experience slower AI adoption, allowing for more gradual transitions and less job displacement. These findings highlight the complex and region-specific effects of AI on unemployment, with the potential for both job creation and job displacement depending on the stage of technological development and the structure of the labor market. The conclusion of this study highlights the complex relationship between AI investments and unemployment in the European Union, revealing significant differences between old and new EU member states. For the overall EU sample, the results show a statistically significant positive impact of AI investments on unemployment, with a coefficient of 0.051. This indicates that, on average, increased AI investments are associated with higher unemployment across the EU, aligning with early predictions by Leontief (1983), who theorized that AI and automation could replace jobs, potentially leading to widespread unemployment unless mitigated by government policies. In the old EU member states, the effect is even more pronounced, with a positive coefficient of 0.149, suggesting that labor markets in these countries may be experiencing greater disruption due to AI adoption. This supports the view of Zeira (1998), who argued that technological innovations reduce the demand for labor, as well as Hirst (2014), who emphasized the vulnerability of low-skilled workers in the face of rapid technological change. The displacement of routine jobs by AI in advanced economies mirrors findings by Acemoglu and Restrepo ( 2020 ), who documented job losses and wage declines in regions of the United States following the introduction of industrial robots. Similarly, Stiglitz (2014) argues that the replacement of workers by AI is driven by decisions made by capital owners, particularly in sectors like manufacturing, where automation is most prevalent. In contrast, the new EU member states exhibit a markedly different trend, with a negative coefficient of -0.496, indicating that AI investments in these countries are associated with a reduction in unemployment. This suggests that AI is playing a more complementary role in these economies, possibly creating new job opportunities and facilitating economic development. This result contrasts with the more disruptive effects observed in old EU member states and could be explained by the earlier stages of AI integration in these countries. The creation of jobs in new industries, supported by foreign investment in AI technologies, may be mitigating the displacement effects seen in more developed economies. These findings are consistent with Roubini’s (2014) observation that high-skilled workers tend to benefit more from automation, while low-skilled workers face greater risks of unemployment. In new member states, the complementary role of AI could be fostering job creation in sectors like IT and software development, where high-skilled workers are in demand. In conclusion, while AI investments contribute to rising unemployment in the more advanced economies of the old EU member states, they appear to reduce unemployment in the newer member states, where job creation in AI-related sectors offsets the displacement effects. These contrasting outcomes highlight the importance of tailored policy responses to address the divergent impacts of AI across different regions. As McKinsey Global Institute (2017) warned, automation could displace up to 800 million jobs globally by 2030, underscoring the need for governments to proactively manage this transition through education, retraining, and redistributive measures. Conclusion The findings of this study reveal a complex relationship between AI investments and unemployment across the 27 EU member states. Overall, the analysis indicates a statistically significant positive effect of AI investments on unemployment in the EU, with a coefficient of 0.051. This suggests that, on a broad scale, increasing AI investments per million people is associated with higher unemployment levels. This outcome highlights the short-term disruptions in the labor market caused by technological advancements, particularly as automation driven by AI displaces workers in vulnerable sectors. When analyzing the old EU member states, the coefficient for AI investments is 0.149, significant at the 10% level. This result indicates that, in these more advanced economies, the integration of AI technologies into established industries leads to more pronounced worker displacement, thereby contributing to higher unemployment rates. The findings suggest that while these economies benefit from technological advancements, the transition may result in significant job losses in sectors susceptible to automation. In contrast, the analysis of new EU member states reveals a different narrative. Here, the coefficient for AI investments is -0.496, which is both negative and statistically significant. This indicates that AI investments in these countries are associated with a reduction in unemployment. The findings suggest that in these emerging economies, AI may be facilitating job creation in new industries, particularly within the IT and AI sectors. This positive impact may offset the displacement effects observed in more technologically mature economies. The two-step system GMM estimation approach employed in this study effectively addresses potential endogeneity and the dynamic nature of unemployment, enhancing the robustness of the analysis across different country groupings. While the study sheds light on these important dynamics, several limitations should be acknowledged. The dataset covers annual data from 1990 to 2023, with AI investments gaining significance primarily in the latter part of this period. Consequently, the long-term effects of AI investments on unemployment may not be fully captured. Additionally, the heterogeneity among EU member states necessitates more nuanced models to account for differences in digital infrastructure, education, and economic maturity, particularly between old and new member states. Moreover, potential spillover effects from AI investments in neighboring countries are not considered, which could influence unemployment dynamics. The results suggest critical implications for policymakers, especially in new EU member states. To maximize the benefits of AI investments, it is essential to improve conditions for AI adoption and integration within the broader economy. Key areas of focus should include investing in digital infrastructure, enhancing education, and providing workforce training to prepare the labor market for AI-driven transformations. For old EU member states, sustaining the benefits of AI requires promoting innovation, supporting entrepreneurship in AI-related sectors, and ensuring that workers are continuously upskilled to adapt to technological changes. At the EU level, coordinated efforts are vital to bridge the digital divide between old and new member states. This could involve targeted funding and support for AI research, infrastructure development, and skills training in emerging economies. Additionally, policies that foster cross-border collaboration in AI innovation could facilitate knowledge transfer and enhance the overall impact of AI investments throughout the EU. In summary, while AI investments contribute to unemployment trends across the EU, the varying effects highlight the necessity for targeted policies that address regional disparities. Ensuring that all EU countries can leverage the full potential of AI technology will be crucial for fostering sustainable economic growth and technological advancement. Declarations Funding Declaration: The research received no specific grant from any funding agency in the public,commercial, or not-for-profit sectors Author Contribution J.KJ. WROTTE ALL PAPERS References Acemoglu, D., and Restrepo, P. (2018). Artificial Intelligence, Automation, and Work. NBER Working Paper No. 24196 , National Bureau of Economic Research. https://www.nber.org/papers/w24196 Acemoglu, D., and Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. Journal of Economic Perspectives , 33(2), 3-30. Acemoglu, D., and Restrepo, P. (2020). Robots and Jobs: Evidence from US Labor Markets. Journal of Political Economy , 128(6), 2188-2244. https://doi.org/10.1086/705716 Andersen, T. M., and Svarer, M. (2007). Flexicurity—Labour Market Performance in Denmark. CESifo Economic Studies , 53(3), 389-429. https://doi.org/10.1093/cesifo/ifm015 Arntz, M., Gregory, T., and Zierahn, U. (2016). The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis. OECD Social, Employment and Migration Working Papers No. 189 , OECD Publishing. https://doi.org/10.1787/5jlz9h56dvq7-en Auerbach, A. J., and Gorodnichenko, Y. (2012). Measuring the output responses to fiscal policy. American Economic Journal: Economic Policy , 4(2), 1-27. Baldwin, R. (2019). The Globotics Upheaval: Globalization, Robotics, and the Future of Work . Oxford University Press. Ball, L., Leigh, D., and Loungani, P. (2013). Okun’s law: Fit at fifty? NBER Working Paper No. 18668 . Ball, L., Mankiw, N. G., and Romer, D. (1988). The new Keynesian economics and the output-inflation trade-off. Brookings Papers on Economic Activity , 1988(1), 1-82. Basu, S., Fernald, J. G., and Kimball, M. S. (2006). Are technology improvements contractionary? American Economic Review , 96(5), 1418-1448. https://doi.org/10.1257/aer.96.5.1418 Becker, G. S. (1993). Human Capital: A Theoretical and Empirical Analysis, with Special Reference to Education (3rd ed.). University of Chicago Press. Berg, J., Furrer, M., Harmon, E., Rani, U., and Silberman, M. S. (2018). Digital Labour Platforms and the Future of Work: Towards Decent Work in the Online World . International Labour Office. https://www.ilo.org/global/publications/books/WCMS_645337/lang--en/index.htm Berger, T., and Frey, C. B. (2016). Regional Technological Dynamism and the Emergence of New Occupations: Implications for Local Labor Markets. Journal of Economic Geography , 16(6), 1215-1243. https://doi.org/10.1093/jeg/lbv057 Bessen, J. (2019). AI and Jobs: The Role of Demand. NBER Working Paper No. 24235 , National Bureau of Economic Research. https://www.nber.org/papers/w24235 Bessen, J., Goos, M., Salomons, A., and Van den Berge, W. (2020). Firm-Level Automation: Evidence from the Netherlands. A Report for the European Commission . https://data.europa.eu/doi/10.2760/869954 Bloom, D. E., Canning, D., and Sevilla, J. (2003). The Demographic Dividend: A New Perspective on the Economic Consequences of Population Change . Rand Corporation. Bowles, J. (2017). The Computerization of European Jobs. Bruegel , July 24th. Bregman, R. (2017). Utopia for Realists: How We Can Build the Ideal World . Little, Brown and Company. Bresnahan, T. F., Brynjolfsson, E., and Hitt, L. M. (2002). Information Technology, Workplace Organization, and the Demand for Skilled Labor: Firm-Level Evidence. Quarterly Journal of Economics , 117(1), 339-376. https://doi.org/10.1162/003355302753399526 Brynjolfsson, E., and McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies . W. W. Norton and Company. Carbonero, F., Ernst, E., and Weber, E. (2020). Robots worldwide: The impact of automation on employment and trade. Chiacchio, F., Petropoulos, G., and Pichler, D. (2018). The Impact of Industrial Robots on EU Employment and Wages: A Local Labor Market Approach. Bruegel Working Paper No. 02 , Brussels. Choi, Y., Stewart, S., and Ellingrud, K. (2020). The Future of Work in America: People and Places, Today and Tomorrow . McKinsey Global Institute. https://www.mckinsey.com/featured-insights/future-of-work/the-future-of-work-in-america-people-and-places-today-and-tomorrow Cowls, J., and Floridi, L. (2018). Prolegomena to a White Paper on an Ethical Framework for a Good AI Society. SSRN Electronic Journal . https://doi.org/10.2139/ssrn.3288950 De Stefano, V. (2016). The Rise of the "Just-in-time Workforce": On-demand Work, Crowdwork, and Labor Protection in the "Gig-economy". Comparative Labor Law and Policy Journal , 37(3), 471-503. Dutt, P., Mitra, D., and Ranjan, P. (2009). International Trade and Unemployment: Theory and Cross-National Evidence. Journal of International Economics , 78(1), 32-44. Elkeles, T., and Phillips, J. (2017). The Chief Learning Officer: Driving Value Within a Changing Organization Through Learning and Development . Routledge. European Commission. (2020). White Paper on Artificial Intelligence: A European Approach to Excellence and Trust . https://ec.europa.eu/info/publications/white-paper-artificial-intelligence-european-approach-excellence-and-trust_en Felbermayr, G., Prat, J., and Schmerer, H. (2011). Globalisation and Labour Market Outcomes: Wage Bargaining, Search Frictions, and Firm Heterogeneity. Journal of Economic Theory , 146(1), 39-73. Frank, M. R., Autor, D., Bessen, J. E., Brynjolfsson, E., Cebrian, M., Deming, D. J., and Rahwan, I. (2019). Toward Understanding the Impact of Artificial Intelligence on Labor. Proceedings of the National Academy of Sciences , 116(14), 6531. https://doi.org/10.1073/pnas.1900949116 Frey, C. B., and Osborne, M. A. (2017). The Future of Employment: How Susceptible Are Jobs to Computerisation? Technological Forecasting and Social Change , 114, 254-280. https://doi.org/10.1016/j.techfore.2016.08.019 Gallegati, M., Ramsey, J. B., and Semmler, W. (2014). Long-run Waves in Unemployment and Productivity: Evidence from Wavelet Analysis. Journal of Economic Dynamics and Control , 44, 130-149. https://doi.org/10.1016/j.jedc.2014.03.011 Gordon, R. J. (2010). Revisiting the Rise and Fall of American Growth. American Economic Review , 100(2), 71-76. https://doi.org/10.1257/aer.100.2.71 Hanushek, E. A., and Woessmann, L. (2008). The Role of Cognitive Skills in Economic Development. Journal of Economic Literature , 46(3), 607-668. Heimberger, P. (2023). Public Debt and Economic Growth in Advanced Economies: A Survey of Recent Empirical Evidence. Journal of Economic Surveys , 37(1), 1-25. https://doi.org/10.1111/joes.12436 Helpman, E., Itskhoki, O., and Redding, S. (2010). Inequality and Unemployment in a Global Economy. Econometrica , 78(4), 1239-1283. Healy, J., Nicholson, D., and Pekarek, A. (2017). Should We Take the Gig Economy Seriously? Labour and Industry: A Journal of the Social and Economic Relations of Work , 27(3), 232-248. https://doi.org/10.1080/10301763.2017.1377048 Janiak, A. (2006). Does Trade Liberalisation Lead to Unemployment? Theory and Some Evidence. ECARES , Université Libre de Bruxelles. Josten, C., and Lordan, G. (2019). Robots at Work: Automation and the Ageing Workforce. IZA Institute of Labor Economics Discussion Paper No. 12493 . https://www.iza.org/publications/dp/12493/robots-at-work-automation-and-the-ageing-workforce Keynes, J. M. (1936). The General Theory of Employment, Interest, and Money . Macmillan, London. Kenney, M., and Zysman, J. (2019). Work and Value Creation in the Platform Economy: The Case of Artificial Intelligence. Global Perspectives , 1(1), 10-13. https://doi.org/10.1525/gp.2019.12382 Levy, F., and Murnane, R. J. (2003). The Skill Content of Recent Technological Change: An Empirical Exploration. Quarterly Journal of Economics , 118(4), 1279-1333. https://doi.org/10.1162/003355303322552801 Lucas, R. E. (1988). On the Mechanics of Economic Development. Journal of Monetary Economics , 22(1), 3-42. https://doi.org/10.1016/0304-3932(88)90168-7 Lund, S., Madgavkar, A., Manyika, J., and Smit, S. (2021). The Future of Work After COVID-19 . McKinsey Global Institute. https://www.mckinsey.com/featured-insights/future-of-work/the-future-of-work-after-covid-19 Matusz, S. J. (1996). International Trade, the Division of Labor, and Unemployment. International Economic Review , 37(1), 71-83. McAfee, A., and Brynjolfsson, E. (2017). Machine, Platform, Crowd: Harnessing Our Digital Future . W. W. Norton and Company. Mokyr, J., Vickers, C., and Ziebarth, N. L. (2015). The History of Technological Anxiety and the Future of Economic Growth: Is This Time Different? Journal of Economic Perspectives , 29(3), 31-50. https://doi.org/10.1257/jep.29.3.31 Moretti, E. (2012). The New Geography of Jobs . Houghton Mifflin Harcourt. Moore, M. P., and Ranjan, P. (2005). Globalisation vs Skill-Biased Technological Change: Implications for Unemployment and Wage Inequality. Economic Journal , 115(503), 391-422. Neufeind, M., O'Reilly, J., and Ranft, F. (2018). Work in the Digital Age: Challenges of the Fourth Industrial Revolution . Rowman and Littlefield International. Ngai, L. R., and Pissarides, C. A. (2007). Structural Change in a Multisector Model of Growth. American Economic Review , 97(1), 429-443. https://doi.org/10.1257/aer.97.1.429 Nordhaus, W. D. (2007). Two Centuries of Productivity Growth in Computing. Journal of Economic History , 67(1), 185-194. https://doi.org/10.1017/S0022050707000058 Piketty, T. (2014). Capital in the Twenty-First Century . Harvard University Press. Rifkin, J. (1995). The End of Work: The Decline of the Global Labor Force and the Dawn of the Post-Market Era . Putnam Publishing Group. Sachs, J. D., Benzell, S. G., and LaGarda, G. (2015). Robots: Curse or Blessing? A Basic Framework. NBER Working Paper No. 21091 . National Bureau of Economic Research. https://doi.org/10.3386/w21091 Schwab, K. (2016). The Fourth Industrial Revolution . World Economic Forum. Sener, F. (2001). Schumpeterian Unemployment, Trade and Wages. Journal of International Economics , 54(1), 119-148. Solow, R. M. (1956). A Contribution to the Theory of Economic Growth. Quarterly Journal of Economics , 70(1), 65-94. https://doi.org/10.2307/1884513 Steigum, E. (2019). Unemployment and Automation: Long-Term Implications of Technological Innovation. Economic Policy , 34(98), 259-288. https://doi.org/10.1093/epolic/eiz003 Stiglitz, J. E. (2012). The Price of Inequality: How Today's Divided Society Endangers Our Future . W. W. Norton and Company. Summers, L. H. (2014). Reflections on the “New Secular Stagnation Hypothesis.” In C. Teulings & R. Baldwin (Eds.), Secular Stagnation: Facts, Causes, and Cures (pp. 27-40). CEPR Press. Van Reenen, J. (2011). Wage Inequality, Technology and Trade: 21st Century Evidence. Labour Economics , 18(6), 730-741. Wisskirchen, G., Thibault, B., Bormann, U., Muntz, A., Niehaus, G., Soler, G., ... and Schaude, T. (2017). Artificial Intelligence and Robotics and Their Impact on the Workplace . IBA Global Employment Institute. https://www.ibanet.org/document/Wages Footnotes Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, and Sweden. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6641180","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":457874136,"identity":"ef9c85fa-4aad-46d7-975d-4ba676c4e2e3","order_by":0,"name":"Jordan Kjosevski","email":"data:image/png;base64,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","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Jordan","middleName":"","lastName":"Kjosevski","suffix":""}],"badges":[],"createdAt":"2025-05-11 18:53:05","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6641180/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6641180/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":84170814,"identity":"3c106251-0eaa-45d3-aaf0-c97b2a35c203","added_by":"auto","created_at":"2025-06-08 21:46:26","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":703556,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6641180/v1/a9cdb67d-5045-47a0-988e-19b8614e54cf.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial Intelligence and Its Impact on Unemployment: A Comparative Analysis of Old and New EU Member States","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eArtificial Intelligence (AI) has emerged as one of the most transformative technologies of the 21st century, influencing various aspects of human life, particularly in the workplace. As AI continues to evolve and integrate into different sectors, there is growing concern about its impact on employment and unemployment rates. While AI has the potential to enhance productivity and create new job opportunities, it also poses significant risks to existing jobs, particularly those that involve routine and repetitive tasks. This study explores the multifaceted impact of AI on unemployment, analyzing theoretical perspectives and empirical evidence. It also considers the implications of AI-driven changes in the labor market and discusses potential policy responses to mitigate the adverse effects on employment.\u003c/p\u003e \u003cp\u003eAI refers to the simulation of human intelligence in machines programmed to think and learn. These machines can perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision-making, and language translation (Russell and Norvig, 2016). AI can be broadly categorized into two types: narrow AI, which is designed for specific tasks, and general AI, which possesses the ability to perform any intellectual task that a human can do (Goodfellow, Bengio, and Courville, 2016).\u003c/p\u003e \u003cp\u003eThe scope of AI extends across various sectors, including manufacturing, healthcare, finance, and customer service. In manufacturing, AI-powered robots and automation systems are used to enhance production efficiency (Brynjolfsson and McAfee, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). In healthcare, AI is utilized for diagnostic purposes, personalized treatment plans, and robotic surgeries (Topol, 2019). The financial sector leverages AI for algorithmic trading, fraud detection, and customer service chatbots (Baldwin, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These applications illustrate the widespread adoption of AI, which is poised to have profound effects on employment.\u003c/p\u003e \u003cp\u003eThe impact of AI on employment is a subject of considerable debate among economists and policymakers. One of the primary concerns is the potential for AI to displace jobs through automation. The concept of \"technological unemployment\" was first introduced by Keynes (1933), who suggested that technological advancements could lead to a net loss of jobs. AI, with its ability to perform complex tasks that were once the domain of humans, could exacerbate this phenomenon, particularly in sectors where routine and manual tasks are prevalent (Acemoglu and Restrepo, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eHowever, AI also has the potential to create new jobs, particularly in emerging fields such as AI development, data science, and machine learning (Arntz, Gregory, and Zierahn, 2016). The introduction of AI could lead to job creation through increased demand for new products and services, as well as the need for workers to develop, maintain, and supervise AI systems. This dual effect of AI on employment\u0026mdash;job displacement and job creation\u0026mdash;reflects the broader theory of creative destruction, where technological progress leads to the obsolescence of certain jobs while giving rise to new ones (Schumpeter, 1942).\u003c/p\u003e \u003cp\u003eThe concept of skill-biased technological change (SBTC) is also relevant in the context of AI. SBTC refers to technological advancements that favor workers with higher skill levels, leading to increased demand for skilled labor and reduced demand for unskilled labor (Levy and Murnane, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). AI, with its reliance on data and complex algorithms, is likely to exacerbate this trend, widening the gap between skilled and unskilled workers and contributing to rising income inequality (Goos, Manning, and Salomons, 2014).\u003c/p\u003e \u003cp\u003eGiven the diverse perspectives on the effects of AI, a major concern revolves around its impact on unemployment levels. The World Economic Forum's \"Future of Jobs\" report (2018) notes that AI is increasingly being adopted across various economic sectors, potentially altering employment opportunities and skill development. Similarly, the European Commission's \"Artificial Intelligence for Europe\" statement (2018) underscores AI's profound influence on multiple economic sectors, advocating for its responsible use in line with ethical standards.\u003c/p\u003e \u003cp\u003eDespite extensive research on AI's effects on employment and income distribution, there remains a gap in understanding its specific implications for the European economy. Much of the existing literature is either global (Mutascu, 2021; Guliyev, 2023) or country-specific (Acemoglu and Restrepo, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Carbonero, Ernst, and Weber, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), leaving a need for focused investigations into AI's impact on the European Union (except Bowles \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Chiacchio et al. \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). This paper aims to fill this research gap by examining the relationship between AI and unemployment in the 27 EU countries, using worldwide venture capital (VC) investments in AI and AI-related data start-ups as key indicators. This study distinguishes itself by providing a comprehensive analysis of the 27 EU countries and by being among the first to apply a dynamic panel data estimation method to assess AI's impact on unemployment. Furthermore, one of the key contributions of this study is the use of worldwide venture capital (VC) investments in AI and data start-ups as a proxy for AI adoption and innovation within the EU. This approach provides a novel way of measuring AI's influence on the labor market, capturing the intensity and scale of AI-related activities that are likely to affect employment. By focusing on venture capital investments, this study not only considers the current state of AI adoption but also the future potential for AI-driven economic transformation in the region.\u003c/p\u003e \u003cp\u003eThe subsequent sections of this study are structured as follows: The second section provides a review of the relevant academic literature. In the third section, we delve into detail regarding our empirical approach, including the data sources and variables used in our analysis. Next, the methodology and econometric framework are outlined in the subsequent section. Moving forward to the fifth section, we present the empirical findings. Finally, the sixth section concludes the study and discusses policy implications.\u003c/p\u003e"},{"header":"2. Literature review","content":"\u003cp\u003eArtificial intelligence (AI) has profound implications for the socioeconomic landscape, influencing various areas such as unemployment, inequality, and productivity. Over the past few decades, the literature on the impact of AI on unemployment has expanded considerably (Ernst et al., 2019; Martens and Tolan, 2018), though many studies treat AI as part of a broader automation process (Wang and Siau, 2019). Empirical evidence on the impact of AI on unemployment is mixed.\u003c/p\u003e \u003cp\u003eBroadly speaking, this literature can be divided into two main strands: the \"replacement effect\" and the \"displacement effect.\", with outcomes varying across sectors and regions.\u003c/p\u003e \u003cp\u003eThe first strand supports the \"replacement effect,\" which includes both theoretical and empirical studies. This perspective argues that AI negatively impacts the labor market by replacing jobs and increasing unemployment. Leontief (1983) was one of the first to theorize about AI's potential to destroy jobs, predicting that almost all jobs could be replaced by AI in the coming decades, leading to higher unemployment unless offset by government redistributive policies.\u003c/p\u003e \u003cp\u003eBessen (2018, 2020) introduces the concept of demand elasticity, arguing that automation only reduces employment if demand is inelastic; otherwise, there could be positive effects on job creation. Atkinson (2018) describes AI's destructive role in employment, noting that as automation reduces jobs in one sector, such as agriculture, people move to new sectors, but eventually, there will be no new sectors left to absorb the displaced workers. Acemoglu and Restrepo (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) expand on this by modeling the race between humans and machines, showing that automation reduces employment and wages when capital is fixed and technology is exogenous. They later (Acemoglu and Restrepo, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) argue that increased AI use lowers labor demand, national income, and productivity growth while exacerbating inequality. Empirical studies also support this strand, with Frey and Osborne (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) estimating that automation could lead to job losses of 47 percent in the U.S. over the next two decades. Arntz et al. (2017) provide a more conservative estimate of a 9 percent job loss in 21 OECD countries. Acemoglu and Restrepo (2017) find that increased industrial robot usage between 1990 and 2007 negatively impacted U.S. local labor markets, reducing employment and wages. Bowles (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) and Chiacchio et al. (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) find similar negative effects in the EU.\u003c/p\u003e \u003cp\u003eThe second strand of literature promotes the \"displacement effect,\" arguing that AI positively impacts the labor market by creating new jobs and reducing unemployment. Albus (1983) was one of the first to argue that technological progress, including AI, increases productivity and generates new markets, thereby creating more jobs and reducing unemployment. Boden (1987) echoes this sentiment, suggesting that AI will eventually lead to more jobs than before, despite transitional challenges. Dauth et al. (2017) find no net job losses in Germany due to automation. Frey and Osborne (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) also introduce the concept of sectors with varying risks of automation, noting that unemployment increases only in sectors most exposed to automation. Su (2018) argues that AI reduces unemployment by changing job structures, with policymakers needing to stimulate job creation in new sectors to offset job losses. Gries and Naudé (2018) incorporate AI into an economic growth model and find that AI does not immediately increase unemployment. Acemoglu and Restrepo (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) claim that AI improves productivity and creates new jobs through increased capital accumulation.\u003c/p\u003e \u003cp\u003eCautionary perspectives also exist within this strand. Trehan (2003) empirically shows that AI's positive impact on unemployment is only valid under specific conditions and may have a retrograde effect over time. Autor (2015) argues that automation changes the nature of jobs rather than eliminating them, with new jobs replacing old ones. Berriman and Hawksworth (2017) suggest that the net impact of automation on jobs is neutral, as job losses are compensated by newly created jobs. Gordon (2018) finds that automation slowly replaces jobs, mainly in a minority of sectors, while Carbonero et al. (2018) show that robots have a small negative impact on jobs in developing countries.\u003c/p\u003e \u003cp\u003eRecent studies such as Mutascu (2021) and Guliyev (2023) contribute to this field by examining the impact of AI on unemployment in high-tech, developed countries. Mutascu (2021) uses a nonlinear approach with panel threshold and GMM-system estimations to analyze data from 23 countries over the period 1998–2016. The results indicate that AI reduces unemployment at low levels of inflation but has a neutral effect otherwise, with no \"switch effect\" between the \"displacement\" and \"replacement\" effects. Guliyev (2023) extends this analysis by examining the relationship between AI-related Google Trends data and unemployment rates in 24 high-tech countries from 2005 to 2021, using dynamic panel data and GMM-system estimation. The study finds that AI decreases unemployment, validating the \"displacement effect.\"\u003c/p\u003e"},{"header":"3. Variables and Data","content":"\u003cp\u003eOur study dataset consists of a sample of 27 EU member states and uses annual data from 2012 to 2023.\u003csup\u003e1\u003c/sup\u003e The panel is unbalanced given that the study features many “new” EU member states, and not all of the indicators were available for each country over the entire period (Ostrihoň, Siranova, and Workie Tiruneh,2023). On the other hand, there also existed differences among these countries that made them a remarkably heterogeneous group. These differences were mostly reflected in the high disparities observed concerning the level of public debt, GDP growth, unemployment, and so on.\u003c/p\u003e\u003cp\u003eThe core determinants selected in our model have been previously used in academic literature (Mutascu, 2021; Guliyev, 2023 among others). The internationally comparable and reliable data were obtained from the OECD.AI International Labor Organization (ILOSTAT) and the World Bank’s World Development Indicators.\u003c/p\u003e\u003cp\u003eThe dependent variable, expressed as a percentage of the total labor force, represents the unemployment rate, which reflects the portion of the labor force that is unemployed, available for work, and actively seeking employment. This variable was used in studies by Mutascu (2021) and Guliyev (2023) to analyze labor market dynamics and the impact of various economic factors on unemployment. Additionally, the unemployment rate serves as a key indicator for policymakers, helping to guide decisions related to job creation and economic growth strategies. The consistent use of this variable in empirical research underscores its importance in understanding and addressing unemployment issues.\u003c/p\u003e\u003cp\u003eIn this study, we utilize Preqin's data on global venture capital (VC) investments in AI and data startups, as reported by OECD.AI. To provide a more accurate and meaningful comparison across countries, we normalize these investments by calculating the number of investments per million people. Adjusting for population size allows for a standardized metric that better reflects the intensity of AI investment activity across different nations. This approach helps account for variations in market scale and investment capacity, offering a clearer picture of where AI investments are most concentrated and identifying areas of potential growth.\u003c/p\u003e\u003cp\u003eNormalizing investments per capita is particularly valuable for enabling researchers and policymakers to make fair comparisons between countries with differing population sizes. It highlights investment concentration and reveals gaps or opportunities in AI development. Additionally, this metric sheds light on the relationship between investment levels and a country’s technological advancement per capita, offering insights into the nation’s capacity for innovation. By focusing on investments per million people, stakeholders can make more informed decisions regarding resource allocation, strategy planning, and policy development, fostering sustainable growth and competitiveness in the AI sector.\u003c/p\u003e\u003cp\u003eA set of \u003cem\u003econtrol variables\u003c/em\u003e is considered to isolate the effects of interest variables. They are inspired by both the theoretical model and the literature: economic growth, inflation rate, government spending, school enrollment, trade, and population growth.\u003c/p\u003e\u003cp\u003eThe first control variable is GDP growth. Economic growth and the unemployment rate are the key indicators that are simultaneously monitored by both policymakers and the public, as they create a clear picture of the economic development of a country. Moreover, the linkage between the unemployment rate and economic growth as a relevant macroeconomic issue covers a wide area of theoretical and empirical research. It is a widely accepted view in economics that a higher growth rate of the GDP of an economy increases employment and reduces unemployment. This theoretical proposition relating output and unemployment is known as “Okun’s Law.” This relationship is among the most prominent in macroeconomic theory and is held for several countries and regions, mainly in developed countries (Lee, 2000; Farsio and Quade, 2003).\u003c/p\u003e\u003cp\u003eInflation can influence unemployment by affecting real wages and purchasing power. According to the Phillips curve, there is a short-term trade-off between inflation and unemployment, where higher inflation can lower unemployment by reducing real wages and making labor more affordable for employers (Phillips, 1958). However, over the long term, this relationship tends to diminish, as empirical studies indicate that the impact of inflation on unemployment depends on the specific country and its economic conditions (Ball, Mankiw, and Romer, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1988\u003c/span\u003e). Inflation, measured through the consumer price index, reflects the annual rate of price fluctuations in an economy. The theoretical literature often finds a negative correlation between inflation and unemployment (Phillips, 1958). The Phillips curve suggests a stable relationship between inflation and unemployment, arguing that rising inflation can increase employment opportunities and reduce unemployment (Wulandari et al., 2019).\u003c/p\u003e\u003cp\u003eGovernment spending, expressed as a percentage of GDP includes cash payments for operating activities related to providing goods and services. Fiscal actions can have mixed effects on unemployment: while increased spending may stimulate job creation, it can also raise labor supply and unemployment, even if output rises. Conversely, cuts in government spending might lead to higher private-sector output and productivity, potentially causing GDP growth alongside rising unemployment. This ambiguity is reflected in recent research. Some studies, such as Monacelli et al. (2010), Auerbach and Gorodnichenko (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), Ramey (2012), and Holden and Sparrman (2018), find that higher government spending reduces unemployment. However, other studies suggest that large-scale government spending may increase unemployment, with Abrams (1999), Christopoulos et al. (2005), Feldmann (2006), and Brückner and Pappa (2012) indicating that increased government purchases can lead to higher unemployment.\u003c/p\u003e\u003cp\u003eEducation, particularly secondary school enrollment, is a critical determinant of labor market outcomes. Higher enrollment rates can lead to a more skilled workforce, which may reduce unemployment by increasing employability and adaptability to technological changes brought about by AI. The human capital theory posits that investment in education improves individual productivity, thereby lowering unemployment (Becker, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1993\u003c/span\u003e). Empirical evidence suggests that regions with higher secondary school enrollment rates tend to experience lower unemployment rates due to the increased availability of skilled labor (Hanushek and Woessmann, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e\u003cp\u003ePopulation growth affects labor supply, and its impact on unemployment depends on the economy's ability to absorb new entrants into the labor market. High population growth can lead to higher unemployment if job creation does not keep pace with the growing labor force. Malthusian theory suggests that rapid population growth can outstrip economic growth, leading to higher unemployment (Malthus, 1798). However, empirical studies indicate that population growth can have varied effects on unemployment, depending on the structure of the economy and the rate of economic expansion (Bloom, Canning, and Sevilla, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2003\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eTheoretical models on the effect of trade on unemployment offer mixed conclusions, with no consensus on whether trade increases or decreases unemployment. Some argue for a negative association between trade and unemployment, as trade enhances the economy-wide value of the marginal product of labor. Dutt et al. (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2009\u003c/span\u003e) suggest that trade openness improves aggregate labor productivity, fostering job creation and job search, thus reducing unemployment. Similarly, Felbermayr et al. (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) argue that trade liberalization lowers unemployment by reallocating labor to more productive firms after driving out less efficient ones. Matusz (\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e1996\u003c/span\u003e) also claims that trade increases productivity through greater labor division, reducing unemployment. On the other hand, some models suggest trade may increase unemployment. Helpman and Itskhoki (2010) argue that reduced trade barriers boost the profitability of exporting firms, leading to an expansion of the trading sector, which could raise unemployment if this sector faces greater labor market frictions. Janiak (\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) finds that greater trade exposure can lead to more job destruction among small, low-productivity firms than job creation by large firms, increasing unemployment. Finally, other models argue the effect of trade on unemployment is ambiguous. Sener (\u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2001\u003c/span\u003e) and Moore and Ranjan (\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) find that while trade may increase demand for skilled labor, it can also accelerate job destruction for unskilled workers, leading to frictional unemployment. Moore and Ranjan (\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2005\u003c/span\u003e) further note that the impact depends on a country’s labor composition: skilled-labor-abundant countries may see lower unemployment, while unskilled-labor-abundant countries may experience higher unemployment. In conclusion, the overall impact of trade on unemployment remains unclear, with outcomes varying based on factors such as labor market frictions and a country’s labor composition.\u003c/p\u003e\u003cp\u003eLabor productivity, measured as output per worker, is typically derived from the ratio of Gross Domestic Product (GDP) to the total employed population. The relationship between labor productivity and unemployment is complex and can vary depending on the timeframe and context. Namely, Okun's Law traditionally suggests that higher productivity growth should reduce unemployment by boosting economic output, but the relationship is not always straightforward. In this context, some studies suggest a negative correlation between productivity and unemployment. For instance, Basu et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) highlight the adverse effect of technological shocks, which can increase productivity but reduce labor demand, thus raising unemployment in the short term. On the other hand, Gallegati et al. (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) argue that labor productivity can have a positive impact on unemployment, but this effect is more pronounced in the long term, using a wavelet approach to capture these dynamics. In addition, as pointed out by Gordon (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2010\u003c/span\u003e), in periods of rapid technological change, the displacement of labor by technology can create frictional unemployment, even as overall productivity improves. In other words underscores the role of automation in reshaping labor markets, suggesting that while productivity gains benefit the economy, they may contribute to job polarization, leading to varying impacts on unemployment depending on the skill level of workers. Thus, the net effect of labor productivity on unemployment depends on a range of factors, including technological advancements, labor market flexibility, and the adaptability of the workforce. Yet, in the 1990s, Europe was seen to suffer from higher growth rates of productivity that did not show up in the labor market as higher employment Gallegati et al. (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e)\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the descriptive statistics for all the variables used in the regressions\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDefinition of variables\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSymbol\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eUnits\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSource\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVC investments in AI\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAI\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNumber of investments per million people\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eOECD.AI\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnemployment, total\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eUNP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(% of total labor force) (modeled ILO estimate)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGDP growth (annual %)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGDPG\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eClassification code of various exchange regimes\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGeneral government final consumption expenditure (% of GDP)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGGC\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(annual %)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThe sum of exports and imports of goods and services\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTRADE\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(% of GDP)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInflation, consumer prices\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eINF\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(annual %)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSchool enrollment, secondary\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEDU\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(% gross)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePopulation growth\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePOP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(annual %)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eWorld Development Indicators\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnnual growth rate of output per worker\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(GDP constant 2017 international \u003cspan\u003e$\u003c/span\u003e at PPP) (%)\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInternational Labor Organization (ILOSTAT)\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003eWe also present descriptive statistics for all countries.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive statistics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"6\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eObs\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMean\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStd. Dev.\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eMin\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eMax\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e278\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.38\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.89\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e14.09\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUNP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8.07\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.52\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e2.02\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e27.69\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGDPG\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.20\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.77\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-11.17\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e24.48\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGGC\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e19.95\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.19\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e11.38\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e26.47\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTRADE\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e135.81\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e70.42\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e54.87\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e394.22\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eINF\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.66\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.66\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-2.10\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e19.71\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEDU\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e278\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e111.12\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16.38\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e86.20\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e164.08\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePOP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.26\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.98\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-6.19\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e4.08\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e324\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1.12\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.10\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-11.39\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e20.23\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e \u003cb\u003eSource: Authors’ calculations.\u003c/b\u003e \u003c/p\u003e\u003cp\u003eSummary statistics for all variables used in the analysis, presented in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, demonstrate considerable heterogeneity across countries and over time.\u003c/p\u003e\u003cdiv class=\"gridtable\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCorrelation matrix\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003c/colgroup\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAI\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEDU\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eUNP\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eTRADE\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePOP\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eGDPG\u003c/p\u003e \u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eGGC\u003c/p\u003e \u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEDU\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.2557\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUNP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.1712\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0932\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTRADE\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.2359\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.0564\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.3191\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePOP\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.3253\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.23\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.1216\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.5094\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGDPG\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0761\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0037\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.1629\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2402\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.0529\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGGC\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.0332\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.4398\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0476\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.3661\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.0284\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e-0.3315\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eINF\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.269\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.07\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.2823\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.1284\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.0792\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.1263\u003c/p\u003e \u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.034\u003c/p\u003e \u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003c/div\u003e\u003cp\u003e \u003cstrong\u003eSource\u003c/strong\u003e \u003c/p\u003e\u003cp\u003eAuthors’ calculations.\u003c/p\u003e\u003cp\u003eBefore analyzing the regression panel model, a correlation matrix was formed between the dependent and independent variables, and an analysis of Pearson's correlation coefficients was carried out. Namely, we estimate the correlation between selected determinants to check possible problems of multicollinearity between them. We have a multicollinearity problem if the correlation between selected determinants is above 0.80 Gujarati and Porter (2009) and simultaneous inclusion of the variable in the model should be avoided. According to the results from Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e, there are no multicollinearity problems between selected determinants.\u003c/p\u003e"},{"header":"4. Methodology","content":"\u003cp\u003eThis section outlines the econometric method chosen to investigate the relationship between AI and unemployment. Unemployment is typically measured as the number of people without a job over a specific period, such as a month, quarter, or year. It reflects the current state of the labor market by indicating how many individuals are actively seeking work but unable to find it at that time. The dynamic lag effect is crucial in the context of unemployment because economic conditions and policies often take time to influence the labor market. For instance, a government stimulus program might require several months for implementation and for its impact on job creation or retention to become apparent. Similarly, during a recession, the unemployment rate may not immediately reflect economic downturns, as businesses might initially opt to reduce hours or wages rather than lay off workers. Furthermore, even after an economic recovery, the unemployment rate may take time to return to pre-recession levels due to factors like workers leaving the labor force or becoming discouraged during the downturn, which can delay their re-entry into the job market. Therefore, incorporating the dynamic effect when analyzing labor market changes provides a more accurate and comprehensive understanding of unemployment dynamics and contributing factors. To capture these effects, this study employs a dynamic panel data model, which is particularly suited for examining the dynamic nature of unemployment.\u003c/p\u003e \u003cp\u003eBond (2002) emphasizes that dynamic relations in analyzing the base process can be decisive for proper and consistent estimations of parameters of the observed independent variables. Specific characteristics of the sample covering 27 countries over 12 years, or the situation when T\u0026thinsp;\u0026lt;\u0026thinsp;N, is an important argument for choosing a dynamic panel model (Greene, 2008). The dynamic panel model is also a good method of estimation when potential endogeneity is considered, which is the case in our model (Greene, 2008). The dynamic panel model offers the possibility of generating internal instruments (external ones are typically difficult to find), so the treatment of potential endogeneity is comprehensive and the estimations more consistent (Roodman, 2006, 2007; Baum, 2006).\u003c/p\u003e \u003cp\u003eVarious econometric methods are available for estimating dynamic panel data models. The dynamic panel data approach is advantageous as it accounts for the endogeneity of explanatory variables and considers unobservable, time-invariant country effects (Baek, 2016; Yerdelen Tatoglu, 2018). To address the issue of endogeneity, Arellano and Bond (1991) introduced the difference generalized method of moments (difference GMM), which uses instrumental variables to derive GMM estimates of the relevant moment conditions. This method involves taking the first difference of the regression equation to eliminate individual fixed effects, using lagged variables as instruments for endogenous variables in the differenced equation. While this approach effectively mitigates endogeneity concerns, it may suffer from \"weak instruments\" in finite samples, leading to reduced precision (Bond et al., 2001). To overcome this limitation, Arellano and Bover (1995) and Blundell and Bond (1998) proposed the system GMM estimator, which enhances the difference GMM by using lagged variables as instruments in both the differenced and level equations.\u003c/p\u003e\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"586\" height=\"66\"\u003e\u003c/p\u003e \u003cp\u003eWhere the \u003cem\u003eαj\u003c/em\u003e is \u003cem\u003ep\u003c/em\u003e parameters to be estimated, \u003cem\u003eit\u003c/em\u003e is a 1 \u0026times; \u003cem\u003ek\u003c/em\u003e1 vector of strictly exogenous covariates, \u003cem\u003eβ\u003c/em\u003e1 is a \u003cem\u003ek\u003c/em\u003e1 \u0026times; 1 vector of parameters to be estimated, \u003cem\u003eit\u003c/em\u003e is a 1 \u0026times; \u003cem\u003ek\u003c/em\u003e2 vector of predetermined or endogenous covariates,\u003cem\u003eβ\u003c/em\u003e2 is a \u003cem\u003ek\u003c/em\u003e2 \u0026times; 1 vector of parameters to be estimated, \u003cem\u003eνi\u003c/em\u003e are the panel-level effects (which may be correlated with the covariates), and ε\u003cem\u003eit\u003c/em\u003e are \u003cem\u003ei.i. d\u003c/em\u003e. over the whole sample with variance \u003cem\u003eσ\u003c/em\u003e2\u0026isin;. The \u003cem\u003eνi\u003c/em\u003e and the \u0026isin;\u003cem\u003eit\u003c/em\u003e are assumed to be independent for each \u003cem\u003ei\u003c/em\u003e overall \u003cem\u003et\u003c/em\u003e.\u003c/p\u003e \u003cp\u003eThere are several compelling reasons for favoring system GMM over difference GMM in empirical analysis. First, when dealing with variables close to a random walk, lagged levels of the regressors, as utilized in difference GMM, tend to be poor instrumental variables (IVs) for the first-differenced regressors (Arellano and Bover, 1995). Second, unbalanced panels can exacerbate data gaps when employing first differences, a characteristic of difference GMM. Third, the standard errors associated with different GMM are often biased downwards (Blundell and Bond, 1998). Fourth, difference GMM tends to perform inadequately with small sample sizes, whereas system GMM exhibits better properties in such scenarios (Blundell and Bond, 2000). Fifth, system GMM provides additional IVs by incorporating a second equation, wherein variables in levels are instrumented with their first differences (Arellano and Bover, 1995). Sixth, the use of more IVs in system GMM enhances estimation efficiency compared to difference GMM. Seventh, system GMM is better suited for modeling non-stationary data and accommodating predetermined explanatory variables. Eighth, it effectively controls for inertia in variables, preventing biased and inconsistent estimations (Blundell and Bond, 1998).\u003c/p\u003e \u003cp\u003eMost of our System GMM estimations are based on two to three lags of the endogenous variable in our instrument set. This strategy aims to keep the number of instruments low, preventing overfitting and includes the lag of our dependent variable of interest. The latter helps avoid any bias due to the large number of instruments in a relatively small sample (Petkovski et al., 2023). Related to this, we address the downward bias of standard errors in two-step GMM by using the correction proposed by Windmeijer (2005), which is implemented by the \u003cem\u003extabond2\u003c/em\u003e syntax. System GMM has two variants: one-step and two-step estimation methods. The distinction lies in whether the weight matrix is homoscedastic or heteroscedastic. The two-step estimators are often considered more efficient because they reduce bias in the standard errors of estimation values, particularly in small samples (Bond et al., 2001). However, as the number of periods increases, system GMM may generate numerous instrumental variables, which can lead to model overfitting and poor model specification (Roodman, 2009). Thus, a one-step system GMM is recommended for models involving a small number of countries over a longer period, while a two-step system GMM is more suitable for models with a larger number of countries over a shorter period (Teixeira and Queir\u0026oacute;s, 2016). Given that our analysis covers data from 27 EU countries over 11 years, we opt for the two-step system GMM over the one-step approach\u003c/p\u003e \u003cp\u003eTo ensure the consistency of our GMM estimation results, we subject our instruments to two specification tests. Firstly, the application of the Hansen test of over-identifying restrictions provides no evidence to reject the validity of the instruments. Secondly, we employ a second-order autocorrelation test to assess whether our error terms exhibit serial correlation. \u003cem\u003eM1\u003c/em\u003e and \u003cem\u003em2\u003c/em\u003e tests demonstrate a first-order serial correlation in the first-difference equation but detect no evidence of second-order serial correlation.\u003c/p\u003e"},{"header":"5. Empirical Results","content":"\u003cp\u003eThe analysis of the system GMM estimations in Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e3\u003c/span\u003e provides insights into how AI investments, along with other key variables, influence unemployment across different groups of European Union (EU) countries. The estimations are divided into three groups: all EU countries, old EU member states, and new EU member states. This approach allows for a nuanced examination of how AI investments and other economic factors affect unemployment in these regions.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSystem GMM estimations\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEU Countries (all)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eEU Countries (old)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEU Countries (new)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUNP (-1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.935*** (0.018)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.918*** (0.031)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.956*** (0.145)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.051*** (0.022)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.149* (0.089)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.496* (0.396)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEDU\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.001 (0.004)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.019 (0.016)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.136 (0.205)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePOP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.103 (0.141)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.117 (0.173)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.711 (1.156)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTRADE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.008 (0.002)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.003 (0.002)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.004 (0.019)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGDPG\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.359*** (0.062)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.304*** (0.105)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.730*** (0.231)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGGC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.243*** (0.059)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.003 (0.002)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.004 (0.019)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eINF\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.037 (0.028)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.149*** (0.050)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.284 (0.184)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.376*** (0.069)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.109*** (0.045)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.175 (0.124)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConsant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-5.741*** (1.692)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.351*** (0.108)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.742*** (0.260)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of observations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e219\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e135\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e84\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNumber of countries\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArellano-Bond test for AR(1) in first differences\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.083\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.015\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eArellano-Bond test for AR(2) in first differences\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.543\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.471\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.227\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHansen test of the validity of instruments (\u003cem\u003ep\u003c/em\u003e-value)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.711\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.316\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.487\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe results indicate substantial variation in the impact of AI investments on unemployment across the three groups:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003eAll EU Countries: The coefficient for AI investments is 0.051, which is positive and statistically significant. This suggests that an increase in AI investment per million people leads to higher unemployment in the EU as a whole. The positive effect implies that technological advancements, particularly automation driven by AI, are causing short-term disruptions in the labor market, displacing workers in industries susceptible to automation.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eOld EU Countries: The coefficient for AI investments in old member states is 0.149, also positive but significant only at the 10% level. The larger coefficient compared to the overall sample suggests that in advanced economies, where AI technologies are more integrated into traditional industries, the displacement of workers is more pronounced, contributing to higher unemployment.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003eNew EU Countries: In contrast, the AI coefficient for new EU member states is -0.496, which is negative and statistically significant. This finding indicates that AI investments in these countries are associated with a reduction in unemployment. AI may be fostering job creation in new industries, particularly in IT and AI-related sectors, which offsets the displacement effects seen in more technologically mature economies.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eAcross all groups, the lagged unemployment variable is highly significant, with coefficients close to 1, highlighting the persistence of unemployment. The highest persistence is observed in new member states (0.956), indicating slower adjustment in labor markets compared to old member states (0.918), where the slightly lower persistence may reflect more dynamic labor market policies. The effect of school enrollment on unemployment is largely insignificant across the three groups, including in new member states where the coefficient is positive but not statistically significant. This result suggests that secondary school enrollment alone does not have a meaningful impact on unemployment in the short term, possibly due to a mismatch between the education provided and the skills demanded in a rapidly changing labor market. Population growth shows mixed and generally insignificant effects. In the overall EU sample, the coefficient is negative but not significant, while it is positive and insignificant in both old and new member states. These results suggest that population growth does not exert a strong direct influence on unemployment, although it could interact with other economic factors in more complex ways. The results for the trade variable, show positive effect across all three groups of EU countries, but they are not statistically significant, meaning that we cannot confidently conclude that trade has a reliable impact on unemployment in these cases. Nevertheless, the positive coefficients suggest a potential association between increased trade and higher unemployment, although it is important to interpret these results cautiously due to the lack of significance. However, the positive association between trade and unemployment in these results could reflect short-term labor market frictions or structural adjustments as industries adapt to global competition. The fact that the coefficients are not statistically significant suggests that trade's impact on unemployment might be complex and not easily captured in a straightforward positive or negative relationship, especially when considering different dynamics in EU member states. Economic growth is consistently negative and significant in all groups, confirming its expected role in reducing unemployment. The strongest effect is found in new member states (-0.730), where GDP growth has a particularly pronounced impact on lowering unemployment, likely reflecting the ongoing process of economic convergence with older EU members. In old EU countries, the coefficient is smaller (-0.304), possibly due to more stable and developed labor markets where growth translates less directly into job creation. General government consumption has a positive and significant effect on unemployment in the overall EU sample, suggesting that higher government spending may not be effectively targeted at reducing unemployment in the short term. However, the effect is insignificant in both old and new member states, indicating that government consumption may not be a major factor in determining unemployment at the regional level. Inflation has a positive and significant effect on unemployment in old EU countries, suggesting that rising prices may be contributing to unemployment, possibly through mechanisms like cost-push inflation. In new member states, however, the effect is insignificant, pointing to different inflation dynamics in these countries, which are likely still adjusting to the economic integration within the EU. The effect of labor productivity on unemployment is positive and significant in the overall sample, which may seem counterintuitive. One possible explanation is that productivity improvements, often driven by technological advances like AI, lead to job displacement as firms automate production processes. In old EU countries, the coefficient is negative and significant, indicating that productivity gains help reduce unemployment, likely due to better labor market adaptation. In new member states, the positive and insignificant coefficient may reflect the early stages of productivity growth, where the displacement effects of automation are more apparent.\u003c/p\u003e \u003cp\u003eThe contrasting effects of AI investments on unemployment in old and new EU member states can be explained by several factors:\u003c/p\u003e \u003cp\u003e \u003cul\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eStage of Technological Adoption\u003c/b\u003e: Old EU member states are more advanced in integrating AI technologies, leading to higher displacement of routine jobs in sectors like manufacturing and services. This is reflected in the positive AI coefficient (0.149) for these countries, where unemployment rises as firms adopt labor-saving technologies.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eJob Creation vs. Displacement\u003c/b\u003e: In new EU member states, AI investments appear to be generating more jobs, especially in emerging industries such as IT and software development. The negative AI coefficient (-0.496) suggests that job creation outweighs displacement in these countries, where the AI sector is still in a growth phase.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eLabor Market Structure\u003c/b\u003e: Old EU countries generally have more rigid labor markets, with higher wages and stronger protections for workers, making it harder for labor markets to adjust to AI-driven changes. In contrast, new member states have more flexible labor markets, allowing for easier transitions to new industries created by AI investments.\u003c/p\u003e \u003c/li\u003e \u003cli\u003e \u003cp\u003e \u003cb\u003eSectoral Composition\u003c/b\u003e: The economic composition of old and new member states differs, with old EU countries having a larger share of high-tech industries where AI adoption is faster and more disruptive. New EU countries, with a larger share of labor-intensive sectors, experience slower AI adoption, allowing for more gradual transitions and less job displacement.\u003c/p\u003e \u003c/li\u003e \u003c/ul\u003e \u003c/p\u003e \u003cp\u003eThese findings highlight the complex and region-specific effects of AI on unemployment, with the potential for both job creation and job displacement depending on the stage of technological development and the structure of the labor market.\u003c/p\u003e \u003cp\u003eThe conclusion of this study highlights the complex relationship between AI investments and unemployment in the European Union, revealing significant differences between old and new EU member states. For the overall EU sample, the results show a statistically significant positive impact of AI investments on unemployment, with a coefficient of 0.051. This indicates that, on average, increased AI investments are associated with higher unemployment across the EU, aligning with early predictions by Leontief (1983), who theorized that AI and automation could replace jobs, potentially leading to widespread unemployment unless mitigated by government policies.\u003c/p\u003e \u003cp\u003eIn the old EU member states, the effect is even more pronounced, with a positive coefficient of 0.149, suggesting that labor markets in these countries may be experiencing greater disruption due to AI adoption. This supports the view of Zeira (1998), who argued that technological innovations reduce the demand for labor, as well as Hirst (2014), who emphasized the vulnerability of low-skilled workers in the face of rapid technological change. The displacement of routine jobs by AI in advanced economies mirrors findings by Acemoglu and Restrepo (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), who documented job losses and wage declines in regions of the United States following the introduction of industrial robots. Similarly, Stiglitz (2014) argues that the replacement of workers by AI is driven by decisions made by capital owners, particularly in sectors like manufacturing, where automation is most prevalent.\u003c/p\u003e \u003cp\u003eIn contrast, the new EU member states exhibit a markedly different trend, with a negative coefficient of -0.496, indicating that AI investments in these countries are associated with a reduction in unemployment. This suggests that AI is playing a more complementary role in these economies, possibly creating new job opportunities and facilitating economic development. This result contrasts with the more disruptive effects observed in old EU member states and could be explained by the earlier stages of AI integration in these countries. The creation of jobs in new industries, supported by foreign investment in AI technologies, may be mitigating the displacement effects seen in more developed economies.\u003c/p\u003e \u003cp\u003eThese findings are consistent with Roubini\u0026rsquo;s (2014) observation that high-skilled workers tend to benefit more from automation, while low-skilled workers face greater risks of unemployment. In new member states, the complementary role of AI could be fostering job creation in sectors like IT and software development, where high-skilled workers are in demand.\u003c/p\u003e \u003cp\u003eIn conclusion, while AI investments contribute to rising unemployment in the more advanced economies of the old EU member states, they appear to reduce unemployment in the newer member states, where job creation in AI-related sectors offsets the displacement effects. These contrasting outcomes highlight the importance of tailored policy responses to address the divergent impacts of AI across different regions. As McKinsey Global Institute (2017) warned, automation could displace up to 800\u0026nbsp;million jobs globally by 2030, underscoring the need for governments to proactively manage this transition through education, retraining, and redistributive measures.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThe findings of this study reveal a complex relationship between AI investments and unemployment across the 27 EU member states. Overall, the analysis indicates a statistically significant positive effect of AI investments on unemployment in the EU, with a coefficient of 0.051. This suggests that, on a broad scale, increasing AI investments per million people is associated with higher unemployment levels. This outcome highlights the short-term disruptions in the labor market caused by technological advancements, particularly as automation driven by AI displaces workers in vulnerable sectors.\u003c/p\u003e \u003cp\u003eWhen analyzing the old EU member states, the coefficient for AI investments is 0.149, significant at the 10% level. This result indicates that, in these more advanced economies, the integration of AI technologies into established industries leads to more pronounced worker displacement, thereby contributing to higher unemployment rates. The findings suggest that while these economies benefit from technological advancements, the transition may result in significant job losses in sectors susceptible to automation.\u003c/p\u003e \u003cp\u003eIn contrast, the analysis of new EU member states reveals a different narrative. Here, the coefficient for AI investments is -0.496, which is both negative and statistically significant. This indicates that AI investments in these countries are associated with a reduction in unemployment. The findings suggest that in these emerging economies, AI may be facilitating job creation in new industries, particularly within the IT and AI sectors. This positive impact may offset the displacement effects observed in more technologically mature economies.\u003c/p\u003e \u003cp\u003eThe two-step system GMM estimation approach employed in this study effectively addresses potential endogeneity and the dynamic nature of unemployment, enhancing the robustness of the analysis across different country groupings.\u003c/p\u003e \u003cp\u003eWhile the study sheds light on these important dynamics, several limitations should be acknowledged. The dataset covers annual data from 1990 to 2023, with AI investments gaining significance primarily in the latter part of this period. Consequently, the long-term effects of AI investments on unemployment may not be fully captured. Additionally, the heterogeneity among EU member states necessitates more nuanced models to account for differences in digital infrastructure, education, and economic maturity, particularly between old and new member states. Moreover, potential spillover effects from AI investments in neighboring countries are not considered, which could influence unemployment dynamics.\u003c/p\u003e \u003cp\u003eThe results suggest critical implications for policymakers, especially in new EU member states. To maximize the benefits of AI investments, it is essential to improve conditions for AI adoption and integration within the broader economy. Key areas of focus should include investing in digital infrastructure, enhancing education, and providing workforce training to prepare the labor market for AI-driven transformations. For old EU member states, sustaining the benefits of AI requires promoting innovation, supporting entrepreneurship in AI-related sectors, and ensuring that workers are continuously upskilled to adapt to technological changes.\u003c/p\u003e \u003cp\u003eAt the EU level, coordinated efforts are vital to bridge the digital divide between old and new member states. This could involve targeted funding and support for AI research, infrastructure development, and skills training in emerging economies. Additionally, policies that foster cross-border collaboration in AI innovation could facilitate knowledge transfer and enhance the overall impact of AI investments throughout the EU.\u003c/p\u003e \u003cp\u003eIn summary, while AI investments contribute to unemployment trends across the EU, the varying effects highlight the necessity for targeted policies that address regional disparities. Ensuring that all EU countries can leverage the full potential of AI technology will be crucial for fostering sustainable economic growth and technological advancement.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eDeclaration: The research received no specific grant from any funding agency in the public,commercial, or not-for-profit sectors\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.KJ. WROTTE ALL PAPERS\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAcemoglu, D., and Restrepo, P. (2018). Artificial Intelligence, Automation, and Work. \u003cem\u003eNBER Working Paper No. 24196\u003c/em\u003e, National Bureau of Economic Research. https://www.nber.org/papers/w24196\u003c/li\u003e\n\u003cli\u003eAcemoglu, D., and Restrepo, P. (2019). Automation and new tasks: How technology displaces and reinstates labor. \u003cem\u003eJournal of Economic Perspectives\u003c/em\u003e, 33(2), 3-30.\u003c/li\u003e\n\u003cli\u003eAcemoglu, D., and Restrepo, P. (2020). Robots and Jobs: Evidence from US Labor Markets. \u003cem\u003eJournal of Political Economy\u003c/em\u003e, 128(6), 2188-2244. https://doi.org/10.1086/705716\u003c/li\u003e\n\u003cli\u003eAndersen, T. M., and Svarer, M. (2007). Flexicurity\u0026mdash;Labour Market Performance in Denmark. \u003cem\u003eCESifo Economic Studies\u003c/em\u003e, 53(3), 389-429. https://doi.org/10.1093/cesifo/ifm015\u003c/li\u003e\n\u003cli\u003eArntz, M., Gregory, T., and Zierahn, U. (2016). The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis. \u003cem\u003eOECD Social, Employment and Migration Working Papers No. 189\u003c/em\u003e, OECD Publishing. https://doi.org/10.1787/5jlz9h56dvq7-en\u003c/li\u003e\n\u003cli\u003eAuerbach, A. J., and Gorodnichenko, Y. (2012). Measuring the output responses to fiscal policy. \u003cem\u003eAmerican Economic Journal: Economic Policy\u003c/em\u003e, 4(2), 1-27.\u003c/li\u003e\n\u003cli\u003eBaldwin, R. (2019). \u003cem\u003eThe Globotics Upheaval: Globalization, Robotics, and the Future of Work\u003c/em\u003e. Oxford University Press.\u003c/li\u003e\n\u003cli\u003eBall, L., Leigh, D., and Loungani, P. (2013). Okun\u0026rsquo;s law: Fit at fifty? \u003cem\u003eNBER Working Paper No. 18668\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eBall, L., Mankiw, N. G., and Romer, D. (1988). The new Keynesian economics and the output-inflation trade-off. \u003cem\u003eBrookings Papers on Economic Activity\u003c/em\u003e, 1988(1), 1-82.\u003c/li\u003e\n\u003cli\u003eBasu, S., Fernald, J. G., and Kimball, M. S. (2006). Are technology improvements contractionary? \u003cem\u003eAmerican Economic Review\u003c/em\u003e, 96(5), 1418-1448. https://doi.org/10.1257/aer.96.5.1418\u003c/li\u003e\n\u003cli\u003eBecker, G. S. (1993). \u003cem\u003eHuman Capital: A Theoretical and Empirical Analysis, with Special Reference to Education\u003c/em\u003e (3rd ed.). University of Chicago Press.\u003c/li\u003e\n\u003cli\u003eBerg, J., Furrer, M., Harmon, E., Rani, U., and Silberman, M. S. (2018). \u003cem\u003eDigital Labour Platforms and the Future of Work: Towards Decent Work in the Online World\u003c/em\u003e. International Labour Office. https://www.ilo.org/global/publications/books/WCMS_645337/lang--en/index.htm\u003c/li\u003e\n\u003cli\u003eBerger, T., and Frey, C. B. (2016). Regional Technological Dynamism and the Emergence of New Occupations: Implications for Local Labor Markets. \u003cem\u003eJournal of Economic Geography\u003c/em\u003e, 16(6), 1215-1243. https://doi.org/10.1093/jeg/lbv057\u003c/li\u003e\n\u003cli\u003eBessen, J. (2019). AI and Jobs: The Role of Demand. \u003cem\u003eNBER Working Paper No. 24235\u003c/em\u003e, National Bureau of Economic Research. https://www.nber.org/papers/w24235\u003c/li\u003e\n\u003cli\u003eBessen, J., Goos, M., Salomons, A., and Van den Berge, W. (2020). Firm-Level Automation: Evidence from the Netherlands. \u003cem\u003eA Report for the European Commission\u003c/em\u003e. https://data.europa.eu/doi/10.2760/869954\u003c/li\u003e\n\u003cli\u003eBloom, D. E., Canning, D., and Sevilla, J. (2003). \u003cem\u003eThe Demographic Dividend: A New Perspective on the Economic Consequences of Population Change\u003c/em\u003e. Rand Corporation.\u003c/li\u003e\n\u003cli\u003eBowles, J. (2017). The Computerization of European Jobs. \u003cem\u003eBruegel\u003c/em\u003e, July 24th.\u003c/li\u003e\n\u003cli\u003eBregman, R. (2017). \u003cem\u003eUtopia for Realists: How We Can Build the Ideal World\u003c/em\u003e. Little, Brown and Company.\u003c/li\u003e\n\u003cli\u003eBresnahan, T. F., Brynjolfsson, E., and Hitt, L. M. (2002). Information Technology, Workplace Organization, and the Demand for Skilled Labor: Firm-Level Evidence. \u003cem\u003eQuarterly Journal of Economics\u003c/em\u003e, 117(1), 339-376. https://doi.org/10.1162/003355302753399526\u003c/li\u003e\n\u003cli\u003eBrynjolfsson, E., and McAfee, A. (2014). \u003cem\u003eThe Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies\u003c/em\u003e. W. W. Norton and Company.\u003c/li\u003e\n\u003cli\u003eCarbonero, F., Ernst, E., and Weber, E. (2020). Robots worldwide: The impact of automation on employment and trade.\u003c/li\u003e\n\u003cli\u003eChiacchio, F., Petropoulos, G., and Pichler, D. (2018). The Impact of Industrial Robots on EU Employment and Wages: A Local Labor Market Approach. \u003cem\u003eBruegel Working Paper No. 02\u003c/em\u003e, Brussels.\u003c/li\u003e\n\u003cli\u003eChoi, Y., Stewart, S., and Ellingrud, K. (2020). \u003cem\u003eThe Future of Work in America: People and Places, Today and Tomorrow\u003c/em\u003e. McKinsey Global Institute. https://www.mckinsey.com/featured-insights/future-of-work/the-future-of-work-in-america-people-and-places-today-and-tomorrow\u003c/li\u003e\n\u003cli\u003eCowls, J., and Floridi, L. (2018). Prolegomena to a White Paper on an Ethical Framework for a Good AI Society. \u003cem\u003eSSRN Electronic Journal\u003c/em\u003e. https://doi.org/10.2139/ssrn.3288950\u003c/li\u003e\n\u003cli\u003eDe Stefano, V. (2016). The Rise of the \u0026quot;Just-in-time Workforce\u0026quot;: On-demand Work, Crowdwork, and Labor Protection in the \u0026quot;Gig-economy\u0026quot;. \u003cem\u003eComparative Labor Law and Policy Journal\u003c/em\u003e, 37(3), 471-503.\u003c/li\u003e\n\u003cli\u003eDutt, P., Mitra, D., and Ranjan, P. (2009). International Trade and Unemployment: Theory and Cross-National Evidence. \u003cem\u003eJournal of International Economics\u003c/em\u003e, 78(1), 32-44.\u003c/li\u003e\n\u003cli\u003eElkeles, T., and Phillips, J. (2017). \u003cem\u003eThe Chief Learning Officer: Driving Value Within a Changing Organization Through Learning and Development\u003c/em\u003e. Routledge.\u003c/li\u003e\n\u003cli\u003eEuropean Commission. (2020). \u003cem\u003eWhite Paper on Artificial Intelligence: A European Approach to Excellence and Trust\u003c/em\u003e. https://ec.europa.eu/info/publications/white-paper-artificial-intelligence-european-approach-excellence-and-trust_en\u003c/li\u003e\n\u003cli\u003eFelbermayr, G., Prat, J., and Schmerer, H. (2011). Globalisation and Labour Market Outcomes: Wage Bargaining, Search Frictions, and Firm Heterogeneity. \u003cem\u003eJournal of Economic Theory\u003c/em\u003e, 146(1), 39-73.\u003c/li\u003e\n\u003cli\u003eFrank, M. R., Autor, D., Bessen, J. E., Brynjolfsson, E., Cebrian, M., Deming, D. J., and Rahwan, I. (2019). Toward Understanding the Impact of Artificial Intelligence on Labor. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, 116(14), 6531. https://doi.org/10.1073/pnas.1900949116\u003c/li\u003e\n\u003cli\u003eFrey, C. B., and Osborne, M. A. (2017). The Future of Employment: How Susceptible Are Jobs to Computerisation? \u003cem\u003eTechnological Forecasting and Social Change\u003c/em\u003e, 114, 254-280. https://doi.org/10.1016/j.techfore.2016.08.019\u003c/li\u003e\n\u003cli\u003eGallegati, M., Ramsey, J. B., and Semmler, W. (2014). Long-run Waves in Unemployment and Productivity: Evidence from Wavelet Analysis. \u003cem\u003eJournal of Economic Dynamics and Control\u003c/em\u003e, 44, 130-149. https://doi.org/10.1016/j.jedc.2014.03.011\u003c/li\u003e\n\u003cli\u003eGordon, R. J. (2010). Revisiting the Rise and Fall of American Growth. \u003cem\u003eAmerican Economic Review\u003c/em\u003e, 100(2), 71-76. https://doi.org/10.1257/aer.100.2.71\u003c/li\u003e\n\u003cli\u003eHanushek, E. A., and Woessmann, L. (2008). The Role of Cognitive Skills in Economic Development. \u003cem\u003eJournal of Economic Literature\u003c/em\u003e, 46(3), 607-668.\u003c/li\u003e\n\u003cli\u003eHeimberger, P. (2023). Public Debt and Economic Growth in Advanced Economies: A Survey of Recent Empirical Evidence. \u003cem\u003eJournal of Economic Surveys\u003c/em\u003e, 37(1), 1-25. https://doi.org/10.1111/joes.12436\u003c/li\u003e\n\u003cli\u003eHelpman, E., Itskhoki, O., and Redding, S. (2010). Inequality and Unemployment in a Global Economy. \u003cem\u003eEconometrica\u003c/em\u003e, 78(4), 1239-1283.\u003c/li\u003e\n\u003cli\u003eHealy, J., Nicholson, D., and Pekarek, A. (2017). Should We Take the Gig Economy Seriously? \u003cem\u003eLabour and Industry: A Journal of the Social and Economic Relations of Work\u003c/em\u003e, 27(3), 232-248. https://doi.org/10.1080/10301763.2017.1377048\u003c/li\u003e\n\u003cli\u003eJaniak, A. (2006). Does Trade Liberalisation Lead to Unemployment? Theory and Some Evidence. \u003cem\u003eECARES\u003c/em\u003e, Universit\u0026eacute; Libre de Bruxelles.\u003c/li\u003e\n\u003cli\u003eJosten, C., and Lordan, G. (2019). Robots at Work: Automation and the Ageing Workforce. \u003cem\u003eIZA Institute of Labor Economics Discussion Paper No. 12493\u003c/em\u003e. https://www.iza.org/publications/dp/12493/robots-at-work-automation-and-the-ageing-workforce\u003c/li\u003e\n\u003cli\u003eKeynes, J. M. (1936). \u003cem\u003eThe General Theory of Employment, Interest, and Money\u003c/em\u003e. Macmillan, London.\u003c/li\u003e\n\u003cli\u003eKenney, M., and Zysman, J. (2019). Work and Value Creation in the Platform Economy: The Case of Artificial Intelligence. \u003cem\u003eGlobal Perspectives\u003c/em\u003e, 1(1), 10-13. https://doi.org/10.1525/gp.2019.12382\u003c/li\u003e\n\u003cli\u003eLevy, F., and Murnane, R. J. (2003). The Skill Content of Recent Technological Change: An Empirical Exploration. \u003cem\u003eQuarterly Journal of Economics\u003c/em\u003e, 118(4), 1279-1333. https://doi.org/10.1162/003355303322552801\u003c/li\u003e\n\u003cli\u003eLucas, R. E. (1988). On the Mechanics of Economic Development. \u003cem\u003eJournal of Monetary Economics\u003c/em\u003e, 22(1), 3-42. https://doi.org/10.1016/0304-3932(88)90168-7\u003c/li\u003e\n\u003cli\u003eLund, S., Madgavkar, A., Manyika, J., and Smit, S. (2021). \u003cem\u003eThe Future of Work After COVID-19\u003c/em\u003e. McKinsey Global Institute. https://www.mckinsey.com/featured-insights/future-of-work/the-future-of-work-after-covid-19\u003c/li\u003e\n\u003cli\u003eMatusz, S. J. (1996). International Trade, the Division of Labor, and Unemployment. \u003cem\u003eInternational Economic Review\u003c/em\u003e, 37(1), 71-83.\u003c/li\u003e\n\u003cli\u003eMcAfee, A., and Brynjolfsson, E. (2017). \u003cem\u003eMachine, Platform, Crowd: Harnessing Our Digital Future\u003c/em\u003e. W. W. Norton and Company.\u003c/li\u003e\n\u003cli\u003eMokyr, J., Vickers, C., and Ziebarth, N. L. (2015). The History of Technological Anxiety and the Future of Economic Growth: Is This Time Different? \u003cem\u003eJournal of Economic Perspectives\u003c/em\u003e, 29(3), 31-50. https://doi.org/10.1257/jep.29.3.31\u003c/li\u003e\n\u003cli\u003eMoretti, E. (2012). \u003cem\u003eThe New Geography of Jobs\u003c/em\u003e. Houghton Mifflin Harcourt.\u003c/li\u003e\n\u003cli\u003eMoore, M. P., and Ranjan, P. (2005). Globalisation vs Skill-Biased Technological Change: Implications for Unemployment and Wage Inequality. \u003cem\u003eEconomic Journal\u003c/em\u003e, 115(503), 391-422.\u003c/li\u003e\n\u003cli\u003eNeufeind, M., O\u0026apos;Reilly, J., and Ranft, F. (2018). \u003cem\u003eWork in the Digital Age: Challenges of the Fourth Industrial Revolution\u003c/em\u003e. Rowman and Littlefield International.\u003c/li\u003e\n\u003cli\u003eNgai, L. R., and Pissarides, C. A. (2007). Structural Change in a Multisector Model of Growth. \u003cem\u003eAmerican Economic Review\u003c/em\u003e, 97(1), 429-443. https://doi.org/10.1257/aer.97.1.429\u003c/li\u003e\n\u003cli\u003eNordhaus, W. D. (2007). Two Centuries of Productivity Growth in Computing. \u003cem\u003eJournal of Economic History\u003c/em\u003e, 67(1), 185-194. https://doi.org/10.1017/S0022050707000058\u003c/li\u003e\n\u003cli\u003ePiketty, T. (2014). \u003cem\u003eCapital in the Twenty-First Century\u003c/em\u003e. Harvard University Press.\u003c/li\u003e\n\u003cli\u003eRifkin, J. (1995). \u003cem\u003eThe End of Work: The Decline of the Global Labor Force and the Dawn of the Post-Market Era\u003c/em\u003e. Putnam Publishing Group.\u003c/li\u003e\n\u003cli\u003eSachs, J. D., Benzell, S. G., and LaGarda, G. (2015). Robots: Curse or Blessing? A Basic Framework. \u003cem\u003eNBER Working Paper No. 21091\u003c/em\u003e. National Bureau of Economic Research. https://doi.org/10.3386/w21091\u003c/li\u003e\n\u003cli\u003eSchwab, K. (2016). \u003cem\u003eThe Fourth Industrial Revolution\u003c/em\u003e. World Economic Forum.\u003c/li\u003e\n\u003cli\u003eSener, F. (2001). Schumpeterian Unemployment, Trade and Wages. \u003cem\u003eJournal of International Economics\u003c/em\u003e, 54(1), 119-148.\u003c/li\u003e\n\u003cli\u003eSolow, R. M. (1956). A Contribution to the Theory of Economic Growth. \u003cem\u003eQuarterly Journal of Economics\u003c/em\u003e, 70(1), 65-94. https://doi.org/10.2307/1884513\u003c/li\u003e\n\u003cli\u003eSteigum, E. (2019). Unemployment and Automation: Long-Term Implications of Technological Innovation. \u003cem\u003eEconomic Policy\u003c/em\u003e, 34(98), 259-288. https://doi.org/10.1093/epolic/eiz003\u003c/li\u003e\n\u003cli\u003eStiglitz, J. E. (2012). \u003cem\u003eThe Price of Inequality: How Today\u0026apos;s Divided Society Endangers Our Future\u003c/em\u003e. W. W. Norton and Company.\u003c/li\u003e\n\u003cli\u003eSummers, L. H. (2014). Reflections on the \u0026ldquo;New Secular Stagnation Hypothesis.\u0026rdquo; In C. Teulings \u0026amp; R. Baldwin (Eds.), \u003cem\u003eSecular Stagnation: Facts, Causes, and Cures\u003c/em\u003e (pp. 27-40). CEPR Press.\u003c/li\u003e\n\u003cli\u003eVan Reenen, J. (2011). Wage Inequality, Technology and Trade: 21st Century Evidence. \u003cem\u003eLabour Economics\u003c/em\u003e, 18(6), 730-741.\u003c/li\u003e\n\u003cli\u003eWisskirchen, G., Thibault, B., Bormann, U., Muntz, A., Niehaus, G., Soler, G., ... and Schaude, T. (2017). \u003cem\u003eArtificial Intelligence and Robotics and Their Impact on the Workplace\u003c/em\u003e. IBA Global Employment Institute. https://www.ibanet.org/document/Wages\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, and Sweden.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6641180/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6641180/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study examines the impact of venture capital (VC) investments in artificial intelligence (AI) on unemployment rates across 27 EU member states, distinguishing between old and new EU countries. Utilizing annual data from 2012 to 2023, we explore whether AI investments significantly influence unemployment and how these effects vary between advanced economies and those still developing their digital infrastructure. Employing the two-step system Generalized Method of Moments (GMM), we effectively address endogeneity and the dynamic nature of unemployment, making this method well-suited for our panel dataset covering 27 countries over 12 years. Our findings reveal that AI investments correlate with higher unemployment in old EU countries while positively impacting job creation in new EU member states. Based on these results, we recommend targeted policies to enhance AI adoption, improve digital infrastructure, and promote workforce training, particularly in new member states, to optimize the benefits of AI investments and mitigate potential job displacement.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eJEL Codes: O1, O3\u003c/strong\u003e\u003c/p\u003e","manuscriptTitle":"Artificial Intelligence and Its Impact on Unemployment: A Comparative Analysis of Old and New EU Member States","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-18 16:49:44","doi":"10.21203/rs.3.rs-6641180/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"f92cc0d8-3f1a-4929-9543-e0c8d60c54f8","owner":[],"postedDate":"May 18th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-06-08T21:38:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-18 16:49:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6641180","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6641180","identity":"rs-6641180","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}