An inter-regional input-output table series of China from 1987–2017 with integrated carbon emission data Authors: Qiong Zhao, Chen Pan, Zengkai Zhang, Xuefan Guo, Kunfu Zhu, Jianwu He, Shantong Li Published: 2024 ABSTRACT Inter-regional input-output (IRIO) tables are essential for socioeconomic and environmental analysis. This paper compiled a continuous time series of Chinese IRIO tables with a detailed regional and sectoral classification, covering a longer period from 1987 to 2017 than existing Chinese IRIO tables. Additionally, we integrated the China provincial CO2 emission inventory data (1987–2017) to analyze trends and identify the main driving factors behind regional economic linkages and carbon emissions changes. Specifically, we adjusted the initial values of the international and interprovincial imports and exports using customs data. Subsequently, we employed the minimum cross-entropy model to balance the adjusted input-output tables (IOTs) from both row and column perspectives. Following this, we estimated the interprovincial trade flow for each sector using the gravity model. Then, we balanced these estimates with the minimum cross-entropy model to form the multi-regional input-output (MRIO) table. Finally, we constructed the IRIO table based on the proportion assumption. The compiled IRIO tables, integrated with carbon emission data, have significant applications in regional economic analysis, policy evaluation, and environmental studies. Background & Summary With the acceleration of global economic integration and the increasingly prominent problem of climate change, understanding interregional economic linkages and their environmental impact, particularly the interregional embodied carbon emissions, has become an essential topic in global sustainable development research. China is the largest contributor to global CO2 emissions, responsible for almost all emissions growth between 2019 and 20211,2, as reported by the International Energy Agency (IEA). In 2021, China’s CO2 emissions reached 12.8 billion tons, accounting for 31.5% of the total global emissions1. While China has committed to ambitious carbon reduction targets, namely peaking emissions by 2030 and becoming carbon neutral by 2060, socioeconomic disparities among regions3–5 lead to varying carbon emissions6,7. Economic links between provinces have also resulted in carbon emissions transfer8, which will affect the scale and distribution of carbon emissions in provinces5,9 and hinder the attainment of national carbon reduction goals10,11. Therefore, understanding China’s interregional economic linkages and carbon emissions is crucial for grasping regional disparities and achieving ecological and economic balance. Inter-regional input-output (IRIO) tables are a powerful tool for analyzing economic flows and capturing transactions and embodied carbon emissions between regions12,13. Many studies have explored carbon transfer within China based on the IRIO table14–20. They found that the developed provinces often shift carbon-intensive industries to less developed provinces and import carbon-intensive products from these regions to meet emission targets21–23. For instance, Mi (2017)24 investigated carbon transfers and their main drivers in 2012, concluding that less-developed regions experienced the highest net carbon inflows. Similarly, Meng et al.16 discovered that developed coastal regions’ international exports resulted in emissions from less-developed inland regions between 2002–2007. Several driving factors contribute to carbon emission transfers among Chinese provinces25–28, such as population differences15, production structures24,29, energy efficiency25,30, and trade activities29,31, as well as economic growth deceleration15,25 and changes in consumption patterns25,32. Additionally, some researchers have used multi-regional input-output (MRIO) to assess emission responsibilities33,34, outline carbon transfer in international trade35,36, and analyze the equity of interprovincial trade in China5,37. However, there is still the problem of short time series, hindering the comprehensive understanding of long-term trends and cyclical changes. Compiling various IRIOs depends on diverse data sources, statistical criteria, and time frames, posing challenges for cross-reference tables and impacting data consistency and accuracy12,38. To conduct an in-depth analysis of China’s regional economic structure, industrial linkages, and their impact on carbon emissions, compiling a comprehensive, accurate, and time-spanning set of China IRIO with integrated carbon emission data is essential. Several organizations in China have linked provincial input-output tables (IOTs) based on socioeconomic datasets to compile the China Multi-Regional Input-Output Table (MRIO). Existing studies primarily use two trade data estimation strategies: direct methods based on surveys and indirect methods based on models. For example, the State Information Center (SIC) utilized a combination of survey-based direct methods and non-survey methods to compile MRIO for the years 1997, 2002, and 2007, covering 17 sectors (30 sectors in 1997) across 8 regions (Northeast, Beijing Tianjin, northern coastal, eastern coastal, southern coastal, central, northwest and southwest)39. Most studies primarily rely on indirect methods, particularly gravity models. For instance, the Development Research Center (DRC) of the State Council released the 1987 MRIO for 7 regions and 9 sectors in 200340, followed by a series of MRIO publications from 1987 to 201741–45. The Key Laboratory of Regional Sustainable Development Modeling at the Chinese Academy of Sciences (CAS-RSDM) has sequentially published China MRIOs for 200746, 201047, and 201248. Additionally, the Research Centre on Fictitious Economy and Data Science of the Chinese Academy of Sciences (CAS-FEDS) has compiled China IRIO for 2002, covering 60 sectors in 30 provinces, making it the largest number of sectors in the existing interregional input-output database of China49. Carbon Emission Accounts and Datasets (CEADs) compiled IRIO for 42 sectors in 31 provinces in 2012, 2015, and 201750,51. To visualize these differences, we have created Table S1 to summarize the comparisons. Compared to the aforementioned datasets, the IRIO table compiled in this study has a long time coverage (1987–2017), achieving time-series continuity and offering a more detailed categorization of regions and industries. Importantly, the dataset integrates carbon emission accounts, enabling a holistic perspective encompassing economic and environmental dimensions. This comprehensive and meticulously categorized long-term dataset is crucial for conducting in-depth analyses of the evolving trends within China’s economic structure, assessing the impacts of policy measures, and elucidating the dynamic interconnections among regional economies and carbon emissions changes. Furthermore, our methodological innovation in the compilation the IRIO tables underscores three key aspects: Firstly, we ensure methodological consistency when compiling IRIO across various years. This ensures the comparability and reliability of time series data. Although the compilation method remains consistent annually, adjustments are occasionally necessary based on the specific data situation of each year. Secondly, we aim to retain as much information as possible by maintaining the total output constant and avoiding modifications to the intermediate use matrix and the value-added matrix. Finally, we adopt a ‘bottom-up’ approach, which prioritizes high-quality provincial data over national data tables and aligns with other provincial datasets, such as energy data, to improve the accuracy of regional research analysis. Methods Data source 1987–2017 provincial single region input-output table The provincial single region input-output table (IOT) is the foundational data for constructing the China IRIO model. This data is sourced from the National Bureau of Statistics of China (https://data.stats.gov.cn/) and its publications related to input-output tables52–55. However, discrepancies exist across different years regarding the number of regions and sectors and the breakdowns of value-added and final-use data. For instance, some provinces lack detailed value-added data for 1987, and the value-added entries based on the income approach for 1987 and 1992 are inconsistent with subsequent years. Additionally, the final-use breakdowns in the regional IOT for 1987 differ from those in later years. This also explains the annual variation in the number of sectors and the breakdown of value added. Trade data also exhibit inconsistencies. Some provinces provide comprehensive trade data, including exports, imports, interprovincial exports, and interprovincial imports. Others offer only total exports and total imports or merely net exports. Even among provinces with complete trade data, significant imbalances exist between interprovincial exports and interprovincial imports. This imbalance arises from independent provincial tabulations and the lack of a statistical or survey basis for quantifying interprovincial trade. Theoretically, the sum of interprovincial exports of a product within a country should equal the sum of its interprovincial imports. Therefore, we subsequently adjusted inter-provincial trade flows using customs import and export data and railway transportation data. 1987–2017 customs import and export data The customs import and export (IMEX) data used in this study is the value of customs IMEX goods, calculated based on domestic source/destination statistics, classified according to the Harmonized Commodity Description and Coding System (HS). Subscription access to detailed customs data is available through guidance on the official website (http://www.customs.gov.cn/customs/302249/zfxxgk/2799825/302274/tjfwzn/2319672/index.html). Some years are based on the Standard International Trade Classification (SITC) standard. The statistics include data from mainland provinces, cities, and districts, consistent with the provincial administrative divisions of the respective years. The concordance of customs IMEX data to IRIO is shown in Table 1. In practical applications, it is essential to preprocess the customs IMEX data annually. The first step involves merging the customs IMEX data at the IO sector level based on the correspondence between the IO sector and the customs commodity code. Although the correspondence between customs commodity codes and IO sectors is generally consistent each year, slight variations necessitate annual adjustments based on customs data. Therefore, we provide examples of each type for reference: 1987 SITC codes matched to 33 sectors of the 1987 IOT refer to SITCto1987Iomap.xlsx file. For 2017, we utilized the correspondence provided by Chen et al.56. The next step involves converting prices into producer prices. Precisely, the value of exported goods by customs is calculated based on offshore prices, from which circulation costs are deducted to convert them into domestic producer prices. Conversely, the value of imported goods by customs is calculated based on the landed price, adding import tariffs, consumption taxes, or special taxes.Table 1A summary of the concordance of customs IMEX data to IRIO.YearCoding ruleHS/SITC codeNumber of provincesIRIO1987SITC628331992SITC2(Reference to the 1993 4-bit code for further splitting)29331997HS830402002HS830422007HS830422012HS631422017HS83142 Table 1A summary of the concordance of customs IMEX data to IRIO.YearCoding ruleHS/SITC codeNumber of provincesIRIO1987SITC628331992SITC2(Reference to the 1993 4-bit code for further splitting)29331997HS830402002HS830422007HS830422012HS631422017HS83142 A summary of the concordance of customs IMEX data to IRIO. 1987–2017 railway transport data of nine categories of commodities between provinces This study used railway transportation data for grain, coal, petroleum, coke, metal ores, non-metallic ores, mineral building materials, iron and steel, fertilizer, and pesticides to construct a gravity equation model for estimating the inter-provincial trade flow. The railway transportation data for 1992, 1997, and 2002 were sourced from the Beijing Jiaotong University Library, while data for the remaining years were obtained through collaboration and interviews with research institutions related to the National Railway Administration of the People’s Republic of China (see Table 2 for references).Table 2List of railway transportation data source used to compile the Chinese IRIOs.NameSourceYearCompliation of National Railway Statistics 1987 (1987年全国铁路统计资料汇编)Ministry of Railways1988Compliation of National Railway Statistics 1992 (1992年全国铁路统计资料汇编)Ministry of Railways1993Compliation of National Railway Statistics 1997 (1997年全国铁路统计资料汇编)Ministry of Railways1998Compliation of National Railway Statistics 2002 (2002年全国铁路统计资料汇编)Ministry of Railways2003Compliation of National Railway Statistics 2007 (2007年全国铁路统计资料汇编)Ministry of Railways2008Compliation of National Railway Statistics 2012 (2012年全国铁路统计资料汇编)Ministry of Railways2013Compliation of National Railway Statistics 2017 (2017年全国铁路统计资料汇编)China Railway2018 Table 2List of railway transportation data source used to compile the Chinese IRIOs.NameSourceYearCompliation of National Railway Statistics 1987 (1987年全国铁路统计资料汇编)Ministry of Railways1988Compliation of National Railway Statistics 1992 (1992年全国铁路统计资料汇编)Ministry of Railways1993Compliation of National Railway Statistics 1997 (1997年全国铁路统计资料汇编)Ministry of Railways1998Compliation of National Railway Statistics 2002 (2002年全国铁路统计资料汇编)Ministry of Railways2003Compliation of National Railway Statistics 2007 (2007年全国铁路统计资料汇编)Ministry of Railways2008Compliation of National Railway Statistics 2012 (2012年全国铁路统计资料汇编)Ministry of Railways2013Compliation of National Railway Statistics 2017 (2017年全国铁路统计资料汇编)China Railway2018 List of railway transportation data source used to compile the Chinese IRIOs. 2017 Inter-provincial trade survey data of beijing and tianjin-hebei Some interregional trade survey data, sourced from the Beijing Municipal Bureau of Statistics (https://tjj.beijing.gov.cn/), were also available in 2017 as a supplement. Beijing conducted a trade statistics survey with Tianjin and Hebei in 2017 to better understand the interregional trade dynamics. This survey provided detailed data on the sources and destinations of products from industrial enterprises. Specifically, it included information on the quantities of agricultural and industrial products purchased from local sources, Tianjin, Hebei, other domestic provinces, and foreign countries. It also detailed the quantities of these products sold to local markets, Tianjin, Hebei, other domestic provinces, and international destinations. This study will utilize this data to refine Beijing’s inter-provincial trade statistics. The compilation method of IRIO tables This paper complied the Chinese IRIO tables spanning from 1987 to 2017. These IRIO tables illustrate the interprovince economic structure of China over a period of three decades, including the intermediate use matrix, final demand matrix, export and import information for each province. The layout follows the standard IRIO table (Fig. S2). The compilation method remains consistent annually, this paper takes 2017 as an example to illustrate the six steps involved in compiling IRIO, offering readers a representative reference (Fig. 1). Firstly, based on the official provincial IOTs (Original IOTs) of 2017 published by the National Bureau of Statistics of China (NBS), we use the customs data from the General Administration of Customs of the People’s Republic of China to adjust the initial values of the international and interprovincial imports and exports. Secondly, we used the minimum cross-entropy model to balance the adjusted IOTs from both row and column perspectives. Third, we estimate the interprovincial trade flow for each sector using the railway transport data provided by the Compliation of National Railway Statistics and then using the minimum cross-entropy model to balance the interprovincial trade flows at the sectoral level to form the MRIO table of China. Finally, we constructed the IRIO table of 2017 using the proportionality assumption. Based on data availability, we compared our IRIO tables (DRC) of 2012 and 2017 with that compiled by the Research Centre on Fictitious Economy and Data Science of the Chinese Academy of Sciences (CAS-FEDS), Key Laboratory of Regional Sustainable Development Modeling at the Chinese Academy of Sciences (CAS-RSDM), and Carbon Emission Accounts and Datasets (CEADs). The comparison results for 2017 are shown in the Supporting Information.Fig. 1Flowchart of IRIO table compilation. Flowchart of IRIO table compilation. Adjust international and interprovincial trade data In the original IOTs, the sum of export and interprovincial export is greater than total exports due to entrepot trade. To solve this problem, we need to make a preliminary adjustment to international and interprovincial trade data. First, we collect the import and export (IMEX) data at the sectoral level based on the customs data and then use it to replace the corresponding IMEX data in the original IOTs of each province. Next, we need to adjust interprovincial trade data based on the relationships between two data relationships: the relationship between IMEX data and interprovincial trade under entrepot trade, the relationship between IMEX data in customs and IMEX data in the original IOTs). We will replace the negative value cross interprovincial trade with zero. Balance the adjusted IOTs In the adjusted IOTs, a discrepancy arises between the total interprovincial exports and imports at the sectoral level. To solve this problem, we need to balance the adjusted IOTs with less change to the original data. This involves two main perspectives: 1) balancing by row, which means that the total output should equal the row sum of intermediate use and final demand, and 2) balancing by column, which requires that the column sum of intermediate use plus value-added is equal to the total output. Balance the adjusted IOTs from the row perspective To balance the adjusted IOTs from a row perspective, we ensured that the row sum of intermediate use plus final demand equaled total output. The adjustment terms in our model included t the row sum of intermediate use, consumption (including rural household consumption, urban household consumption, and government consumption), gross capital formation (gross fixed capital formation, change in inventories), exports, imports, interprovincial exports and interprovincial imports. To accomplish this, we employed the minimum cross-entropy model, which balanced the adjusted IOTs in all provinces and sectors simultaneously.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\min (\mathop{\sum }\limits_{i}^{42}\mathop{\sum }\limits_{j}^{9}\mathop{\sum }\limits_{k}^{31}{h}_{ijk}(\mathrm{ln}\,{h}_{ijk}-\,\mathrm{ln}\,{\bar{h}}_{ijk}))$$\end{document}min(∑i42∑j9∑k31hijk(lnhijk−lnh¯ijk))1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$s.t.\{\begin{array}{c}{H}_{k}.{q}_{ctr{l}_{k}}+fu{202}_{k}+pe{x}_{k}-pi{m}_{k}+{r}_{k}={x}_{k}\\ \sum {ih}_{ijk}=1\\ 0\le {h}_{ijk}\le 1\\ |{r}_{k}|\le 0.05\times {x}_{k}\\ |\,fu{202}_{k}|\le 0.01\times {x}_{k}\\ e{x}_{k}+pe{x}_{k}\le {x}_{k}\\ 0.5\times outr\le (\hat{{x}_{k}}{)}^{-1}\times ({ex}_{k}+{pex}_{k})\le 1.5\times outr\\ pe{x}_{k}-pi{m}_{k}=0\end{array}$$\end{document}s.t.{Hk.qctrlk+fu202k+pexk−pimk+rk=xk∑ihijk=10≤hijk≤1|rk|≤0.05×xk|fu202k|≤0.01×xkexk+pexk≤xk0.5×outr≤(xk^)−1×(exk+pexk)≤1.5×outrpexk−pimk=0where the matrix H is the column normalization of the adjusted matrix Q to be balanced, and the element \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}_{{ijk}}={q}_{{ijk}}/\,\sum _{i}{q}_{{ijk}}$$\end{document}hijk=qijk/∑iqijk; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{h}}_{{ijk}}$$\end{document}h®ijk is the initial values of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}_{{ijk}}$$\end{document}hijk, Q represents a 31 × 9-dimensional matrix corresponding to a sector (31 is the number of provinces, 9 is the number of items to be balanced, including the sum of intermediate use, rural household consumption, urban household consumption, and government consumption, gross fixed capital formation, change in inventories, exports, imports, interprovincial exports and interprovincial imports). \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${q}_{{ctrl}}$$\end{document}qctrl is the constraint of the sum of the columns of the matrix \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q$$\end{document}Q to be balanced, the total imports and exports of each province are constrained by the total imports and exports in Original IOTs, and the constraints of other items are equal to those in Original IOTs (Recorded as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{IO}}_{{ini}}$$\end{document}IOini). There are eight constraints in the Eq. (1). The first one in Eq. (1) is the constraint on row balance, where fu202, pex, pim, r and x represent the change in inventories, interprovincial exports, interprovincial imports, error items and total outputs, respectively. The 2–5th equations ensure that each variable in the model remains within its specific range of variation. The 6th equation stipulates that no “entrepot trade” is allowed within any sector or province. The 7th equation further limits the range of product outflow rates for various sectors in Beijing, where 𝑜𝑢𝑡𝑟 is calculated based on trade survey data between Beijing, Tianjin and Hebei. The 8th equation establishes a constraint on interprovincial trade balance by indicating that the total interprovincial exports of various sectors within China should equal the total of interprovincial imports. Balance the adjusted IOTs from the column perspective We move further to balance the adjusted IOTs from the column perspective by utilizing the intermediate use matrix and the value-added matrix of each province as specific balancing items. We also leverage the minimum cross-entropy model to construct the single province balance model.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\min (\mathop{\sum }\limits_{i}^{46}\mathop{\sum }\limits_{j}^{42}{p}_{ij}(\mathrm{ln}\,{p}_{ij}-\,\mathrm{ln}\,{\bar{p}}_{ij}))$$\end{document}min(∑i46∑j42pij(lnpij−lnp¯ij))2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$s.\,t.\{\begin{array}{c}P.x={(tz,v)}^{T}\\ \sum _{i}{p}_{ij}=1\\ 0\le {p}_{ij}\le 1\\ 0.9.vr\le \mathop{\sum }\limits_{i=42}^{46}{p}_{ij}\le 1.1\cdot vr\end{array}$$\end{document}s.t.{P.x=(tz,v)T∑ipij=10≤pij≤10.9.vr≤∑i=4246pij≤1.1·vrwhere the matrix P is the column normalization of the matrix w to be adjusted, and the element \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${p}_{ij}={w}_{ij}/{\sum }_{i}{w}_{ij}$$\end{document}pij=wij/∑iwij; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{p}}_{{ij}}$$\end{document}p®ij is the initial values of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${p}_{{ij}}$$\end{document}pij; W represents a 46 × 42-dimensional matrix corresponding to a province (46 is the sum of the number of sectors and value-added, 42 is the number of sectors). The multiplication of P and the total output column vector x is the sum of the row of the intermediate use matrix and the value-added matrix, tz is the total intermediate use and v is the value-added. The 2–3th equations specify the constraint on the sum of pij and its range of variation. Meanwhile, the 4th equation is a constraint on the range of variation of the value-added rate by sector, where vr is the value-added rate derived from the original IOTs (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{IO}}_{{ori}}$$\end{document}IOori). Once these calculations have been completed, we have obtained the balanced IOTs of the interprovincial trade balance from both row and column perspectives. Estimate interprovincial trade of nine commodities The trade flow matrix of each sector among 31 regions is not directly available; therefore, this study employs an indirect estimation method based on the Gravity model to derive the interprovincial trade flows at the sectoral level. Specifically, we first gathered the interprovincial railway transportation data of grain, coal, petroleum, coke, metal ores, non-metallic ores, mineral building materials, steel, fertilizer, and pesticides from the Compliation of National Railway Statistics in 2017 and then used the gravity model to estimate the main factors that influence the trade flows of these nine products. By matching the 42 sectors in the IOTs and the nine products in the interprovincial railway transportation, we estimated the interprovincial trade flows of corresponding sectors. The gravity model that we utilized is expressed as follows:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{f}}_{s.r}^{k}={e}^{\alpha +\varepsilon }{(S{P}_{s}^{k})}^{{\beta }_{1}}{(D{M}_{r}^{k})}^{{\beta }_{2}}\frac{{(G{S}_{s})}^{{\beta }_{3}}{(G{S}_{r})}^{{\beta }_{4}}}{{{d}_{s,r}}^{{\beta }_{5}}}$$\end{document}f¯s.rk=eα+ε(SPsk)β1(DMrk)β2(GSs)β3(GSr)β4ds,rβ5where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{f}}_{s.r}^{k}$$\end{document}f®s.rk represents the initial trade flow estimate of the product k from region s to region r; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{SP}}_{s}^{k}$$\end{document}SPsk represents the total supply of product k in region s; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{DM}}_{r}^{k}$$\end{document}DMrk represents the total demand of product k in region r; GSs and GSr represent the proportion of the GDP in region s and region r to the domestic GDP, respectively; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${d}_{s,r}$$\end{document}ds,r is the distance from region s to region r. This model can reflect the influence of distance and other factors on interprovincial trade flows. Through econometric regression analysis, we obtain the estimated value of α, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{1}$$\end{document}β1, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{2}$$\end{document}β2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{3}$$\end{document}β3, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{4}$$\end{document}β4, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\beta }_{5}$$\end{document}β5, and then calculate the initial trade flows matrix of 31 regions. Equation (3) can be transformed into:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ln{\bar{f}}_{s.r}^{k}=\alpha +{\beta }_{1}(lnS{P}_{s}^{k})+{\beta }_{2}(lnD{M}_{r}^{k})+{\beta }_{3}(lnG{S}_{s})+{\beta }_{4}(G{S}_{r})-{\beta }_{5}(ln{d}_{s,r})$$\end{document}lnf¯s.rk=α+β1(lnSPsk)+β2(lnDMrk)+β3(lnGSs)+β4(GSr)−β5(lnds,r) Gravity models, based on railway transportation data from the Compliation of National Railway Statistics, estimate initial trade flows of products \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{f}}_{s.r}^{k}$$\end{document}f®s.rk between provinces for each sector. We used the data from the balanced IOTs to calculate the total supply \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{SP}}_{s}^{k}$$\end{document}SPsk and demand \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{DM}}_{r}^{k}$$\end{document}DMrk for each product, and the corresponding relationship is shown in Table 3. Specifically, the total product supply \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{SP}}_{s}^{k}$$\end{document}SPsk can be defined as the difference between the total output and exports. The total product demand \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{DM}}_{r}^{k}$$\end{document}DMrk can be calculated by adding intermediate use, final consumption, and capital formation and subtracting imports. We used data from China Statistical Yearbook to calculate the GDP share \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{GS}}_{s}$$\end{document}GSs and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{GS}}_{r}$$\end{document}GSr, and the distance between regions \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\mathrm{ln}{d}_{s,r}$$\end{document}lnds,r was based on the shortest railway transportation distance between provincial capital cities in 2007. This phenomenon can be attributed to data limitations, but it also possesses practical rationality. Despite the rapid development of railway construction in China since 2007, most of the newly added railway mileage has been dedicated to passenger transport, thereby having a minimal effect on freight mileage.Table 3The corresponding relationships between nine products and sectors in IOTs.Commodity categoryCorresponding input-output sectorGrainAgricultural, forestry, animal husbandry and fishery products and servicesCoalMining and washing of coalPetroleumExtraction of petroleum and natural gasCokeProcessing of petroleum, coking, processing of nuclear fuelMetal oresMining and processing of metal oresNon-metallic oresMining and processing of nonmetal and other oresMineral building materialsManufacture of non-metallic mineral productsIron and steelSmelting and processing of metalsFertilizer and pesticideManufacture of chemical products Table 3The corresponding relationships between nine products and sectors in IOTs.Commodity categoryCorresponding input-output sectorGrainAgricultural, forestry, animal husbandry and fishery products and servicesCoalMining and washing of coalPetroleumExtraction of petroleum and natural gasCokeProcessing of petroleum, coking, processing of nuclear fuelMetal oresMining and processing of metal oresNon-metallic oresMining and processing of nonmetal and other oresMineral building materialsManufacture of non-metallic mineral productsIron and steelSmelting and processing of metalsFertilizer and pesticideManufacture of chemical products The corresponding relationships between nine products and sectors in IOTs. Estimate of interprovincial trade flow We used three different methods for various sectors to estimate the interprovincial trade flows preliminarily. For the sectors that match railway freight transport (as presented in Table 3), we used the first method, which utilized the railway freight as the initial matrix. The estimation method is as follows:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{s.r}^{k}=pi{m}_{r}^{k}\frac{t{r}_{s,r}^{k}}{{\sum }_{i}t{r}_{i,r}^{k}}$$\end{document}fs.rk=pimrktrs,rk∑itri,rkwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{pim}}_{r}^{k}$$\end{document}pimrk represents the interprovincial import of sector k in regional r, which was taken from the balanced IOTs; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{tr}}_{s,r}^{k}$$\end{document}trs,rk represents the corresponding quantity of the product imported from region s to region r through railway transportation in sector k; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{tr}}_{i,r}^{k}$$\end{document}tri,rk denotes the corresponding quantity of the product imported from all other regions to region r through railway transportation in sector k. We estimated interprovincial trade flows in mining and manufacturing sectors without directly matching railway transportation products by applying the gravity model to nine sectoral products with similar characteristics at the sectoral level. The relationship between the sectors and their related products is presented in Table 4.Table 4The relationship between railway transportation products and mining and manufacturing sectors in IOTs.Commodity categoryCorresponding input-output sectorGrainAgricultural, forestry, animal husbandry and fishery products and services; Food and tobacco processingCoalMining and washing of coalPetroleumExtraction of petroleum and natural gasCokeProcessing of petroleum, coking, processing of nuclear fuelMetal oresMining and processing of metal oresNon-metallic oresMining and processing of nonmetal and other oresMineral building materialsManufacture of non-metallic mineral products; Textile industry; Manufacture of leather, fur, feather and related products; Processing of timber and furniture; Manufacture of paper, printing and articles for culture, education and sport activity; Other manufacturing and waste resources; Repair of metal products, machinery and equipment;SteelSmelting and processing of metals; Manufacture of metal products; Manufacture of general purpose machinery; Manufacture of special purpose machinery; Manufacture of transport equipment; Manufacture of electrical machinery and equipment; Manufacture of communication equipment, computers and other electronic equipment; Manufacture of measuring instruments;Fertilizer and pesticideManufacture of chemical products Table 4The relationship between railway transportation products and mining and manufacturing sectors in IOTs.Commodity categoryCorresponding input-output sectorGrainAgricultural, forestry, animal husbandry and fishery products and services; Food and tobacco processingCoalMining and washing of coalPetroleumExtraction of petroleum and natural gasCokeProcessing of petroleum, coking, processing of nuclear fuelMetal oresMining and processing of metal oresNon-metallic oresMining and processing of nonmetal and other oresMineral building materialsManufacture of non-metallic mineral products; Textile industry; Manufacture of leather, fur, feather and related products; Processing of timber and furniture; Manufacture of paper, printing and articles for culture, education and sport activity; Other manufacturing and waste resources; Repair of metal products, machinery and equipment;SteelSmelting and processing of metals; Manufacture of metal products; Manufacture of general purpose machinery; Manufacture of special purpose machinery; Manufacture of transport equipment; Manufacture of electrical machinery and equipment; Manufacture of communication equipment, computers and other electronic equipment; Manufacture of measuring instruments;Fertilizer and pesticideManufacture of chemical products The relationship between railway transportation products and mining and manufacturing sectors in IOTs. The third method is to directly use the interprovincial import and export data to estimate trade flow, mainly targeting sectors such as construction, electricity, gas, water, and services industry. The calculation equation of the initial trade flow matrix of sector k is as follows:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{s.r}^{k}={{pex}}_{s}^{k}\frac{{{pim}}_{r}^{k}}{\sum _{i}{{pim}}_{i}^{k}}(s\ne r)$$\end{document}fs.rk=pexskpimrk∑ipimik(s≠r)where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{pex}}_{s}^{k}$$\end{document}pexsk indicates the interprovincial export of sector k in region s, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{pim}}_{r}^{k}$$\end{document}pimrk indicates the interprovincial import of sector k in region r. Form MRIO table To ensure total interprovincial imports and exports were equal at the sectoral level, we utilized the minimum cross-entropy model to balance the initial interprovincial trade flow matrix. This approach is applied under the constraint of the balanced interprovincial trade data of each province in balanced IOTs. This model aims to regulate interprovincial trading flows to prevent potential inequal due to uneven imports and exports. The model for sector k is as follows, and we applied similar procedures to the other sectors.\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\min (\sum _{s}\sum _{r}{h}_{s,r}^{k}(\mathrm{ln}\,{h}_{s,r}^{k}-\,\mathrm{ln}\,{\bar{h}}_{s,r}^{k}))$$\end{document}min(∑s∑rhs,rk(lnhs,rk−lnh¯s,rk))7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{c}s.t.\{\begin{array}{c}{H}^{k}pi{m}^{k}=pe{x}^{k}\\ \sum _{s}{h}_{s,r}^{k}=1\\ 0\le {h}_{s,r}^{k}\le 1\end{array}\end{array}$$\end{document}s.t.{Hkpimk=pexk∑shs,rk=10≤hs,rk≤1where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\bar{h}}_{s,r}^{k}={\bar{f}}_{s.r}^{k}/\sum _{s}{\bar{\,f}}_{s.r}^{k}$$\end{document}h®s,rk=f®s.rk/∑sf®s.rk is the initial value of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}_{s,r}^{k}$$\end{document}hs,rk. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{pim}}^{k}$$\end{document}pimk and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{pex}}^{k}$$\end{document}pexk represent the interprovincial import and export of sector k, respectively. Through the above series of work, we finally obtained the trade flow matrix among 31 provinces in China in 2017. Table 5 shows the estimated interprovincial trade flow matrix of sector k.Table 5Schematic diagram of interprovincial trade flow for sector k.BeijingTianjinHebeiXinjiangTotal interprovincial exportBeijing—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,2}^{k}$$\end{document}f1,2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,3}^{k}$$\end{document}f1,3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,31}^{k}$$\end{document}f1,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{1}^{k}$$\end{document}pex1kTianjin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,1}^{k}$$\end{document}f2,1k—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,3}^{k}$$\end{document}f2,3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,31}^{k}$$\end{document}f2,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{2}^{k}$$\end{document}pex2kHebei\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,1}^{k}$$\end{document}f3,1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,2}^{k}$$\end{document}f3,2k—…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,31}^{k}$$\end{document}f3,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{3}^{k}$$\end{document}pex3k⋮⋮⋮⋮⋮⋮⋮Xinjiang\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,1}^{k}$$\end{document}f31,1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,2}^{k}$$\end{document}f31,2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,3}^{k}$$\end{document}f31,3k…—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{31}^{k}$$\end{document}pex31kTotal interprovincial import\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{1}^{k}$$\end{document}pim1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{2}^{k}$$\end{document}pim2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{3}^{k}$$\end{document}pim3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{31}^{k}$$\end{document}pim31k— Table 5Schematic diagram of interprovincial trade flow for sector k.BeijingTianjinHebeiXinjiangTotal interprovincial exportBeijing—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,2}^{k}$$\end{document}f1,2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,3}^{k}$$\end{document}f1,3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{1,31}^{k}$$\end{document}f1,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{1}^{k}$$\end{document}pex1kTianjin\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,1}^{k}$$\end{document}f2,1k—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,3}^{k}$$\end{document}f2,3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{2,31}^{k}$$\end{document}f2,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{2}^{k}$$\end{document}pex2kHebei\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,1}^{k}$$\end{document}f3,1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,2}^{k}$$\end{document}f3,2k—…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{3,31}^{k}$$\end{document}f3,31k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{3}^{k}$$\end{document}pex3k⋮⋮⋮⋮⋮⋮⋮Xinjiang\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,1}^{k}$$\end{document}f31,1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,2}^{k}$$\end{document}f31,2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${f}_{31,3}^{k}$$\end{document}f31,3k…—\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pe{x}_{31}^{k}$$\end{document}pex31kTotal interprovincial import\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{1}^{k}$$\end{document}pim1k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{2}^{k}$$\end{document}pim2k\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{3}^{k}$$\end{document}pim3k…\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$pi{m}_{31}^{k}$$\end{document}pim31k— Schematic diagram of interprovincial trade flow for sector k. Construct IRIO table We divided the intermediate use and final demand of the balanced IOTs into local, interprovincial and international imports by estimating detailed trade flows between sectors and provinces. We adopted the proportionality assumption, calculating the share of interprovincial and international imports in each province’s total local use, as demonstrated in Eqs. (8, 9):8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{s,r}=\frac{pi{m}_{k}^{s,r}}{{z}_{k}^{r}+{y}_{k}^{r}}$$\end{document}αks,r=pimks,rzkr+ykr9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{g,r}=\frac{i{m}_{k}^{r}}{{z}_{k}^{r}+{y}_{k}^{r}}$$\end{document}αkg,r=imkrzkr+ykrwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{s,r}$$\end{document}αks,r represents the proportion of the interprovincial imports products of sector k that flow from province s to province r in total local use; \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{g,r}$$\end{document}αkg,r is the share of the international imports of province r in its total local use for sector k. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${z}_{k}^{r}$$\end{document}zkr and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${y}_{k}^{r}$$\end{document}ykr indicates the intermediate use and final demand of sector k in province r, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${{im}}_{k}^{r}$$\end{document}imkr represents the international imports of sector k in province s. Combined with \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{s,r}$$\end{document}αks,r and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }_{k}^{g,r}$$\end{document}αkg,r, and use Eqs. (10, 11), intermediate use and final demand can be divided into interprovincial, international import and local province according to product source.10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}^{{\rm{s}},{\rm{r}}}={\rm{diag}}({\alpha }^{{\rm{s}},{\rm{r}}}){Z}^{{\rm{r}}},{Z}^{{\rm{m}},{\rm{r}}}={\rm{diag}}({\alpha }^{{\rm{m}},{\rm{r}}}){Z}^{{\rm{r}}},{Z}^{{\rm{r}},{\rm{r}}}={Z}^{{\rm{r}}}-\sum _{s}{Z}^{{\rm{s}},{\rm{r}}}-{Z}^{{\rm{m}},{\rm{r}}}$$\end{document}Zs,r=diag(αs,r)Zr,Zm,r=diag(αm,r)Zr,Zr,r=Zr−∑sZs,r−Zm,r11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}^{{\rm{s}},{\rm{r}}}={\rm{diag}}({\alpha }^{s,r}){Y}^{{\rm{r}}},{Y}^{{\rm{m}},{\rm{r}}}={\rm{diag}}({\alpha }^{{\rm{m}},{\rm{r}}}){Y}^{{\rm{r}}},{Y}^{{\rm{r}},{\rm{r}}}={Y}^{{\rm{r}}}-\sum _{s}{Y}^{{\rm{s}},{\rm{r}}}-{Y}^{{\rm{m}},{\rm{r}}}$$\end{document}Ys,r=diag(αs,r)Yr,Ym,r=diag(αm,r)Yr,Yr,r=Yr−∑sYs,r−Ym,rwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}^{{\rm{s}},{\rm{r}}}$$\end{document}Zs,r, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}^{{\rm{m}},{\rm{r}}}$$\end{document}Zm,r and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}^{{\rm{r}},{\rm{r}}}$$\end{document}Zr,r represent the intermediate use from the interprovincial imports, international import and local province, respectively. Similarly, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}^{{\rm{s}},{\rm{r}}}$$\end{document}Ys,r, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}^{{\rm{m}},{\rm{r}}}$$\end{document}Ym,r and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}^{{\rm{r}},{\rm{r}}}$$\end{document}Yr,r represent the final demand from the interprovincial imports, international import and the local province, respectively. The \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${diag}({\alpha }^{s,r})$$\end{document}diag(αs,r) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${diag}({\alpha }^{m,r})$$\end{document}diag(αm,r) are the diagonal matrices of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }^{s,r}$$\end{document}αs,r and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\alpha }^{m,r}$$\end{document}αm,r, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Z}^{r}$$\end{document}Zr and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Y}^{r}$$\end{document}Yr represent intermediate use and final demand (excluding international exports and interprovincial exports) in province r, respectively. Calculation inter-provincial carbon emission transfer in China The carbon transfer is determined by using the IRIO tables in conjunction with the carbon emission data of provinces and sectors. The data on carbon emissions is obtained from the China Emission Accounts Datasets (CEADs), which provide a comprehensive inventory of carbon emissions for 30 provinces for two decades, spanning from 1997 to 2017. The emission data for 1987 and 1992 are sourced from the Global Carbon Plan (GCP) because it provides more comprehensive information on total CO2 emissions from fossil fuels and cement. Additionally, it provides more precise data on CO2 emissions from petroleum, coal, and natural gas combustion. The fundamental equation of the IRIO model is shown in Eq. (12).12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X={(I-A)}^{-1}y$$\end{document}X=(I−A)−1yWhere X is the total output matrix; y represents to the final demand, which includes rural household consumption, urban household consumption, government consumption, gross fixed capital formation, and change in inventories; A is the direct consumption coefficient; i represents the identity matrix, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${(I-A)}^{-1}$$\end{document}(I−A)−1 is the Leontief inverse matrix, which refers to the direct and indirect consumption of commodity i in region r required by sector j in region s to produce one unit of the final product. Define E as the carbon intensity matrix, which is a diagonalized matrix. The carbon emissions induced by final demand can be expressed as13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$C=E{(I-A)}^{-1}y$$\end{document}C=E(I−A)−1y From 1987 to 2017, China’s final demand-driven carbon transfers, as illustrated in Fig. 2, have significantly increased by nearly 7 times, soaring from 710.6 Mt to 4735.8 Mt. This substantial growth can be attributed to the rapid economic expansion during this period. Notably, the interprovincial carbon outflow pattern in China has undergone a noteworthy reversal in the last three decades, as depicted in the same figure. Specifically, Northern China has transformed into a net carbon outflow region, marking a significant shift brought about by variations in Chinese production and consumption behaviors14. Concurrently, less economically developed regions, such as northwest China, have transitioned from being net emission inflows to net emission outflows, indicating a broader structural transformation in China’s carbon flows.Fig. 2Change in inter-provincial net carbon flows from 1987 to 2017. The backgrounds are colored by net carbon outflows, with red indicating net outflow and green indicating net inflow. The sizes of the arrows correspond to the amount of net emission flows. There are no data for the blank regions. Change in inter-provincial net carbon flows from 1987 to 2017. The backgrounds are colored by net carbon outflows, with red indicating net outflow and green indicating net inflow. The sizes of the arrows correspond to the amount of net emission flows. There are no data for the blank regions. The CO2 emissions from consuming products and services in affluent eastern coastal provinces, such as Guangdong and Zhejiang, mostly originate from the less prosperous areas in central and western China. In 1987, carbon outflows were distributed in several central areas (Red color), such as Beijing, Hebei, Hunan, and Hubei, as well as in regions along the eastern coast, such as Guangdong and Fujian. In 1992, the most active emission transfers occurred between eastern provinces, with net carbon outflows concentrated in the developed eastern coastal provinces such as Guangdong, Fujian, Jiangsu, and Shandong. Moreover, the major shift is from the developed southern provinces to the northern coastal regions. For example, Guangdong was a net carbon outflow province in 1992, with 11.1 Mt net emission outflows to Hebei. Since 1997, the northern provinces have shown a reversal from net carbon inflow provinces to net carbon outflow provinces, while carbon flow has been transferred from the north to the south. The North China Plain area, including Inner Mongolia, Hebei and Shanxi, saw the most significant net carbon outflow in 1997 and 2002. In 2007, 38 million metric tons (MT) of carbon flowed to Zhejiang, equivalent to 10% of Hebei’s overall carbon emissions. Consequently, the northern provinces significantly contribute to the economic consumption of both intermediate and final items in the southern provinces. Second, China’s net carbon outflow provinces have expanded to inland northwestern regions like Xinjiang, particularly since 2002. Conversely, Guangdong, Zhejiang, and other southern provinces, as well as Beijing, Tianjin, and other economically developed provinces, have become the primary net carbon inflow destinations. This tendency mostly arises from the inherent resource advantages in different places. The underdeveloped Northwest region has been pivotal in supporting the economic growth of the southern provinces by offering high-carbon-intensive and low-value-added products. For instance, Inner Mongolia, a significant energy supplier24, contributed 68% of its local coal production (23.711 million tons) to other provinces in 2012. In 2012, Fig. 2 illustrates that Inner Mongolia’s carbon flows to Hebei, Beijing, and Heilongjiang amounted to 28.6Mt, 34.2Mt, and 26.4Mt, respectively. By contrast, the more developed eastern provinces have completed industrialization and enjoy a prosperous tertiary sector. In addition, environmental restrictions require the elimination of obsolete industrial capacity, which necessitates the importation of energy-intensive items from other regions. Guangdong recorded the highest net carbon inflow among provinces, amounting to 177.2 million tons in 2012. Structural decomposition analysis Structural decomposition analysis (SDA) is widely employed to estimate the driving factors of changes in carbon emissions and energy consumption by combining with IO tables24,57. This method distinguishes the direct and indirect effects of changes in final demand and intermediate inputs on the economy58. SDA decomposes the changes in dependent variables within the economic system into the sum of corresponding driving factors, thereby enabling the measurement of the contribution of each driving factor. To conduct this study effectively, the first step involved adjusting the IRIO tables to constant prices, aligning with the requirements of SDA to ensure the comparability of economic data. The methodology outlined by Pan et al.59 was employed to obtain sector-level price indices from the China Statistical Yearbook. Then, we adopted SDA to analyze the driving factors of the current carbon transfer pattern. Among various decomposition techniques available60, the study utilized the two-polar decomposition method61 to decompose CBC into four factors: carbon intensity (CI), production structure (PS), total consumption (TC), and gross capital formation (GCF), as presented below:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\begin{array}{c}\triangle C=(\triangle CI{L}_{0}(T{C}_{0}+GC{F}_{0})+\triangle CI{L}_{1}(T{C}_{1}+GC{F}_{1}))/2+C{I}_{1}\triangle L(T{C}_{0}+GC{F}_{0})/2\,+\,\\ C{I}_{0}\triangle L(T{C}_{1}+GC{F}_{1})/2+(C{I}_{0}{L}_{0}\triangle TC+C{I}_{1}{L}_{1}\triangle TC)/2+(C{I}_{0}{L}_{0}\triangle GCF+C{I}_{1}{L}_{1}\triangle GCF)/2\end{array}$$\end{document}∆C=(∆CIL0(TC0+GCF0)+∆CIL1(TC1+GCF1))/2+CI1∆L(TC0+GCF0)/2+CI0∆L(TC1+GCF1)/2+(CI0L0∆TC+CI1L1∆TC)/2+(CI0L0∆GCF+CI1L1∆GCF)/2 Where Δ represents the change in a factor. Each of the seven terms in Eq. (14) denotes the contributions to carbon transfer changes that are triggered by one driving force if other variables are kept constant. The carbon transfers from the northern and northwest regions (NN) to the central and south coast regions (CS) showed a significant increase from 34.1 MtCO2 in 1987 to 329.4 MtCO2 in 2017, as shown in Fig. 3(a). During the first subperiod from 1987 to 1992, there was a 78.6% rise in carbon transfers from NN to CS. Additionally, carbon outflow rose in all eight provinces, with the highest increase in the northern region of China (Hebei and Shandong). Since 1992, Inner Mongolia has increasingly emerged as a significant region for the outflow of carbon transfers, mainly owing to its abundant natural resources. During the third sub-period from 1997 to 2002, Inner Mongolia’s carbon transfer contribution increased significantly, surpassing other provinces by 8.2%. After 2002, carbon outflows from the entire northwest region gradually increased and surpassed the northern region, becoming the primary contributor to carbon transfer to CS after 2007. Particularly noteworthy is the significant role played by Xinjiang in supporting the economic development of CS from 2012 to 2017. This trend underscores the interdependence between the economic development of CS and the changes in carbon outflow from the northwest region, highlighting the crucial impact of economic advancements on carbon transfer dynamics.Fig. 3Changes in carbon transfers from the north and northwest regions to the central and south coast regions in 1987–2017. (a) Changes in carbon transfers from the north and northwest regions to the central and south coast regions. (b) Contributions of driving factors to change in transfers from the north and northwest regions to the central and south coast regions. HB: Hebei; IM: Inner Mongolia; SD: Shandong; SX: Shaanxi; GS: Gansu; QH: Qinghai; NX: Ningxia; XJ: Xinjiang; CI: Carbon Intensity; PS: Production Structure; TC: Total Consumption; GCF: Gross Capital Formation. Changes in carbon transfers from the north and northwest regions to the central and south coast regions in 1987–2017. (a) Changes in carbon transfers from the north and northwest regions to the central and south coast regions. (b) Contributions of driving factors to change in transfers from the north and northwest regions to the central and south coast regions. HB: Hebei; IM: Inner Mongolia; SD: Shandong; SX: Shaanxi; GS: Gansu; QH: Qinghai; NX: Ningxia; XJ: Xinjiang; CI: Carbon Intensity; PS: Production Structure; TC: Total Consumption; GCF: Gross Capital Formation. Figure 3(b) illustrates the changes in driving factors for carbon transfers from NN to CS in China. The increase in carbon transfer from NN to CS is mainly attributed to the consumption growth in CS, including changes in gross capital formation (GCF) and total consumption (TC). For the initial subperiod (1987–1992), GCF and TC resulted in respective increases of 47.4% and 44.5% in carbon transfer from NN to CS, without accounting for other influencing factors. The decline in carbon intensity (CI) resulted in a 55.5% drop in carbon transfers from NN to CS. Furthermore, the production structure (PS) alteration accounted for 44.5%. The deceleration in carbon transfers from NN to CS in China during 1992–2002 was primarily driven by shifts in production structure (PS) and carbon intensity (CI). The trend of carbon transfers from NN to CS between 2002 and 2007 aligns with the findings of Pan et al.62. Although the carbon transfers continued to rise, the growth rates slowed in the fifth (2007–2012) and sixth (2012–2017) subperiods, recording increments of 7.9% and 20.9% respectively. Notably, during 2007–2012, carbon intensity was dominant role in reducing carbon transfers from NN to CS in China, potentially leading to an 18.9% decrease. Between 2012 and 2017, changes in production structure were more important than carbon intensity in counteracting China’s carbon transfers from NN to CS. These data indicate that adjusting the production structure was vital in decreasing carbon transfer from NN to CS in China between 2012 and 2017. Therefore, production structure changes will remain crucial in minimizing carbon transfers in the future. Data Records All IRIO tables data files are available for public access at figshare63, providing comprehensive insights into China’s regional economic structure across provinces and sectors from 1987 to 2017 (named (1987–2017)-IRIO.xlsx file). These tables include the intermediate use matrix, final demand matrix, and export and import information for each province, following the standard layout of IRIO tables (Fig. S2). The number of provinces included fluctuates year, with a range of 28 to 31 provinces (28 provinces in 1987, 29 provinces in 1992, 30 provinces in 1997, 2002, and 2007, and 31 provinces in 2012 and 2017). Furthermore, the sector classification may slightly differ due to using different national sector classifications (Table S2). The final demand comprises five categories: rural household consumption, urban household consumption, government consumption, fixed capital formation, and inventory change. However, in 1987, rural and urban household consumption was merged into household consumption. The value-added data includes employee compensation, net production tax, fixed capital depreciation, and operating surplus for each year, except for 1987, when only total value-added data is provided. In addition, the integrated carbon emission data are China provincial CO2 emission inventory from 1987 to 2017. Technical Validation Compare with other IRIO datasets Based on data availability, we compared our IRIO tables (DRC) of 2012 and 2017 with those compiled by the Research Centre on Fictitious Economy and Data Science of the Chinese Academy of Sciences (CAS-FEDS), Key Laboratory of Regional Sustainable Development Modeling at the Chinese Academy of Sciences (CAS-RSDM), and carbon emission accounts and datasets (CEADs). The IRIO tables obtained from these sources comprise 42 sectors in 31 provinces and are compatible with our IRIO tables. We used the mean absolute deviation (MAD), Isard-Romanoff similarity index (DSIM), and absolute entropy distance (AED) indicators, as recommended by Zheng et al.3, to assess the similarity between the intermediate demand matrices. The results of these comparisons are shown in Table S3. In 2012, the IRIO constructed in the present study showed higher similarity to that of CEADs when compared with CAS-FEDS and CAS-RSDM, with a marginally smaller mean absolute distance (MAD). The average MAD between our IRIO and CEADs is 0.24, while 0.40 for CAS-FEDS and 0.26 for CAS-RSDM. The MAD measure, weighted according to the distance, can be more sensitive to larger and relatively richer provinces. Developed provinces such as Shanghai, Jiangsu, Beijing, Zhejiang, Hebei, and Shandong, indicated lower similarity with our IRIO, as shown by larger MAD indicators. Conversely, less developed provinces like Qinghai and Xinjiang displayed significantly smaller MAD indicators, indicating higher similarity. However, the results shown in Table 5 indicate that the DSIM, employed to evaluate relative distance, is more similar to CAS-FEDS, with an average DSIM of 0.56 (compared to 0.62 for CEADs and 0.61 for CAS-RSDM). As for AED, it displays higher similarity to CEADs. Figure 4 compared the percentage of intermediate use and final demand to total output in different provinces using various IRIO tables in 2012. The IRIO constructed in this study is more similar to CAS-RSDM and CEADs regarding the proportion of total intermediate use. However, variability was observed among provinces. For instance, Beijing had the largest difference compared to CAS-RSDM, with the proportion of intermediate use in total output decreasing by 13%. Meanwhile, Chongqing had the largest difference compared to CEADs, at 10%. The standard deviation (SD) analysis shows that the deviation between the IRIO table constructed in this study and CAS-RSDN (0.068) was smaller than that between CEADs (0.077) and CAS-FEDS (0.118).Fig. 4The comparison in the proportion of total intermediate use in the total input, proportion of the final demands in the total output in 2012. The numbers shown in the figure are the standard deviation between the IRIO constructed in the present study with CAS-FEDS, CEADs and CAS-RSDM, respectively. It is worth noting that there is no Tibet in the IRIO table of CAS-FEDS, so the proportion of total intermediate use and final demands in total output is 0. The comparison in the proportion of total intermediate use in the total input, proportion of the final demands in the total output in 2012. The numbers shown in the figure are the standard deviation between the IRIO constructed in the present study with CAS-FEDS, CEADs and CAS-RSDM, respectively. It is worth noting that there is no Tibet in the IRIO table of CAS-FEDS, so the proportion of total intermediate use and final demands in total output is 0. As for the proportion of final demand in total output, our IRIO table aligned more closely with CEADs. The most significant gaps were in Shanghai, with 28% compared to CAS and 18% to CAS-FEDS. But when compared to CEAD, Guangdong has the largest disparity, at around 18%. The standard deviation analysis reveals that our IRIO table had a lower deviation with CAS-RSDM, indicating a similar proportion of final demand in total output. Supplementary information An inter-regional input-output table series of China from 1987-2017 with Integrated Carbon Emission Data An inter-regional input-output table series of China from 1987-2017 with Integrated Carbon Emission Data