中国科学院数学与系统科学研究院期刊网

28 May 2025, Volume 5 Issue 3
    

  • Select all
    |
  • Yuzhi HAO, Danyang XIE
    China Journal of Econometrics. 2025, 5(3): 615-630. https://doi.org/10.12012/CJoE2025-0089
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    This paper pioneers a novel approach to economic and public policy analysis by leveraging multiple large language models (LLMs) as heterogeneous artificial economic agents. We first evaluate five LLMs' economic decision-making capabilities in solving two-period consumption allocation problems under two distinct scenarios: With explicit utility functions and based on intuitive reasoning. While previous research has often simulated heterogeneity by solely varying prompts, our approach harnesses the inherent variations in analytical capabilities across different LLMs to model agents with diverse cognitive traits. Building on these findings, we construct a multi-LLM-agent-based (MLAB) framework by mapping these LLMs to specific educational groups and corresponding income brackets. Using interest income taxation as a case study, we demonstrate how the MLAB framework can simulate policy impacts across heterogeneous agents, offering a promising new direction for economic and public policy analysis by leveraging LLMs' human-like reasoning capabilities and computational power.
  • Xianchun XU, Daokui ZHANG, Ya TANG
    China Journal of Econometrics. 2025, 5(3): 631-664. https://doi.org/10.12012/CJoE2025-0163
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Since reform and opening up in 1978, China has achieved a remarkable economic growth miracle. For a socialist market economy with Chinese characteristics, the task of accurately summarizing its developmental experiences and elevating them into a broadly applicable theory of economic growth necessitates the measurement of both publicly owned and non-publicly owned segments of the economy. The Third Plenary Session of the 20th Central Committee of the Communist Party of China has explicitly called for "the calculation of value added in the state-owned economy". Drawing on the officially published data of the National Bureau of Statistics, this paper measures the value added of the public-ownership economy, primarily composed of state-owned entities, and the value added of the non-public economy, chiefly represented by private enterprises, by examining various sectors of the national economy. In doing so, it provides preliminary supplementary historical data for both publicly owned and non-publicly owned economies dating back to 1978. The results of this study shed light on the shifts in the overall ownership structure of China's economy, as well as the changes within industry-specific ownership structures since the beginning of reform and opening-up, thus offering a critical foundation for deeper investigations into China's economic growth miracle.
  • Dong QIU, Tongji GUO
    China Journal of Econometrics. 2025, 5(3): 665-693. https://doi.org/10.12012/CJoE2025-0016
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    This paper proposes the "Chapter One Paradox in Macroeconomics", and organizing and revealing the explicit and implicit theoretical principles of economic statistical within the gross domestic product (GDP) accounting by conducting a systematic deconstruction of eleven embedded measurement paradoxes. The study seeks to dispel two prevalent misconceptions regarding GDP accounting: First, "political arithmetic" is not simplistic merely due to its computational nature, when in reality, achieving "additivity" and "comparability" with socioeconomic significance proves extraordinarily challenging; second, the international statistical standards are often based on the provisional epistemological frameworks of the discipline, and "operation with inherent flaws" is unavoidable, thus requiring thorough examination. Economics statistics is a fundamental discipline that requires continuous in-depth study, otherwise, our judgments on "national significant issues" may be prone to bias, and the development of economic theory itself could face disruptive risks.
  • Guangzhong LI, Yingtao TANG, Qing GAO
    China Journal of Econometrics. 2025, 5(3): 694-722. https://doi.org/10.12012/CJoE2024-0338
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    This paper utilizes macroeconomic indicators from 18 Belt and Road Initiative (BRI) countries from 2012 to 2023 to construct a quantile factor-augmented vector autoregression (QFAVAR) model. This model captures the median and tail quantiles of various macroeconomic indicators to analyze the impact of shocks from China and other global factors on the macroeconomic variables of BRI countries. The findings reveal that: 1) A decline in China's economic policy uncertainty and the volatility of the RMB exchange rate helps reduce uncertainty in prices and output in BRI countries; 2) following the COVID-19 pandemic, the inclusion of China's macroeconomic variables enhances the ability to predict the short-term left-tail risks of interest rates, CPI, industrial output, and exchange rates in BRI countries; 3) after joining the BRI, most participating countries experience improved predictability of both left-tail and right-tail risks, leading to reduced uncertainty. This study provides valuable insights for a deeper understanding of the macroeconomic effects of the BRI and contributes to fostering higher-level cooperation between China and BRI countries.
  • Junjie MA, Jing GU, Xiaoguang YANG
    China Journal of Econometrics. 2025, 5(3): 723-743. https://doi.org/10.12012/CJoE2024-0205
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Financialization of physical enterprises will lead to the weakening of their main business, reduced motivation for innovation, and even leading to the hollowing out of industries and turbulence in the financial market, which is one of the most important problems that require addressing. In this digital era, the public, as the important supervisor of enterprise operation and the ultimate purchaser of enterprise products or services, may have a governance role in the excessive financialization of physical enterprises. This paper empirically examines the impact of public attention on the financialization of physical enterprises using the data of A-share listed companies in Shanghai and Shenzhen from 2011 to 2022. We note that public attention inhibits financialization of physical enterprises through supervisory pressure and business expansion, and this inhibitory effect is more pronounced in firms that are non-state-owned and have high proportions of institutional shareholding, with better financial information quality, and low financing constraints.
  • Lin ZHANG, Jian LI, Jiyun HOU, Han ZHANG
    China Journal of Econometrics. 2025, 5(3): 744-757. https://doi.org/10.12012/CJoE2024-0060
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Stock price fluctuations are crucial for the healthy and stable development of the stock market. However, the econometrics methods, which have traditionally demonstrated strong performance in addressing nearly all economic issues, exhibit limitations specifically in stock-related analyses—A gap that has only begun to be addressed with the emergence of artificial intelligence methods as cutting-edge applications in this field. Building on this foundation, this study integrates theoretical economic frameworks with empirical methodologies across disciplines to undertake a comparative investigation into the impact of monetary policy on stock prices and its associated asymmetric effects in China. The findings reveal that econometric models (VAR and MS-VAR) yield only partially satisfactory results in analyzing this issue, whereas the artificial intelligence-driven LSTM model demonstrates good explanatory power. Monetary policy exerts asymmetric effects on stock prices: Expansionary policies during bear markets show greater efficacy than contractionary measures during bull markets. Additionally, the transmission effect of rising prices on stock price is stronger, and the drag of economic recession on stock price is more pronounced. This study provides insights for the monetary policy interventions and offers practical guidance for the application of interdisciplinary methodological approaches in addressing complex economic issues.
  • Jia YU, Yuying SUN
    China Journal of Econometrics. 2025, 5(3): 758-787. https://doi.org/10.12012/CJoE2025-0131
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The advent of large language models (LLMs) represents a paradigm shift that has profoundly affected market risk spillover mechanisms. Technological breakthroughs, such as the release of ChatGPT-3.5, trigger changes within the technology industries, but also indirectly impact energy markets by increasing the demand for computational resources. We develop an interval-valued vector autoregressive model and propose an interval-based risk spillover matrix and total spillover index (TSI). Utilizing daily stock price data from 2022 to 2025, we quantify the dynamic impact of LLM development on risk spillovers between the technology and energy industry sectors. Our findings indicate that the release of ChatGPT-3.5 amplified risk spillovers among leading technology companies, such as NVIDIA. As various technology giants announced the construction of new data centers, risk transmission emerged between the energy and technology industries. Moreover, Trump's election victory exacerbated risk spillovers both within and across these industries. Interestingly, our proposed TSI index exhibits a rapid response to significant events and circumvents the discontinuity issues inherent in traditional Diebold-Yilmaz spillover indices based on point-valued volatility estimates, thereby demonstrating enhanced robustness.
  • Jing WANG, Yi FU, Xiaorui WANG, Jian SONG
    China Journal of Econometrics. 2025, 5(3): 788-817. https://doi.org/10.12012/CJoE2024-0277
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The current external environment of China is undergoing profound and complex changes. Under the dual pressures of "low-end diversion" and "high-end reshoring,'' the over-reliance on imported intermediate products has increasingly become a barrier to the development of China's manufacturing industry. This study uses industry-level data from the 2016 World Input-Output Database (WIOD) to calculate the import dependency of intermediate products in China's manufacturing sector as a core indicator. It systematically examines the impact mechanism of robot application on the domestic value-added rate of exported products. Empirical results show that excessive reliance on imported intermediate products has a significant inhibitory effect on the domestic value-added rate of products. However, the implementation of industrial automation, especially the application of robots, can effectively weaken this negative relationship, and this conclusion remains consistent across multiple robustness tests. Mechanism analysis further reveals that automation achieves this through three pathways: First, by standardizing production processes to improve product quality; second, by promoting production technology innovation to generate spillover effects; and third, by optimizing the allocation of production factors to create a labor substitution effect. This study not only expands the research boundaries of the economic effects of robot technology application but also deepens the understanding of the impact mechanisms of industrial intelligence on the quality of participation in the global value chain from a micro perspective. Against the backdrop of the "Made in China 2025" strategy, exploring the impact of intermediate product import dependency on Chinese export enterprises holds significant practical importance. This research provides practical reference value for the transformation and upgrading of China's manufacturing industry under the current dual pressures.
  • Yong HE, Yi YANG, Yan CHEN, Mingzhu HU
    China Journal of Econometrics. 2025, 5(3): 818-841. https://doi.org/10.12012/CJoE2024-0422
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    The rise of large language models has injected fresh vigor into the development of Robo-advisors in China and promoted the innovation of financial technology. In this context, this paper constructs an AI agent based on the domestic generative language model framework, from sentiment analysis, market prediction, factor indicators and other dimensions, in-depth mining of alternative data and traditional financial data in stock trading signals, and then constructs the Chinese stock market investment and trading strategies. Empirical studies show that AI agent based on the domestic large models have the potential to perform quantitative analysis, and that the investment returns under the combined multiple dimensions will be significantly better than the results of a single dimension. This suggests that by utilizing powerful natural language processing capabilities and data analysis capabilities, the application of domestic large models in Robo-advisors is promising, providing new ideas and methods for the continuous development of quantitative investment. With the evolution of technology, the future AI agent will be able to understand market dynamics and investor needs more deeply, thus providing more targeted support for investment decisions, enhancing overall investment returns and creating more value.
  • Xianbo ZHOU, Hujie BAI
    China Journal of Econometrics. 2025, 5(3): 842-874. https://doi.org/10.12012/CJoE2024-0406
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Educational equity is a fundamental guarantee to accelerate the modernization of education and build a strong education country. How to alleviate and eliminate the negative impact of the intense education competition caused by superior and limited educational resources is an issue with little research in current theory and practice. This paper applies data from the China Family Panel Studies (CFPS) in 2018 and 2020 to investigate the spatial spillover effects and threshold social interactions of education expenditures among the Chinese households by the spatial autoregressive model and the social threshold model. We first provide a model specification test for the social threshold regression and show by simulation that the test statistic performs well in finite samples. The empirical study shows that within the same province (county), there is a significant spatial positive correlation in household education expenditure. For every 1% increase in other similar household education expenditure in the same province (county/district), household education expenditures increase by 0.32% (0.152%), showing that there is a spatial spillover effect among family education expenditures. The estimation results based on the social threshold regression show that the endogenous interaction effect of family education expenditure is significantly positive among families with high children's education expectation, families with low income, and families with high information asymmetry, who are more likely to fall into educational entanglement. This study is the first to empirically test the education spatial competition, which essentially reflects issues such as educational equity and competition for educational resources in China. The study provides practical guidance and management basis for educational policy making and educational resource planning.
  • Jinxin CUI
    China Journal of Econometrics. 2025, 5(3): 875-915. https://doi.org/10.12012/CJoE2024-0335
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Exploring the risk spillover effects between energy and metal futures has important practical significance for improving the quality and efficiency of systemic risk supervision and ensuring the smooth operation of commodity futures markets. However, most existing studies are limited to the low-order moment level and fail to fully reveal the cross-market risk transmission mechanism. Given this, this paper integrates the autoregressive conditional density model and the time-varying parameter vector autoregressive extended joint connectedness approach to explore the higher-order moment and cross-moment risk spillover effects between China's energy and metal futures markets; secondly, the nonparametric causality-in-quantile test is used to study the Granger causality relationship between geopolitical risks and total spillovers. Empirical results show that risk spillovers between energy and metal markets show significant differences under different moments, and the total spillover of high-order moments is lower than the total volatility spillover. Copper dominates the volatility and skewness spillover, while fuel oil dominates the kurtosis spillover. Copper skewness, zinc kurtosis, and fuel oil skewness dominate the three cross-moment spillover effects, respectively. The dynamic total spillover and net spillover indexes show significant time-varying characteristics and have risen sharply after the outbreak of major crises such as the COVID-19 epidemic and the Russia-Ukraine war. Geopolitical risk is a key driver of energy-metal spillovers, and its predictive power on cross-moment total spillovers is significantly higher than conditional volatility and higher-order moment total spillovers.
  • Huanyu ZHAO, Nuo XU, Fukang ZHU
    China Journal of Econometrics. 2025, 5(3): 916-940. https://doi.org/10.12012/CJoE2024-0256
    Abstract ( ) Download PDF ( )   Knowledge map   Save
    Integer-valued generalized autoregressive conditional heteroscedasticity (INGARCH) models often use Poisson distribution as the conditional distribution, but Poisson distribution cannot describe overdispersion. In order to handle over-dispersed and under-dispersed data at the same framework, this paper studies the INGARCH transfer function model based on the mean parameterized Conway-Maxwell Poisson (CMP) distribution, i.e., the CMP INGARCH (1,1) transfer function model. Afterwards, the softplus function is used as the link function, and the softplus CMP INGARCH (1,1) transfer function model is proposed, which avoids the shortcoming of the rectified linear unit (ReLU) function being non-differentiable at the zero point. This paper adopts the adaptive MCMC algorithm, and conducts MCMC iterative sampling of the parameter group, and provides numerical simulations for four intervention types of the two models. All results have been effectively detected. The example uses sexual crime data from Albury, New South Wales, Australia, from February 1995 to September 2023 for intervention analysis. The final results show that the new model is superior.