Economics Research

Abstract  In recent years, artificial intelligence (AI) has become one of the principal drivers of economic and social change worldwide, yet concerns have been raised about its social and economic consequences—particularly with respect to income distribution and inequality. AI can play a dual role: on the one hand, it may boost productivity and create new employment opportunities; on the other hand, because it is capital-intensive and requires specific skills, the returns to AI may accrue disproportionately to particular groups and thereby exacerbate inequality. This study examines the effect of AI investment on income inequality in countries that are at the forefront of AI deployment. Using panel quantile regression on a sample of 20 countries leading in AI investment over the period 2017–2023, the analysis reveals that the impact of such investment is heterogeneous and stage-dependent. In the early stages of technological diffusion, higher-income groups benefit disproportionately and inequality increases; however, as AI adoption broadens into smaller firms and the services sector, this pattern reverses and inequality declines. The findings suggest that to harness AI’s potential for reducing inequality, complementary policies are necessary—among them large-scale skills training, expansion of public health services, support for small enterprises, and transparent regulation of data ownership and market competition—so that AI contributes to inclusive and sustainable growth. Introduction Artificial intelligence has become one of the most important drivers of economic and social change in the world in recent years. With its ability to analyze vast data and make rapid decisions, this technology has enabled economies to improve productivity and service quality, and has created new opportunities for growth and development. The main objective of this study is to examine the impact of investment in artificial intelligence on income inequality in leading countries in this field. In particular, the study seeks to answer the question of whether investment in artificial intelligence leads to increased inequality in the early stages and whether this trend is reversed as the technology spreads to different sectors of the economy. Given the rapid pace of investment in artificial intelligence and its potential effects on economic structure and income distribution, it is of great importance to examine the effects of this technology on income inequality in different countries. Many previous studies have examined the general relationship between technology and inequality, but few have examined the heterogeneity of these effects across income distributions, particularly using appropriate econometric methods such as quantile panel analysis. The main innovation of this study is to use data from the 20 leading countries in AI investment over the period 2017–2023 and to analyze the staged and heterogeneous effects of this investment on income inequality. By focusing on the effects of trade, human capital, and public health alongside investment in AI, this study provides a comprehensive view of the factors affecting inequality and helps policymakers plan for the optimal use of technology’s potential to reduce inequality. Methods and Materials In order to examine the impact of artificial intelligence on income inequality in leading countries in the field of artificial intelligence investment, the following econometric model has been specified, based on theoretical literature:   (1) , the inequality of income distribution, is calculated using the share of the top 20 percent of the population, and the Gini coefficient is used to examine the robustness of the results. is the artificial intelligence index, which uses private sector investment in artificial intelligence in million dollars. The data for the AI index are from Stanford University.  is economic growth or real GDP growth, which is extracted from World Bank data. is labor productivity, which shows the annual growth rate of output (constant 2021 GDP, international dollars based on purchasing power parity) per worker, whose data is extracted from the International Labor Organization.  is the human capital index, which uses the World Bank's secondary enrollment rate. is the openness index of the economy, which is obtained from the sum of exports and imports over GDP, and the source of this data is also the World Bank. Finally, is a health index that uses the life expectancy at birth index to indicate the health status of the community. The quantile method was used to estimate Model 1 in order to examine the impact of AI on different levels of income inequality at the level of each decile of the inequality index. The sample studied includes 20 leading countries in private sector investment in AI in the period 2017-2023. Results and Discussion Quantile regression estimates show that the impact of AI technology on income distribution, contrary to conventional linear analyses, is significantly circular along the inequality distribution, rather than being constant. In the lower deciles of inequality, the significant positive coefficient of AI (especially in the 10th to 30th percentiles) suggests a strengthening of the “skill bias” of technology; just as the neoclassical literature on skill-based technological change predicts, the gains from digitization and machine learning initially benefit groups with complementary capabilities (human capital and digital infrastructure), thus increasing the share of the top 20. Beyond the middle of the distribution and entering the 80th and 90th percentiles, the sign of the coefficient reverses, and AI becomes a factor that reduces the weight of the rich deciles; a phenomenon that could reflect the second stage of the “digital Kuznets curve”: when technology penetrates smaller firms, service sectors, and data-driven welfare programs, the initial increasing returns cease and its redistributive effect is activated.  Figure 1. Effects of explanatory variables on income distribution inequality among different deciles (inequality index: share of the top 20 percent of the population) Source: Research Calculations       Conclusion This study showed that the effect of investment in AI on income inequality is heterogeneous and staged: at low percentiles of inequality, the technological shock increases the share of the upper income classes and, following a “digital Kuznets”-like path after passing through the midpoints of the distribution, reverses, and at high percentiles of inequality, leads to a decrease in the share of the rich and a decrease in the Gini coefficient. This finding confirms, on the one hand, the logic of the “skill bias” of technology and, on the other hand, the hypothesis of widespread spillovers in the stages of technological maturity.

Abstract Based on economic structure of a country, defense spending and security costs as a public good, can have a positive or negative impact on private sector activities. This research applies a partial coherence approach for the period 1959-2017 to evaluate the relationship between defense spending and security costs as two important parts of the state budget, with the formation of a fixed private capital and oil revenues with a constant tax effect. The results showed that by keeping the effect of tax revenues as fixed in four-year time scales, and in the period 1969-1978, defensive expenditures and the formation of fixed private capital were not co-phase and defense expenditure was a leading variable for the formation of fixed capital, which shows a confirmation of crowding-out effect in Iranian economy. In contrast, in period 1989-1997, these two variables are co-phase and no crowding-out effect has occurred. Also, in short- and long-term, the formation of private fixed capital is a leading variable that justifies the need for security and increasing its costs. During period 1969-1978, in the short-term, two variables of defense spending and oil revenues are in opposite phase. The two variables of security costs and oil revenues are in co-phase in the short run and the phasic difference arrows indicate the causality of oil revenues to security costs during the years 1986-1999. Based on the structure of Iranian economy and the fact that Iran does not export military equipment, it is suggested that defense spending be optimized in order to control the crowding-out effect and security costs to be increased to the point that they have a positive impact on the formation of private equity.