Artificial Intelligence and Sectoral Job Market Nexus in U.S.A: An Empirical Study

Abstract
Head of the Laboratory for International and Regional Economics, Academic Head of the PhD Programme in Economics, Associate Professor at the Graduate School of Economics and Management Ural Federal University, Russian Federation Artificial Intelligence and Sectoral Job Market Nexus in U.S.A: An Empirical Study The rise of AI has significant implications for job markets. This research aims to measure the impact of artificial intelligence on the U.S. labour market across var ious sectors. To achieve this, a cross-quantilogram approach is employed, analysing daily data fr om February 2, 2020, to June 6, 2024. The findings reveal that AI positively influences the U.S. job markets, particularly in Scientific Research and Development, Software Development, Education, and Culture and Recreation sec tors. Moreover, the time-frequency analysis confirms that the association between AI development and job opportunities in these relevant fields has become more pronounced after the release of ChatGPT. Consequently, this study proposes sig nificant policy implications aimed at balancing the sector-wise job market in the United States. It emphasizes the importance of leveraging human labour while simultaneously promoting industry growth through the application of AI.
Keywords: Artificial intelligence, Sectoral job markets, Employment, Human labour, Cross- quantilogram approach, U.S.A

1. Introduction
Technological progress is a key driver of long-term economic growth and labour productivity, with the evolution of technology creating new opportunities for workflow enhancement and automation. However, the rise of artificial intelligence (AI), capable of self-training, decision- making, and executing tasks tradition ally requiring human intelligence, has sparked concerns about job displacement (Lane et al., 2021). Despite these concerns, AI has limitations, particularly in areas requiring creative or social intelligence, wh ere humans remain essential. AI doesn't replace human labour entirely but aids in addressing work challenges and boosting productivity (Dwivedi et al., 2021). Simultaneously, AI's evolution has created new jobs in the engineering, software, and IT sectors in the U.S (Cockburn et al., 2018). Amid this debate, our study aims to determine whether AI-based automation leads to a decrease or increase in labour participation in U.S. sectoral markets. Driven by several propositions, our investigation delves into the impact of AI technology on the U.S. job market. First, the Fourth Industrial Revolution (4IR) has brought about automation in manufacturing and services, with sectors such as high-tech, automotive assembly, telecommunications, transport and logistics, f inancial services, retail, and healthcare leading the charge in AI adoption (Bessen et al., 2018). Despite AI technologies offering potential benefits, they also disrupt labour markets, posing a considerable threat. In fact, AI technologies are already implemented to perform human tasks in sectors including agriculture (Javaid et al., 2023), transport (Wu et al., 2022), food industry (Esmaeily et al., 2024), manu facturing (Kim et al., 2022; Nti et al., 2022), medicine (Reddy, 2022), and finance (Ahmed et al., 2022). However, the risks of job substitution due to AI technol ogy adoption vary across different sectors, occupations and functionalities. For instance, the introduction of generative AI models like ChatGPT has been shown to adversely affect employment outcomes for knowledge workers, particularly in tasks that are easily automated, such as content creation and editing (Hui et al., 2024). Besides, while some occupations may experience job losses, others might see a transformation rather than complete automation, allowing workers to benefit fr om increased productivity through close interaction with AI technologies (Car bonero et al., 2023). Therefore, while AI holds the potential to enhance productiv ity and create new job opportunities, its adoption also poses significant risks of job substitution, particularly in roles that involve routine and easily automatable tasks. Fr om this perspective our analysis aims to examine the impact of AI on the labour markets, specifically identifying wh ere AI creates new job demand and where it results in substitution. Secondly, countries actively involved in the global AI race, such as the United States, are more susceptible to the effects of AI technology. Specifically, the United States has undoubtedly become a leader in AI development, and technology giants such as Google, Facebook, and Microsoft remain key players in AI research (Rikap, 2024). Besides, Maximize Market Research predicts (2023) that the U.S. AI market will reach an impressive size of U.S.$223 billion by 2029. Although the U.S. stands to gain fr om its leadership in AI, it is essential to consider the potential risks and vulnerabilities associated with this technology. While the U.S. artificial intelligence market was estimated to be U.S.$ 31 billion by 2023, the American labour market is already facing the challenges posed by the rapid development and implementation of artificial intelligence technologies. Therefore, we intend to examine the impact of the development of artificial intelligence on the U.S. labour market. Thirdly, influenced significantly by structural shifts fr om new technologies such as AI, the U.S. labour market is experiencing a transformation in production processes. For instance, contrary to global fears of technological unemployment, post-Great Recession employment in the U.S. manufacturing sector actually saw a boost due to robots (Leigh et al., 2020), and it's been found that increased AI usage can even reduce unemployment (Mutascu, 2021). In a move towards AI advancement, the U.S., in 2020, joined the Global Partnership on Artificial Intel ligence (GPAI) (Artificial Intelligence (AI) - United States Department of State), and enacted the National Artificial Intelligence Initiative Act (NAIIA) in 2021 to ensure its leadership in AI and its integration into all societal and economic sectors (National Artificial Intelligence Initiative Act of 2020, 2020). Despite major U.S. tech firms actively exploiting AI, national adoption is still in its early stages (Beede et al., 2020). Specifically, McElheran et al. (2024) argue that fewer than 6% of U.S. f irms had adopted AI-related technologies, with adoption rates slightly higher when weighted by employment, reaching just over 18%. Ethical concerns, data pri vacy issues, and the potential for job displacement also pose significant obstacles to AI acceptance (Rane et al., 2024). Even with a declining unemployment rate (3.6%) and reduced initial jobless claims in 2023 (U.S. Bureau of Labour Statistics; U.S. Department of Labour), there's a looming threat fr om computerization and automation for about 25% of total employment, or 36 million Americans (Muro et al., 2019). Predictions indicate that up to 70% of their tasks could be automated, with job losses potentially reaching 73 million by 2030 (Manyika et al., 2017), a scenario that could be further amplified by AI and neural networks. Thus, the true impact of AI expansion on the U.S. job market requires further detailed evaluation. Given these motivations, our research aims to evaluate the impact of AI development on various U.S. labour market sectors. Our study contributes to the existing literature along different dimensions. Firstly, it is the first empirical attempt to investigate the early response of the U.S. job market to the rise of AI, using high-frequency data and disaggregating different job market sectors. To this end, we utilize AI-INDEX 15, which captures the performance of U.S. companies engaged in artificial intelligence by designing, creating, integrating or providing AI in the form of product, software or systems. AI-INDEX 15 allows to account for the development of artificial intelligence and can serve as a proxy to capture the interest of population towards the AI technologies. To represent the U.S. labour market, we use percent change of job posting on American worldwide employ ment platform Indeed in the United States by sectors. The Indeed combines job listings fr om thousands of websites, therefore this indicator can globally represent employers' demand for labour in the U.S.. Besides, to conduct a more sophisticated analysis we examine the industry level by grouping U.S. labour market sectors into economic industries, such as Accommodation and Food service, Administration, Cultural and recreational services, Engineering/Technicians, Finance and Bank ing, Personal and other services, Scientific Research and Development, Education and Instruction, Information Design & Documentation and technological indus tries, namely IT Operations and Helpdesk, Software Development. Secondly, our analysis captures extreme conditions and the dynamic impact over time, showing how new job creation in different sectors responds to AI augmentation. The ana lyzed period encompasses the last fourth years fr om February 2, 2020, to June 6, 2024. This time period is marked by significant progress in the field of artificial intelligence technologies. Thirdly, we apply several advanced methods with time series analysis features. For instance, the cross-quantile (CQ) method is employed to ascertain the influence of AI advancement on demand in a particular industry, considering different quantile combinations to evaluate the response of U.S. indus tries to varying degrees of AI development. By using CQ approach we can conclude whether AI generates new job demand or results in substitution in specific indus try. Besides, we apply a time-varying vector autoregression (TVP- VAR) model to study the impact of AI development on U.S. labour market sectors, considering dif ferent phases of AI market development and periods of high economic uncertainty. Fourthly, our findings provide new insights into how AI development influences the U.S. job market and affirm that AI development significantly promotes job opportunities in certain sectors, including Scientific Research and Development, Software Development, Education, and Culture and Recreation. Conversely, AI development has either a negative or no significant association with less relevant sectors like Accommodation, Finance and Banking, Personal Services, and others. Finaly, the study's findings can assist policymakers in formulating pragmatic poli cies that effectively combine human labour with AI implementation across diverse market scenarios. This enables a better understanding of the extent to which AI engagement can contribute to the profit maximization of U.S. industries. The remaining part of the study is organized as follows. Section 2 presents the literature review on exposure of job market to AI technology development. Section 3 provides a description of the research methodology and the data utilized. Sec tion 4 presents the descriptive statistics and empirical results. Section 5 contains a discussion of the results including comparisons with previous studies. Section 6 summarizes the findings of the study and discusses policy implications. Finally, Section 7 emphasizes the limitations of the study and provides further research opportunities.

2. Literature review
The development of artificial intelligence (AI) is significantly reshaping the job market, presenting both challenges and opportunities across various sectors. The impact of AI technologies adoption in different economic sectors is increasingly discussed in scientific literature. One strand of the literature highlights the positive impact of the widespread adoption of AI technologies. It is noted that AI technologies are enhancing productivity and efficiency in such industries as finance, healthcare, and manufacturing (Zhang & Lu, 2021). In the financial sector, AI has been instrumental in automating administrative tasks, thereby improving efficiency and reducing human error, which leads to optimized processes and cost reductions (Brynjolfssonet al., 2019). In healthcare, AI is used to improve diagnostic accuracy and treatment precision, contributing to safer and more effective patient care (Wu et al., 2021). The manufacturing industry has also seen substan tial benefits from AI, particularly through the automation of production processes using robots, which reduces labour costs and increases production efficiency (Wang et al., 2024). Besides, by analyzing the impact of patented AI-related inven tions across 3,500 companies Damioli et al. (2023) observe that AI patents have a positive and significant impact on employment, which suggests that AI innova tion may be beneficial for the workforce. Furthermore, AI's ability to analyse big data allows for real-time monitoring and optimization of operations, which is cru cial in maintaining high productivity levels (Fisher et al., 2023). However, while AI can enhance productivity and efficiency, it also requires careful strategic plan ning and investment in both technology and human resources to ensure success ful implementation (Yang, 2022). Education and training programs are essential to bridge the skills gap and ensure that the workforce can adapt to new AI-driven job requirements (Sorensen et al., 2021). Moreover, the adoption of patent innovation in the field of AI also increases worker productivity (Alderucci et al., 2020; Damioli et al., 2021). Therefore, some occupations can experience transformation rather than complete automation, allowing workers to benefit from productivity gains by using AI technologies in their professional activities. Another body of literature emphasizes that AI technologies may lead to job displacement, especially in positions that involve routine and manual tasks. AI can perform repetitive and predictable tasks, thus making some occupations more sus ceptible to automation (Zarifhonarvar et al., 2024). Automation of routine tasks is a common problem across industries, raising concerns about job stability and the potential for large-scale unemployment. However, the risks of substitution caused by technological innovation, especially the widespread adoption of artificial intelligence, vary in the type and functionality of occupations and sectoral specifics. Therewith, high-skilled occupations, such as those involving cognitive and ana lytical skills, can face pressure and be susceptible to substitute, compared to low skill jobs that require manual labour (Zarifhonarvar, 2024). Besides, Pizzinelli et al. (2023) argue that high-skill jobs, while exposed to AI, often have a higher potential for complementarity, meaning AI can augment rather than replace these roles. Hence, Acemoglu & Restrepo (2019) argue that the effect of automation on the labour market can be represented through the productivity effect, which increases labour demand for non- automated tasks, and the displacement effect, resulting in a reduction of labour share. A number of economic sectors, e.g. healthcare and legal services, face significant risks of substitution due to AI's ability to perform complex tasks such as diagnostics and legal research (Qin et al., 2024). Contrary, industries involving routine manual tasks, such as manufacturing, are also at risk, but the nature of AI's impact may differ, focusing more on automation of repetitive tasks (Fisher et al., 2023) However, while the adoption of AI may displace human labour it simultaneously creates new opportunities, particularly in fields that require specialized skills such as AI engineering, machine learning, data science, and AI regulation (Liu et al., 2024). Liu et al. (2024) also argue that the demand for AI-related skills is grow ing rapidly, with significant increases in job postings for job positions that integrate AI into habitual roles. Besides, the number of sectors characterised by low value added and high labour intensity is expected to decrease. Conversely, technology- intensive and knowledge-based sectors are expected to develop, leading to a shift in employment patterns (Liang & Tan, 2024). Tolan et al. (2021) argues that some occupations that had been less affected by previous waves of computerization and automation may now be in a risky position. However, the impact of AI does not necessarily mean automation, but rather a restructuring of some tasks (Tolan et al., 2021). The most related to our study works also investigate the effects of the devel opment of AI technology on the labour market in terms of online job vacancies (Acemoglu et al., 2022; Goldfarb et al., 2020; Lyu et al., 2021). These studies were conducted on the basis of the U.S. vacancy database provided by Burning Glass Technologies (BGT). The data set consists of job title, firm name, required level of education, work experience and skills. Lyu & Liu (2021) use BGT data to identify the most widely adopted digital technologies from 2010 to 2019 in the U.S. energy sector. Their results show that the share of jobs requiring AI and robotics skills is highest among new digital technologies. Goldfarb et al. (2020) classify posted vacancies in the field of artificial intelligence for hospital jobs between 2015 and 2018 to reveal the level of adoption of AI in healthcare. Acemoglu et al. (2022) report that the share of AI vacancies in the U.S. labour market has increased rapidly over the past decade across sectors, with an especially pronounced growth in sec tors like information, professional services, finance, and manufacturing. Moreover, U.S. companies have significantly increased their AI-related postings, which indi cates a strong link between AI exposure and the demand for AI skills. Nevertheless, despite the reduction in non-AI vacancies, the overall impact on labour force at the industry level remains undetectable, indicating that AI's effects might not be widespread across the whole labour market up to 2018. Despite the diversity of existing literature, a precise answer on how AI has affected the labour market across sectors and how this impact has changed over time remains to be found. Furthermore, the previous literature has mainly focused on retrospective data in order to predict the impact of AI on the labour force. There fore, our study extends the existing literature in several ways. For example, related studies mainly focus on the level of AI adoption by U.S. market and mainly discuss the susceptibility of occupations to the impact of the AI technologies by analyz ing the functional characteristics of occupations, while a direct assessment of the impact of AI development on labour demand across different economic sectors has been overlooked. Our research aims to address this gap by conducting a com prehensive exploration of the role of AI in the sectoral marketing landscapes of the United States. We specifically aim to identify the suitability of AI applications for different job sectors using a meticulous and methodical approach. Since the global adoption of AI is still at an early stage, from this point of view we consider the anal ysis of the impact of AI technology on the labour market over the period of rapid widespread adoption of AI technologies from 2020 to 2024 by using high-fre quency and up-to-date data. This allows us to make a more detailed assessment of the impact of AI over time and by level of AI development. Tolan et al. (2021) ranked occupations in terms of their potential to be affected by AI. Although Tolan et al. (2021) present a ranked from 0 to 1 mathematically calculated values for specific occupation types according to their exposure to artificial intelligence, we specify their classification simplifying it to three groups accord ing to ‘high’, ‘medium’ and ‘low’ degrees of exposure to artificial intelligence. We use the numerical values for different types of occupations reported in Tolan et al. (2021), grouping them according to their belonging to the industries represented in our study. Values less than 0.33 fall into the group of ‘low’ propensity to be affected by AI, between 0.33 and 0.66 into the group with ‘medium’ AI impact on sector performance, and finally between 0.66 and above into the group characterized by ‘high’ AI market pressure on the U.S. labour market. 

3. Methodology and data description
3.1 Artificial intelligence (AI) usage in the labour markets
Artificial intelligence (AI) has been integrated into various sectors of the labour market, each with its unique applications and impacts. First, in the infor mation design, IT operations and helpdesk and public administration, AI is used to automate tasks, analyze large volumes of data, and improve decision-making pro cesses. It also necessitates new skills to overcome the dynamic challenges posed by AI in these sectors (Kuziemski & Misuraca, 2020). Secondly, the healthcare, edu cation, and social services sectors are less likely to be automated as they require more human interaction and empathy. AI in these sectors is used to enhance ser vice delivery, such as AI-assisted diagnosis in healthcare or personalized learn ing in education (Kerasidou, 2020). Third, in the construction, engineering, and technicians’ sectors, AI is used to address operational and productivity challenges, safety concerns, labour shortages, and cost and schedule overruns. For instance, AI can help in planning and designing construction projects, predicting potential structural issues, and automating routine tasks (Abioye et al., 2021). Fourth, the impact of AI in the trade sector can vary depending on market conditions. AI can help businesses optimize their operations and reduce costs. However, the integra tion of AI might lead to job displacement and other challenges (Ernst et al., 2019). Fifth, in the task-based roles, AI has a significant impact on task-based roles, wh ere tasks can be automated or assisted by AI. This includes roles in manufacturing, logistics, and other sectors wh ere routine tasks are prevalent (Cheng et al., 2019). Finally, in the entertainment industry, AI is becoming increasingly adept at imitating human work, raising questions about the rights and roles of human actors, writers, and other creators (Han, 2021). Moreover, the impact of AI on the labour market includes changes in employment and wages, transformation of jobs and skill needs, and alterations to the work environment. The adoption and integration of AI into the workplace can lead to increased productivity, new job creation, and economic growth, but it also presents challenges such as potential job losses and rising inequalities.
3.2 Data and Sources
This study employs high-frequency data (5 days a week) fr om February 2, 2020 to June 6, 2024 to evaluate the impact of artificial intelligence development on the U.S. labour market. Specifically, to capture the progress of AI technology we utilize AI-INDEX 15, which tracks the performance of U.S. companies operating in the artificial intelligence. The AI-INDEX 15 includes 15 companies that design, create, integrate or provide artificial intelligence (AI) in the form of prod ucts, software or systems and is calculated by assigning weights to shares of these companies. The rapid development of AI in the last few years may have diminished the creative and intellectual skills of the population, since AI is capable of data analysis, idea creation, problem solving, etc. Thus AI-INDEX 15 allows to account for the development of artificial intelligence and can serve as a proxy to capture the interest of population towards the AI technologies. Moreover, to represent the U.S. labour market, we rely on the Indeed Job Postings Index provided by the Indeed Hiring Lab. We focus on seasonally adjusted sector-level daily data specific to the United States. The analyzed period encompasses the last fourth years, which have witnessed significant advancements in artificial intelligence technology.
3.2 Econometric methods
Our investigation utilizes two state-of-the-art econometric analysis procedures, namely the cross- quantilogram and TVP-VAR Model with stochastic volatility approaches. These are suitable for investigating the role of AI in the U.S. labour market due to their ability to capture dynamic relationships and directional predictability across different time horizons and market conditions.
3.2.1 Cross-quantilogram (CQ) dependency approach
The CQ approach is particularly adept at measuring the cross-quantile dependence across time series without any moment condition requirement. In this way, this technique is useful for identifying the dependence between AI development and labour market variables across different quantiles of each variable's distribution. The CQ approach proposed by Han et al. (2016) enables the assessment of a non-linear bivariate dependence between the development of AI technology and the U.S. labour market, considering different conditional levels of quantiles.
3.2.2 Time-Varying Vector Autoregression model TVP-VAR Model with stochastic volatility
The TVP-VAR approach facilitates the examination of dynamic connections across an array of time horizons, encompassing short-, medium-, and long-term perspectives. This method is useful when investigating the impact of AI develop ment volatility on the U.S. labour market. The Time- Varying Vector Autoregression model (TVP-VAR) devised by Nakajima et al. (2011) considers that the variance of a random process is itself randomly distributed, i.e., it has stochastic volatility distribution. Therefore, the TVP-VAR approach is enforced to measure the degree and sign of volatility transmission between two variables over time.

4. Empirical results
4.1 Descriptive statistics
Table 4 provides the descriptive statistics for demand on labour force among U.S. economic industries constructed using the Indeed Job Postings Index data as well as for Artificial intelligence index 15, which represents the state of AI develop ment. We conduct Skewness (Skew), Kurtosis, and Jarque-Bera (J-B) tests to exam ine the statistical properties of our variables. The results indicate that all variables exhibit non-normal distributions, except for AI-INDEX 15. To ensure the station arity of our variables, we employ Augmented Dickey-Fuller (ADF) Unit Root Tests and confirm that all variables are stationary in the first difference.  Furthermore, we utilize the slope heterogeneity test proposed by Pesaran & Yamagata (2008) to assess the heterogeneity of slope coefficients. The results, represented in Appendix A, indicate that the null hypothesis (H0) can be rejected based on the p-value, suggesting that the sectors respond differently to AI in general. However, it should be noted that we assume homogeneity of the sector at low levels of AI adoption. In the case of high levels of AI adoption, the effect may be different. Figure 2 depicts the evolution of stock price index of companies engaged in the field of artificial intelligence technologies development, which can stand as an appropriate benchmark for assessing the development of artificial intelligence and the growing interest of the population towards AI. Given the weekly specificity of the data, the AI index experiences a number of prominent periods of ups and downs over the 4-year period, beginning with the Covid-19 pandemic.  Thus, the stock prices of AI companies enjoyed rapid growth in 2020; the composite AI-INDEX reached a high of more than 3000 basis points. Although the index was more volatile in 2021 the consistent growth illustrates the undisputed leadership of AI companies in the stock market due to the potential of AI-driven business models. Amid heightened uncertainty in the global financial systems in 2022, the subsequent rise in inflation has led to higher key interest rates, which have particularly affected venture capital-dependent technology stocks and resulted in a sharp fall in the AI-INDEX 15. Having recovered fr om the blows of the coronavirus pandemic and geopolitical risks, markets have been able to recover on the back of a technological upswing and investor optimism regarding the devel opment of AI systems and software applications, including for example ChatGpt. The explosive development of artificial intelligence technologies and its integra tion into routine and professional activities has made the potential of AI tangible for everyone. The United States labour market is represented in our study through 11 specific economic sectors, which are differently exposed to the impact of artificial intelli gence. Meanwhile, the demand for labour in these economic sectors varies due to distinct industry structures, the degree of labour intensity of production, the pro ficiency level of employees, the economic environment, the state tax policy, etc. Besides, the subject of artificial intelligence continues to dominate in the econ omy, revealing more and more frontiers for its widespread adoption. However, the question regarding the actual impact of artificial intelligence on the demand for labour resources in different sectors of the economy remains open. Figure 3 illustrates the evolution of job posting indices in the U.S. labour market by economic sector. We observe a certain resemblance in the dynamics of indicators specifically in the period of labour constraints amid the coronavirus pandemic crisis and the subsequent surge in labour demand in 2021 post-crisis time. Interestingly, having reached a certain peak in job openings growth in late 2021 and early 2022, some of the sectoral indices experienced a strongly down ward trend, another on the contrary a steady rise, while the performance of the rest almost returned to pre-crisis levels. For instance, Information Design & Documen tation, IT Operations and Helpdesk, and Software Development show the sharpest decline during 2023-2024, which can be attributed to the promotion and applica tion of artificial intelligence systems and reduction in labour intensity. Conversely, such economic industries as Engineering/Technicians, Education demonstrated a steady rise in labour demand in 2023-2024 compared to the pre-pandemic Covid- 19 period. The remaining part of the sectors demonstrated more resistant to labour market turbulence dynamics, having returned to the pre-crisis level. However, without empirical evaluation we can only hypothesize sectoral differences in the response of U.S. labour market sectors to the development and adoption of AI technologies, according to their potential differences in susceptibility to substitution by AI technologies. Therefore, the following section presents a detailed assessment of the impact of artificial intelligence on the U.S. labour market across various sectors using econometric modeling.
3.2 Quantile Response of U.S. sectoral labour market to the artificial intel ligence (AI) development
This section explores the interconnection between the U.S. labour market and the development of the artificial intelligence market using the cross-quantilogram approach. The results obtained are visualized through heat map matrices displayed in Figures 4-5. In these heat maps, the vertical axis represents the values of the AI index, while the horizontal axis corresponds to the quantile distribution associated with each sector of the labour market. The color scale employed, ranging from blue to red, indicates the magnitude and direction of the response between the variables. Blue represents negative values of cross-quantile dependence, while red signifies positive values. We set the lag number at 66, which effectively captures the response of the U.S. labour market to changes in the artificial intelligence market after a single quarter. Results with fewer lags were found to be statistically insignificant, which is logical considering the labour market's limited ability to swiftly and noticeably react. Notably, positive reactions are observed across all sectors of the economy at the low quantiles of the AI-INDEX 15. However, the previously conducted slope het erogeneity test confirmed the variation in economic sectors' responses to artificial intelligence. Consequently, our focus primarily lies on reactions occurring at high levels of artificial intelligence, thus considering the impact of advanced technologi cal development on different sectors of the U.S. labour market. Figures 4-5 demonstrate the quantile response of labour market conditions to changes in the artificial intelligence industry. To provide a better perception of the results, the obtained heatmap matrices are organized according to the degree to which the labour market are related to the AI industry. Therefore, Figure 4 show cases the outcomes derived from the cross-quantilogram approach applied to the relevant job sectors. The findings reveal a critical landscape within the U.S. job market, as it encompasses both positive and negative responses to AI technologies. We observe a strong positive reaction at low and middle quan tiles (0.2-0.65 quantiles) of the AI index and IT-related sectors, including Soft ware Development and Scientific Research and Development. This response is attributed to high interest in the labour force in this sector during the time of moderate development of AI, which required large recruitment of specialists for the development of IT, including the creation of AI-based applications and ser vices, and the integration of AI into various areas of human activity. Conversely, our results reveal a slightly negative response in the context of IT Operations and Helpdesk and Scientific Research and Development. This discovery paints a crit ical picture of the U.S. sectoral landscape wh ere the influence of AI applications is significant. Additionally, among such industries as Information Design & Doc umentation and Scientific Research and Development exhibit a strong negative reaction on the high quantiles of both indicators, implying that the development of AI has lessened the demand for human resources in these sectors. However, in the areas of information, design and documentation, and software development, AI  applications appear to foster a beneficial equilibrium between human labour and AI operations (Acemoglu & Restrepo, 2018), thereby contributing to the continued growth and development of U.S. tech industries. Intriguingly, sectors with elevated communication requirements, such as Cul tural and recreational services, Education and Instruction exhibit a modest positive response to the advancements in AI development. Consequently, these sectors, characterized by a lower level of exposure to AI, can be classified as having a higher positive impact of AI. Besides, a weak positive response to high values of the AI index is observed in such sectors as Engineering/Technicians, wh ere the workflow requires a high level of professional skills in both cognitive and manual tasks. The rapid integration of AI into production and technological processes increases the demand for specialists capable of skillfully using AI technologies to automate work projects and increase productivity. Figure 5 unveils the outcomes derived fr om the cross-quantile dependency analysis for sectors characterized by low relevance to AI. We observe that when the market for AI technologies is in a bearish condition (low quantiles), the sectors dis played in Figure 5 exhibit moderately positive reactions. Of particular significance is the notable negative response, denoted by a significance sign (*), observed in the Accommodation and Food services, Personal and other services. The adoption of artificial intelligence in daily life simplifies the process of using accommodation, food services and personal services, which determines the negative response of Accommodation and Food services and Personal and other services sectoral indi ces at high quantiles of the artificial intelligence index (0.8-0.95 quantiles). Inter estingly, we find a lack of response in Administration and Finance and Banking of the U.S. labour market. Consequently, this sector can be classified as having a comparatively low impact of AI at the current stage. This could be due to the com plex nature of these sectors, wh ere human expertise and personal touch are still crucial and cannot be fully replaced by AI. It might also be a result of these sectors being in the early stages of AI adoption, with the full impact of AI yet to be realized. These findings underscore the multifaceted impact of AI across different sectors, highlighting the need for tailored strategies to harness the benefits of AI while mitigating its potential negative effects.  The impact of AI development on different sectors of the U.S. labour market at different quantiles can also be estimated by using cross-quantilogram based on rolling-window technique. Setting the size of the rolling window equal to 252 days (1 year) we study the time-varying linkage between the two timeseries to make a robustness check for our main results. The results of rolling- window cross-quanti logram are presented in Appendix B. We provide results on the dynamic impact of AI development on shifts in demand in the U.S. labour market by considering this relationship at the middle quartiles, i.e., given that Tau is equal to 0.5. The blue line indicates how the level online postings respond to the AI development over time, while the red lines mark the borders of the 90% confidence interval. Therefore, we observe a positive shift in labour demand in industries such as Scientific Research and Software Development and Education especially fr om 2022 amid the rapid development of AI technologies. Service-oriented industries exhibit a smooth decline in response to AI, highlighting their inability to be replaced at the current stage. The obtained results are consistent with prior findings.
3.3 Dynamic Response of U.S. sectoral labour market to the artificial intel ligence (AI) development by using TVP-VAR Model with Stochastic Volatility
To estimate the impact of AI development on U.S. labour market sectors, considering different phases of AI market development and periods of high eco nomic uncertainty, including the global lockdown during the COVID-19 pandemic, we apply a time-varying vector autoregression (TVP- VAR) model. This model uses rolling windows to capture the time dynamics and stochastic volatility of the parameters. A Markov Chain Monte Carlo (MCMC) simulation technique with 1,000 samples is employed to evaluate the model. The results of quantile dependence provide a general overview of the labour market's exposure to AI, while the TVP-VAR approach measures the variation of this response over time. Figures 6-7 display the impulse response functions of sectoral new job posting indices in response to changes in the AI market. During the COVID-19 pandemic, although the demand for labour in all sectors of the U.S. labour market plunged, this may not be directly associated with the low level of AI market development. However, almost all economic sectors experienced rela tively small declines in industry-specific new job postings indices, driven by the AI industry's development in early 2021. The impulse response functions depicted in Figure 6 indicate a strong surge in labour demand in such AI related sectors as Information Design & Documentation and Software Development, wh ere the creation of GhatGPT and enhancements to other machine learning algorithms appear to have boosted the demand for spe cialists capable of using AI algorithms in order to automate workflows, as well as being able to enhance existing technologies. Conversely, a reverse dynamic is evi dent in such sectors as IT Operations and Helpdesk and Scientific Research and Development, implying that the development of AI has lessened the demand for human resources in these sectors. Moreover, as for the Education and Instruction sector, we observe a negative response during the period of economic slowdown in the U.S. economy in 2022, while a positive response is evident during 2023-2024, indicating that AI stimulates new openings in the education market due to the integration of AI and machine learning technologies into the educational process. AI technologies like ChatGPT have also been used to facilitate automated con versations and generate human-like text, which has broad implications for teach ing and learning (Jarrahi, 2018). Similar trends can be observed in the Engineering related market, wh ere the rapid integration of AI into production and technological processes increases the demand for specialists capable of skillfully using AI tech nologies to automate work projects and increase productivity. Figure 7 illustrates the impulse responses of low relevant job sectors to changes in the AI market. Among sectors with potentially medium exposure to AI, includ ing Accommodation and Food services, Administration, Finance and Personal and other services, a strong negative reaction is observed since early 2023, which cor responds to a period of rapid development and widespread use of AI technologies. Thus, these U.S. labour sectors, wh ere the adoption of artificial intelligence greatly simplifies workflows, tend to have a negative trend of decreasing labour demand as a result of AI development.

4. Discussion
Our research provides a comprehensive analysis of the impact of AI develop ment on various sectors of the U.S. labour market, revealing several crucial insights that have significant implications for both policymakers and industry stakehold ers. Therefore, our findings indicate that the integration of AI technologies can on the one hand yields a reduction in human employment levels, and on the other hand have the opposite effect, wherein positive influences of AI development are observed in enhancing labour potentials, irrespective of the market conditions. Based on obtained results, we introduce a classification of U.S. labour market sec tors according to their empirically validated exposure to AI development. We compare the findings on the impact of artificial intelligence on the U.S. labour market by sector, comparing with the classification presented in the study by Tolan et al. (2021). Our results suggest the varying degrees and directions of sectroral job indices exposure to the AI market, including both positive and negative reactions, as well as no meaningful response at all. Overall, our results are consistent with the f indings of Tolan et al. (2021) but highlight the direction of AI's impact on the U.S. labour market.  This study builds on and extends the existing literature on the effects of AI and other technological developments on social and economic processes. For instance, Webb (2019) predicts the impact of AI technologies on occupations by analyzing job task descriptions and patents. Webb's assertion that AI handles tasks involving judgment and decision optimization, distinct fr om previous technologies, is con sistent with our finding that AI significantly promotes job opportunities in sectors like Scientific Research and Development, Software Development, Education. These sectors benefit fr om AI's ability to enhance productivity and innovation, supporting the notion that AI complements rather than replaces human labour in these fields. The widespread incorporation of AI across industries has concurrently catalyzed new opportunities for human labour within various markets' production processes (Tschang & Almirall, 2021). Fr om this perspective our findings are consistent with Liu et al. (2024) and Liang & Tan (2024) that AI development creates new opportunities in sectors that require specialized skills of AI engineering, machine learning, data science, and AI regulation. Thus, we confirm hypothesis 1 and state that Software Development and Scientific Research and Development industries respond positively to the AI development due to the increased demand for highly skilled labour force. Our study also aligns with Alderucci et al. (2020) and Damioli et al. (2021), who found that AI-related patent innovations increase worker productivity. By using high-frequency data, our research captures the dynamic impact of AI over time, reinforcing the idea that AI can enhance job creation in certain sectors. Addi tionally, our research extends the seminal work of Autor et al. (2003) and Brynjolfs son & McAfee (2012), who distinguish between cognitive and manual tasks, and routine and non-routine tasks. While AI can substitute routine tasks, our findings show that sectors that require predominantly non-routine tasks, especially those requiring creativity and complex decision-making, benefit from AI augmentation. In particular, our results indicate the positive impact of AI development in educa tion and engineering sectors by increasing efficiency and productivity, which is in line with Damioli et al. (2023). Moreover, in the education sector, AI technologies like ChatGPT have been used to facilitate automated conversations and generate human-like text, which has broad implications for teaching and learning (Jarrahi, 2018). Thus, we also confirm the second hypothesis. Hence, such industries as Engineering/Technicians and Education and Instruction tend to increase demand for labour force due to the ability of workers to use AI technologies to increase per sonal efficiency and labour productivity. Our findings also reveal a negative response of a number of sectors to the AI development, indicating the probability of displacement, especially in such sec tors as IT Operations and Helpdesk and Information Design & Documentation. This can be attributed to the pronounced role of AI in displacing human labour, wh ere human workers face challenges in competing with the efficiency and capabilities offered by AI-powered operations, which is in line with Minevich (2023). Moreover, our findings on the negative impact on these sectors are consistent with Huang and Rust (2018), who argue that the disruption caused by AI might lead to job displacement. This highlights the need for targeted policies to support work force transitions in these vulnerable sectors, as AI-driven automation may lead to job displacement without corresponding new job opportunities. Therefore, we also confirm the third hypothesis related to the negative impact of AI development on the industries that involve the use of cognitive skills and routine tasks. Our results also highlight the positive response of demand for working force to AI development among service-related sector as Cultural and recreational ser vices. Wh ere the exposure to AI is relatively lower, the integration of AI technologies complements human labour by fostering collabouration and enhancing human skills, thereby establishing a working environment that synergistically combines the efforts of human labour and AI technologies in the United States (Tschang & Almirall, 2021). Based on this we can’t accept the fourth hypothesis and argue that service-related industries are also susceptible to the impact of AI development. Our study is also closely related to the works of Acemoglu et al. (2022), Gold farb et al. (2020), and Lyu et al. (2021), who investigate the effects of AI on the labour market through online job vacancies. Unlike these studies, which primarily focus on the level of AI adoption, our research directly assesses the impact of AI development on job creation across different sectors. By using high frequency and up-to-date data, our study provides a more detailed and timely assessment of AI's impact, addressing the gap in the literature regarding the sector-specific effects of AI over time. In general, our research contributes to the existing body of knowledge by providing a nuanced understanding of how AI development influences the U.S. labour market. By disaggregating job market sectors and using high-frequency data, we offer valuable insights into nonlinear impacts of AI in terms of the degree of development of technology and economic sectors. This is further supported by Nguyen & Vo (2022), who document a non-linear relationship between AI and unemployment, suggesting that AI's impact varies with economic conditions, which our analysis also confirms. Our findings can inform targeted policy interventions, promote economic equity, and support workforce resilience in the face of rapid technological change. 

5. Conclusion and policy implications
The exponential rise in AI adoption within the labour market has ignited a pro found and ongoing debate surrounding the implications of AI implementation for the human labour force. Some proponents argue that the integration of AI leads to a reduction in human labour, while others assert that it enhances their capabilities by seam lessly integrating advanced technological operations, especially within the realm of marketing. Amid this debate, this analysis delves into the relationship between AI and the U.S. sectoral job markets, employing a cross-quantilogram approach and utilizing daily data spanning fr om February 2, 2020, to June 6, 2024. The findings of this study remain robust across the TVP-VAR model with stochastic volatility. Remarkably, the findings of this study shine light predominantly on the positive cross- quantilogram dependency between AI adoption and U.S. sectoral labour mar kets, particularly in sectors with higher, medium, and lower exposure to AI, regardless of prevailing market conditions. This suggests that the application of AI in these sectors is yet to complement human labour, resulting in the presence of collabourations between AI and human workers. Conversely, it is observed that all levels (higher, medium, and lower) of AI involvement in U.S. job markets contribute to a slight reduction in job opportunities for human labour, as demonstrated by the negative cross-quantilogram dependency between AI adoption and U.S. sectoral labour markets. This phenome non can be attributed to the replacement of human workers by AI implementation in intricate technological processes that require sophisticated handling. Industries grap pling with high-volume computational processes and complex algorithmic tasks often f ind human labour insufficient to cope with the demands, leading to the displacement of human workers in favour of AI systems. However, it is crucial to recognize that the impact of AI on employment is not solely determined by the technology itself. Rather, it is shaped by a multitude of factors including industry dynamics, workforce skills, and policy frameworks. Based on these findings, this research provides some policy implications. First, policymakers should focus on increasing investments in sectors wh ere AI development has been shown to significantly promote job opportunities, such as Scientific Research and Development, Software Development, Education, and Culture and Recreation. This could include funding for AI research, subsidies for companies innovating with AI, and training programs to prepare workers for jobs in these fields. Secondly, for sec tors wh ere AI development has a negative or negligible impact on job creation, such as Accommodation, Finance, and Personal Services, policymakers should implement support programs to aid workers transitioning to more AI-resilient jobs. This might include retraining initiatives, unemployment benefits, and incentives for companies to diversify their business models to integrate AI more effectively. Thirdly, reform educa tional curricula to include AI-related skills and promote lifelong learning programs to help workers continually adapt to technological changes. This includes integrating AI and data science courses in higher education and providing continuous professional development opportunities. Finally, ensure that AI development and implementation strategies are inclusive and consider the socio-economic disparities across different regions and communities. Policies should aim to mitigate the risks of job displacement and ensure that the benefits of AI are widely distributed, promoting economic equity. 

6. Limitations and further research opportunities
Like any academic endeavour, this study is not without its limitations. Firstly, the primary analytical framework of the study hinges on a cross-quantilogram-based dependency analysis, which diverges from a causal inference approach. This methodological choice may lim it the depth of conclusions drawn regarding cause-ef fect relationships between AI adoption and labour market outcomes. Secondly, the geographical scope of the study is confined to the United States, representing one of the high robot density countries. However, the study does not extend its analy sis to other countries with significant robot density such as Singapore, Japan, and South Korea. While this focus allows for an in-depth exploration of the U.S. con text, it may lim it the applicability of the findings to other national contexts. Given these limitations, there are several promising avenues for further research. Future studies could benefit from the implementation of causal inferen tial techniques to deepen our understanding of the causal relationships between AI adoption and labour market outcomes. Additionally, adopting a cross-sectional study design under a comparative analysis framework could enhance the general izability of the findings. This would allow for a broader exploration of AI's impact on labour markets across different national contexts, further enriching the discourse on this critical topic.