Environmental Regulation, Rural Residents’ Health Investment, and Agricultural Eco-Efficiency: An Empirical Analysis Based on 31 Chinese Provinces
Abstract
:1. Introduction
2. Research Mechanism and Hypothesis
2.1. The Role of ER on AEE
2.2. ER, RRHI, and AEE
3. Research Methodology and Data Sources
3.1. Model Construction
3.1.1. Super-SBM Model Based on the Undesired Output
3.1.2. Benchmark Regression Model
3.2. Selection of Variables
4. Regression Analysis
4.1. Benchmark Regression Analysis
4.2. Endogeneity Problem
4.3. Robustness Test
5. ER, RRHI, and AEE
5.1. Analysis of Mediating Effects
5.1.1. Model Construction
5.1.2. Empirical Analysis
5.2. Heterogeneity Analysis
6. Conclusions and Recommendations
7. Limitations and Perspectives
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
Appendix A
Year | City | AEE | City | AEE | City | AEE | City | AEE | City | AEE | City | AEE | City | AEE |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2009 | Anhui | 0.2400 | Beijing | 0.6212 | Chongqing | 0.3093 | Fujian | 0.4146 | Gansu | 0.2339 | Guangdong | 0.3693 | Guangxi | 0.4694 |
2010 | Anhui | 0.2766 | Beijing | 0.6733 | Chongqing | 0.3469 | Fujian | 0.4898 | Gansu | 0.2827 | Guangdong | 0.4169 | Guangxi | 0.5406 |
2011 | Anhui | 0.2993 | Beijing | 0.7297 | Chongqing | 0.4027 | Fujian | 0.5642 | Gansu | 0.3018 | Guangdong | 0.4855 | Guangxi | 0.6207 |
2012 | Anhui | 0.3179 | Beijing | 0.8089 | Chongqing | 0.4467 | Fujian | 0.6179 | Gansu | 0.3410 | Guangdong | 0.5088 | Guangxi | 0.6833 |
2013 | Anhui | 0.3391 | Beijing | 0.8979 | Chongqing | 0.4701 | Fujian | 0.6778 | Gansu | 0.3736 | Guangdong | 0.5445 | Guangxi | 0.7482 |
2014 | Anhui | 0.3595 | Beijing | 0.8861 | Chongqing | 0.4912 | Fujian | 0.7595 | Gansu | 0.3882 | Guangdong | 0.5691 | Guangxi | 0.7935 |
2015 | Anhui | 0.3510 | Beijing | 1.0039 | Chongqing | 0.4805 | Fujian | 0.6614 | Gansu | 0.3089 | Guangdong | 0.6010 | Guangxi | 0.7428 |
2016 | Anhui | 0.3795 | Beijing | 1.0159 | Chongqing | 0.5846 | Fujian | 1.0442 | Gansu | 0.4549 | Guangdong | 0.6621 | Guangxi | 1.0329 |
2017 | Anhui | 0.3873 | Beijing | 1.0132 | Chongqing | 0.5877 | Fujian | 0.8613 | Gansu | 0.3891 | Guangdong | 0.7594 | Guangxi | 0.9137 |
2018 | Anhui | 0.3893 | Beijing | 1.1030 | Chongqing | 0.6472 | Fujian | 1.0011 | Gansu | 0.4296 | Guangdong | 1.0010 | Guangxi | 1.0351 |
2009 | Guizhou | 0.2863 | Hebei | 0.2967 | Henan | 0.3412 | Heilongjiang | 0.2942 | Hainan | 0.4901 | Hubei | 0.3362 | Hunan | 0.3382 |
2010 | Guizhou | 0.3164 | Hebei | 0.3844 | Henan | 0.4469 | Heilongjiang | 0.3176 | Hainan | 0.5169 | Hubei | 0.4227 | Hunan | 0.4096 |
2011 | Guizhou | 0.3091 | Hebei | 0.4461 | Henan | 0.4493 | Heilongjiang | 0.4109 | Hainan | 0.5590 | Hubei | 0.4932 | Hunan | 0.4810 |
2012 | Guizhou | 0.4021 | Hebei | 0.5146 | Henan | 0.5060 | Heilongjiang | 0.5445 | Hainan | 0.6090 | Hubei | 0.5525 | Hunan | 0.5233 |
2013 | Guizhou | 0.4497 | Hebei | 0.5948 | Henan | 0.5373 | Heilongjiang | 0.6916 | Hainan | 0.6399 | Hubei | 0.5655 | Hunan | 0.5796 |
2014 | Guizhou | 0.5760 | Hebei | 0.5881 | Henan | 0.5892 | Heilongjiang | 0.7370 | Hainan | 0.7216 | Hubei | 0.6059 | Hunan | 0.5987 |
2015 | Guizhou | 0.7529 | Hebei | 0.4494 | Henan | 0.5922 | Heilongjiang | 0.7947 | Hainan | 0.7511 | Hubei | 0.4405 | Hunan | 0.5737 |
2016 | Guizhou | 0.8215 | Hebei | 0.6269 | Henan | 0.6156 | Heilongjiang | 0.6777 | Hainan | 1.0055 | Hubei | 0.7318 | Hunan | 0.6826 |
2017 | Guizhou | 0.8967 | Hebei | 0.4988 | Henan | 0.6232 | Heilongjiang | 0.9133 | Hainan | 1.0262 | Hubei | 0.5255 | Hunan | 0.7180 |
2018 | Guizhou | 1.1513 | Hebei | 0.5740 | Henan | 0.7319 | Heilongjiang | 1.0607 | Hainan | 1.1096 | Hubei | 0.5504 | Hunan | 0.7899 |
2009 | Jilin | 0.2982 | Jiangsu | 0.3953 | Jiangxi | 0.2440 | Liaoning | 0.3214 | Neimenggu | 0.2579 | Ningxia | 0.3299 | Qinghai | 0.4870 |
2010 | Jilin | 0.3263 | Jiangsu | 0.4655 | Jiangxi | 0.2598 | Liaoning | 0.3825 | Neimenggu | 0.2991 | Ningxia | 0.4215 | Qinghai | 0.5758 |
2011 | Jilin | 0.3671 | Jiangsu | 0.5879 | Jiangxi | 0.2911 | Liaoning | 0.4286 | Neimenggu | 0.3486 | Ningxia | 0.4369 | Qinghai | 0.5620 |
2012 | Jilin | 0.4242 | Jiangsu | 0.6997 | Jiangxi | 0.3110 | Liaoning | 0.5167 | Neimenggu | 0.3659 | Ningxia | 0.4595 | Qinghai | 0.7784 |
2013 | Jilin | 0.4574 | Jiangsu | 0.7637 | Jiangxi | 0.3707 | Liaoning | 0.5790 | Neimenggu | 0.4168 | Ningxia | 0.5087 | Qinghai | 0.7817 |
2014 | Jilin | 0.4914 | Jiangsu | 0.8269 | Jiangxi | 0.3921 | Liaoning | 0.6019 | Neimenggu | 0.4378 | Ningxia | 0.5407 | Qinghai | 1.0036 |
2015 | Jilin | 0.3813 | Jiangsu | 0.9393 | Jiangxi | 0.4616 | Liaoning | 0.6292 | Neimenggu | 0.4576 | Ningxia | 0.5873 | Qinghai | 0.7799 |
2016 | Jilin | 0.4345 | Jiangsu | 0.9623 | Jiangxi | 0.4972 | Liaoning | 0.6835 | Neimenggu | 0.4200 | Ningxia | 0.6560 | Qinghai | 0.8399 |
2017 | Jilin | 0.3076 | Jiangsu | 1.0035 | Jiangxi | 0.5116 | Liaoning | 0.5508 | Neimenggu | 0.4243 | Ningxia | 0.6581 | Qinghai | 0.9003 |
2018 | Jilin | 0.3488 | Jiangsu | 1.0135 | Jiangxi | 0.5510 | Liaoning | 0.6296 | Neimenggu | 0.4733 | Ningxia | 1.0195 | Qinghai | 1.0226 |
2009 | Sichuan | 0.3629 | Shandong | 0.3816 | Shanghai | 0.7482 | Shanxi | 0.2561 | Shaanxi | 0.3683 | Tianjin | 0.4153 | Xinjiang | 0.3701 |
2010 | Sichuan | 0.4054 | Shandong | 0.4393 | Shanghai | 0.8272 | Shanxi | 0.2976 | Shaanxi | 0.4760 | Tianjin | 0.5124 | Xinjiang | 0.6440 |
2011 | Sichuan | 0.4870 | Shandong | 0.4654 | Shanghai | 1.0191 | Shanxi | 0.3286 | Shaanxi | 0.5739 | Tianjin | 0.5465 | Xinjiang | 0.6241 |
2012 | Sichuan | 0.5719 | Shandong | 0.4893 | Shanghai | 1.0055 | Shanxi | 0.3532 | Shaanxi | 0.6310 | Tianjin | 0.6129 | Xinjiang | 0.7077 |
2013 | Sichuan | 0.6089 | Shandong | 0.5880 | Shanghai | 1.0055 | Shanxi | 0.3826 | Shaanxi | 0.7100 | Tianjin | 0.7072 | Xinjiang | 0.7326 |
2014 | Sichuan | 0.6566 | Shandong | 0.6503 | Shanghai | 1.0116 | Shanxi | 0.3973 | Shaanxi | 0.7938 | Tianjin | 0.7855 | Xinjiang | 0.7128 |
2015 | Sichuan | 0.7203 | Shandong | 0.6281 | Shanghai | 0.9685 | Shanxi | 0.3410 | Shaanxi | 0.7807 | Tianjin | 0.6200 | Xinjiang | 0.7446 |
2016 | Sichuan | 0.8454 | Shandong | 0.6444 | Shanghai | 0.8624 | Shanxi | 0.4222 | Shaanxi | 0.8787 | Tianjin | 1.0180 | Xinjiang | 0.7749 |
2017 | Sichuan | 0.9383 | Shandong | 0.6067 | Shanghai | 0.8989 | Shanxi | 0.3999 | Shaanxi | 0.9128 | Tianjin | 0.7470 | Xinjiang | 0.8694 |
2018 | Sichuan | 1.0666 | Shandong | 0.7218 | Shanghai | 1.0875 | Shanxi | 0.4190 | Shaanxi | 1.0782 | Tianjin | 1.1089 | Xinjiang | 1.0665 |
2009 | Tibet | 1.0608 | Yunnan | 0.2409 | Zhejiang | 0.3755 | ||||||||
2010 | Tibet | 0.9996 | Yunnan | 0.2402 | Zhejiang | 0.4540 | ||||||||
2011 | Tibet | 0.9986 | Yunnan | 0.2776 | Zhejiang | 0.5085 | ||||||||
2012 | Tibet | 1.0287 | Yunnan | 0.3291 | Zhejiang | 0.5508 | ||||||||
2013 | Tibet | 0.7977 | Yunnan | 0.3735 | Zhejiang | 0.6140 | ||||||||
2014 | Tibet | 0.8106 | Yunnan | 0.4023 | Zhejiang | 0.6525 | ||||||||
2015 | Tibet | 0.7639 | Yunnan | 0.3918 | Zhejiang | 0.6577 | ||||||||
2016 | Tibet | 0.5616 | Yunnan | 0.4226 | Zhejiang | 0.9094 | ||||||||
2017 | Tibet | 0.8866 | Yunnan | 0.4276 | Zhejiang | 0.9169 | ||||||||
2018 | Tibet | 1.0982 | Yunnan | 0.5483 | Zhejiang | 1.0276 |
References
- Salahuddin, M.; Gow, J.; Vink, N. Effects of Environmental Quality on Agricultural Productivity in Sub Saharan African Countries: A Second Generation Panel Based Empirical Assessment. Sci. Total Environ. 2020, 741, 140520. [Google Scholar] [CrossRef] [PubMed]
- Su, Y.; He, S.; Wang, K.; Shahtahmassebi, A.R.; Zhang, L.; Zhang, J.; Zhang, M.; Gan, M. Quantifying the Sustainability of Three Types of Agricultural Production in China: An Emergy Analysis with the Integration of Environmental Pollution. J. Clean. Prod. 2020, 252, 119650. [Google Scholar] [CrossRef]
- Koondhar, M.A.; Udemba, E.N.; Cheng, Y.; Khan, Z.A.; Koondhar, M.A.; Batool, M.; Kong, R. Asymmetric Causality among Carbon Emission from Agriculture, Energy Consumption, Fertilizer, and Cereal Food Production—A Nonlinear Analysis for Pakistan. Sustain. Energy Technol. Assess. 2021, 45, 101099. [Google Scholar] [CrossRef]
- Fang, L.; Hu, R.; Mao, H.; Chen, S. How Crop Insurance Influences Agricultural Green Total Factor Productivity: Evidence from Chinese Farmers. J. Clean. Prod. 2021, 321, 128977. [Google Scholar]
- Jones, A.D.; Ngure, F.M.; Pelto, G.; Young, S.L. What Are We Assessing When We Measure Food Security? A Compendium and Review of Current Metrics. Adv. Nutr. 2013, 4, 481–505. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Guo, J. Factors Affecting Green Agricultural Production Financing Behavior in Heilongjiang Family Farms: A Structural Equation Modeling Approach. Front. Psychol. 2021, 12, 13. [Google Scholar]
- Deng, X.; Gibson, J. Improving Eco-Efficiency for the Sustainable Agricultural Production: A Case Study in Shandong, China. Technol. Forecast. Soc. Change 2019, 144, 394–400. [Google Scholar] [CrossRef]
- Akbar, U.; Li, Q.L.; Akmal, M.A.; Shakib, M.; Iqbal, W. Nexus between agro-ecological efficiency and carbon emission transfer: Evidence from China. Environ. Sci. Pollut. Res. 2021, 28, 18995–19007. [Google Scholar]
- Xu, X.; Huang, X.; Huang, J.; Gao, X.; Chen, L. Spatial-Temporal Characteristics of Agriculture Green Total Factor Productivity in China, 1998–2016: Based on More Sophisticated Calculations of Carbon Emissions. Int. J. Environ. Res. Public Health 2019, 16, 3932. [Google Scholar] [CrossRef] [Green Version]
- Li, M.; Wang, J.; Zhao, P.; Chen, K.; Wu, L. Factors Affecting the Willingness of Agricultural Green Production from the Perspective of Farmers’ Perceptions. Sci. Total Environ. 2020, 738, 140289. [Google Scholar] [CrossRef]
- Lu, W.; Wu, H.; Geng, S. Heterogeneity and threshold effects of environmental regulation on health expenditure: Considering the mediating role of environmental pollution. J. Environ. Manag. 2021, 297, 113276. [Google Scholar] [CrossRef] [PubMed]
- Shuai, S.; Fan, Z. Modeling the Role of Environmental Regulations in Regional Green Economy Efficiency of China: Empirical Evidence from Super Efficiency DEA-Tobit Model. J. Environ. Manag. 2020, 261, 110227. [Google Scholar] [CrossRef] [PubMed]
- Wang, L.; Wang, Z.; Ma, Y. Does environmental regulation promote the high-quality development of manufacturing? A quasi-natural experiment based on China’s carbon emission trading pilot scheme. Socio-Econ. Plan. Sci. 2021, 10, 101216. [Google Scholar] [CrossRef]
- Zhou, A.; Li, J. Impact of Anti-Corruption and Environmental Regulation on the Green Development of China’s Manufacturing Industry. Sustain. Prod. Consum. 2021, 27, 1944–1960. [Google Scholar] [CrossRef]
- Song, Y.; Zhang, X.; Zhang, M. The Influence of Environmental Regulation on Industrial Structure Upgrading: Based on the Strategic Interaction Behavior of Environmental Regulation among Local Governments. Technol. Forecast. Soc. Change 2021, 170, 120930. [Google Scholar] [CrossRef]
- Wang, Q. Do Land Price Variation and Environmental Regulation Improve Chemical Industrial Agglomeration? A Regional Analysis in China. Land Use Policy 2020, 94, 104568. [Google Scholar] [CrossRef]
- Zhou, X. Does Environmental Regulation Induce Improved Financial Development for Green Technological Innovation in China? J. Environ. Manag. 2021, 300, 113685. [Google Scholar] [CrossRef]
- Du, K. Environmental Regulation, Green Technology Innovation, and Industrial Structure Upgrading: The Road to the Green Transformation of Chinese Cities. Energy Econ. 2021, 98, 205247. [Google Scholar] [CrossRef]
- Wu, X.; Deng, H.; Li, H.; Guo, Y. Impact of Energy Structure Adjustment and Environmental Regulation on Air Pollution in China: Simulation and Measurement Research by the Dynamic General Equilibrium Model. Technol. Forecast. Soc. Change 2021, 172, 121010. [Google Scholar] [CrossRef]
- Wu, H.; Hao, Y.; Ren, S. How do environmental regulation and environmental decentralization affect green total factor energy efficiency: Evidence from China. Energy Econ. 2020, 9883, 104880. [Google Scholar] [CrossRef]
- Bonde, T.A. Environmental Regulations of Agriculture in the Baltic Sea Catchment Areas, with Reference to the European Union and the North Sea. Mar. Pollut. Bull. 1994, 29, 491–499. [Google Scholar] [CrossRef]
- Gonzalez-Martinez, A.R.; Jongeneel, R.; Kros, H.; Lesschen, J.P.; de Vries, M.; Reijs, J.; Verhoog, D. Aligning Agricultural Production and Environmental Regulation: An Integrated Assessment of the Netherlands. Land Use Policy 2021, 105, 105388. [Google Scholar] [CrossRef]
- Lutz, E. Integration of Environmental Concerns into Agricultural Policies of Industrial and Developing Countries. World Dev. 1992, 20, 241–253. [Google Scholar] [CrossRef]
- Victor, M.; Margarita, R.; Pedro, M. Economic-Environmental Efficiency of European Agriculture—A Generalized Maximum Entropy Approach. Agric. Econ. 2018, 64, 423–435. [Google Scholar] [CrossRef] [Green Version]
- Weltin, M.; Hüttel, S. Farm Eco-Efficiency: Can Sustainable Intensification Make the Difference? Agric. Land Mark. –Effic. Regulation. 2019, 10, 1–26. [Google Scholar]
- Rybaczewska-Błażejowska, M.; Gierulski, W. Eco-Efficiency Evaluation of Agricultural Production in the EU-28. Sustainability 2018, 10, 4544. [Google Scholar] [CrossRef] [Green Version]
- Pang, J.; Chen, X.; Zhang, Z.; Li, H. Measuring Eco-Efficiency of Agriculture in China. Sustainability 2016, 8, 398. [Google Scholar] [CrossRef] [Green Version]
- Dai, S.H.; Xu, J.B.; Wang, Y. The impacts of farmers’ perceived benefits on green fertilization behaviors under environmental regulations. Res. Agric. Mod. 2021, 42, 880–888. (In Chinese) [Google Scholar]
- Zhu, H.; Ge, J. A Theoretical and Empirical Analysis on the Impact of Government Regulation on Green Entrepreneurship of Agricultural Enterprises—A Case of Agricultural Leading Enterprises in Jiangxi Province. East China Econ. Manag. 2018, 32, 30–36. (In Chinese) [Google Scholar]
- Alberini, A.; Cropper, M.; Fu, T.-T.; Krupnick, A.; Liu, J.-T.; Shaw, D.; Harrington, W. Valuing Health Effects of Air Pollution in Developing Countries—The Case of Taiwan. J. Environ. Econ. Manag. 1997, 34, 107–126. [Google Scholar] [CrossRef] [Green Version]
- Jerrett, M.; Eyles, J.; Dufournaud, C.; Birch, S. Environmental Influences on Healthcare Expenditures: An Exploratory Analysis from Ontario, Canada. J. Epidemiol. Community Health 2003, 57, 334–338. [Google Scholar] [CrossRef] [PubMed] [Green Version]
- Zheng, J. Environmental regulation, what is the impact? J. Transl. Foreign Lit. Econ. 2021, 2, 64–71. (In Chinese) [Google Scholar]
- Tanaka, S. Environmental Regulations on Air Pollution in China and Their Impact on Infant Mortality. J. Health Econ. 2015, 42, 90–103. [Google Scholar] [CrossRef] [PubMed]
- Yang, J.; Gong, L.; Zhang, Q. Human Capital Formation land its Effects on Economic Growth. Manag. World 2006, 5, 10–18. [Google Scholar]
- Zivin, J.G.; Neidell, M. The Impact of Pollution on Worker Productivity. Am. Econ. Rev. 2012, 102, 3652–3673. [Google Scholar] [CrossRef] [Green Version]
- Fisher, S.; Bellinger, D.C.; Cropper, M.L.; Kumar, P.; Binagwaho, A.; Koudenoukpo, J.B.; Park, Y.; Taghian, G.; Landrigan, P.J. Air Pollution and Development in Africa: Impacts on Health, the Economy, and Human Capital. Lancet Planet. Health 2021, 5, e681–e688. [Google Scholar] [CrossRef]
- Grossman, M. On the Concept of Health Capital and the Demand for Health. J. Political Econ. 1972, 80, 223–255. [Google Scholar] [CrossRef] [Green Version]
- Jahanshahloo, G.R.; Hosseinzadeh, F.; Shoja, N.; Sanei, M.; Tohidi, G. Sensitivity and Stability Analysis in DEA. Appl. Math. Comput. 2005, 169, 897–904. [Google Scholar] [CrossRef]
- Tone, K. A Slacks-Based Measure of Efficiency in Data Envelopment Analysis. Eur. J. Oper. Res. 2001, 130, 498–509. [Google Scholar] [CrossRef] [Green Version]
- Tone, K. A Slacks-Based Measure of Super-Efficiency in Data Envelopment Analysis. Eur. J. Oper. Res. 2002, 143, 32–41. [Google Scholar] [CrossRef] [Green Version]
- West, T.O.; Marland, G. A synthesis of carbon sequestration, carbon emissions, and net carbon flux in agriculture: Comparing tillage practices in the United States. Agric. Ecosyst. Environ. 2002, 16, 217–232. [Google Scholar] [CrossRef]
- Wang, B.; Zhang, W. Cross-provincial Differences in Determinants of Agricultural Eco-efficiency in China: An Analysis Based on Panel Data from 31 Provinces in 1996–2015. Chin. Rural Econ. 2018, 1, 46–62. (In Chinese) [Google Scholar]
- Zeng, F.; Li, D.; Tan, Y. Impact of hog industry transfer on the adjustment of agricultural structure under the background of environmental regulations. China Popul. Resour. Environ. 2021, 31, 158–166. (In Chinese) [Google Scholar]
- Wang, X.; Gu, G. Spatial Pattern and Spillover Effects of Urban Economic Growth Based on Market Potential. Sci. Geogr. Sin. 2017, 37, 1617–1623. (In Chinese) [Google Scholar]
- Shen, K.; Jin, G.; Fang, X. Does Environmental Regulation Cause Pollution to Transfer Nearby? Econ. Res. J. 2017, 52, 44–59. (In Chinese) [Google Scholar]
- Li, G.; Li, Y.; Zhou, X. Agricultural Mechanization, Labor Transfer and the Growth of Farmer’s Income: A Re-examination of Causality. Chin. Rural Econ. 2017, 11, 112–127. (In Chinese) [Google Scholar]
- Song, L.; Cui, F. Environmental Regulation, Environmental Pollution and the Health of Residents—The Analysis of Mediating Effect and Spatial Spillover Effect. J. Xiangtan Univ. (Philos. Soc. Sci.) 2019, 43, 60–68. (In Chinese) [Google Scholar]
Variables | Variable Specific Definition | Mean | SD | Minimum | Maximum |
---|---|---|---|---|---|
Agricultural electricity consumption (AEC) | Agricultural electricity consumption | 266.3 | 397.2 | 0.800 | 1933 |
Agricultural labor force (ALF) | Agriculture, forestry, animal husbandry and fishery employees × agriculture GDP/agriculture, forestry and fishing GDP | 945.1 | 681.9 | 33.38 | 2765 |
Sown area (SA) | Total crop sown area | 5292 | 3777 | 103.8 | 14,903 |
The use of water in agriculture (IA) | Irrigated area | 2067 | 1611 | 109.7 | 6120 |
Total agricultural machinery power (TAMP) | Total mechanical power | 3228 | 2923 | 94 | 13,353 |
Fertilizer (Fert) | Fertilizer input | 186.9 | 148.1 | 4.700 | 716.1 |
Agricultural film (AF) | Agricultural film input | 78,074 | 66,961 | 441 | 322,965 |
Diesel (Ds) | Diesel input | 67.40 | 60.46 | 1.900 | 301.9 |
Pesticide (Ptc) | Pesticide input | 55,971 | 43,650 | 921 | 169,043 |
Agricultural output (Agr-GDP) | Agricultural GDP | 1600 | 1193 | 39.10 | 4974 |
Carbon emissions (CO2-E) | Carbon emissions from agricultural production processes | 350.9 | 250.7 | 13.91 | 1049 |
Fertilizer and film residues (FFR) | Agricultural film and fertilizer residue | 18,154 | 15,569 | 102.6 | 75,094 |
Agriculture eco-efficiency (AEE) | agro-ecological efficiency | 0.6007 | 0.2311 | 0.23386 | 1.1512 |
Environment regulation (LnER) | Ln (regional GDP × (2/3(area of regional jurisdiction × 1/circumference)1/2)−1) | 3.941 | 0.575 | 2.029 | 5.081 |
Industiral structure (IS) | Agriculture GDP/agriculture, forestry and fishery GDP | 0.53 | 0.0881 | 0.302 | 0.899 |
The level of agricultural mechanization (LAM) | Total agricultural machinery power/total crop sown area | 0.669 | 0.347 | 0.25 | 2.451 |
The sown area per capita (SAPC) | Total crop area sown/rural population | 6.155 | 3.222 | 1.422 | 19.92 |
Unit area labor Inputs (LI) | Employees in the primary sector/total area sown to crops | 0.203 | 0.102 | 0.050 | 0.703 |
Rural residents’ health investment (Lnmedical) | Ln (rural residents’ health care expenditure) | 2.773 | 0.281 | 1.786 | 3.299 |
Variables | Model (1) | Model (2) | Model (3) | Model (4) |
---|---|---|---|---|
AEE | AEE | AEE | AEE | |
LnER | 0.95173 *** | 0.990 *** | 0.427 ** | 0.879 *** |
(21.69) | (23.01) | (2.18) | (4.39) | |
IS | 1.267 *** | 1.373 *** | ||
(6.96) | (7.07) | |||
LAM | −0.289 *** | −0.244 *** | ||
(−5.27) | (−4.06) | |||
SAPC | 0.0299 *** | 0.0256 ** | ||
(3.25) | (2.53) | |||
LI | 1.491 *** | 1.328 *** | ||
(5.80) | (4.66) | |||
cons | −3.149 | −4.267 *** | −1.193 | −3.856 *** |
(−18.20) | (−22.03) | (−1.65) | (−4.97) | |
Times-fixed | NO | NO | YES | YES |
Province-fixed | YES | YES | YES | YES |
R2 | 0.6285 | 0.7268 | 0.6641 | 0.7441 |
N | 310 | 310 | 310 | 310 |
Variables | Model (5) AEE | Model (6) AEE |
---|---|---|
L.LnER | 0.8918 *** | - |
(0.088) | ||
LnER | - | 1.042 *** |
(0.0568) | ||
IS | −0.0408 | 1.325 *** |
(0.029) | (0.2632) | |
LAM | −0.015 | −0.2483 ** |
(0.0136) | (0.0823) | |
SAPC | −0.0016 | 0.0286 *** |
(0.0015) | (0.0121) | |
LI | 0.072 | 1.453 *** |
(0.033) | (0.2901) | |
N | 279 | 279 |
Underidentification test (Kleibergen-Paaprk LM statistic) | 97.88, p = 0.0000 | |
Weak identification test (Cragg–Donald Wald F statistic): | 12,695.96 | |
(Kleibergen–Paap rk Wald statistic): | 10,259.08 |
Variables | Model (7) | Model (8) |
---|---|---|
AEE | Nonoutput−1 | |
L.AEE | 0.128 | - |
(1.64) | - | |
LnER | 0.920 *** | 0.00626 ** |
(11.26) | (2.06) | |
IS | 1.060 *** | −0.0016 |
(6.20) | (−0.53) | |
SAPC | −0.302 *** | 0.00018 |
(−5.19) | (0.35) | |
LAM | 0.0246 ** | 0.00006 |
(2.09) | (0.63) | |
LI | 1.683 *** | −0.0012 *** |
(4.94) | (−0.48) | |
cons | −3.951 *** | −0.0221 *** |
(−12.80) | (−3.05) | |
Times-fixed | YES | YES |
Province-fixed | YES | YES |
Sargan test | 0.819 | - |
AR (1) | 0.0369 | - |
R2 | - | 0.226 |
N | 279 | 310 |
Variables | Model (9) | Model (10) | Model (11) |
---|---|---|---|
Lnmedical | AEE | AEE | |
Lnmedical | - | 0.172 ** | 0.133 * |
(2.13) | (1.69) | ||
LnER | 0.297 * | - | 0.839 *** |
(1.91) | (4.17) | ||
IS | 0.216 | 1.401 *** | 1.344 *** |
(1.43) | (7.01) | (6.91) | |
LAM | −0.0524 | −0.137 ** | −0.237 *** |
(−1.12) | (−2.42) | (−3.95) | |
SAPC | 0.000853 | 0.0101 | 0.0255 ** |
(0.11) | (1.05) | (2.53) | |
LI | −0.495 ** | 0.969 *** | 1.394 *** |
(−2.24) | (3.51) | (4.87) | |
cons | 1.414 ** | 0.945 *** | −4.04 *** |
(2.34) | (−3.78) | (−5.18) | |
R2 | 0.9057 | 0.7302 | 0.7469 |
Times-fixed | YES | YES | YES |
Province-fixed | YES | YES | YES |
N | 310 | 310 | 310 |
Variables | Model (12) | Model (13) | Model (14) |
---|---|---|---|
AEE | AEE | AEE | |
LnER | 1.424 *** | −0.189 | 1.285 *** |
(3.12) | (−0.54) | (3.46) | |
CYJG | 0.928 *** | 1.938 *** | 1.145 *** |
(2.67) | (6.54) | (3.47) | |
LAM | −0.1912 | 0.196 ** | −0.556 *** |
(−1.52) | (2.30) | (−5.85) | |
SAPC | 0.0582 *** | −0.017 | 0.0883 ** |
(3.43) | (−1.01) | (2.50) | |
LI | 1.305 *** | 0.89 | 1.138 * |
(3.24) | (1.06) | (1.83) | |
cons | −6.546 ** | −0.105 | −4.772 *** |
(−3.38) | (−0.07) | (−3.93) | |
R2 | 0.864 | 0.8159 | 0.7789 |
Times-fixed | YES | YES | YES |
Province-fixed | YES | YES | YES |
N | 90 | 110 | 110 |
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. |
© 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Zhou, K.; Zheng, X.; Long, Y.; Wu, J.; Li, J. Environmental Regulation, Rural Residents’ Health Investment, and Agricultural Eco-Efficiency: An Empirical Analysis Based on 31 Chinese Provinces. Int. J. Environ. Res. Public Health 2022, 19, 3125. https://doi.org/10.3390/ijerph19053125
Zhou K, Zheng X, Long Y, Wu J, Li J. Environmental Regulation, Rural Residents’ Health Investment, and Agricultural Eco-Efficiency: An Empirical Analysis Based on 31 Chinese Provinces. International Journal of Environmental Research and Public Health. 2022; 19(5):3125. https://doi.org/10.3390/ijerph19053125
Chicago/Turabian StyleZhou, Kun, Xingqiang Zheng, Yan Long, Jin Wu, and Jianqiang Li. 2022. "Environmental Regulation, Rural Residents’ Health Investment, and Agricultural Eco-Efficiency: An Empirical Analysis Based on 31 Chinese Provinces" International Journal of Environmental Research and Public Health 19, no. 5: 3125. https://doi.org/10.3390/ijerph19053125
APA StyleZhou, K., Zheng, X., Long, Y., Wu, J., & Li, J. (2022). Environmental Regulation, Rural Residents’ Health Investment, and Agricultural Eco-Efficiency: An Empirical Analysis Based on 31 Chinese Provinces. International Journal of Environmental Research and Public Health, 19(5), 3125. https://doi.org/10.3390/ijerph19053125