Does artificial intelligence improve or reduce the level of regional savings? Evidence from China

Does artificial intelligence improve or reduce the level of regional savings? Evidence from China | 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 Does artificial intelligence improve or reduce the level of regional savings? Evidence from China Hongwei Zhang, Xiaohui Luo, Xiaohui Chen This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6702658/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 Artificial intelligence(AI) is causing concern among many who fear becoming increasingly redundant, prompting them to save more as a precaution. This paper conducts a qualitative analysis based on theories such as precautionary saving motives. Subsequently, employing a city and year two-way fixed effects model, a qualitative analysis was conducted using data from Chinese cities from 2008 to 2020. The results show, first, that there is an 'inverted U' nonlinear relationship between AI and regional savings levels. Second, AI promotes the development of the digital economy, which in turn exhibits an 'inverted U' nonlinear relationship with regional savings levels. Third, the impact of AI on regional savings levels is heterogeneous: in first-tier cities, the critical inflection point of AI development occurs at a lower level. This study provides new evidence on the economic impact of AI and serves as a reference for government departments to formulate policies that promote economic development. JEL: O33, E21, R11 Artificial intelligence regional savings digital economy digital entrepreneurship employment level Figures Figure 1 Figure 2 1. Introduction Savings are an important source of investment funds, and increasing savings capacity can enhance the role of investment in promoting economic growth (Makori et al., 2022 ) China, the world's largest developing country, has maintained high economic growth for more than three decades since the reform and opening up, and this growth has been mainly driven by investment. Investment comes from savings, and China's high savings over the past three decades have provided an endless source of investment funds for China's high economic growth (Zhao et al., 2017 ). In recent years, China's savings have continued to rise (Chen et al., 2019 ; Baker et al., 2023 ). Can this continued increase be sustained? Over the past decade, artificial intelligence (AI) has developed rapidly with the development of related technologies such as machine learning, robotics, and neural networks (Yang, 2022 ). With the disappearance of the demographic dividend due to aging and the need for high-quality economic development under the new normal, the application of AI, such as industrial robots, has grown rapidly in the Chinese market (Han, 2022 ). AI has changed almost every aspect of people's lives (Haseeb et al., 2019 ) and profoundly affected their behavior. Saving is an indispensable and important behavior in people's lives. China is a vast country, and regions rely on local savings to drive their economies. How exactly will AI affect regional savings in China? This is the topic of this paper. AI makes capital substitutable for labor (Acemoglu & Restrepo, 2019 ), creating new jobs while eliminating existing ones (Acemoglu et al., 2018; Acemoglu & Restrepo, 2019 ). This raises concerns that labor will become increasingly redundant (Acemoglu et al., 2018; Acemoglu et al., 2020 ), thereby exposing people to uncertainty. In response to future uncertainty and precautionary motives, people increase their extra savings (Leland, 1968 ). Digital technologies such as AI create many entrepreneurial opportunities, leading to the emergence of digital entrepreneurship (Nambisan, 2017 ), which may increase the consumption of people's savings. Concerns arising from AI are always present, and digital entrepreneurship requires many conditions. Therefore, we arrive at the research hypothesis that AI first increases the level of regional savings and then gradually reduces the level of regional savings. In addition, digital technologies such as AI and big data drive the digital economy (Chen et al., 2022 ; Chen et al., 2022 ). The digital economy can distribute digital information without restriction (Kajtazi, 2010 ), thereby reducing information asymmetry and intensifying competition for jobs. This, in turn, encourages people to save more to cope with employment uncertainty. Based on the reduction of information asymmetry, the digital economy also reduces people's liquidity and borrowing constraints, which in turn decreases savings (Wen, 2009 ; Bussière et al., 2013 ). For this reason, we propose that AI also indirectly affects regional savings levels by promoting the digital economy. China is a vast country with large disparities in the development level of each city. For this reason, this paper relies on the practice of Chinese scholars such as Hong& Yang ( 2021 ); and Xia et al. ( 2023 ) to obtain AI and rAI as proxy variables for the level of AI development by dividing the number of AI patents filed and granted each year in each city by the total population of the city, respectively. China's financial system is underdeveloped, and Chinese people's savings vehicles are relatively limited, with bank deposits being the main way for Chinese people to save. In this paper, we obtain SAVE by dividing the balance of urban and rural residents' savings deposits by the total population of the city. Similarly, rSAVE is calculated by dividing the balance of all types of deposits, including those of individuals and other enterprises, by the city's total population, serving as proxy variables for the level of regional savings. Currently, there is no public data on the level of digital economy development in each city. Therefore, this paper uses factor analysis to derive the city's digital economy development index DECO as a proxy variable, based on 11 data items including the number of data centers and digitized enterprises across three industries. Based on the construction of these variables, we adopt the city and year two-way fixed effects model to empirically test the impact of AI on the regional savings level, with the city's regional savings level as the dependent variable and the city's level of AI development as the independent variable. Based on the existing literature (Chen et al., 2022 ; Chen et al., 2022 ), we empirically test the transmission mechanism of AI indirectly affecting the regional savings level based on the digital economy by using the city's digital economy development level as the mediating variable. Finally, we conduct a heterogeneity test based on first-tier and non-first-tier cities. Based on the above research design, we first observe that there is an 'inverted U' nonlinear relationship between AI and regional savings levels. There is an inflection point in the level of AI development, and on both sides of the inflection point, AI raises and lowers the level of regional savings. From a national perspective, the average value of the level of AI development is still below the inflection point, and it is still in the stage of raising the level of regional savings. However, in 2020, the development level of AI in Beijing and Shanghai crossed the inflection point and entered the stage of reducing the level of regional savings. Second, AI promotes the development of the digital economy, and the latter has an "inverted U" type nonlinear relationship with the level of regional savings. Therefore, AI indirectly affects the regional savings level by promoting the development of the digital economy. Third, the impact of AI on the regional savings level is heterogeneous. In the four first-tier cities of Beijing, Shanghai, Guangzhou, and Shenzhen, the inflection point of the level of AI development is lower, which means that from the perspective of development level, AI in first-tier cities will reduce the regional savings level earlier than in non-first-tier cities. The remainder of the paper is organized as follows: Section 2 articulates our testable hypothesis. Section 3 describes our econometric model, variables, and data. Section 4 discusses our main empirical results with robustness tests, conducts tests of the transmission mechanism heterogeneity analysis. Finally, we summarize the entire paper. 2. Testable Hypotheses 2.1 Direct Impact of AI on the Level of Regional Savings Based on the existing results, we believe that AI has both positive and negative effects on the level of regional savings. AI can increase the level of regional savings or decrease the level of regional savings. 2.1.1 AI Increases the Level of Regional Savings On the one hand, AI increases precautionary savings among young people. First, the prevention of lower incomes increases precautionary savings among young people. AI forces young workers to compete with machines, increasingly diminishing their paychecks in relative or even absolute terms (Acemoglu et al., 2018). AI reduces young people's wages (Acemoglu et al., 2018). For example, one additional robot per 1,000 workers reduces average wages by 0.77% relative to areas without robots (Acemoglu et al., 2020 ). China is a developing country where wage income is the main source of income for young people. To cope with the decrease in wage income due to AI, young people need to save more to prevent insecurity. Second, preventing unemployment increases young people's precautionary savings. Routine tasks, defined as tasks that follow clear rules and can be performed by machines, are easily replaced by AI (Frey & Osborne, 2017 ). This increases the likelihood of unemployment among young people. AI provides greater output at lower cost to the extent that machines displace people (Haseeb et al., 2019 ), increasing the pressure on young people to become unemployed. For example, AI-based banks spend less time processing bill payments, accounts receivable, and other accounting tasks, reducing the number of accountants (Haseeb et al., 2019 ). Firms using AI are differentially eliminating job postings that list a range of previously advertised skills, while at the same time posting skill requirements that were not previously listed. To prevent unemployment, young people need to save more. Finally, improving employability skills increases precautionary savings among young people. AI favors high-skilled workers, changing the structure of firms' labor demand (Yang, 2022 ). In the case of China, research shows that the country has adopted more aggressive policies in AI development, but its overall employment structure is more vulnerable to AI disruption (Haseeb et al., 2019 ). AI eliminates existing jobs and adds new ones; for example, AI replaces the existing workforce while creating new jobs such as data managers and analysts (Acemoglu et al., 2018). To improve their employability skills, young people will need to save more to improve their employability skills in the future. On the other hand, AI increases precautionary savings for the elderly. First, it increases precautionary savings for their protection. The Chinese are deeply influenced by Confucianism. According to Confucian classics, children have an unconditional obligation to support their parents; therefore, older people in developing countries like China rely heavily on intergenerational transfers from adult children to their parents as a source of old age (Chen et al., 2019 ). AI reduces the incomes of young people, increases unemployment and pressure to upgrade employment skills, and leads to income volatility for adult children. Savings can help people manage income fluctuations and cushion against unexpected losses (Benami & Carter, 2021 ). Therefore, to mitigate fluctuations in intergenerational transfers due to the varying incomes of adult children, older adults need to increase their precautionary savings. Second, the desire to provide for the love and care of their children motivates older adults to increase their precautionary savings. Chinese people have a strong sense of family, and older people (including the elderly) love and care for their children throughout their lives. AI, such as automation, will eliminate low-skill jobs and create new high-skill jobs (Acemoglu & Restrepo, 2019 ). To prepare children for a high-skill future, older adults will boost their savings to better fund the advanced skill development of their children. When children are of working age, older adults may also continue to save to enhance their children's employability. This is exemplified by the widespread phenomenon of "old-age nibbling" in China. After their children are employed, the elderly will continue to save to avoid the possibility of their children losing their jobs, or even leave their savings as an inheritance to their children. In summary, AI can increase precautionary savings among young and old people. This means that AI will increase the regional savings level in China. 2.1.2 AI Reduces the Level of Regional Savings AI promotes digital entrepreneurship, which, in turn, leads to reduced savings. The deep integration of digital technologies such as AI with the economy has created the digital economy (Chen et al., 2022 ; Chen et al., 2022 ). In the process of deep integration with the economy, the application of digital technologies such as AI has generated digital entrepreneurship (Nambisan, 2017 ). Digital entrepreneurship is important for driving economic growth and is now one of the key priorities advocated by many national and international institutions, including the European Commission, the World Bank, and the Organization for Economic Cooperation and Development (Leong et al., 2022 ). Digital entrepreneurship, defined as the intersection of digital technologies like AI and traditional entrepreneurship (Nambisan, 2017 ), involves pursuing business opportunities that leverage these technologies (Davidson & Vaast, 2010 ). AI, along with other digital technologies, has led to the rapid digitization of products and services across industries (Nambisan, 2017 ) and the emergence of digital markets. Digital markets reduce information asymmetries, facilitate information exchange, democratize entrepreneurship, and promote digital entrepreneurship (Aldrich, 2014 ; Nambisan, 2017 ). AI, along with other digital technologies, facilitates new digital infrastructures, such as digital marketplaces and 3D printing, which accelerate the iterative cycle of shaping, implementing, and modifying product ideas and business models, thereby facilitating digital entrepreneurship (Ries, 2011 ; Nambisan, 2017 ). Additionally, AI and other digital technologies have spawned new business models in sectors such as smart healthcare, bike sharing, and swipe-and-pay. These new businesses also contain many opportunities for digital entrepreneurship and are conducive to digital entrepreneurship. Under the "mass entrepreneurship and innovation" strategy, digital entrepreneurship has become an important choice for Chinese entrepreneurs. For most entrepreneurs, obtaining credit to start a business is a challenge (Mair & Marti, 2009 ; Bradley et al., 2012 ). They face greater income and expenditure uncertainty and higher credit constraints due to unstable careers and limited collateral, such as real estate, necessitating more precautionary savings to smooth these fluctuations (Guo & Gao, 2018 ). AI weakens rural labor migration in China and reduces savings. Starting about three decades ago, the accelerated pace of industrialization led to rapid urbanization and large-scale population mobility (Chen et al., 2019 ), and massive rural-urban labor migration emerged in China (Yin et al., 2020 ). China has a typical urban-rural "dualistic" structure. Although rural workers migrate to cities for employment opportunities, they are often excluded from local public services such as education, health care, and housing. They face more income and expenditure uncertainty and higher credit constraints with unstable careers and less collateral such as real estate, so they need more precautionary savings to smooth income and expenditure uncertainty (Guo & Gao, 2018 ). Rural workers who migrate to cities still have their rural hukou, belong to the rural household registration, and still enjoy only a lower level of medical social security, so they have a higher propensity to save than urban residents; at the same time, their income in the city is close to or even exceeds that of some urban household residents, which is several times the income they received when they were originally employed in agriculture, so they have gradually freed themselves from the minimum level of consumption. Their real savings are substantial (Zhang et al., 2014 ). Consequently, the migration of rural workers to urban areas has significantly increased savings across China (Yin et al., 2020 ). AI is changing the labor demand structure of firms (Yang, 2022 ), and eliminating low-skilled workers (Acemoglu & Restrepo, 2019 ; Yang, 2022 ). Due to limited education and other factors, many rural workers migrating to urban areas are employed in low-skilled jobs that are highly susceptible to automation by AI. This suggests that AI's impact on the job market will diminish the migration of rural labor to urban areas in China, which in turn will reduce savings. In summary, AI reduces savings by promoting digital entrepreneurship and weakening labor migration from rural China, thereby reducing regional savings levels. 2.1.3 Research Hypothesis Overall, AI impacts regional savings through two opposing forces: one that increases savings and another that decreases them. The public's awareness of AI's potential, marked by the notable human-computer chess match between Deep Blue and Kasparov in February 1996, sparked early concerns about its implications. Since then, academics and the media have extensively covered the substitution effects of AI on labor. As a result, Chinese people's concern about the negative impact of AI on themselves and their children is much earlier. This means that AI has started to increase the regional savings level of Chinese people at the early stage of its development. In contrast, the entry of AI into generalized commercial applications started around 2008 (Chen et al., 2022 ; Chen et al., 2022 ; Chen, 2023 ). AI-generated digital entrepreneurship and weakened rural labor migration, which itself existed as a process later than 2008. As a result, the effect of AI in reducing the level of regional savings is delayed and is likely to become apparent as the level of AI development continues to increase. If this effect exceeds the effect of increasing per capita savings, AI may reduce the level of regional savings (Fig. 1 ). Therefore, we propose the following hypothesis: H1: The relationship between AI and regional savings level is "inverted U" non-linear, there is a turning point in the level of AI development, and on both sides of the turning point, AI increases and decreases the level of regional savings, respectively. 2.2 Analysis of Action Channels Based on the Digital Economy In 1996, Tapscott first proposed the concept of 'digital economy' in his book "The Digital Economy: Hopes and Risks in the Age of Networked Intelligence", defining it as the 'Internet economy,' a result of integrating Internet technology with the economy". After 2008, AI and other digital technologies became more mature, and enterprises in the three major industries gradually integrated these digital technologies into their production and operation activities, giving rise to the digital economy that is currently being discussed by scholars, especially in China. Thus, the digital economy has taken on a new connotation, referring to the result of the deep integration of AI and other digital technologies with the economy. Digital technologies like AI are the technological drivers behind the digital economy. As a cutting-edge technology, AI is constantly evolving and rapidly iterating, significantly influencing the pace of digital economic development. The higher the level of AI development, the higher the level of development of the digital economy. Therefore, AI can promote the development of the digital economy. 2.2.1 AI Promotes the Development of the Digital Economy and Increases Regional Savings AI is driving the digital economy, and the latter can increase regional savings. First, the change in the employment environment increases savings. The digital economy spurs enterprises towards digital transformation, enhancing R&D, automating production, improving management efficiency, and refining services (Yang et al., 2023 ). As digital transformation progresses, there is a rising demand for technical and service-oriented, high-skilled employees, significantly displacing many production-oriented, low-skilled roles. This inevitably creates income uncertainty for production-oriented, low-skilled workers, prompting these workers to increase their savings to prevent the unexpected. On the other hand, the digital economy creates, stores, and distributes digital information through unlimited access (Kajtazi, 2010 ). In the digital economy environment, companies can distribute recruitment-related digital information, and workers can obtain appropriate and timely employment information through digital platforms, which reduces the information asymmetry in the labor market and increases the intensity of competition in the labor market. To cope with the uncertainty caused by the pressure of employment competition, people will increase their savings. Furthermore, digital platforms in the digital economy compile diverse economic data, broaden information channels, reduce transaction costs, and enhance labor mobility, contributing to more dynamic market conditions. This increased competition in the labor market consequently encourages greater savings among workers. Second, expanding the boundaries of financial services increases savings. China's development is unbalanced, with large development gaps between urban and rural areas. Financial institutions are often reluctant to serve people in remote and poor areas, and have few branches in underdeveloped areas, which lack convenient financial services such as cash deposit and withdrawal (Xie et al., 2018 ). In the digital economy environment, financial institutions are providing online financial services based on digital technologies such as AI to overcome traditional distance limitations (Li et al., 2022 ). For example, financial institutions use facial recognition and other technologies for user identity authentication, providing online payment services to residents in remote and underdeveloped areas. China traditionally relies on cash transactions. Without online payment services, residents in remote, poor, and underdeveloped areas must rely solely on cash for daily transactions. The digital economy's expansion into these areas, providing online payment services, encourages residents to deposit their cash in financial institutions for future use. The cash they collect in their daily transactions will also be deposited in financial institutions, which will increase savings deposits. Traditionally, financial institutions have primarily served wealthy customers, who conduct numerous small transactions at a high cost per transaction (Feng & Guo, 2019 ). These "long tail" individuals are often excluded from the financial system. The digital economy, characterized by its low threshold and inclusivity, promotes online financial services, significantly extending the reach and benefiting 'long-tail' groups traditionally excluded from the financial system (Jiang et al., 2022 ). The traditional cash transactions of these groups are also completed through the financial system, thereby increasing savings deposits in the financial system. Finally, increased competition among commercial banks increases savings. In the digital economy, the convergence of digital technology with traditional finance has altered the competitive landscape for banks, expanding beyond the market reach of traditional financial institutions. Emerging digital economy startups (e.g., nonbank payment institutions), with the potential advantages of wide coverage, low cost, and high efficiency, mainly serve micro and small enterprises and individual lending areas, eroding banks' lending market share, and the shrinking of banks' lending market share exacerbates competition for commercial banks' credit business (Feng & Guo, 2019 ). Commercial banks are information-intensive enterprises (Berger, 2003 ), and the reduction of information asymmetry breaks the private information monopoly of commercial banks, reduces relationship-based lending based on formerly private information, reduces the monopoly power of banks, and increases the degree of competition among commercial banks (Berger, 2003 ; Song et al., 2023 ). Furthermore, the digital economy has digitized commercial banking services, reducing the physical and temporal barriers to financial transactions, broadening the competitive landscape, and intensifying bank competition (Jiang et al., 2022 ). China's financial system is underdeveloped, and bank deposits are the main source of savings for Chinese people. China's financial sector is compartmentalized, and commercial banks absorb savings deposits as their main source of funding. As deposit interest rates are already marketized, increased competition among commercial banks will encourage them to raise interest rates on savings deposits, which will increase people's propensity to save and thus increase savings deposits. In summary, the digital economy changes the employment environment, expands the boundaries of financial services, increases competition among commercial banks, and thus increases savings, and the digital economy ultimately increases the level of regional savings. 2.2.2 AI Promotes the Development of the Digital Economy and Reduces Regional Savings AI is driving the digital economy, and the latter could also reduce regional savings. First, encouraging consumption reduces savings. First, the digital economy improves consumption. The digital economy moves transactions online, promoting e-commerce, which overcomes spatial and temporal transaction barriers and enhances information availability. This enables consumers to easily access and utilize information when interested in specific products or services. Overall, the digital economy broadens consumption channels and enhances convenience, thereby improving the consumption experience and ultimately reducing savings (Dijk et al., 2007 ), which in turn improves the consumption experience promotes the consumption of residents, and reduces savings. Second, the digital economy reduces transaction costs for consumers. Transaction costs, which include counterparty search and information verification, account for about 35%-40% of total transaction activities (Dyer & Chu, 2003 ). The digital economy has the inherent advantages of intertemporal information dissemination and information sharing, which can reduce counterparty search costs and information verification costs, reduce consumers' transaction costs, and increase the marginal utility of consumption, thus promoting residents' consumption and reducing savings. Third, the digital economy brings enterprises and consumers closer together. The digital economy can improve information transparency (D'souza & Williams, 2017 ), strengthen the relationship between enterprises and consumers (Carlsson, 2004 ), enhance communication between enterprises and consumers, and encourage enterprises to better understand consumers' personalized consumption needs. Accordingly, enterprises can provide consumers with personalized products and services to maximize consumer satisfaction and increase consumers' marginal utility, thereby promoting consumption and reducing savings. In the digital economy, enterprises engage consumers through recommendation systems and precision marketing, effectively bridging the gap between them. Visual communication tools and situational consumption further enrich interactions between buyers and sellers, enhancing their willingness to consume (Yang et al., 2023 ), promoting residents' consumption, and reducing savings. Second, the expansion of social networks reduces savings. China is a traditional relational and human society (Bian, 1997 ; Yang et al., 2023 ), social networks, together with social rules and trust, are considered to belong to the category of social capital, which is crucial in determining people's socioeconomic status (Ma & Yang, 2011 ), and social networks can transmit a variety of information (Gompers et al., 2005 ), which has a significant impact on Chinese economic behavior (Yang et al., 2023 ). On one hand, social networks, connected by geography and blood ties, facilitate risk coordination through gift-giving and income transfer. This coordination helps mitigate the shocks from income and expenditure uncertainties, thereby weakening the incentive for precautionary savings and ultimately reducing overall savings(Yi et al., 2012 ). On the other hand, social networks have the effect of sharing information and reducing opportunism; people in a social network have more ex-ante information about each other's credit and are more likely to lend money to creditworthy people; ex-post, people in a social network are more likely to monitor the borrower's behavior after obtaining funds, which can solve the problem of hidden behavior. Thus, credit obtained through social networks can complement formal credit (Yang et al., 2023 ), and people with more social networks have more access to private loans (Ma & Yang, 2011 ), reducing people's borrowing constraints and thus reducing savings. The online, digitized, and data-enabled digital economy compresses spatial and temporal distances in daily activities, thereby strengthening social networks. This enhancement weakens precautionary saving incentives, eases borrowing constraints, and lowers the need for savings. Finally, improved financial services reduce savings. Lack of credit or other financial buffers is an important cause of high savings (Baker et al., 2023 ). The digital economy environment has enabled the development of mobile payments. Mobile payments on platforms such as WeChat and Alipay not only offer microcredit services but also enable financial institutions to gather users' credit information (Yin et al., 2022 ), thereby providing Chinese residents with credit and financial buffers that reduce the need for savings. As a key feature of the digital economy, mobile payments often involve small direct transfers that facilitate broader financial transactions, helping Chinese residents manage risks more effectively (Benami & Carter, 2021 ). At the same time, health risks, medical risks, unemployment risks, and income risks faced by households are major sources of uncertainty, and mobile payment platforms not only provide households with comprehensive services such as credit, insurance, and healthcare but also promote self-employment and help them become profitable. Credit services alleviate households' liquidity constraints, insurance services enable the settlement of claims after a household incurs a risk, healthcare services reduce the probability of health risks and medical risks, and employment enables households to gain more security (Yin et al., 2022 ). In summary, as a crucial component of the digital economy, mobile payments enhance households' capacity to diversify risks (Jack & Suri, 2014 ; Riley, 2018 ; Yin et al., 2022 ), thereby reducing the need for precautionary savings. Overall, by increasing consumption, expanding social networks, and strengthening financial services, the digital economy effectively reduces savings. 2.2.3 Research Hypothesis Overall, AI promotes the digital economy, which impacts regional savings through two opposing forces: one that increases savings and another that decreases them. The role of the digital economy on the level of regional savings depends on the confrontation of the two forces. Enterprises in the three industries are undergoing digital transformation based on digital technologies such as AI, thereby driving the digital economy. Digital transformation first creates demand for highly skilled personnel who are proficient in digital technologies such as AI. In this way, the digital economy will be the first to trigger changes in the employment environment and increase the level of regional savings. As the digital economy develops, the information environment will improve. Only when the digital economy promotes consumption, expands social networks, and strengthens financial services does it begin to reduce the level of regional savings. If the reduction in savings outweighs the increase, the digital economy will ultimately lead to lower regional savings levels (Fig. 2 ). For this purpose, we propose the following hypothesis: H2: AI indirectly influences regional savings by fostering the digital economy, which exhibits an 'inverted U' non-linear relationship with regional savings, mirroring the relationship between AI itself and regional savings. 3. Methodology 3.1 Models 3.1.1 Model of Direct Impact Chinese scholars typically prefer city-level empirical analysis when studying regional phenomena in China. We investigate how AI affects the level of regional savings at the city level. To do so, we refer to the literature on regional savings (Jia & Han, 2023 ) and design the following year and city two-way fixed effects model to test the direct effect of AI on regional savings levels: $$\:{SAVE}_{it}={\alpha\:}_{0}+{\beta\:}_{1}\ast\:{AI}_{it}+{\beta\:}_{2}\ast\:{AI2}_{it}+\eta\:\ast\:X+{\alpha\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 1 where i and t are the subscripts of city and time, respectively; \(\:{\alpha\:}_{i}\) is to capture the city FE; \(\:{\lambda\:}_{t}\) is the capture year FE, and \(\:{\epsilon\:}_{it}\) is the random error term. \(\:{SAVE}_{it}\) is the dependent variable, that is, the level of regional savings of the city i in year t. \(\:{AI}_{it}\) is an independent variable, that is, the development level of AI in the ith city in the tth year; \(\:{AI2}_{it}\) is the quadratic term and \(\:{\beta\:}_{2}\) is the coefficient of the quadratic line; if it is significantly negative, AI has an inverted U-shaped nonlinear effect on the level of regional savings. X is the control variable, as shown below. 3.1.2 Model of Channel of Action To test whether AI indirectly affects regional savings levels through the digital economy, we designed the following model based on the existing literature that studies the transmission mechanism (Chen et al., 2022 ; Chen, 2023 ): $$\:{SAVE}_{it}={\alpha\:}_{0}+{\beta\:}_{1}\ast\:{AI}_{it}+{\beta\:}_{2}\ast\:{AI2}_{it}+\eta\:\ast\:X+{\alpha\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 2 $$\:{DECO}_{it}={\alpha\:}_{0}+{\beta\:}_{1}\ast\:{AI}_{it}+\eta\:\ast\:Z+{\alpha\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 3 $$\:{SAVE}_{it}={\alpha\:}_{0}+{\beta\:}_{1}\ast\:{AI}_{it}+{\beta\:}_{2}\ast\:{AI2}_{it}+{\beta\:}_{3}\ast\:{DECO}_{it}+{\beta\:}_{4}\ast\:{DECO2}_{it}+\eta\:\ast\:X+{\alpha\:}_{i}+{\lambda\:}_{t}+{\epsilon\:}_{it}$$ 4 \(\:{DECO}_{it}\) is the mediating variable (the digital economy level of the ith city in the tth year), and \(\:{DECO2}_{it}\) is a quadratic term for \(\:{DECO}_{it}\) . X in Equations ( 2 ) and ( 4 ) is the control variable, which is the same as that in Eq. ( 1 ). Z in Eq. ( 3 ) is the control variable, and financial efficiency is also controlled. The test procedure is as follows: First, without adding intermediary variables, we estimated Eq. ( 2 ); if the coefficient of the level of AI development \(\:{\beta\:}_{1}\) is significant, it shows that AI has a total effect on the level of regional savings, and we continue with follow-up analysis; otherwise, it is a masking effect. Second, we estimated Eq. ( 3 ) to determine the impact of AI on the mediating variables. Third, after adding the intermediary variable, we estimated Eq. ( 4 ). If the coefficient of the level of AI development in Eq. ( 3 ) is \(\:{\beta\:}_{1}\) , and the coefficient of the intermediary variable in Eq. ( 4 ) is \(\:{\beta\:}_{3}\) , it shows that the intermediary effect exists. Fourth, if \(\:{\beta\:}_{1}\) in Eq. ( 3 ) and \(\:{\beta\:}_{3}\) in Eq. ( 4 ) are significant, the mediating effect must be tested using the Sobel test. 3.2 Variables Based on the existing literature, this study's independent, dependent, mediating, and control variables are presented in Appendix 1. 3.2.1 Dependent Variables The dependent variable in this paper is the regional level of savings ( SAVE ). Saving is a concept corresponding to residents' income and residents' consumption, and some scholars study the savings level based on household microdata, usually calculating the residents' savings level according to (disposable income - consumption)/disposable income (Chen et al., 2019 ; Wu et al., 2021 ; Baker et al., 2023 ; Choukhmane et al., 2023 ). This paper examines regional savings at the macro level. Given the underdevelopment of China's financial system, Chinese residents have limited savings options, primarily relying on bank deposits. When Chinese scholars study the macro-level saving behavior of Chinese residents, they use per capita savings deposit balances to measure regional savings levels (Luo et al., 2016 ; Luo & Wen, 2016 ; Wang & Liao, 2020 ; Gao & Shi, 2023 ). Therefore, in this paper, SAVE is calculated by dividing the household savings deposit balances by the total population in each city, serving as a proxy for regional savings levels. For robustness testing, rSAVE is obtained by dividing all types of deposits (including self-employed and other business deposits) by the total population in each city as another proxy variable for the regional savings level. 3.2.2 Independent Variables The independent variable in this paper is the level of AI development ( AI ). Referring to the existing literature by Hong& Yang ( 2021 ), and Hong& Yang ( 2021 ); Xia et al. ( 2023 ), The level of AI development ( AI ) is measured by dividing the number of AI patent applications by the total population in each city, serving as a proxy for this variable. 3.2.3 Intermediary Variables The mediating variable in this paper is the level of digital economy development ( DECO ). At present, there is no publicly available official data on the level of digital economy development in cities. This study constructed the indicator system shown in Appendix 2, collecting data from various cities via the State Administration of Market Supervision and Administration. It then used factor analysis to calculate the city digital economy development index score, normalizing it by (Score-Min)/(Max-Min) to obtain DECO as a proxy for the development of the digital economy. 3.2.4 Control Variables Referring to the existing literature (Chen et al., 2019 ; Chen et al., 2022 ; Makori et al., 2022 ; Zhang et al., 2022 ), this paper includes control variables such as the level of economic development, the rate of economic growth, the level of foreign investment, the level of industrial structure, fiscal expenditures on science and technology, the rate of urbanization, population size, population density, and the level of financial development. 3.3 Data 3.3.1 Data Source Since the global financial crisis in 2008, digital technologies such as AI and the economy have begun to be deeply integrated (Chen et al., 2022 ; Chen, 2023 ), and AI has gradually entered large-scale commercial applications. China is a vast country with unbalanced development, and it is the preferred choice of Chinese scholars to study Chinese issues at the city level. The data of the China City Statistical Yearbook will be updated to 2020. Therefore, this paper empirically analyzes Chinese city data from 2008 to 2020. In this paper, the data are processed as follows: (1) missing samples are eliminated; (2) the China City Statistical Yearbook no longer reports foreign direct investment ( FDI ) in cities in 2020, so FDI in 2020 is linearly interpolated; (3) household savings deposits in 2010 are linearly interpolated. Subsequently, the 2010 and 2020 data are excluded for robustness testing. We end up with 3,362 year-city observations. The number of AI patent applications and the number of approved patents are from the State Intellectual Property Office (SIPO), and we extracted them with the keyword "AI", while the other data are from China City Statistical Yearbook. In addition, taking the natural logarithm can eliminate the influence of outliers, to eliminate the influence of outliers, this paper takes the natural logarithm of other continuous variables other than the upper and lower 1% Winsorize shrinkage treatment. 3.3.2 Summary Statistics The descriptive statistics of variables are shown in Appendix 3. The gap between the minimum and the maximum of each variables is significant, consistent with China's primary national conditions of unbalanced development. 4. Results 4.1 Direct Impact 4.1.1 Univariable Analysis The results of univariable analysis are shown in Appendix 4.1. The results shown that there is a relatively obvious "inverted U" nonlinear relationship between AI and regional savings level. 4.1.2 Multivariable Regression Controlling for city and year fixed effects, with SAVE as the dependent variable and AI and rAI as the independent variables, respectively, we estimate Eq. ( 1 ) with the FE fixed effects model without adding the control variables, resulting in Columns (1) and (2) of Table 1 . Re-estimating Eq. ( 1 ) with all the control variables added results in Columns (3) and (4) of Table 1 . From Columns (1)-(4) of Table 1 , the coefficients of the quadratic terms of the level of AI development are all significantly negative at the 1% significance level, while the coefficients of the primary terms are all significantly positive at the 1% significance level, and there is an "inverted U" nonlinear relationship between AI and the level of regional savings. These empirical results support Hypothesis H1 and align with the theoretical analysis presented in the previous section. AI has both positive and negative effects on the level of regional savings, and AI has been used in large-scale general commercial applications since 2008 (Chen et al. 2022 ; Chen et al. 2022 ; Chen 2023 ), while the negative effect is later, resulting in an "inverted U" nonlinear relationship between AI and the level of regional savings. The relationship between AI and regional savings levels is an "inverted U" nonlinear. Table 1 FE estimation results of Eq. ( 1 ) (1) (2) (3) (4) Variables SAVE SAVE SAVE SAVE AI 0.8830*** 0.9935*** (0.1009) (0.0801) AI2 -0.0809*** -0.0797*** (0.0130) (0.0086) rAI 4.0186*** 4.0481*** (0.4064) (0.3359) rAI2 -1.8517*** -1.4667*** (0.3027) (0.2064) PGDP -1.6968*** -1.6086*** (0.3863) (0.3675) GGDP 0.0114 0.0108 (0.0085) (0.0080) FDI -0.9400*** -0.8907*** (0.2786) (0.2596) INDS -0.0717 -0.0401 (0.1350) (0.1319) SCIP -0.0107** -0.0144*** (0.0051) (0.0051) CITY -0.0472 -0.2201 (0.5780) (0.5351) POP -4.3912*** -4.1473*** (0.8149) (0.7747) PDEN -1.1610* -1.2070* (0.6639) (0.6686) FSIZ 0.0042** 0.0043** (0.0021) (0.0021) Constant 1.6856*** 1.6836*** 40.7377*** 39.0235*** (0.0515) (0.0518) (4.8636) (4.6069) City FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes Observations 3,362 3,362 3,362 3,362 R-squared 0.8029 0.8075 0.8539 0.8553 N 282 282 282 282 Note: Robust standard errors are in parentheses; * p < 0.1, * * p < 0.05, * p < 0.01, same as below. 4.1.3 Robustness Checks The results of robustness checks are shown in Appendix 4.2. In summary, the conclusion that H1 is established is robust under the conditions of controlling endogeneity, changing the measurement of regional savings level, excluding interpolated data, using double clustering robust standard errors, and controlling for the level of digital economy development. 4.2 Conduction Mechanism 4.2.1 Univariate Analysis The results of univariable analysis are shown in Appendix 5. The results shown that As the level of AI development increases, the level of digital economic development improves; As the level of digital economy development increases, the level of regional savings first increases and then decreases. 4.2.2 Multiple Regression Equations ( 2 )-( 4 ) are estimated using FE with SAVE as the dependent variable and DECO as the mediator variable, and the results are shown in Columns (1)-(3) of Table 2 , Panel A. In Column (1) of Table 2 Panel A, the coefficient on the level of AI development ( AI ) is significant at the 1% significance level, indicating the existence of the overall effect; the coefficient on the level of AI development ( AI ) is significant at the 1% significance level in Column (2), and the coefficient on the mediator variable, the level of development of the digital economy ( DECO ), is significant at the 1% significance level in Column (3), so there is a mediating effect. Combining the sign and significance of the level of AI development ( AI ) and the level of digital economy development ( DECO ) in Columns (2) and (3), it can be seen that there is an "inverted U" shaped non-linear relationship between AI and the level of regional savings by promoting the development of digital economy. Therefore, the research hypothesis H2 is established. In estimating Eq. ( 3 ) with the level of digital economic development ( DECO ) as the dependent variable and the level of AI development ( AI ) as the independent variable, similar to section 4.4.1, the level of AI development ( AI ) is endogenous, and using the same instrumental variable ivAI as in 4.4.1, we reestimate Eq. ( 3 ) using IV, and the results are shown in Table 2 , Panel A, Column (4). In estimating Eq. ( 4 ), the level of digital economic development ( DECO ) is also likely to be endogenous due to measurement error, following Laeven& Levine (2007), Laeven& Levine (2009), Faccio et al. (2011), Chen et al. ( 2022 ), and Chen ( 2023 ). Calculate the mean of the digital economy development index of other cities for the same year and its quadratic term to get ivDECO and ivDECO2 as instrumental variables. Taking ivAI , ivAI2 , ivDECO , and ivDECO2 as instrumental variables, we re-estimate Eq. ( 4 ) using IIV, and the results are shown in Column (5) of Table 2 Panel A. The results are summarized as follows. From Columns (4) and (5) of Panel A of Table 2 , we can see that AI forms an "inverted U" nonlinear relationship with regional savings levels by promoting the development of the digital economy. Therefore, the conclusion that the research hypothesis H2 holds is robust after excluding endogeneity. Replacing the independent and dependent variables with rSAVE and rAI , respectively, and estimating Equations ( 2 )-( 4 ) in the above order, the results are shown in Table 2 Panels B and C. From Panels B and C, AI forms an "inverted U" type nonlinear relationship with regional savings levels by promoting the development of the digital economy. Therefore, the conclusion that research hypothesis H2 is valid is robust. Table 2 Estimation results of the conduction mechanism Panel A (1) (2) (3) (4) (5) Variables SAVE DECO SAVE DECO SAVE AI 0.9935*** 0.0535*** 0.7411*** 0.0506*** 1.8992*** (0.0801) (0.0080) (0.0936) (0.0053) (0.1710) AI2 -0.0797*** -0.0610*** -0.1643*** (0.0086) (0.0097) (0.0198) DECO 5.0878*** 4.3383** (1.0669) (1.9738) DECO2 -3.8566*** -5.3214** (1.0389) (2.3473) City FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Control Yes Yes Yes Yes Yes Observations 3,362 3,362 3,362 3,362 3,362 R-squared 0.8539 0.6819 0.8598 0.6811 0.9548 N 282 282 282 282 282 Panel B (1) (2) (3) (4) (5) Variables rSAVE DECO rSAVE DECO rSAVE AI 3.8188*** 0.0535*** 2.6763*** 0.0506*** 7.6254*** (0.2839) (0.0080) (0.4322) (0.0053) (0.7711) AI2 -0.2625*** -0.1622*** -0.6256*** (0.0343) (0.0484) (0.0938) DECO 24.6907*** 32.8475*** (4.7595) (7.8951) DECO2 -22.1700*** -38.5346*** (5.3403) (9.4705) City FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Control Yes Yes Yes Yes Yes Observations 3,362 3,362 3,362 3,362 3,362 R-squared 0.7069 0.6819 0.7308 0.6811 0.9348 N 282 282 282 282 282 Panel C (1) (2) (3) (4) (5) Variables SAVE DECO SAVE DECO SAVE rAI 4.0481*** 0.2518*** 3.1642*** 0.2937*** 8.9577*** (0.3359) (0.0370) (0.3633) (0.0355) (0.8023) rAI2 -1.4667*** -1.2309*** -3.7352*** (0.2064) (0.2221) (0.4955) DECO 5.4012*** 4.0063** (1.0322) (1.8929) DECO2 -3.8739*** -4.4664* (1.1082) (2.4570) City FE Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Control Yes Yes Yes Yes Yes Observations 3,362 3,362 3,362 3,362 3,362 R-squared 0.8553 0.6444 0.8632 0.6384 0.9507 N 282 282 282 282 282 Note: All instrumental variables have passed validity tests. 4.3 Regional Heterogeneity Analysis Previous analyses indicate that AI exerts both positive and negative effects on regional savings levels. Chinese people worry about the negative impact of AI on themselves and their children much earlier. This suggests that AI began to increase regional savings levels in China early in its development. In contrast, the entry of AI into large-scale generalized commercial applications started around 2008 (Chen et al., 2022 ; Chen et al., 2022 ; Chen, 2023 ). There is a process of digital entrepreneurship and weakened rural labor migration generated by AI itself, later than 2008. As a result, the role of AI in reducing the regional savings level is later. Therefore, the relationship between AI and regional savings level is "inverted U" non-linear. China is a huge country with unbalanced development. Beijing, Shanghai, Guangzhou, and Shenzhen are first-tier cities, while the rest are non-first-tier cities. First-tier cities are economically developed and have more job opportunities. As a result, people in first-tier cities are relatively less concerned about the negative impact of AI on themselves and their children. Consequently, the positive impact of AI is relatively weaker in first-tier cities. At the same time, first-tier cities are rich in human and economic resources and have more entrepreneurial opportunities, and more rural workers have migrated to first-tier cities. This means that the negative force of AI is relatively stronger in first-tier cities. Weaker positive effects and stronger negative effects suggest a lower inflection point in the level of AI development. Therefore, we hypothesize that there is heterogeneity in the role of AI in influencing regional savings levels and that the inflection point of the level of AI development is lower in first-tier cities. To test this speculation, we estimate Eq. ( 1 ) using a variable coefficient individual fixed effects model with SAVE and rSAVE as the dependent variables and AI as the independent variable, and the results are in Columns (1) and (2) of Table 3 . Eq. ( 1 ) is re-estimated by replacing the independent variables with rAI , and the results are shown in Columns (3) and (4) of Table 3 . From the estimation results, first, the coefficients of the quadratic terms of the level of AI development are all significantly negative at the 1% significance level. Second, the inflection point of the level of AI development is lower in first-tier cities than in non-first-tier cities (Table 4 ). Table 3 Heterogeneity of first-tier and not-first- tier cities (1) (2) (3) (4) Variables SAVE rSAVE SAVE rSAVE AI (not first tier) 0.9230*** 3.4845*** (0.0775) (0.2711) AI (first tier) 1.5646*** 7.4092*** (0.2296) (1.3046) AI2 (not first tier) -0.0695*** -0.2198*** (0.0092) (0.0467) AI2 (first tier) -0.1389*** -0.6156*** (0.0221) (0.1130) rAI (not first tier) 3.7310*** 13.2891*** (0.3391) (1.4410) rAI (first tier) 7.4256*** 35.4233*** (1.8745) (8.9790) rAI2 (not first tier) -1.2253*** -3.7437*** (0.2337) (1.2542) rAI2 (first tier) -3.2722*** -14.4583*** (0.8641) (3.8345) City FE Yes Yes Yes Yes Year FE Yes Yes Yes Yes Control Yes Yes Yes Yes Observations 3,362 3,362 3,362 3,362 R-squared 0.8551 0.7127 0.8571 0.6954 N 282 282 282 282 Table 4 The inflection point of AI in first-tier cities and not-first-tier cities Panel A SAVE is an independent variable rSAVE is an independent variable Not first tier First tier Not first tier First tier Coefficient of AI 0.9230 1.5646 3.4845 7.4092 Coefficient of AI2 -0.0695 -0.1389 -0.2198 -0.6156 The inflection point of AI 6.6403 5.6321 7.9265 6.0179 Panel B SAVE is an independent variable rSAVE is an independent variable Not first tier First tier Not first tier First tier Coefficient of rAI 3.7310 7.4256 13.2891 35.4233 Coefficient of rAI2 -1.2253 -3.2722 -3.7437 -14.4583 The inflection point of rAI 1.5225 1.1346 1.7749 1.2250 5. Conclusion Our paper investigates AI's impact on regional savings levels, analyzing both the direct impacts and action channels theoretically, then empirically testing these theories using a sample from Chinese cities between 2008 and 2020 with city and year fixed effects. We find, firstly, an 'inverted U' nonlinear relationship between AI and regional savings levels; as AI technology improves, regional savings initially increase and then decrease. Overall, AI is still in the stage of increasing regional savings, but some cities have entered the stage of decreasing regional savings. Second, AI promotes the development of a digital economy, and the latter has an "inverted U" nonlinear relationship with regional savings level. Third, the impact of AI on the regional savings level is heterogeneous, and the inflection point of AI technology is lower in first-tier cities. Based on the fact that savings are the financial driver of economic development and AI is a brand new driver of economic growth, our paper examines the impact of AI on regional savings levels. In addition, there are the following aspects that are worth studying in depth. First, the impact of AI on the lending behavior of financial institutions. The deep integration of digital technologies such as AI and the economy has created a digital economy (Chen et al., 2022 ; Chen, 2023 ), which can reduce information asymmetry and improve the quality of credit assets of financial institutions (Chen et al., 2022 ). How does this affect the lending behavior of financial institutions, and what mechanisms are involved? These questions need to be studied in depth. Second, the impact of AI on household microbehavior. In this paper, we have focused on AI's impact on city-level savings from a macro perspective. At the micro level, how does AI influence family fertility behaviors? Will families choose to have fewer children as AI adoption accelerates and machine replacement of human labor becomes more common, subsequently impacting their savings behaviors? These questions have important implications for long-term economic growth. The fact that this question has not been explored in this paper is a shortcoming of this paper and one of the directions for future research. Declarations Author contributions: Conceptualization: Hongwei Zhang, Xiaohui Chen, Xiaohui Luo；Methodology: Xiaohui Luo Investigation: Xiaohui Luo Visualization: Xiaohui Chen Supervision: Xiaohui Luo, Hongwei Zhang, Xiaohui Chen Writing—original draft: Hongwei Zhang Writing—review & editing: Hongwei Zhang, Xiaohui Luo, Xiaohui Chen Funding: none. Clinical trial number: not applicable. References Acemoglu & Daron & Restrepo & Pascual (2018). "The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment." American Economic Review . Acemoglu, D. & D. H. Autor & J. Hazell & P. Restrepo (2020). "Ai and Jobs: Evidence from Online Vacancies." Social Science Electronic Publishing . Acemoglu, D. & P. Restrepo (2019). "Automation and New Tasks: How Technology Displaces and Reinstates Labor." Journal of Economic Perspectives 33 (2): 3-30. Aldrich, H. E. (2014). The Democratization of Entrepreneurship? Hackers, Makerspaces, and Crowdfunding . Academy of Management Meeting. Baker, S. R. & E. Benmelech & Z. Yang & Q. Zhang (2023). "The Determinants of The Chinese Household Savings Rate1." Benami, E. & M. R. Carter (2021). "Can digital technologies reshape rural microfinance? Implications for savings, credit, & insurance." Applied Economic Perspectives and Policy . Berger, A. N. (2003). "The economic effects of technological progress: Evidence from the banking industry." Journal of Money, credit and Banking : 141-176. Bian, Y. (1997). "Bringing strong ties back in: Indirect ties, network bridges, and job searches in China." American sociological review : 366-385. Bradley, S. W. & J. S. Mcmullen & K. Artz & E. M. Simiyu (2012). "Capital Is Not Enough: Innovation in Developing Economies." Journal of Management Studies 49 (4). Bussière, M. & Y. Kalantzis & R. Lafarguette & T. Sicular (2013). "Understanding household savings in China: the role of the housing market and borrowing constraints." MPRA Paper . Cao, W. & G. Liu & L. Zeng & Y. Feng (2023). "The Impact of the Integration of Family Pension and Social Pension on Residents’ Savings Rate." Economic Research Journal 58 (3): 172-190. Carlsson, B. (2004). "The Digital Economy: what is new and what is not?" Structural change and economic dynamics 15 (3): 245-264. Chen, X. (2023). "Information moderation principle on the regulatory sandbox." Economic Change and Restructuring 56 (1): 111-128. Chen, X. & L. Teng & W. Chen & H. Beladi (2022). "How does FinTech affect the development of the digital economy? Evidence from China." The North American Journal of Economics and Finance 61 . Chen, X. & D. Yan & W. Chen (2022). "Can the digital economy promote FinTech development?" Growth and Change 53 . Chen, X. & H. Zhang & Y. Wu (2020). "How does the digital economy affect the level of industrial structure?" Securities Market Herald (7): 20-29. Chen, Y. J. & Z. Chen & S. He (2019). "Social Norms and Household Savings Rates in China." Choukhmane, T. & N. Coeurdacier & K. Jin (2023). "The one-child policy and household saving." Journal of the European Economic Association : jvad001. Cui, F. & J. Yang & K. Fu (2019). "The Heterogeneous Influence of Rural Labor Transferon Household Saving Rate." Journal of Statistics and Information 34 (12): 40-49. D'souza, C. & D. Williams (2017). "The digital economy." Bank of Canada Review 2017 (Spring): 5-18. Davidson, E. & E. Vaast (2010). Digital entrepreneurship and its sociomaterial enactment . 2010 43rd Hawaii International Conference on System Sciences, IEEE. Dijk, G. V. & S. Minocha & A. Laing (2007). "Consumers, channels and communication: Online and offline communication in service consumption." Interacting with Computers 19 (1): 7-19. Dyer, J. H. & W. Chu (2003). "The Role of Trustworthiness in Reducing Transaction Costs and Improving Performance: Empirical Evidence from the United States, Japan, and Korea." Organization Science 14 (1): 57-68. Feng, S. & R. Guo (2019). "Digital Finance, Bank Competition and Bank Efficiency." Reform (11): 75-89. Frey, C. B. & M. A. Osborne (2017). "The future of employment: How susceptible are jobs to computerisation?" Technological Forecasting and Social Change 114 : 254-280. Gao, J. & Q. Shi (2023). "Financial Behavior Changes of Ｒural Households in China under the Background of Urbanization: Based on the Data of Ten Villages in Zhejiang." Issues in Agricultural Economy (02): 89-106. Gompers, P. & J. Lerner & D. Scharfstein (2005). "Entrepreneurial spawning: Public corporations and the genesis of new ventures, 1986 to 1999." The journal of Finance 60 (2): 577-614. Guo, Q. & X. Gao (2018). "Multidimensional Examination of Migrant Workers’ Semi-urbanization: Influencing Factors and Basic Outlet." Journal of Northwest A&F University(Social Science Edition) 18 (03): 22-30. Han, Y. (2022). "The Impact of Industrial Robots on the Skill-Based Wage Gap." American Journal of Industrial and Business Management 12 (4): 571-602. Haseeb, M. & S. Sasmoko & L. Mihardjo & K. Jermsittiparsert (2019). "Economic Impact of Artificial Intelligence: New Look for the Macroeconomic Assessment in Asia-Pacific Region." International Journal of Computational Intelligence Systems 12 (2): 1295-1310. Hojeghan, S. B. & A. N. Esfangareh (2011). "Digital economy and tourism impacts, influences and challenges." Procedia - Social and Behavioral Sciences 19 (none): 308-316. Hong, J. & Z. Yang (2021). "Dialect Boundaries，Communication Barriers and Innovation Performances." Nankai Economic Studies (02): 46-66. Jack, W. & T. Suri (2014). "Risk Sharing and Transactions Costs: Evidence from Kenya's Mobile Money Revolution." American Economic Review 104 (1): 183-223. Jia, D. & H. Han (2023). "Fintech, Bank Liabilities and Competition." The Journal of World Economy 46 (02): 183-208. Jiang, E. X. & G. Y. Yu & J. Zhang (2022). "Bank Competition amid Digital Disruption: Implications for Financial Inclusion." Available at SSRN 4178420 . Kajtazi (2010). Information asymmetry in the digital economy . International Conference on Information Society. Kehal, H. S. & V. P. Singh (2005). Digital Economy: impacts, influences, and challenges , IGI Global. Leland, H. E. (1968). "Saving and Uncertainty: The Precautionary Demand for Saving." Uncertainty in Economics 82 (3): 465-473. Leong, C. & F. T. C. Tan & B. Tan & F. Faisal (2022). "The emancipatory potential of digital entrepreneurship: A study of financial technology-driven inclusive growth." Information & Management 59 (3): 103384. Li, Y. & J. Tan & B. Wu & J. Yu (2022). "Does digital finance promote entrepreneurship of migrant? Evidence from China." Applied Economics Letters 29 (19): 1829-1832. Luo, C. & Z. Zhang & C. Wu & Q. Qu (2016). "Financial Support for Supply-side Structural Reform: A Reanalysis of Savings and Investment Conversion Efficiency." Macroeconomics (03): 8-23. Luo, J. & Q. Wen (2016). "An Empirical Study on the Influence of Household Financial Asset Allocation on Household Consumption." Consumer Economics 32 (01): 18-22. Ma, G. & E. Yang (2011). "Economies of Scale， Returns to Scale and the Problem of Optimum-scale Farm Management: An Empirical Study Based on Grain Production in China." Economic Research Journal 46 (03): 83-96. Mair, J. & I. Marti (2009). "Entrepreneurship in and around institutional voids: A case study from Bangladesh." Journal of Business Venturing 24 (5): 419-435. Makori, E. & E. Matundura & N. Mose (2022). "Effect of fiscal and monetary policy on gross domestic savings in Kenya." Makori, E., Matundura, E., and Mose (2022): 44-52. Nambisan, S. (2017). "Digital entrepreneurship: Toward a digital technology perspective of entrepreneurship." Entrepreneurship theory and practice 41 (6): 1029-1055. Rayna, T. (2008). "Understanding the Challenges of the Digital Economy: The Nature of Digital Goods." Communications & Strategies 1 (71): 13-36. Ries, E. (2011). The lean startup: How today's entrepreneurs use continuous innovation to create radically successful businesses , Currency. Riley, E. (2018). "Mobile money and risk sharing against village shocks." Song, K. & Z. Li & J. Yang (2023). "FinTech and Banking Behavior: A Liquidity Creation Perspective." Journal of Financial Research (02): 60-77. Wang, L. & G. Liao (2020). "Regional Linkage of Poverty and the Spatial Spillover Effects on Poverty Reduction：Taking the Ethnic Areas of Sichuan Province as Examples." Northwest Population Journal 41 (01): 83-92. Wen, Y. (2009). "Saving and Growth Under Borrowing Constraints Explaining the 'High Saving Rate' Puzzle." Working Papers 35 (3): 935. Wu, W. & X. Zhang & K. Wu (2021). "Financial Literacy and the Household Savings Ｒate: The Role of Financial Planning and Borrowing Constraints." Journal of Financial Research (08): 119-137. Xia, J. & P. Sun & B. Li (2023). "Local Government R&D Investment, Technological Innovation and Optimization and Upgrading of Tourism Industry Structure." Study & Exploration (03): 93-104. Xie, X. & Y. Shen & H. Zhang & F. Guo (2018). "Can Digital Finance Promote Entrepreneurship?—Evidence from China." China Economic Quarterly 17 (4). Yang, B. & Z. Guo & X. Yi & L. Zhang (2023). "The Micro Effect of Digital Economy on Household Consumption Upgrade: Empirical Evidence from CHFS." Economic Review (03): 31-47. Yang, B. & Z. Yang & Y. Guo (2023). "Employment structure effect in enterprises digital transformation: Empirical analysis based on text mining of annual reports of manufacturing industry listed enterprises." China Soft Science (04): 141-150. Yang, C. H. (2022). "How Artificial Intelligence Technology Affects Productivity and Employment: Firm-level Evidence from Taiwan." Research Policy 51 . Yang, F. & H. Zhang & X. Lliu (2023). "Digital economy driving household consumption: Mechanism and spatial effects." Progress in Geography 42 (05): 837-851. Yi, X. & B. Zhang & R. Yang & B. Yang (2012). "The Family Social Network and the Rural Household Saving Behavior: a Case Study Based on Chinaꆯ s Villages." Journal of Management World (05): 43-51. Yin, Z. & T. Liu & C. Zhang (2020). "The Influence of Rural Labor Migration on Household Saving Rate." China Industrial Economics (01): 24-42. Yin, Z. & Z. Wu & J. Jiang (2022). " The Effect of Mobile Payment on the Household Savings Rate in China." Journal of Financial Research (09): 57-74. Zhang, H. & C. Xiong & X. Chen & Y. Wang (2022). "The Influence of Financial Investment of Science and Technology on the Development of Digital Economy." Finance & Economics (05): 135-148. Zhang, X. & X. Liu & G. Fan (2014). "Rural Labor Migration and Household Saving Rate: The Perspective of Discrepancy in Income and Social Security." Economic Research Journal 49 (04): 130-142. Zhao, X. & H. Wang & T. Liu (2017). "Population Aging, Endowment Insurance and Household Saving Rate." China Soft Science (08): 156-165. Appendix Appendix 1 to 5 are not available with this version Additional Declarations The authors declare no competing interests. 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-6702658","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":459015143,"identity":"f47b6775-696b-4e92-b97a-3ae6c4051745","order_by":0,"name":"Hongwei Zhang","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Hongwei","middleName":"","lastName":"Zhang","suffix":""},{"id":459015144,"identity":"4d4e25c6-5c27-45a0-95e9-c95903e20c8e","order_by":1,"name":"Xiaohui Luo","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzklEQVRIiWNgGAWjYDACZjCSADEZHyRU1JCmhdngwZljRFsEBmySD1uY8aoEA4PjzA8fF7ZZMMi3nzGrSGxgY+Bv707Aq0Wymc3YeGabBIPBmRyzG4k7ZBgkzpzdgFcLPzODmTQvSIsED1DLGTYgIxe/FjZm9m9gLfIzeMwKEtuYCWvhZ+aB2MJwg8eMgSgtks08xcY850B+SSuWSDhzjIegXwzOH9/4mKesDhhihzd+/FFRI8ff3otfCwzUN0AZPEQpHwWjYBSMglGAHwAASkY7Sf70NxAAAAAASUVORK5CYII=","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Xiaohui","middleName":"","lastName":"Luo","suffix":""},{"id":459015145,"identity":"20826c5c-0828-4b8d-bb23-5f58e7af5b5a","order_by":2,"name":"Xiaohui Chen","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Xiaohui","middleName":"","lastName":"Chen","suffix":""}],"badges":[],"createdAt":"2025-05-20 01:50:45","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6702658/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6702658/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83311600,"identity":"9c79c60d-d755-45d9-8f82-c1f9c693038f","added_by":"auto","created_at":"2025-05-22 19:36:22","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":286026,"visible":true,"origin":"","legend":"\u003cp\u003eLogic diagram of AI directly affecting the level of regional savings\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6702658/v1/9c8bbfec3a4197dbc36f3a2f.png"},{"id":83311398,"identity":"cf55dc84-1266-4f90-9d91-686c2ddb5d99","added_by":"auto","created_at":"2025-05-22 19:28:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":391484,"visible":true,"origin":"","legend":"\u003cp\u003eLogic diagram of AI indirectly affecting the level of regional savings\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6702658/v1/b68579c5470add0faf4e667c.png"},{"id":83311970,"identity":"ebb1e6fa-e8fa-41de-9072-a348bd3635b8","added_by":"auto","created_at":"2025-05-22 19:52:23","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2086691,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6702658/v1/39f12c72-7889-43ae-9dad-73cbb4236cba.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eDoes artificial intelligence improve or reduce the level of regional savings? Evidence from China\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eSavings are an important source of investment funds, and increasing savings capacity can enhance the role of investment in promoting economic growth (Makori et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) China, the world's largest developing country, has maintained high economic growth for more than three decades since the reform and opening up, and this growth has been mainly driven by investment. Investment comes from savings, and China's high savings over the past three decades have provided an endless source of investment funds for China's high economic growth (Zhao et al., \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). In recent years, China's savings have continued to rise (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Baker et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Can this continued increase be sustained? Over the past decade, artificial intelligence (AI) has developed rapidly with the development of related technologies such as machine learning, robotics, and neural networks (Yang, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). With the disappearance of the demographic dividend due to aging and the need for high-quality economic development under the new normal, the application of AI, such as industrial robots, has grown rapidly in the Chinese market (Han, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). AI has changed almost every aspect of people's lives (Haseeb et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and profoundly affected their behavior. Saving is an indispensable and important behavior in people's lives. China is a vast country, and regions rely on local savings to drive their economies. How exactly will AI affect regional savings in China? This is the topic of this paper.\u003c/p\u003e \u003cp\u003eAI makes capital substitutable for labor (Acemoglu \u0026amp; Restrepo, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), creating new jobs while eliminating existing ones (Acemoglu et al., 2018; Acemoglu \u0026amp; Restrepo, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). This raises concerns that labor will become increasingly redundant (Acemoglu et al., 2018; Acemoglu et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), thereby exposing people to uncertainty. In response to future uncertainty and precautionary motives, people increase their extra savings (Leland, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e1968\u003c/span\u003e). Digital technologies such as AI create many entrepreneurial opportunities, leading to the emergence of digital entrepreneurship (Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), which may increase the consumption of people's savings. Concerns arising from AI are always present, and digital entrepreneurship requires many conditions. Therefore, we arrive at the research hypothesis that AI first increases the level of regional savings and then gradually reduces the level of regional savings. In addition, digital technologies such as AI and big data drive the digital economy (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The digital economy can distribute digital information without restriction (Kajtazi, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2010\u003c/span\u003e), thereby reducing information asymmetry and intensifying competition for jobs. This, in turn, encourages people to save more to cope with employment uncertainty. Based on the reduction of information asymmetry, the digital economy also reduces people's liquidity and borrowing constraints, which in turn decreases savings (Wen, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Bussi\u0026egrave;re et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). For this reason, we propose that AI also indirectly affects regional savings levels by promoting the digital economy.\u003c/p\u003e \u003cp\u003eChina is a vast country with large disparities in the development level of each city. For this reason, this paper relies on the practice of Chinese scholars such as Hong\u0026amp; Yang (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e); and Xia et al. (\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) to obtain AI and rAI as proxy variables for the level of AI development by dividing the number of AI patents filed and granted each year in each city by the total population of the city, respectively. China's financial system is underdeveloped, and Chinese people's savings vehicles are relatively limited, with bank deposits being the main way for Chinese people to save. In this paper, we obtain \u003cem\u003eSAVE\u003c/em\u003e by dividing the balance of urban and rural residents' savings deposits by the total population of the city. Similarly, \u003cem\u003erSAVE\u003c/em\u003e is calculated by dividing the balance of all types of deposits, including those of individuals and other enterprises, by the city's total population, serving as proxy variables for the level of regional savings. Currently, there is no public data on the level of digital economy development in each city. Therefore, this paper uses factor analysis to derive the city's digital economy development index \u003cem\u003eDECO\u003c/em\u003e as a proxy variable, based on 11 data items including the number of data centers and digitized enterprises across three industries. Based on the construction of these variables, we adopt the city and year two-way fixed effects model to empirically test the impact of AI on the regional savings level, with the city's regional savings level as the dependent variable and the city's level of AI development as the independent variable. Based on the existing literature (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), we empirically test the transmission mechanism of AI indirectly affecting the regional savings level based on the digital economy by using the city's digital economy development level as the mediating variable. Finally, we conduct a heterogeneity test based on first-tier and non-first-tier cities.\u003c/p\u003e \u003cp\u003eBased on the above research design, we first observe that there is an 'inverted U' nonlinear relationship between AI and regional savings levels. There is an inflection point in the level of AI development, and on both sides of the inflection point, AI raises and lowers the level of regional savings. From a national perspective, the average value of the level of AI development is still below the inflection point, and it is still in the stage of raising the level of regional savings. However, in 2020, the development level of AI in Beijing and Shanghai crossed the inflection point and entered the stage of reducing the level of regional savings. Second, AI promotes the development of the digital economy, and the latter has an \"inverted U\" type nonlinear relationship with the level of regional savings. Therefore, AI indirectly affects the regional savings level by promoting the development of the digital economy. Third, the impact of AI on the regional savings level is heterogeneous. In the four first-tier cities of Beijing, Shanghai, Guangzhou, and Shenzhen, the inflection point of the level of AI development is lower, which means that from the perspective of development level, AI in first-tier cities will reduce the regional savings level earlier than in non-first-tier cities.\u003c/p\u003e \u003cp\u003eThe remainder of the paper is organized as follows: Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e articulates our testable hypothesis. Section \u003cspan refid=\"Sec11\" class=\"InternalRef\"\u003e3\u003c/span\u003e describes our econometric model, variables, and data. Section \u003cspan refid=\"Sec23\" class=\"InternalRef\"\u003e4\u003c/span\u003e discusses our main empirical results with robustness tests, conducts tests of the transmission mechanism heterogeneity analysis. Finally, we summarize the entire paper.\u003c/p\u003e"},{"header":"2. Testable Hypotheses","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Direct Impact of AI on the Level of Regional Savings\u003c/h2\u003e \u003cp\u003eBased on the existing results, we believe that AI has both positive and negative effects on the level of regional savings. AI can increase the level of regional savings or decrease the level of regional savings.\u003c/p\u003e \u003cdiv id=\"Sec4\" class=\"Section3\"\u003e \u003ch2\u003e2.1.1 AI Increases the Level of Regional Savings\u003c/h2\u003e \u003cp\u003eOn the one hand, AI increases precautionary savings among young people. First, the prevention of lower incomes increases precautionary savings among young people. AI forces young workers to compete with machines, increasingly diminishing their paychecks in relative or even absolute terms (Acemoglu et al., 2018). AI reduces young people's wages (Acemoglu et al., 2018). For example, one additional robot per 1,000 workers reduces average wages by 0.77% relative to areas without robots (Acemoglu et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). China is a developing country where wage income is the main source of income for young people. To cope with the decrease in wage income due to AI, young people need to save more to prevent insecurity. Second, preventing unemployment increases young people's precautionary savings. Routine tasks, defined as tasks that follow clear rules and can be performed by machines, are easily replaced by AI (Frey \u0026amp; Osborne, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). This increases the likelihood of unemployment among young people. AI provides greater output at lower cost to the extent that machines displace people (Haseeb et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), increasing the pressure on young people to become unemployed. For example, AI-based banks spend less time processing bill payments, accounts receivable, and other accounting tasks, reducing the number of accountants (Haseeb et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Firms using AI are differentially eliminating job postings that list a range of previously advertised skills, while at the same time posting skill requirements that were not previously listed. To prevent unemployment, young people need to save more. Finally, improving employability skills increases precautionary savings among young people. AI favors high-skilled workers, changing the structure of firms' labor demand (Yang, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In the case of China, research shows that the country has adopted more aggressive policies in AI development, but its overall employment structure is more vulnerable to AI disruption (Haseeb et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). AI eliminates existing jobs and adds new ones; for example, AI replaces the existing workforce while creating new jobs such as data managers and analysts (Acemoglu et al., 2018). To improve their employability skills, young people will need to save more to improve their employability skills in the future.\u003c/p\u003e \u003cp\u003eOn the other hand, AI increases precautionary savings for the elderly. First, it increases precautionary savings for their protection. The Chinese are deeply influenced by Confucianism. According to Confucian classics, children have an unconditional obligation to support their parents; therefore, older people in developing countries like China rely heavily on intergenerational transfers from adult children to their parents as a source of old age (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). AI reduces the incomes of young people, increases unemployment and pressure to upgrade employment skills, and leads to income volatility for adult children. Savings can help people manage income fluctuations and cushion against unexpected losses (Benami \u0026amp; Carter, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Therefore, to mitigate fluctuations in intergenerational transfers due to the varying incomes of adult children, older adults need to increase their precautionary savings. Second, the desire to provide for the love and care of their children motivates older adults to increase their precautionary savings. Chinese people have a strong sense of family, and older people (including the elderly) love and care for their children throughout their lives. AI, such as automation, will eliminate low-skill jobs and create new high-skill jobs (Acemoglu \u0026amp; Restrepo, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). To prepare children for a high-skill future, older adults will boost their savings to better fund the advanced skill development of their children. When children are of working age, older adults may also continue to save to enhance their children's employability. This is exemplified by the widespread phenomenon of \"old-age nibbling\" in China. After their children are employed, the elderly will continue to save to avoid the possibility of their children losing their jobs, or even leave their savings as an inheritance to their children.\u003c/p\u003e \u003cp\u003eIn summary, AI can increase precautionary savings among young and old people. This means that AI will increase the regional savings level in China.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section3\"\u003e \u003ch2\u003e2.1.2 AI Reduces the Level of Regional Savings\u003c/h2\u003e \u003cp\u003eAI promotes digital entrepreneurship, which, in turn, leads to reduced savings. The deep integration of digital technologies such as AI with the economy has created the digital economy (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In the process of deep integration with the economy, the application of digital technologies such as AI has generated digital entrepreneurship (Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Digital entrepreneurship is important for driving economic growth and is now one of the key priorities advocated by many national and international institutions, including the European Commission, the World Bank, and the Organization for Economic Cooperation and Development (Leong et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Digital entrepreneurship, defined as the intersection of digital technologies like AI and traditional entrepreneurship (Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), involves pursuing business opportunities that leverage these technologies (Davidson \u0026amp; Vaast, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). AI, along with other digital technologies, has led to the rapid digitization of products and services across industries (Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) and the emergence of digital markets. Digital markets reduce information asymmetries, facilitate information exchange, democratize entrepreneurship, and promote digital entrepreneurship (Aldrich, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). AI, along with other digital technologies, facilitates new digital infrastructures, such as digital marketplaces and 3D printing, which accelerate the iterative cycle of shaping, implementing, and modifying product ideas and business models, thereby facilitating digital entrepreneurship (Ries, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Nambisan, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Additionally, AI and other digital technologies have spawned new business models in sectors such as smart healthcare, bike sharing, and swipe-and-pay. These new businesses also contain many opportunities for digital entrepreneurship and are conducive to digital entrepreneurship. Under the \"mass entrepreneurship and innovation\" strategy, digital entrepreneurship has become an important choice for Chinese entrepreneurs. For most entrepreneurs, obtaining credit to start a business is a challenge (Mair \u0026amp; Marti, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2009\u003c/span\u003e; Bradley et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). They face greater income and expenditure uncertainty and higher credit constraints due to unstable careers and limited collateral, such as real estate, necessitating more precautionary savings to smooth these fluctuations (Guo \u0026amp; Gao, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eAI weakens rural labor migration in China and reduces savings. Starting about three decades ago, the accelerated pace of industrialization led to rapid urbanization and large-scale population mobility (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), and massive rural-urban labor migration emerged in China (Yin et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). China has a typical urban-rural \"dualistic\" structure. Although rural workers migrate to cities for employment opportunities, they are often excluded from local public services such as education, health care, and housing. They face more income and expenditure uncertainty and higher credit constraints with unstable careers and less collateral such as real estate, so they need more precautionary savings to smooth income and expenditure uncertainty (Guo \u0026amp; Gao, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Rural workers who migrate to cities still have their rural hukou, belong to the rural household registration, and still enjoy only a lower level of medical social security, so they have a higher propensity to save than urban residents; at the same time, their income in the city is close to or even exceeds that of some urban household residents, which is several times the income they received when they were originally employed in agriculture, so they have gradually freed themselves from the minimum level of consumption. Their real savings are substantial (Zhang et al., \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Consequently, the migration of rural workers to urban areas has significantly increased savings across China (Yin et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). AI is changing the labor demand structure of firms (Yang, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and eliminating low-skilled workers (Acemoglu \u0026amp; Restrepo, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Yang, \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Due to limited education and other factors, many rural workers migrating to urban areas are employed in low-skilled jobs that are highly susceptible to automation by AI. This suggests that AI's impact on the job market will diminish the migration of rural labor to urban areas in China, which in turn will reduce savings.\u003c/p\u003e \u003cp\u003eIn summary, AI reduces savings by promoting digital entrepreneurship and weakening labor migration from rural China, thereby reducing regional savings levels.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section3\"\u003e \u003ch2\u003e2.1.3 Research Hypothesis\u003c/h2\u003e \u003cp\u003eOverall, AI impacts regional savings through two opposing forces: one that increases savings and another that decreases them. The public's awareness of AI's potential, marked by the notable human-computer chess match between Deep Blue and Kasparov in February 1996, sparked early concerns about its implications. Since then, academics and the media have extensively covered the substitution effects of AI on labor. As a result, Chinese people's concern about the negative impact of AI on themselves and their children is much earlier. This means that AI has started to increase the regional savings level of Chinese people at the early stage of its development. In contrast, the entry of AI into generalized commercial applications started around 2008 (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). AI-generated digital entrepreneurship and weakened rural labor migration, which itself existed as a process later than 2008. As a result, the effect of AI in reducing the level of regional savings is delayed and is likely to become apparent as the level of AI development continues to increase. If this effect exceeds the effect of increasing per capita savings, AI may reduce the level of regional savings (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Therefore, we propose the following hypothesis:\u003c/p\u003e \u003cp\u003eH1: The relationship between AI and regional savings level is \"inverted U\" non-linear, there is a turning point in the level of AI development, and on both sides of the turning point, AI increases and decreases the level of regional savings, respectively.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Analysis of Action Channels Based on the Digital Economy\u003c/h2\u003e \u003cp\u003eIn 1996, Tapscott first proposed the concept of 'digital economy' in his book \"The Digital Economy: Hopes and Risks in the Age of Networked Intelligence\", defining it as the 'Internet economy,' a result of integrating Internet technology with the economy\". After 2008, AI and other digital technologies became more mature, and enterprises in the three major industries gradually integrated these digital technologies into their production and operation activities, giving rise to the digital economy that is currently being discussed by scholars, especially in China. Thus, the digital economy has taken on a new connotation, referring to the result of the deep integration of AI and other digital technologies with the economy. Digital technologies like AI are the technological drivers behind the digital economy. As a cutting-edge technology, AI is constantly evolving and rapidly iterating, significantly influencing the pace of digital economic development. The higher the level of AI development, the higher the level of development of the digital economy. Therefore, AI can promote the development of the digital economy.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e \u003ch2\u003e2.2.1 AI Promotes the Development of the Digital Economy and Increases Regional Savings\u003c/h2\u003e \u003cp\u003eAI is driving the digital economy, and the latter can increase regional savings.\u003c/p\u003e \u003cp\u003eFirst, the change in the employment environment increases savings. The digital economy spurs enterprises towards digital transformation, enhancing R\u0026amp;D, automating production, improving management efficiency, and refining services (Yang et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). As digital transformation progresses, there is a rising demand for technical and service-oriented, high-skilled employees, significantly displacing many production-oriented, low-skilled roles. This inevitably creates income uncertainty for production-oriented, low-skilled workers, prompting these workers to increase their savings to prevent the unexpected. On the other hand, the digital economy creates, stores, and distributes digital information through unlimited access (Kajtazi, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). In the digital economy environment, companies can distribute recruitment-related digital information, and workers can obtain appropriate and timely employment information through digital platforms, which reduces the information asymmetry in the labor market and increases the intensity of competition in the labor market. To cope with the uncertainty caused by the pressure of employment competition, people will increase their savings. Furthermore, digital platforms in the digital economy compile diverse economic data, broaden information channels, reduce transaction costs, and enhance labor mobility, contributing to more dynamic market conditions. This increased competition in the labor market consequently encourages greater savings among workers.\u003c/p\u003e \u003cp\u003eSecond, expanding the boundaries of financial services increases savings. China's development is unbalanced, with large development gaps between urban and rural areas. Financial institutions are often reluctant to serve people in remote and poor areas, and have few branches in underdeveloped areas, which lack convenient financial services such as cash deposit and withdrawal (Xie et al., \u003cspan citationid=\"CR55\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). In the digital economy environment, financial institutions are providing online financial services based on digital technologies such as AI to overcome traditional distance limitations (Li et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). For example, financial institutions use facial recognition and other technologies for user identity authentication, providing online payment services to residents in remote and underdeveloped areas. China traditionally relies on cash transactions. Without online payment services, residents in remote, poor, and underdeveloped areas must rely solely on cash for daily transactions. The digital economy's expansion into these areas, providing online payment services, encourages residents to deposit their cash in financial institutions for future use. The cash they collect in their daily transactions will also be deposited in financial institutions, which will increase savings deposits. Traditionally, financial institutions have primarily served wealthy customers, who conduct numerous small transactions at a high cost per transaction (Feng \u0026amp; Guo, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). These \"long tail\" individuals are often excluded from the financial system. The digital economy, characterized by its low threshold and inclusivity, promotes online financial services, significantly extending the reach and benefiting 'long-tail' groups traditionally excluded from the financial system (Jiang et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The traditional cash transactions of these groups are also completed through the financial system, thereby increasing savings deposits in the financial system.\u003c/p\u003e \u003cp\u003eFinally, increased competition among commercial banks increases savings. In the digital economy, the convergence of digital technology with traditional finance has altered the competitive landscape for banks, expanding beyond the market reach of traditional financial institutions. Emerging digital economy startups (e.g., nonbank payment institutions), with the potential advantages of wide coverage, low cost, and high efficiency, mainly serve micro and small enterprises and individual lending areas, eroding banks' lending market share, and the shrinking of banks' lending market share exacerbates competition for commercial banks' credit business (Feng \u0026amp; Guo, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Commercial banks are information-intensive enterprises (Berger, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2003\u003c/span\u003e), and the reduction of information asymmetry breaks the private information monopoly of commercial banks, reduces relationship-based lending based on formerly private information, reduces the monopoly power of banks, and increases the degree of competition among commercial banks (Berger, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Song et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Furthermore, the digital economy has digitized commercial banking services, reducing the physical and temporal barriers to financial transactions, broadening the competitive landscape, and intensifying bank competition (Jiang et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). China's financial system is underdeveloped, and bank deposits are the main source of savings for Chinese people. China's financial sector is compartmentalized, and commercial banks absorb savings deposits as their main source of funding. As deposit interest rates are already marketized, increased competition among commercial banks will encourage them to raise interest rates on savings deposits, which will increase people's propensity to save and thus increase savings deposits.\u003c/p\u003e \u003cp\u003eIn summary, the digital economy changes the employment environment, expands the boundaries of financial services, increases competition among commercial banks, and thus increases savings, and the digital economy ultimately increases the level of regional savings.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e \u003ch2\u003e2.2.2 AI Promotes the Development of the Digital Economy and Reduces Regional Savings\u003c/h2\u003e \u003cp\u003eAI is driving the digital economy, and the latter could also reduce regional savings.\u003c/p\u003e \u003cp\u003eFirst, encouraging consumption reduces savings. First, the digital economy improves consumption. The digital economy moves transactions online, promoting e-commerce, which overcomes spatial and temporal transaction barriers and enhances information availability. This enables consumers to easily access and utilize information when interested in specific products or services. Overall, the digital economy broadens consumption channels and enhances convenience, thereby improving the consumption experience and ultimately reducing savings (Dijk et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), which in turn improves the consumption experience promotes the consumption of residents, and reduces savings. Second, the digital economy reduces transaction costs for consumers. Transaction costs, which include counterparty search and information verification, account for about 35%-40% of total transaction activities (Dyer \u0026amp; Chu, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). The digital economy has the inherent advantages of intertemporal information dissemination and information sharing, which can reduce counterparty search costs and information verification costs, reduce consumers' transaction costs, and increase the marginal utility of consumption, thus promoting residents' consumption and reducing savings. Third, the digital economy brings enterprises and consumers closer together. The digital economy can improve information transparency (D'souza \u0026amp; Williams, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), strengthen the relationship between enterprises and consumers (Carlsson, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), enhance communication between enterprises and consumers, and encourage enterprises to better understand consumers' personalized consumption needs. Accordingly, enterprises can provide consumers with personalized products and services to maximize consumer satisfaction and increase consumers' marginal utility, thereby promoting consumption and reducing savings. In the digital economy, enterprises engage consumers through recommendation systems and precision marketing, effectively bridging the gap between them. Visual communication tools and situational consumption further enrich interactions between buyers and sellers, enhancing their willingness to consume (Yang et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), promoting residents' consumption, and reducing savings.\u003c/p\u003e \u003cp\u003eSecond, the expansion of social networks reduces savings. China is a traditional relational and human society (Bian, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), social networks, together with social rules and trust, are considered to belong to the category of social capital, which is crucial in determining people's socioeconomic status (Ma \u0026amp; Yang, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), and social networks can transmit a variety of information (Gompers et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), which has a significant impact on Chinese economic behavior (Yang et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). On one hand, social networks, connected by geography and blood ties, facilitate risk coordination through gift-giving and income transfer. This coordination helps mitigate the shocks from income and expenditure uncertainties, thereby weakening the incentive for precautionary savings and ultimately reducing overall savings(Yi et al., \u003cspan citationid=\"CR60\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). On the other hand, social networks have the effect of sharing information and reducing opportunism; people in a social network have more ex-ante information about each other's credit and are more likely to lend money to creditworthy people; ex-post, people in a social network are more likely to monitor the borrower's behavior after obtaining funds, which can solve the problem of hidden behavior. Thus, credit obtained through social networks can complement formal credit (Yang et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), and people with more social networks have more access to private loans (Ma \u0026amp; Yang, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2011\u003c/span\u003e), reducing people's borrowing constraints and thus reducing savings. The online, digitized, and data-enabled digital economy compresses spatial and temporal distances in daily activities, thereby strengthening social networks. This enhancement weakens precautionary saving incentives, eases borrowing constraints, and lowers the need for savings.\u003c/p\u003e \u003cp\u003eFinally, improved financial services reduce savings. Lack of credit or other financial buffers is an important cause of high savings (Baker et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The digital economy environment has enabled the development of mobile payments. Mobile payments on platforms such as WeChat and Alipay not only offer microcredit services but also enable financial institutions to gather users' credit information (Yin et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), thereby providing Chinese residents with credit and financial buffers that reduce the need for savings. As a key feature of the digital economy, mobile payments often involve small direct transfers that facilitate broader financial transactions, helping Chinese residents manage risks more effectively (Benami \u0026amp; Carter, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). At the same time, health risks, medical risks, unemployment risks, and income risks faced by households are major sources of uncertainty, and mobile payment platforms not only provide households with comprehensive services such as credit, insurance, and healthcare but also promote self-employment and help them become profitable. Credit services alleviate households' liquidity constraints, insurance services enable the settlement of claims after a household incurs a risk, healthcare services reduce the probability of health risks and medical risks, and employment enables households to gain more security (Yin et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In summary, as a crucial component of the digital economy, mobile payments enhance households' capacity to diversify risks (Jack \u0026amp; Suri, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Riley, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Yin et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), thereby reducing the need for precautionary savings.\u003c/p\u003e \u003cp\u003eOverall, by increasing consumption, expanding social networks, and strengthening financial services, the digital economy effectively reduces savings.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003e2.2.3 Research Hypothesis\u003c/h2\u003e \u003cp\u003eOverall, AI promotes the digital economy, which impacts regional savings through two opposing forces: one that increases savings and another that decreases them. The role of the digital economy on the level of regional savings depends on the confrontation of the two forces. Enterprises in the three industries are undergoing digital transformation based on digital technologies such as AI, thereby driving the digital economy. Digital transformation first creates demand for highly skilled personnel who are proficient in digital technologies such as AI. In this way, the digital economy will be the first to trigger changes in the employment environment and increase the level of regional savings. As the digital economy develops, the information environment will improve. Only when the digital economy promotes consumption, expands social networks, and strengthens financial services does it begin to reduce the level of regional savings. If the reduction in savings outweighs the increase, the digital economy will ultimately lead to lower regional savings levels (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). For this purpose, we propose the following hypothesis:\u003c/p\u003e \u003cp\u003eH2: AI indirectly influences regional savings by fostering the digital economy, which exhibits an 'inverted U' non-linear relationship with regional savings, mirroring the relationship between AI itself and regional savings.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Models\u003c/h2\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e3.1.1 Model of Direct Impact\u003c/h2\u003e \u003cp\u003eChinese scholars typically prefer city-level empirical analysis when studying regional phenomena in China. We investigate how AI affects the level of regional savings at the city level. To do so, we refer to the literature on regional savings (Jia \u0026amp; Han, \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) and design the following year and city two-way fixed effects model to test the direct effect of AI on regional savings levels:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:{SAVE}_{it}={\\alpha\\:}_{0}+{\\beta\\:}_{1}\\ast\\:{AI}_{it}+{\\beta\\:}_{2}\\ast\\:{AI2}_{it}+\\eta\\:\\ast\\:X+{\\alpha\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003ewhere i and t are the subscripts of city and time, respectively; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{i}\\)\u003c/span\u003e\u003c/span\u003e is to capture the city FE; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\lambda\\:}_{t}\\)\u003c/span\u003e\u003c/span\u003e is the capture year FE, and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\epsilon\\:}_{it}\\)\u003c/span\u003e\u003c/span\u003e is the random error term. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{SAVE}_{it}\\)\u003c/span\u003e\u003c/span\u003e is the dependent variable, that is, the level of regional savings of the city i in year t. \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{AI}_{it}\\)\u003c/span\u003e\u003c/span\u003e is an independent variable, that is, the development level of AI in the ith city in the tth year; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{AI2}_{it}\\)\u003c/span\u003e\u003c/span\u003e is the quadratic term and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{2}\\)\u003c/span\u003e\u003c/span\u003e is the coefficient of the quadratic line; if it is significantly negative, AI has an inverted U-shaped nonlinear effect on the level of regional savings. \u003cem\u003eX\u003c/em\u003e is the control variable, as shown below.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section3\"\u003e \u003ch2\u003e3.1.2 Model of Channel of Action\u003c/h2\u003e \u003cp\u003eTo test whether AI indirectly affects regional savings levels through the digital economy, we designed the following model based on the existing literature that studies the transmission mechanism (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e):\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ2\" name=\"EquationSource\"\u003e\n$$\\:{SAVE}_{it}={\\alpha\\:}_{0}+{\\beta\\:}_{1}\\ast\\:{AI}_{it}+{\\beta\\:}_{2}\\ast\\:{AI2}_{it}+\\eta\\:\\ast\\:X+{\\alpha\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:{DECO}_{it}={\\alpha\\:}_{0}+{\\beta\\:}_{1}\\ast\\:{AI}_{it}+\\eta\\:\\ast\\:Z+{\\alpha\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:{SAVE}_{it}={\\alpha\\:}_{0}+{\\beta\\:}_{1}\\ast\\:{AI}_{it}+{\\beta\\:}_{2}\\ast\\:{AI2}_{it}+{\\beta\\:}_{3}\\ast\\:{DECO}_{it}+{\\beta\\:}_{4}\\ast\\:{DECO2}_{it}+\\eta\\:\\ast\\:X+{\\alpha\\:}_{i}+{\\lambda\\:}_{t}+{\\epsilon\\:}_{it}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003e \u003cspan class=\"InlineEquation\"\u003e \u003cspan class=\"mathinline\"\u003e\\(\\:{DECO}_{it}\\)\u003c/span\u003e \u003c/span\u003e is the mediating variable (the digital economy level of the ith city in the tth year), and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{DECO2}_{it}\\)\u003c/span\u003e\u003c/span\u003e is a quadratic term for \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{DECO}_{it}\\)\u003c/span\u003e\u003c/span\u003e. \u003cem\u003eX\u003c/em\u003e in Equations (\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e) and (\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) is the control variable, which is the same as that in Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). \u003cem\u003eZ\u003c/em\u003e in Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) is the control variable, and financial efficiency is also controlled.\u003c/p\u003e \u003cp\u003eThe test procedure is as follows: First, without adding intermediary variables, we estimated Eq.\u0026nbsp;(\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e); if the coefficient of the level of AI development \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{1}\\)\u003c/span\u003e\u003c/span\u003e is significant, it shows that \u003cem\u003eAI\u003c/em\u003e has a total effect on the level of regional savings, and we continue with follow-up analysis; otherwise, it is a masking effect. Second, we estimated Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) to determine the impact of \u003cem\u003eAI\u003c/em\u003e on the mediating variables. Third, after adding the intermediary variable, we estimated Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). If the coefficient of the level of AI development in Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) is \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{1}\\)\u003c/span\u003e\u003c/span\u003e, and the coefficient of the intermediary variable in Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) is \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{3}\\)\u003c/span\u003e\u003c/span\u003e, it shows that the intermediary effect exists. Fourth, if \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{1}\\)\u003c/span\u003e\u003c/span\u003e in Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\beta\\:}_{3}\\)\u003c/span\u003e\u003c/span\u003e in Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) are significant, the mediating effect must be tested using the Sobel test.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Variables\u003c/h2\u003e \u003cp\u003eBased on the existing literature, this study's independent, dependent, mediating, and control variables are presented in Appendix 1.\u003c/p\u003e \u003cdiv id=\"Sec16\" class=\"Section3\"\u003e \u003ch2\u003e3.2.1 Dependent Variables\u003c/h2\u003e \u003cp\u003eThe dependent variable in this paper is the regional level of savings (\u003cem\u003eSAVE\u003c/em\u003e). Saving is a concept corresponding to residents' income and residents' consumption, and some scholars study the savings level based on household microdata, usually calculating the residents' savings level according to (disposable income - consumption)/disposable income (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wu et al., \u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Baker et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Choukhmane et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This paper examines regional savings at the macro level. Given the underdevelopment of China's financial system, Chinese residents have limited savings options, primarily relying on bank deposits. When Chinese scholars study the macro-level saving behavior of Chinese residents, they use per capita savings deposit balances to measure regional savings levels (Luo et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Luo \u0026amp; Wen, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2016\u003c/span\u003e; Wang \u0026amp; Liao, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Gao \u0026amp; Shi, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Therefore, in this paper, \u003cem\u003eSAVE\u003c/em\u003e is calculated by dividing the household savings deposit balances by the total population in each city, serving as a proxy for regional savings levels. For robustness testing, \u003cem\u003erSAVE\u003c/em\u003e is obtained by dividing all types of deposits (including self-employed and other business deposits) by the total population in each city as another proxy variable for the regional savings level.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section3\"\u003e \u003ch2\u003e3.2.2 Independent Variables\u003c/h2\u003e \u003cp\u003eThe independent variable in this paper is the level of AI development (\u003cem\u003eAI\u003c/em\u003e). Referring to the existing literature by Hong\u0026amp; Yang (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), and Hong\u0026amp; Yang (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2021\u003c/span\u003e); Xia et al. (\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), The level of AI development (\u003cem\u003eAI\u003c/em\u003e) is measured by dividing the number of AI patent applications by the total population in each city, serving as a proxy for this variable.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section3\"\u003e \u003ch2\u003e3.2.3 Intermediary Variables\u003c/h2\u003e \u003cp\u003eThe mediating variable in this paper is the level of digital economy development (\u003cem\u003eDECO\u003c/em\u003e). At present, there is no publicly available official data on the level of digital economy development in cities. This study constructed the indicator system shown in Appendix 2, collecting data from various cities via the State Administration of Market Supervision and Administration. It then used factor analysis to calculate the city digital economy development index score, normalizing it by (Score-Min)/(Max-Min) to obtain \u003cem\u003eDECO\u003c/em\u003e as a proxy for the development of the digital economy.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section3\"\u003e \u003ch2\u003e3.2.4 Control Variables\u003c/h2\u003e \u003cp\u003eReferring to the existing literature (Chen et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Makori et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), this paper includes control variables such as the level of economic development, the rate of economic growth, the level of foreign investment, the level of industrial structure, fiscal expenditures on science and technology, the rate of urbanization, population size, population density, and the level of financial development.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Data\u003c/h2\u003e \u003cdiv id=\"Sec21\" class=\"Section3\"\u003e \u003ch2\u003e3.3.1 Data Source\u003c/h2\u003e \u003cp\u003eSince the global financial crisis in 2008, digital technologies such as AI and the economy have begun to be deeply integrated (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), and AI has gradually entered large-scale commercial applications. China is a vast country with unbalanced development, and it is the preferred choice of Chinese scholars to study Chinese issues at the city level. The data of the China City Statistical Yearbook will be updated to 2020. Therefore, this paper empirically analyzes Chinese city data from 2008 to 2020. In this paper, the data are processed as follows: (1) missing samples are eliminated; (2) the China City Statistical Yearbook no longer reports foreign direct investment (\u003cem\u003eFDI\u003c/em\u003e) in cities in 2020, so \u003cem\u003eFDI\u003c/em\u003e in 2020 is linearly interpolated; (3) household savings deposits in 2010 are linearly interpolated. Subsequently, the 2010 and 2020 data are excluded for robustness testing. We end up with 3,362 year-city observations.\u003c/p\u003e \u003cp\u003eThe number of AI patent applications and the number of approved patents are from the State Intellectual Property Office (SIPO), and we extracted them with the keyword \"AI\", while the other data are from China City Statistical Yearbook. In addition, taking the natural logarithm can eliminate the influence of outliers, to eliminate the influence of outliers, this paper takes the natural logarithm of other continuous variables other than the upper and lower 1% Winsorize shrinkage treatment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section3\"\u003e \u003ch2\u003e3.3.2 Summary Statistics\u003c/h2\u003e \u003cp\u003eThe descriptive statistics of variables are shown in Appendix 3. The gap between the minimum and the maximum of each variables is significant, consistent with China's primary national conditions of unbalanced development.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Direct Impact\u003c/h2\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003e4.1.1 Univariable Analysis\u003c/h2\u003e \u003cp\u003eThe results of univariable analysis are shown in Appendix 4.1. The results shown that there is a relatively obvious \"inverted U\" nonlinear relationship between AI and regional savings level.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003e4.1.2 Multivariable Regression\u003c/h2\u003e \u003cp\u003eControlling for city and year fixed effects, with \u003cem\u003eSAVE\u003c/em\u003e as the dependent variable and \u003cem\u003eAI\u003c/em\u003e and \u003cem\u003erAI\u003c/em\u003e as the independent variables, respectively, we estimate Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) with the FE fixed effects model without adding the control variables, resulting in Columns (1) and (2) of Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. Re-estimating Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) with all the control variables added results in Columns (3) and (4) of Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e. From Columns (1)-(4) of Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the coefficients of the quadratic terms of the level of AI development are all significantly negative at the 1% significance level, while the coefficients of the primary terms are all significantly positive at the 1% significance level, and there is an \"inverted U\" nonlinear relationship between AI and the level of regional savings. These empirical results support Hypothesis H1 and align with the theoretical analysis presented in the previous section. AI has both positive and negative effects on the level of regional savings, and AI has been used in large-scale general commercial applications since 2008 (Chen et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al. \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), while the negative effect is later, resulting in an \"inverted U\" nonlinear relationship between AI and the level of regional savings. The relationship between AI and regional savings levels is an \"inverted U\" nonlinear.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\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\u003eFE estimation results of Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8830***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.9935***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.1009)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0801)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0809***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0797***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0130)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0086)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.0186***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4.0481***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.4064)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.3359)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-1.8517***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.4667***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.3027)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.2064)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ePGDP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.6968***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.6086***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.3863)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.3675)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eGGDP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0114\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0108\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0085)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0080)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eFDI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.9400***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.8907***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.2786)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.2596)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eINDS\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0717\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.0401\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.1350)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.1319)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eSCIP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0107**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.0144***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0051)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0051)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eCITY\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0472\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.2201\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.5780)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.5351)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ePOP\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-4.3912***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-4.1473***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.8149)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.7747)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003ePDEN\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.1610*\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-1.2070*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.6639)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.6686)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eFSIZ\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.0042**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0043**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0021)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0021)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConstant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.6856***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.6836***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e40.7377***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e39.0235***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0515)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0518)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(4.8636)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4.6069)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCity FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8029\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.8075\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.8539\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.8553\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"5\"\u003eNote: Robust standard errors are in parentheses; * p\u0026thinsp;\u0026lt;\u0026thinsp;0.1, * * p\u0026thinsp;\u0026lt;\u0026thinsp;0.05, * p\u0026thinsp;\u0026lt;\u0026thinsp;0.01, same as below.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003e4.1.3 Robustness Checks\u003c/h2\u003e \u003cp\u003eThe results of robustness checks are shown in Appendix 4.2. In summary, the conclusion that H1 is established is robust under the conditions of controlling endogeneity, changing the measurement of regional savings level, excluding interpolated data, using double clustering robust standard errors, and controlling for the level of digital economy development.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Conduction Mechanism\u003c/h2\u003e \u003cdiv id=\"Sec29\" class=\"Section3\"\u003e \u003ch2\u003e4.2.1 Univariate Analysis\u003c/h2\u003e \u003cp\u003e \u003cstrong\u003eThe results of univariable analysis are shown in Appendix 5. The results shown that\u003c/strong\u003e \u003cp\u003eAs the level of AI development increases, the level of digital economic development improves; As the level of digital economy development increases, the level of regional savings first increases and then decreases.\u003c/p\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec30\" class=\"Section3\"\u003e \u003ch2\u003e4.2.2 Multiple Regression\u003c/h2\u003e \u003cp\u003eEquations\u0026nbsp;(\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)-(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) are estimated using FE with \u003cem\u003eSAVE\u003c/em\u003e as the dependent variable and \u003cem\u003eDECO\u003c/em\u003e as the mediator variable, and the results are shown in Columns (1)-(3) of Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Panel A. In Column (1) of Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e Panel A, the coefficient on the level of AI development (\u003cem\u003eAI\u003c/em\u003e) is significant at the 1% significance level, indicating the existence of the overall effect; the coefficient on the level of AI development (\u003cem\u003eAI\u003c/em\u003e) is significant at the 1% significance level in Column (2), and the coefficient on the mediator variable, the level of development of the digital economy (\u003cem\u003eDECO\u003c/em\u003e), is significant at the 1% significance level in Column (3), so there is a mediating effect. Combining the sign and significance of the level of AI development (\u003cem\u003eAI\u003c/em\u003e) and the level of digital economy development (\u003cem\u003eDECO\u003c/em\u003e) in Columns (2) and (3), it can be seen that there is an \"inverted U\" shaped non-linear relationship between AI and the level of regional savings by promoting the development of digital economy. Therefore, the research hypothesis H2 is established.\u003c/p\u003e \u003cp\u003eIn estimating Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) with the level of digital economic development (\u003cem\u003eDECO\u003c/em\u003e) as the dependent variable and the level of AI development (\u003cem\u003eAI\u003c/em\u003e) as the independent variable, similar to section 4.4.1, the level of AI development (\u003cem\u003eAI\u003c/em\u003e) is endogenous, and using the same instrumental variable \u003cem\u003eivAI\u003c/em\u003e as in 4.4.1, we reestimate Eq.\u0026nbsp;(\u003cspan refid=\"Equ3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) using IV, and the results are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, Panel A, Column (4). In estimating Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), the level of digital economic development (\u003cem\u003eDECO\u003c/em\u003e) is also likely to be endogenous due to measurement error, following Laeven\u0026amp; Levine (2007), Laeven\u0026amp; Levine (2009), Faccio et al. (2011), Chen et al. (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and Chen (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Calculate the mean of the digital economy development index of other cities for the same year and its quadratic term to get \u003cem\u003eivDECO\u003c/em\u003e and \u003cem\u003eivDECO2\u003c/em\u003e as instrumental variables. Taking \u003cem\u003eivAI\u003c/em\u003e, \u003cem\u003eivAI2\u003c/em\u003e, \u003cem\u003eivDECO\u003c/em\u003e, and \u003cem\u003eivDECO2\u003c/em\u003e as instrumental variables, we re-estimate Eq.\u0026nbsp;(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) using IIV, and the results are shown in Column (5) of Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e Panel A. The results are summarized as follows. From Columns (4) and (5) of Panel A of Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, we can see that AI forms an \"inverted U\" nonlinear relationship with regional savings levels by promoting the development of the digital economy. Therefore, the conclusion that the research hypothesis H2 holds is robust after excluding endogeneity.\u003c/p\u003e \u003cp\u003eReplacing the independent and dependent variables with \u003cem\u003erSAVE\u003c/em\u003e and \u003cem\u003erAI\u003c/em\u003e, respectively, and estimating Equations (\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)-(\u003cspan refid=\"Equ4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) in the above order, the results are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e Panels B and C. From Panels B and C, AI forms an \"inverted U\" type nonlinear relationship with regional savings levels by promoting the development of the digital economy. Therefore, the conclusion that research hypothesis H2 is valid is robust.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\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\u003eEstimation results of the conduction mechanism\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"6\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePanel A\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.9935***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0535***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.7411***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0506***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1.8992***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0801)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0080)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0936)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0053)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.1710)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0797***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.0610***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.1643***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0086)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0097)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.0198)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.0878***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4.3383**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.0669)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(1.9738)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-3.8566***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-5.3214**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.0389)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(2.3473)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCity FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eControl\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8539\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.6819\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.8598\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.6811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.9548\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePanel B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.8188***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.0535***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e2.6763***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.0506***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7.6254***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.2839)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0080)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.4322)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0053)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.7711)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.2625***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.1622***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-0.6256***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0343)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.0484)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.0938)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e24.6907***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e32.8475***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(4.7595)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(7.8951)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-22.1700***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-38.5346***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(5.3403)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(9.4705)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCity FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eControl\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.7069\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.6819\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.7308\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.6811\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.9348\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePanel C\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(5)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.0481***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.2518***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.1642***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.2937***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8.9577***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.3359)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0370)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.3633)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(0.0355)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.8023)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.4667***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.2309***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-3.7352***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.2064)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.2221)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(0.4955)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e5.4012***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4.0063**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.0322)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(1.8929)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eDECO2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-3.8739***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e-4.4664*\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.1082)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e(2.4570)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCity FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eControl\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8553\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.6444\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.8632\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.6384\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.9507\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"6\"\u003eNote: All instrumental variables have passed validity tests.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Regional Heterogeneity Analysis\u003c/h2\u003e \u003cp\u003ePrevious analyses indicate that AI exerts both positive and negative effects on regional savings levels. Chinese people worry about the negative impact of AI on themselves and their children much earlier. This suggests that AI began to increase regional savings levels in China early in its development. In contrast, the entry of AI into large-scale generalized commercial applications started around 2008 (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). There is a process of digital entrepreneurship and weakened rural labor migration generated by AI itself, later than 2008. As a result, the role of AI in reducing the regional savings level is later. Therefore, the relationship between AI and regional savings level is \"inverted U\" non-linear. China is a huge country with unbalanced development. Beijing, Shanghai, Guangzhou, and Shenzhen are first-tier cities, while the rest are non-first-tier cities. First-tier cities are economically developed and have more job opportunities. As a result, people in first-tier cities are relatively less concerned about the negative impact of AI on themselves and their children. Consequently, the positive impact of AI is relatively weaker in first-tier cities. At the same time, first-tier cities are rich in human and economic resources and have more entrepreneurial opportunities, and more rural workers have migrated to first-tier cities. This means that the negative force of AI is relatively stronger in first-tier cities. Weaker positive effects and stronger negative effects suggest a lower inflection point in the level of AI development. Therefore, we hypothesize that there is heterogeneity in the role of AI in influencing regional savings levels and that the inflection point of the level of AI development is lower in first-tier cities.\u003c/p\u003e \u003cp\u003eTo test this speculation, we estimate Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) using a variable coefficient individual fixed effects model with \u003cem\u003eSAVE\u003c/em\u003e and \u003cem\u003erSAVE\u003c/em\u003e as the dependent variables and \u003cem\u003eAI\u003c/em\u003e as the independent variable, and the results are in Columns (1) and (2) of Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) is re-estimated by replacing the independent variables with \u003cem\u003erAI\u003c/em\u003e, and the results are shown in Columns (3) and (4) of Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e. From the estimation results, first, the coefficients of the quadratic terms of the level of AI development are all significantly negative at the 1% significance level. Second, the inflection point of the level of AI development is lower in first-tier cities than in non-first-tier cities (Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\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\u003eHeterogeneity of first-tier and not-first- tier cities\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(1)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(2)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(3)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(4)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariables\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI\u003c/em\u003e (not first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.9230***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3.4845***\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0775)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.2711)\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI\u003c/em\u003e (first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.5646***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.4092***\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.2296)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.3046)\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI2\u003c/em\u003e (not first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0695***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.2198***\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0092)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.0467)\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAI2\u003c/em\u003e (first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.1389***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.6156***\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(0.0221)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.1130)\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 \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI\u003c/em\u003e (not first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.7310***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e13.2891***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.3391)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(1.4410)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI\u003c/em\u003e (first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.4256***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35.4233***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(1.8745)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(8.9790)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI2\u003c/em\u003e (not first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-1.2253***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-3.7437***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.2337)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(1.2542)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003erAI2\u003c/em\u003e (first tier)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-3.2722***\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-14.4583***\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e(0.8641)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e(3.8345)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCity FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYear FE\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eControl\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eObservations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3,362\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eR-squared\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.8551\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.7127\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.8571\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.6954\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eN\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e282\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\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe inflection point of \u003cem\u003eAI\u003c/em\u003e in first-tier cities and not-first-tier cities\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\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 \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003ePanel A\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e is an independent variable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e is an independent variable\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot first tier\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFirst tier\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot first tier\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFirst tier\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient of \u003cem\u003eAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.9230\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.5646\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e3.4845\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7.4092\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient of \u003cem\u003eAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.0695\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-0.1389\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-0.2198\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-0.6156\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThe inflection point of \u003cem\u003eAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.6403\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5.6321\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7.9265\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6.0179\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003ePanel B\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003e\u003cem\u003eSAVE\u003c/em\u003e is an independent variable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c5\" namest=\"c4\"\u003e \u003cp\u003e\u003cem\u003erSAVE\u003c/em\u003e is an independent variable\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot first tier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFirst tier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot first tier\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eFirst tier\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient of \u003cem\u003erAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e3.7310\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e7.4256\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e13.2891\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e35.4233\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoefficient of \u003cem\u003erAI2\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-1.2253\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e-3.2722\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e-3.7437\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e-14.4583\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThe inflection point of \u003cem\u003erAI\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.5225\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.1346\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1.7749\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.2250\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusion","content":"\u003cp\u003eOur paper investigates AI's impact on regional savings levels, analyzing both the direct impacts and action channels theoretically, then empirically testing these theories using a sample from Chinese cities between 2008 and 2020 with city and year fixed effects. We find, firstly, an 'inverted U' nonlinear relationship between AI and regional savings levels; as AI technology improves, regional savings initially increase and then decrease. Overall, AI is still in the stage of increasing regional savings, but some cities have entered the stage of decreasing regional savings. Second, AI promotes the development of a digital economy, and the latter has an \"inverted U\" nonlinear relationship with regional savings level. Third, the impact of AI on the regional savings level is heterogeneous, and the inflection point of AI technology is lower in first-tier cities.\u003c/p\u003e \u003cp\u003eBased on the fact that savings are the financial driver of economic development and AI is a brand new driver of economic growth, our paper examines the impact of AI on regional savings levels. In addition, there are the following aspects that are worth studying in depth. First, the impact of AI on the lending behavior of financial institutions. The deep integration of digital technologies such as AI and the economy has created a digital economy (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Chen, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), which can reduce information asymmetry and improve the quality of credit assets of financial institutions (Chen et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). How does this affect the lending behavior of financial institutions, and what mechanisms are involved? These questions need to be studied in depth. Second, the impact of AI on household microbehavior. In this paper, we have focused on AI's impact on city-level savings from a macro perspective. At the micro level, how does AI influence family fertility behaviors? Will families choose to have fewer children as AI adoption accelerates and machine replacement of human labor becomes more common, subsequently impacting their savings behaviors? These questions have important implications for long-term economic growth. The fact that this question has not been explored in this paper is a shortcoming of this paper and one of the directions for future research.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor contributions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConceptualization: Hongwei Zhang, Xiaohui Chen, Xiaohui Luo；Methodology: Xiaohui Luo\u003c/p\u003e\n\u003cp\u003eInvestigation: Xiaohui Luo\u003c/p\u003e\n\u003cp\u003eVisualization: Xiaohui Chen\u003c/p\u003e\n\u003cp\u003eSupervision: Xiaohui Luo, Hongwei Zhang, Xiaohui Chen\u003c/p\u003e\n\u003cp\u003eWriting\u0026mdash;original draft: Hongwei Zhang\u003c/p\u003e\n\u003cp\u003eWriting\u0026mdash;review \u0026amp; editing: Hongwei Zhang, Xiaohui Luo, Xiaohui Chen\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003enone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number:\u003c/strong\u003e not applicable.\u003c/p\u003e\n"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAcemoglu \u0026amp; Daron \u0026amp; Restrepo \u0026amp; Pascual (2018). \u0026quot;The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment.\u0026quot; \u003cu\u003eAmerican Economic Review\u003c/u\u003e.\u003c/li\u003e\n\u003cli\u003eAcemoglu, D. \u0026amp; D. H. Autor \u0026amp; J. Hazell \u0026amp; P. Restrepo (2020). \u0026quot;Ai and Jobs: Evidence from Online Vacancies.\u0026quot; \u003cu\u003eSocial Science Electronic Publishing\u003c/u\u003e.\u003c/li\u003e\n\u003cli\u003eAcemoglu, D. \u0026amp; P. Restrepo (2019). \u0026quot;Automation and New Tasks: How Technology Displaces and Reinstates Labor.\u0026quot; \u003cu\u003eJournal of Economic Perspectives\u003c/u\u003e \u003cstrong\u003e33\u003c/strong\u003e(2): 3-30.\u003c/li\u003e\n\u003cli\u003eAldrich, H. E. (2014). \u003cu\u003eThe Democratization of Entrepreneurship? Hackers, Makerspaces, and Crowdfunding\u003c/u\u003e. Academy of Management Meeting.\u003c/li\u003e\n\u003cli\u003eBaker, S. R. \u0026amp; E. Benmelech \u0026amp; Z. Yang \u0026amp; Q. Zhang (2023). \u0026quot;The Determinants of The Chinese Household Savings Rate1.\u0026quot;\u003c/li\u003e\n\u003cli\u003eBenami, E. \u0026amp; M. R. Carter (2021). \u0026quot;Can digital technologies reshape rural microfinance? Implications for savings, credit, \u0026amp; insurance.\u0026quot; \u003cu\u003eApplied Economic Perspectives and Policy\u003c/u\u003e.\u003c/li\u003e\n\u003cli\u003eBerger, A. N. (2003). \u0026quot;The economic effects of technological progress: Evidence from the banking industry.\u0026quot; \u003cu\u003eJournal of Money, credit and Banking\u003c/u\u003e: 141-176.\u003c/li\u003e\n\u003cli\u003eBian, Y. (1997). \u0026quot;Bringing strong ties back in: Indirect ties, network bridges, and job searches in China.\u0026quot; \u003cu\u003eAmerican sociological review\u003c/u\u003e: 366-385.\u003c/li\u003e\n\u003cli\u003eBradley, S. W. \u0026amp; J. S. Mcmullen \u0026amp; K. Artz \u0026amp; E. M. Simiyu (2012). \u0026quot;Capital Is Not Enough: Innovation in Developing Economies.\u0026quot; \u003cu\u003eJournal of Management Studies\u003c/u\u003e \u003cstrong\u003e49\u003c/strong\u003e(4).\u003c/li\u003e\n\u003cli\u003eBussi\u0026egrave;re, M. \u0026amp; Y. Kalantzis \u0026amp; R. Lafarguette \u0026amp; T. Sicular (2013). \u0026quot;Understanding household savings in China: the role of the housing market and borrowing constraints.\u0026quot; \u003cu\u003eMPRA Paper\u003c/u\u003e.\u003c/li\u003e\n\u003cli\u003eCao, W. \u0026amp; G. Liu \u0026amp; L. Zeng \u0026amp; Y. Feng (2023). \u0026quot;The Impact of the Integration of Family Pension and Social Pension on Residents\u0026rsquo; Savings Rate.\u0026quot; \u003cu\u003eEconomic Research Journal\u003c/u\u003e \u003cstrong\u003e58\u003c/strong\u003e(3): 172-190.\u003c/li\u003e\n\u003cli\u003eCarlsson, B. (2004). \u0026quot;The Digital Economy: what is new and what is not?\u0026quot; \u003cu\u003eStructural change and economic dynamics\u003c/u\u003e \u003cstrong\u003e15\u003c/strong\u003e(3): 245-264.\u003c/li\u003e\n\u003cli\u003eChen, X. (2023). \u0026quot;Information moderation principle on the regulatory sandbox.\u0026quot; \u003cu\u003eEconomic Change and Restructuring\u003c/u\u003e \u003cstrong\u003e56\u003c/strong\u003e(1): 111-128.\u003c/li\u003e\n\u003cli\u003eChen, X. \u0026amp; L. Teng \u0026amp; W. Chen \u0026amp; H. Beladi (2022). \u0026quot;How does FinTech affect the development of the digital economy? Evidence from China.\u0026quot; \u003cu\u003eThe North American Journal of Economics and Finance\u003c/u\u003e \u003cstrong\u003e61\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eChen, X. \u0026amp; D. Yan \u0026amp; W. Chen (2022). \u0026quot;Can the digital economy promote FinTech development?\u0026quot; \u003cu\u003eGrowth and Change\u003c/u\u003e \u003cstrong\u003e53\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eChen, X. \u0026amp; H. Zhang \u0026amp; Y. Wu (2020). \u0026quot;How does the digital economy affect the level of industrial structure?\u0026quot; \u003cu\u003eSecurities Market Herald\u003c/u\u003e(7): 20-29.\u003c/li\u003e\n\u003cli\u003eChen, Y. J. \u0026amp; Z. Chen \u0026amp; S. He (2019). \u0026quot;Social Norms and Household Savings Rates in China.\u0026quot;\u003c/li\u003e\n\u003cli\u003eChoukhmane, T. \u0026amp; N. Coeurdacier \u0026amp; K. Jin (2023). \u0026quot;The one-child policy and household saving.\u0026quot; \u003cu\u003eJournal of the European Economic Association\u003c/u\u003e: jvad001.\u003c/li\u003e\n\u003cli\u003eCui, F. \u0026amp; J. Yang \u0026amp; K. Fu (2019). \u0026quot;The Heterogeneous Influence of Rural Labor Transferon Household Saving Rate.\u0026quot; \u003cu\u003eJournal of Statistics and Information\u003c/u\u003e \u003cstrong\u003e34\u003c/strong\u003e(12): 40-49.\u003c/li\u003e\n\u003cli\u003eD\u0026apos;souza, C. \u0026amp; D. Williams (2017). \u0026quot;The digital economy.\u0026quot; \u003cu\u003eBank of Canada Review\u003c/u\u003e \u003cstrong\u003e2017\u003c/strong\u003e(Spring): 5-18.\u003c/li\u003e\n\u003cli\u003eDavidson, E. \u0026amp; E. Vaast (2010). \u003cu\u003eDigital entrepreneurship and its sociomaterial enactment\u003c/u\u003e. 2010 43rd Hawaii International Conference on System Sciences, IEEE.\u003c/li\u003e\n\u003cli\u003eDijk, G. V. \u0026amp; S. Minocha \u0026amp; A. Laing (2007). \u0026quot;Consumers, channels and communication: Online and offline communication in service consumption.\u0026quot; \u003cu\u003eInteracting with Computers\u003c/u\u003e \u003cstrong\u003e19\u003c/strong\u003e(1): 7-19.\u003c/li\u003e\n\u003cli\u003eDyer, J. H. \u0026amp; W. Chu (2003). \u0026quot;The Role of Trustworthiness in Reducing Transaction Costs and Improving Performance: Empirical Evidence from the United States, Japan, and Korea.\u0026quot; \u003cu\u003eOrganization Science\u003c/u\u003e \u003cstrong\u003e14\u003c/strong\u003e(1): 57-68.\u003c/li\u003e\n\u003cli\u003eFeng, S. \u0026amp; R. Guo (2019). \u0026quot;Digital Finance, Bank Competition and Bank Efficiency.\u0026quot; \u003cu\u003eReform\u003c/u\u003e(11): 75-89.\u003c/li\u003e\n\u003cli\u003eFrey, C. B. \u0026amp; M. A. Osborne (2017). \u0026quot;The future of employment: How susceptible are jobs to computerisation?\u0026quot; \u003cu\u003eTechnological Forecasting and Social Change\u003c/u\u003e \u003cstrong\u003e114\u003c/strong\u003e: 254-280.\u003c/li\u003e\n\u003cli\u003eGao, J. \u0026amp; Q. Shi (2023). \u0026quot;Financial Behavior Changes of Ｒural Households in China under the Background of Urbanization: Based on the Data of Ten Villages in Zhejiang.\u0026quot; \u003cu\u003eIssues in Agricultural Economy\u003c/u\u003e(02): 89-106.\u003c/li\u003e\n\u003cli\u003eGompers, P. \u0026amp; J. Lerner \u0026amp; D. Scharfstein (2005). \u0026quot;Entrepreneurial spawning: Public corporations and the genesis of new ventures, 1986 to 1999.\u0026quot; \u003cu\u003eThe journal of Finance\u003c/u\u003e \u003cstrong\u003e60\u003c/strong\u003e(2): 577-614.\u003c/li\u003e\n\u003cli\u003eGuo, Q. \u0026amp; X. Gao (2018). \u0026quot;Multidimensional Examination of Migrant Workers\u0026rsquo; Semi-urbanization: Influencing Factors and Basic Outlet.\u0026quot; \u003cu\u003eJournal of Northwest A\u0026amp;F University(Social Science Edition)\u003c/u\u003e \u003cstrong\u003e18\u003c/strong\u003e(03): 22-30.\u003c/li\u003e\n\u003cli\u003eHan, Y. (2022). \u0026quot;The Impact of Industrial Robots on the Skill-Based Wage Gap.\u0026quot; \u003cu\u003eAmerican Journal of Industrial and Business Management\u003c/u\u003e \u003cstrong\u003e12\u003c/strong\u003e(4): 571-602.\u003c/li\u003e\n\u003cli\u003eHaseeb, M. \u0026amp; S. Sasmoko \u0026amp; L. Mihardjo \u0026amp; K. Jermsittiparsert (2019). \u0026quot;Economic Impact of Artificial Intelligence: New Look for the Macroeconomic Assessment in Asia-Pacific Region.\u0026quot; \u003cu\u003eInternational Journal of Computational Intelligence Systems\u003c/u\u003e \u003cstrong\u003e12\u003c/strong\u003e(2): 1295-1310.\u003c/li\u003e\n\u003cli\u003eHojeghan, S. B. \u0026amp; A. N. Esfangareh (2011). \u0026quot;Digital economy and tourism impacts, influences and challenges.\u0026quot; \u003cu\u003eProcedia - Social and Behavioral Sciences\u003c/u\u003e \u003cstrong\u003e19\u003c/strong\u003e(none): 308-316.\u003c/li\u003e\n\u003cli\u003eHong, J. \u0026amp; Z. Yang (2021). \u0026quot;Dialect Boundaries，Communication Barriers and Innovation Performances.\u0026quot; \u003cu\u003eNankai Economic Studies\u003c/u\u003e(02): 46-66.\u003c/li\u003e\n\u003cli\u003eJack, W. \u0026amp; T. Suri (2014). \u0026quot;Risk Sharing and Transactions Costs: Evidence from Kenya\u0026apos;s Mobile Money Revolution.\u0026quot; \u003cu\u003eAmerican Economic Review\u003c/u\u003e \u003cstrong\u003e104\u003c/strong\u003e(1): 183-223.\u003c/li\u003e\n\u003cli\u003eJia, D. \u0026amp; H. Han (2023). \u0026quot;Fintech, Bank Liabilities and Competition.\u0026quot; \u003cu\u003eThe Journal of World Economy\u003c/u\u003e \u003cstrong\u003e46\u003c/strong\u003e(02): 183-208.\u003c/li\u003e\n\u003cli\u003eJiang, E. X. \u0026amp; G. Y. Yu \u0026amp; J. Zhang (2022). \u0026quot;Bank Competition amid Digital Disruption: Implications for Financial Inclusion.\u0026quot; \u003cu\u003eAvailable at SSRN 4178420\u003c/u\u003e.\u003c/li\u003e\n\u003cli\u003eKajtazi (2010). \u003cu\u003eInformation asymmetry in the digital economy\u003c/u\u003e. International Conference on Information Society.\u003c/li\u003e\n\u003cli\u003eKehal, H. S. \u0026amp; V. P. Singh (2005). \u003cu\u003eDigital Economy: impacts, influences, and challenges\u003c/u\u003e, IGI Global.\u003c/li\u003e\n\u003cli\u003eLeland, H. E. (1968). \u0026quot;Saving and Uncertainty: The Precautionary Demand for Saving.\u0026quot; \u003cu\u003eUncertainty in Economics\u003c/u\u003e \u003cstrong\u003e82\u003c/strong\u003e(3): 465-473.\u003c/li\u003e\n\u003cli\u003eLeong, C. \u0026amp; F. T. C. Tan \u0026amp; B. Tan \u0026amp; F. Faisal (2022). \u0026quot;The emancipatory potential of digital entrepreneurship: A study of financial technology-driven inclusive growth.\u0026quot; \u003cu\u003eInformation \u0026amp; Management\u003c/u\u003e \u003cstrong\u003e59\u003c/strong\u003e(3): 103384.\u003c/li\u003e\n\u003cli\u003eLi, Y. \u0026amp; J. Tan \u0026amp; B. Wu \u0026amp; J. Yu (2022). \u0026quot;Does digital finance promote entrepreneurship of migrant? Evidence from China.\u0026quot; \u003cu\u003eApplied Economics Letters\u003c/u\u003e \u003cstrong\u003e29\u003c/strong\u003e(19): 1829-1832.\u003c/li\u003e\n\u003cli\u003eLuo, C. \u0026amp; Z. Zhang \u0026amp; C. Wu \u0026amp; Q. Qu (2016). \u0026quot;Financial Support for Supply-side Structural Reform: A Reanalysis of Savings and Investment Conversion Efficiency.\u0026quot; \u003cu\u003eMacroeconomics\u003c/u\u003e(03): 8-23.\u003c/li\u003e\n\u003cli\u003eLuo, J. \u0026amp; Q. Wen (2016). \u0026quot;An Empirical Study on the Influence of Household Financial Asset Allocation on Household Consumption.\u0026quot; \u003cu\u003eConsumer Economics\u003c/u\u003e \u003cstrong\u003e32\u003c/strong\u003e(01): 18-22.\u003c/li\u003e\n\u003cli\u003eMa, G. \u0026amp; E. Yang (2011). \u0026quot;Economies of Scale， Returns to Scale and the Problem of Optimum-scale Farm Management: An Empirical Study Based on Grain Production in China.\u0026quot; \u003cu\u003eEconomic Research Journal\u003c/u\u003e \u003cstrong\u003e46\u003c/strong\u003e(03): 83-96.\u003c/li\u003e\n\u003cli\u003eMair, J. \u0026amp; I. Marti (2009). \u0026quot;Entrepreneurship in and around institutional voids: A case study from Bangladesh.\u0026quot; \u003cu\u003eJournal of Business Venturing\u003c/u\u003e \u003cstrong\u003e24\u003c/strong\u003e(5): 419-435.\u003c/li\u003e\n\u003cli\u003eMakori, E. \u0026amp; E. Matundura \u0026amp; N. Mose (2022). \u0026quot;Effect of fiscal and monetary policy on gross domestic savings in Kenya.\u0026quot; \u003cu\u003eMakori, E., Matundura, E., and Mose\u003c/u\u003e(2022): 44-52.\u003c/li\u003e\n\u003cli\u003eNambisan, S. (2017). \u0026quot;Digital entrepreneurship: Toward a digital technology perspective of entrepreneurship.\u0026quot; \u003cu\u003eEntrepreneurship theory and practice\u003c/u\u003e \u003cstrong\u003e41\u003c/strong\u003e(6): 1029-1055.\u003c/li\u003e\n\u003cli\u003eRayna, T. (2008). \u0026quot;Understanding the Challenges of the Digital Economy: The Nature of Digital Goods.\u0026quot; \u003cu\u003eCommunications \u0026amp; Strategies\u003c/u\u003e \u003cstrong\u003e1\u003c/strong\u003e(71): 13-36.\u003c/li\u003e\n\u003cli\u003eRies, E. (2011). \u003cu\u003eThe lean startup: How today\u0026apos;s entrepreneurs use continuous innovation to create radically successful businesses\u003c/u\u003e, Currency.\u003c/li\u003e\n\u003cli\u003eRiley, E. (2018). \u0026quot;Mobile money and risk sharing against village shocks.\u0026quot;\u003c/li\u003e\n\u003cli\u003eSong, K. \u0026amp; Z. Li \u0026amp; J. Yang (2023). \u0026quot;FinTech and Banking Behavior: A Liquidity Creation Perspective.\u0026quot; \u003cu\u003eJournal of Financial Research\u003c/u\u003e(02): 60-77.\u003c/li\u003e\n\u003cli\u003eWang, L. \u0026amp; G. Liao (2020). \u0026quot;Regional Linkage of Poverty and the Spatial Spillover Effects on Poverty Reduction：Taking the Ethnic Areas of Sichuan Province as Examples.\u0026quot; \u003cu\u003eNorthwest Population Journal\u003c/u\u003e \u003cstrong\u003e41\u003c/strong\u003e(01): 83-92.\u003c/li\u003e\n\u003cli\u003eWen, Y. (2009). \u0026quot;Saving and Growth Under Borrowing Constraints Explaining the \u0026apos;High Saving Rate\u0026apos; Puzzle.\u0026quot; \u003cu\u003eWorking Papers\u003c/u\u003e \u003cstrong\u003e35\u003c/strong\u003e(3): 935.\u003c/li\u003e\n\u003cli\u003eWu, W. \u0026amp; X. Zhang \u0026amp; K. Wu (2021). \u0026quot;Financial Literacy and the Household Savings Ｒate: The Role of Financial Planning and Borrowing Constraints.\u0026quot; \u003cu\u003eJournal of Financial Research\u003c/u\u003e(08): 119-137.\u003c/li\u003e\n\u003cli\u003eXia, J. \u0026amp; P. Sun \u0026amp; B. Li (2023). \u0026quot;Local Government R\u0026amp;D Investment, Technological Innovation and Optimization and Upgrading of Tourism Industry Structure.\u0026quot; \u003cu\u003eStudy \u0026amp; Exploration\u003c/u\u003e(03): 93-104.\u003c/li\u003e\n\u003cli\u003eXie, X. \u0026amp; Y. Shen \u0026amp; H. Zhang \u0026amp; F. Guo (2018). \u0026quot;Can Digital Finance Promote Entrepreneurship?\u0026mdash;Evidence from China.\u0026quot; \u003cu\u003eChina Economic Quarterly\u003c/u\u003e \u003cstrong\u003e17\u003c/strong\u003e(4).\u003c/li\u003e\n\u003cli\u003eYang, B. \u0026amp; Z. Guo \u0026amp; X. Yi \u0026amp; L. Zhang (2023). \u0026quot;The Micro Effect of Digital Economy on Household Consumption Upgrade: Empirical Evidence from CHFS.\u0026quot; \u003cu\u003eEconomic Review\u003c/u\u003e(03): 31-47.\u003c/li\u003e\n\u003cli\u003eYang, B. \u0026amp; Z. Yang \u0026amp; Y. Guo (2023). \u0026quot;Employment structure effect in enterprises digital transformation: Empirical analysis based on text mining of annual reports of manufacturing industry listed enterprises.\u0026quot; \u003cu\u003eChina Soft Science\u003c/u\u003e(04): 141-150.\u003c/li\u003e\n\u003cli\u003eYang, C. H. (2022). \u0026quot;How Artificial Intelligence Technology Affects Productivity and Employment: Firm-level Evidence from Taiwan.\u0026quot; \u003cu\u003eResearch Policy\u003c/u\u003e \u003cstrong\u003e51\u003c/strong\u003e.\u003c/li\u003e\n\u003cli\u003eYang, F. \u0026amp; H. Zhang \u0026amp; X. Lliu (2023). \u0026quot;Digital economy driving household consumption: Mechanism and spatial effects.\u0026quot; \u003cu\u003eProgress in Geography\u003c/u\u003e \u003cstrong\u003e42\u003c/strong\u003e(05): 837-851.\u003c/li\u003e\n\u003cli\u003eYi, X. \u0026amp; B. Zhang \u0026amp; R. Yang \u0026amp; B. Yang (2012). \u0026quot;The Family Social Network and the Rural Household Saving Behavior: a Case Study Based on Chinaꆯ s Villages.\u0026quot; \u003cu\u003eJournal of Management World\u003c/u\u003e(05): 43-51.\u003c/li\u003e\n\u003cli\u003eYin, Z. \u0026amp; T. Liu \u0026amp; C. Zhang (2020). \u0026quot;The Influence of Rural Labor Migration on Household Saving Rate.\u0026quot; \u003cu\u003eChina Industrial Economics\u003c/u\u003e(01): 24-42.\u003c/li\u003e\n\u003cli\u003eYin, Z. \u0026amp; Z. Wu \u0026amp; J. Jiang (2022). \u0026quot; The Effect of Mobile Payment on the Household Savings Rate in China.\u0026quot; \u003cu\u003eJournal of Financial Research\u003c/u\u003e(09): 57-74.\u003c/li\u003e\n\u003cli\u003eZhang, H. \u0026amp; C. Xiong \u0026amp; X. Chen \u0026amp; Y. Wang (2022). \u0026quot;The Influence of Financial Investment of Science and Technology on the Development of Digital Economy.\u0026quot; \u003cu\u003eFinance \u0026amp; Economics\u003c/u\u003e(05): 135-148.\u003c/li\u003e\n\u003cli\u003eZhang, X. \u0026amp; X. Liu \u0026amp; G. Fan (2014). \u0026quot;Rural Labor Migration and Household Saving Rate: The Perspective of Discrepancy in Income and Social Security.\u0026quot; \u003cu\u003eEconomic Research Journal\u003c/u\u003e \u003cstrong\u003e49\u003c/strong\u003e(04): 130-142.\u003c/li\u003e\n\u003cli\u003eZhao, X. \u0026amp; H. Wang \u0026amp; T. Liu (2017). \u0026quot;Population Aging, Endowment Insurance and Household Saving Rate.\u0026quot; \u003cu\u003eChina Soft Science\u003c/u\u003e(08): 156-165.\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Appendix","content":"\u003cp\u003eAppendix 1 to 5 are not available with this version\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"Artificial intelligence, regional savings, digital economy, digital entrepreneurship, employment level","lastPublishedDoi":"10.21203/rs.3.rs-6702658/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6702658/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eArtificial intelligence(AI) is causing concern among many who fear becoming increasingly redundant, prompting them to save more as a precaution. This paper conducts a qualitative analysis based on theories such as precautionary saving motives. Subsequently, employing a city and year two-way fixed effects model, a qualitative analysis was conducted using data from Chinese cities from 2008 to 2020. The results show, first, that there is an 'inverted U' nonlinear relationship between AI and regional savings levels. Second, AI promotes the development of the digital economy, which in turn exhibits an 'inverted U' nonlinear relationship with regional savings levels. Third, the impact of AI on regional savings levels is heterogeneous: in first-tier cities, the critical inflection point of AI development occurs at a lower level. This study provides new evidence on the economic impact of AI and serves as a reference for government departments to formulate policies that promote economic development.\u003c/p\u003e\n\u003cp\u003eJEL: O33, E21, R11\u003c/p\u003e","manuscriptTitle":"Does artificial intelligence improve or reduce the level of regional savings? Evidence from China","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-22 19:28:17","doi":"10.21203/rs.3.rs-6702658/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":"f6101bdc-0284-4f44-8cb7-6b47c7e66563","owner":[],"postedDate":"May 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-05-22T19:28:17+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-22 19:28:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6702658","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6702658","identity":"rs-6702658","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}