Divergent Impacts and Policy Implications of Rural Shrinkage on Carbon Intensity in the Yellow River Basin

Abstract

The Yellow River Basin (YRB), a vital region for agricultural production in China, is currently grappling with severe rural population shrinkage and variations in the carbon emission intensity across the basin. Based on census data from 2010 to 2020, this study categorized 320 counties by population shrinkage type and applied baseline regression and upper–middle–lower reach heterogeneity analysis to explore population shrinkage’s impact on carbon intensity. The results indicated that population shrinkage in the Yellow River Basin during 2010–2020 was primarily characterized by a rural population decline, which exerted divergent impacts on carbon emissions across the basin. Consequently, the upper reaches were identified as a critical problem area where severe population shrinkage coexisted with a high carbon emission intensity. Based on these findings, targeted and region-specific strategies and policies are proposed. Specifically, High Shrinkage-High Emission (H-H) regions need to focus on promoting ecological migration and the coordinated transformation of industries; High Shrinkage-Low Emission (H-L) regions should strengthen policy coordination in the border areas of the middle and upper reaches; Low Shrinkage-High Emission (L-H) regions should promote the low-carbon technological transformation of traditional industries in downstream counties; and Low Shrinkage-Low Emission (L-L) regions should refine the low-carbon transformation model in the core downstream areas.

1. Introduction

Global climate anomalies are intensifying, making climate change one of the most pressing challenges for humanity [1,2,3]. As the world’s largest developing country, China currently leads globally in terms of energy consumption and carbon dioxide emissions [4,5,6,7]. To address carbon emission challenges, Chinese leaders proposed the goals of “achieving carbon peak by 2030 and carbon neutrality by 2060” at the 75th United Nations General Assembly in September 2020 [8]. Meanwhile, census data show that significant population shrinkage is occurring across China, with 26.71% of prefecture-level and higher administrative units and 37.16% of counties, cities, and districts experiencing shrinkage between 2000 and 2010; from 2010 to 2020, 146 out of 337 prefecture-level and higher units faced shrinkage, indicating varying degrees of urban shrinkage throughout the country. Previous studies have mostly conducted their analyses from the perspective of urban population shrinkage, showing an evolutionary trend of “urban leadership–rural expansion.” The early findings mainly focused on the carbon emission characteristics of shrinking cities, such as old industrial cities and resource-based cities, and the relationship between population loss and urban carbon emission intensity [9,10,11,12,13,14]. In terms of research methods, benchmark regression and fixed-effect models have become mainstream tools, and studies on ecologically sensitive areas and economic transition areas have gradually increased [15,16].

As an important economic zone and densely populated area in China, the Yellow River Basin has seen changes in its population structure that have affected the region’s development. From 2000 to 2020, the permanent rural population in the Yellow River Basin decreased by more than 45.6 million, and the rate of rural population decline has been increasing. This indicates that rural population shrinkage and urban population growth in the Yellow River Basin are prominent phenomena [17,18,19,20]. From the perspective of population and carbon emission intensity, inter-regional population mobility affects carbon emissions in terms of quantity and patterns. Rural population shrinkage may trigger a series of chain reactions, which could have an impact on carbon emission intensity [21,22], and this impact is likely to be complex. In addition, the upper, middle, and lower reaches of the Yellow River Basin exhibit significant spatial differences in natural environment, human culture, and economy, with distinct basin-level variations in rural hollowing. The relationship between population shrinkage and carbon emission intensity may differ between different regions. Heterogeneity analyses have been conducted on the ecologically fragile areas in the upper reaches, energy-rich areas in the middle reaches, and major agricultural production areas in the lower reaches. However, the regions have not been precisely categorized into high-carbon–high-shrinkage areas and low-carbon–low-shrinkage areas, making it difficult to support the formulation of region-specific carbon reduction policies [23,24,25,26,27,28]. Therefore, to achieve the “dual carbon goals” [29,30,31], we must reduce the dependence of economic development on energy consumption amid the dynamic decline in rural populations and realize low-carbon circular development to advance sustainable development [32,33,34].

In summary, domestic and international studies have confirmed that there is a strong correlation between population shrinkage and carbon emission intensity, which is regulated by multiple factors. However, there are still deficiencies in the research: first, there is a lack of studies on major strategic regions such as the Yellow River Basin and a targeted analytical framework combining the industrial and population characteristics of the upper, middle, and lower reaches of the basin. Second, there is insufficient research on the relationship between population shrinkage and carbon emission intensity, making it difficult to form a basis for precise regional classification and policy implementation. Therefore, to address the research gaps, this study seeks to answer the following three research questions:

Q1: How is population shrinkage manifested in the Yellow River Basin, and what is the central role of rural population shrinkage in this process?

Q2: What is the overall impact of rural population shrinkage on carbon emission intensity at the county level in the Yellow River Basin?

Q3: How does the impact of rural population shrinkage on carbon emission intensity vary heterogeneously across the upper, middle, and lower reaches of the Yellow River Basin?

Based on the above, this study systematically identified the impact of rural population shrinkage on carbon emissions using county-level data from 2010 to 2020 and conducted heterogeneity analysis by dividing the study area into the upper, middle, and lower reaches. It aimed to reveal the relationship between rural population shrinkage and carbon emission intensity in the Yellow River Basin and identify the different types of regions. To achieve this goal, this study (1) analyzed the main manifestations of population shrinkage in the Yellow River Basin; (2) clarified the divergent impacts of rural population shrinkage on carbon emission intensity; and (3) using heterogeneity analysis, identified the counties with a negative correlation, which are the regions where the carbon emission intensity increases when the rural population shrinks (Figure 1).

2. Research Hypotheses

2.1. The Central Role of Rural Population Shrinkage in the Yellow River Basin’s Overall Population Shrinkage

As a traditional agricultural region, the Yellow River Basin has had a relatively high rural population, which accounts for the main portion of the total population [35]. Due to the continuous increase in urbanization, the rural population within the basin continues to flow to cities. Coupled with factors such as the decline in rural fertility and population aging, the rural population shows a shrinking trend, and some counties have entered a stage of weak or strong shrinkage [36]. Meanwhile, relying on advantages such as employment opportunities and public services, cities attract population inflows, driving steady growth in the urban population and forming a rural shrinkage–urban growth pattern. This pattern is reflected by the changes in the rural and urban populations in the basin’s census data from 2010 to 2020. The rural population, due to its large initial numbers and its peoples’ willingness to migrate, has seen a scale and speed of shrinkage far exceeding the average levels of urban and total population shrinkage [37,38]. The Yellow River Basin does not exhibit urban population shrinkage and rural population growth. This is consistent with the national population evolution law pointed out by Liu Zhen [39], confirming that the urban–rural differentiation of population shrinkage in the basin is in line with China’s overall spatial population change trend.

Therefore, the study proposed the following hypothesis:

H1. 

The shrinkage of the rural population, by virtue of its substantial proportion of the total population, predominantly drives the overall population decrease in the Yellow River Basin despite urban growth.

2.2. Rural Population Shrinkage Exerts a Significant Influence on Carbon Emission Intensity in the Yellow River Basin

The decrease in the number of residents living in rural areas of the Yellow River Basin inherently and inevitably impacts the carbon emission intensity of this region. The underlying logic is that the rural population is not only a count of individuals but also a fundamental element that constitutes the scale and structure of the rural economy. Therefore, population shrinkage directly and simultaneously alters both the numerator carbon emissions and the denominator economic output in the carbon emission intensity ratio [40]. First, a decline in the rural population essentially means a reduction in the total level of human-driven activities that consume energy while generating economic output. This creates a composite effect on carbon emission intensity, which results from the interaction of two simultaneous changes. A smaller resident population directly translates to a lower baseline for energy demand. This includes reduced residential energy demand for heating, cooking and lighting, as well as a contraction in demand for transportation, public services and the energy required to maintain these services [41]. This exerts direct downward pressure on total carbon emissions. Second, population outflow directly weakens the foundation of local economic production. A smaller population size means a shrinking local labor force, a contracted local consumer market and a reduced ability to sustain the output of local enterprises and agriculture [42]. This will lead to the contraction or stagnation of local gross domestic product. The net effect on carbon emission intensity the ratio of carbon emissions to GDP is determined by the relative magnitude of these two inter-twined contraction processes. It does not mean that population changes need to act through other intermediary factors; instead, the reduction in population size itself triggers a fundamental readjustment of the relationship between emissions and the economy. Therefore, the study proposed the following hypothesis:

H2. 

Rural population shrinkage, through its composite effect, significantly influences the carbon emission intensity in the Yellow River Basin.

2.3. Regional Heterogeneity in the Impact of Rural Population Shrinkage on Carbon Emission Intensity in the Yellow River Basin

There is a marked heterogeneity in the impact of population shrinkage on carbon emission intensity within the Yellow River Basin, which is driven by the significant differences in resource endowments and economic structures across the upper, middle, and lower reaches. The upper reaches, as a critical ecological barrier and energy base, face severe industrial structure lock-in effects. Population shrinkage here fails to drive low-carbon industrial transformation, while labor loss further exacerbates decreases in agricultural production and increases carbon emission intensity [43]. In the middle reaches, which are dominated by energy and chemical industries, labor shortages resulting from population shrinkage initially suppress emissions by improving energy efficiency. However, industrial scale contraction may trigger economic recession risks, reducing energy efficiency and increasing emissions [44]. In contrast, in the lower reaches, which are an agglomeration of agriculture and manufacturing, population shrinkage promotes large-scale agricultural operations and industrial green transformation, cutting emissions through optimized land use and industrial structures [45]. Previous research on the environmental and economic heterogeneities in the Yellow River Basin predominantly focused on single-region analyses or broad north–south comparisons, with insufficient focus on systematic differences across the upper, middle, and lower reaches. Investigating such heterogeneities is essential for providing targeted policy references for local governments to balance population dynamics and low-carbon development. It can also help in identifying regions where the emission reduction is lagging, enabling them to learn from advanced experiences and leverage late-mover advantages in coordinating population changes and carbon emission control.

Therefore, we proposed the following hypothesis:

H3. 

There is regional heterogeneity in the impact of rural population shrinkage on carbon emission intensity in the Yellow River Basin.

3. Research Design

3.1. Research Methods

3.1.1. Measurement and Classification of Population Growth and Shrinkage

This study used data from the 2010 and 2020 census periods to determine the counties experiencing population growth versus shrinkage using the formula:

$${PI}_i=\left({\displaystyle \frac{{pop}_{i2020}-{pop}_{i2010}}{{pop}_{i2010}}}\right)\times 100\%$$

where ${PI}_i$ represents the population change rate for the total, urban, and rural populations in county i; ${pop}_{i2020}$ represents the total, urban, and rural populations of county i in 2020; and ${pop}_{i2010}$ represents the total, urban, and rural populations of county i in 2010. Counties with a ${PI}_i$ < 0 were judged to be experiencing population shrinkage and vice versa. The counties showing population shrinkage were categorized as counties with weak or strong population shrinkage based on their degree of population shrinkage relative to the median population shrinkage.

3.1.2. Benchmark Model

The fixed-effects model was used to explore the impact of population growth and shrinkage on carbon emission intensity, and the heteroskedasticity file standard error was used; the specific formula is as follows:

$${cei}_{it}={\alpha }_0+{\alpha }_1{PGS}_{it}+{\alpha }_2{X}_{it}+{\mu }_i+e_{it}$$

where ${cei}_{it}$ represents the carbon emission intensity of county i in year t, ${PGS}_{it}$ represents the total population shrinkage rate or rural population shrinkage rate of county i in year t, $X_{it}$ represents the vector of the control variables, $e_{it}$ is the error term, and ${\mu }_i$ represents the county-level fixed effect. To address the autocorrelation of observations in different years within the same county and the heteroscedasticity issue in the panel data, the model’s standard errors were used to calculate the robust standard errors clustered at the county level. County-level fixed effects cannot resolve endogeneity problems such as bidirectional causality. Therefore, this study used lag term analysis, lagging the core explanatory variables (-tpsr and -rpsr) by one period and applying the population shrinkage rates from 2009 to 2019 to correspond to the carbon emission intensity values from 2010 to 2020. Population shrinkage in the earlier period is an established historical fact and was not affected by the reverse impact of carbon emissions in the later period, which effectively breaks the immediate causal path. In addition, to comprehensively address the heteroscedasticity and autocorrelation in the panel data, this study compared two standard error processing methods: calculating heteroscedasticity-robust standard errors vce (robust) and robust standard errors clustered at the county level vce (cluster id). The results showed that, due to the weak autocorrelation of observations in different years within the same county in the sample, the standard errors, coefficients, and significance using the two methods were completely consistent.

3.2. Variable Selection

3.2.1. Explanatory Variables

The explanatory variables were the total population shrinkage rate (-tpsr) and the rural population shrinkage rate (-rpsr). This study strictly referred to China’s population census system and the Provisions on the Statistical Division of Urban and Rural Areas and defined the rural population as the permanent population living in township-level administrative areas within the counties of the Yellow River Basin (including areas outside the township seats and rural residential areas). This definition of rural population is consistent with the one in the 2010 and 2020 national population censuses, and the data were directly obtained from the county-level summary tables of the two censuses [46]. To calculate population shrinkage, the following formula was used:

$$-{PGS}_{it}=\mathit{ln}\left({\displaystyle \frac{{pop}_{it}}{{pop}_{{it}_0}}}\right)\times 100\mathrm{\%}$$

where ${pop}_{it}$ represents the total and rural populations of county i in year t and ${pop}_{{it}_0}$ represents the total and rural populations of county i in the base year (2009). If ${PGS}_{it}$ > 0, it indicates population shrinkage; the smaller this indicator, the higher the degree of shrinkage. Conversely, if ${PGS}_{it}$ < 0, it indicates population growth; the larger this indicator, the higher the degree of growth.

3.2.2. Explained Variable

The explained variable was carbon emission intensity, which is commonly measured as the carbon dioxide emissions generated per unit of GDP. The carbon emission data were derived from DMSP/OLS and NPP/VIIRS nighttime remote sensing data inversion, which can accurately project the carbon emissions at the city and county scales, and have the advantages of a consistent statistical caliber and strong continuity from 2001 to 2017. Carbon emissions were extrapolated to 2020 based on the interannual change rate of NPP/VIIRS nighttime light data.

3.2.3. Control Variable

According to the existing literature [47,48], there are many factors that influence carbon emission intensity. To avoid the effects of omitted variable errors, the following control variables were introduced based on the literature (

Table 1. Descriptive statistics of variables.

Category	Variable Name	Symbol	Observations	Average	Standard Deviation	Min	Max
Explained Variables	Carbon emissions intensity (%)	cei	960	3.554	2.943	0.146	27.466
Explanatory Variables	Total population shrinkage rate (%)	-tpsr	960	−0.070	0.186	−0.988	0.959
	Rural population shrinkage rate (%)	-rpsr	960	−0.305	0.315	−1.454	1.441
Variables	Industrial upgrading (%)	ins	960	1.281	1.592	0.094	22.078
	Medical level (bed)	med	960	14.891	12.670	0.440	79.170
	Education level (person)	edu	960	49.895	39.047	1.688	250.856
	Financial level (million CNY)	loan	960	87.815	138.300	0.016	2196.770
	Government financial capacity (million CNY)	gov	960	8.986	12.753	0.033	91.615

). Industrial upgrading (ins) was measured as the ratio of the added value of the tertiary industry to the added value of the secondary industry (%), which directly reduces the regional carbon emission intensity by shifting economic activities from energy-intensive secondary industries to the service industry with a lower energy intensity [49]. The level of medical resource allocation (med) was represented by the number of hospital beds. A higher level of medical resources usually reflects more complete social infrastructure and better public welfare, which helps improve the health of the human capital and may affect the carbon emission path by promoting green technological innovation [50]. The level of basic education (edu) was measured as the number of primary and secondary school students. The level of basic education determines the quality of the future labor force, and a higher reserve of human capital helps promote the adoption and innovation of technology, thereby inhibiting the growth of carbon emission intensity in the long term [51]. The level of financial development (loan) is the balance of various loans of financial institutions at the end of the year (10 million CNY). A developed financial system can provide the necessary financial support for green technological innovation and clean energy projects, and guide capital to flow from high-carbon to low-carbon industries, thereby promoting carbon emission reduction [24]. The government’s financial capacity (gov) is the general budget revenue of the local government (10 million CNY). A stronger government financial capacity enables it to increase environmental protection expenditures and directly provide financial support for low-carbon policies and green infrastructure, effectively reducing the regional carbon emission intensity [52].

3.3. Study Area

The Yellow River Basin spans nine provinces and autonomous regions in northern China, including Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shanxi, Shaanxi, Henan, and Shandong. It is geographically located between 32°10′ N–41°50′ N and 96°00′ E–119°15′ E, covering a total area of approximately 795,000 square kilometers (Figure 2). The terrain exhibits a three-step ladder pattern with higher elevations in the west and lower elevations in the east and primarily consists of plateaus, mountains, and plains. Land use is dominated by cultivated land, grassland, and forestland. The region has a temperate continental monsoon climate with distinct seasons and features sensitive and fragile ecosystems. It also serves as a crucial energy base and agricultural production area in China.

The selection of the Yellow River Basin as the study area was primarily based on the spatial distribution characteristics of the rural population shrinkage in China. The areas experiencing rural population shrinkage in China are mainly concentrated in the Yangtze River Basin, the Yellow River Basin, and the Heilongjiang region [53]. Meanwhile, the region’s economic structure, which relies on traditional energy sources and agricultural production, has also drawn significant attention due to its carbon emission issues. During 2010–2020, the county-level units in the Yellow River Basin faced the dual pressures of continuous rural population reduction and persistently high carbon emissions [54]. This makes the Yellow River Basin a representative area and urgent focus for studying rural population shrinkage and carbon emission intensity.

This study defined the scope of the study based on the natural watershed units and administrative divisions of the basin. Municipal districts were excluded because they are subordinate to established cities and are generally under the direct control of the provincial government, with prefectural-level cities (states) acting on their behalf. Thus, the focus of municipal districts is on “urban areas,” while the focus of counties is on “villages” [55]. Although both municipal districts and counties are administrative units at the county level, there is a large difference between them in name and in practice. In order to highlight the “counties” in this paper, we selected 51 prefectural-level cities in the eight provinces of Qinghai, Gansu, Ningxia, Inner Mongolia, Shaanxi, Shanxi, Henan, and Shandong, through which the Yellow River flows, with 320 counties selected as the study subject. Regarding the adequacy of the 320 counties and 960 balanced panel observations, our sample size is consistent with and falls within the standard range for empirical studies conducted at the county level within the Yellow River Basin. For instance, a study by Cheng effectively analyzed the multidimensional development levels across 318 counties in this region [56], demonstrating the validity of this sample scale for revealing spatial patterns and typologies. Similarly, research by Han utilized a sample of 302 county-level units within the Yellow River Basin for their analysis of agricultural green development [57], covering a similar timeframe (2011–2020).

3.4. Data Sources and Descriptive Statistics

Conducting research on the Yellow River Basin during the 2010–2020 period holds significant practical significance. This period represents a critical phase of rapid urbanization in China, characterized by dramatic shifts in rural population mobility patterns. As a typical region experiencing rural population shrinkage, the Yellow River Basin witnessed particularly pronounced rural population losses during this time. During 2000–2010, the proportion of areas in the basin showing population shrinkage was relatively low, with most counties showing population growth (Figure 3). Between 2010 and 2020, the proportion of areas showing population shrinkage within the basin rose significantly. During this period, the shrinkage intensity was predominantly high. Meanwhile, studies have found that urban populations began to grow after 2010 while rural populations showed a significant downward trend [58]. Thus, 2010 serves as a temporal turning point for urban–rural population changes.

The carbon emission data used in this study were from the China Carbon Accounting Databases (CEADs) (https://www.ceads.net.cn/ accessed on 28 February 2025). The data for the socio-economic variables were obtained from the 2010–2020 China County Statistical Yearbook and each county’s statistical yearbook and statistical bulletin (https://www.stats.gov.cn/sj/ndsj/ accessed on 31 January 2025). Since the county-level socioeconomic data of the Yellow River Basin showed an overall steady growth trend from 2010 to 2020 with no extreme fluctuations or structural mutations, this study used linear interpolation to fill in missing values. This method can maximize the retention of the temporal trend characteristics of the data, avoid artificial biases caused by complex interpolation methods, and conform to the objective laws of the data. To ensure the reliability of the research results, this study carefully handled the missing data values. There were no missing values in the core explanatory variables (-tpsr and -rpsr), which ensures that the core causal link identified was not affected by missing data. Among the control variables, only ins and loan had a small number of missing values (17/960 (1.8%) and 11/960 (1.2%), respectively) while the other control variables (edu, gov, and med) had no missing values. Given that less than 2% of the data for the variables was missing, the impact of interpolation on the sample representativeness, data distribution, and regression results was negligible. This study selects 2010–2020 as the research period and 320 counties in the Yellow River Basin as the sample, giving 960 balanced panel observations. The natural logarithm of the data for all the variables was calculated to eliminate the effect of the variables’ scale. Previous studies found that the pandemic only delayed the outflow of rural populations in the short term and did not change the long-term migration trend [59]; therefore, the 2020 pandemic did not change the overall pattern of continuous rural population outflow in the Yellow River Basin, and the reliability of the population data was not affected. The 2020 pandemic had no significant impact on the core conclusions of this study.

4. Results

4.1. Total, Urban, and Rural Population Shrinkage Trends in the Yellow River Basin

In 2010, the rural population share of the total population in the Yellow River Basin stood at 58.3%, indicating its absolute dominance. By 2020, this proportion had decreased to 46.5%, signaling a fundamental shift in the basin’s demographic structure over the decade—from a predominantly rural society to a near balance between urban and rural populations. The upper reaches maintained the highest rural population share at 55.4%, reflecting their stronger rural character. In contrast, the middle reaches underwent the most dramatic structural transformation, with a notable decline of 12.3 percentage points. The lower reaches, with the highest level of urbanization, saw their rural population share drop to 39.5% by 2020, marking their entry into a development stage primarily led by the urban population.

Rural population shrinkage dominates, characterized by a distinct pattern of urban increase rural decrease. During the study period, there were only eight counties that showed rural population growth compared to 312 counties showing rural population shrinkage, including 156 counties with weak shrinkage and 156 with strong shrinkage (Figure 4). The counties with weak rural population shrinkage were mainly concentrated in Gansu, Qinghai, Shandong, and Henan in the upper and lower reaches. The counties with strong rural population shrinkage were mainly distributed in the upper and middle reaches, mainly in Gansu, Inner Mongolia, Shaanxi, and Shanxi. Regarding the urban population, 287 counties showed population growth, spanning nearly the entire Yellow River Basin. There were 21 counties showing urban population shrinkage, including 10 counties with weak population shrinkage, which were mainly concentrated in the upper reaches, and 11 counties with strong population shrinkage, which were primarily concentrated in the lower reaches (Figure 4). Despite the high degree of rural population shrinkage, urban population growth did not result in total population growth at the county level. In terms of total population, a total of 228 counties showed total population shrinkage. Among them, 114 counties showed weak population shrinkage, which formed clusters in the middle and lower reaches, while the 114 counties that showed strong population shrinkage were concentrated in Shaanxi, Shanxi, and other areas. Thus, the total population of the Yellow River Basin is dominated by the rural population and during the study period, the Yellow River Basin population was characterized by rural population shrinkage and urban population growth (

Table 2. 2010–2020 Rural share of total population.

Category	Total Population	Rural Population	Rural Population Share
YRB	11,200.6	6531.8	58.3%
	11,309.5	5253.7	46.5%
Upper	2768.8	1840.9	66.5%
	2778.5	1539.5	55.4%
Middle	3256.8	2009.4	61.7%
	3076.8	1519.8	49.4%
Lower	5175.0	2661.5	51.5%
	5454.2	2154.4	39.5%

).

4.2. Spatial and Temporal Changes in Carbon Emission Intensity in the Yellow River Basin

4.2.1. Temporal Changes in Carbon Emission Intensity in the Yellow River Basin

In 2010, the mean carbon emission intensity in the Yellow River Basin was 3.51 with a median of 2.68, while in 2020, these figures decreased to 2.4 (mean) and 1.76 (median). This indicates a gradual decline in carbon emission intensity in the counties of the Yellow River Basin over this decade. Additionally, the narrowing gap between the mean and median values reflects shrinking regional disparities in carbon emission intensity, with the county-level emissions becoming more similar and balanced by 2020 (Figure 5).

The peak of the carbon emission intensity curve for the Yellow River Basin continuously shifted leftward, with the peak height first rising and then stabilizing and the peak width continuously narrowing (Figure 6). This indicates that the overall county-level carbon emission intensity declined steadily in 2010, 2015, and 2020, which was accompanied by narrowing spatial disparities and convergence.

The regions with varying degrees of population shrinkage exhibited distinct patterns in their carbon emission intensity trend. In the counties with weak population shrinkage, the total population and rural population curves shifted leftward, with the peak height continuously rising, reflecting an overall downward trend in emission intensity (Figure 6). However, the counties with weak rural population shrinkage showed a small secondary peak in 2020, suggesting that there were internal disparities in emission intensity improvements, which first narrowed and then widened, forming a two-tiered structure. For the counties with strong population shrinkage, the total population and rural population curves continuously shifted leftward, with a gradually increasing peak height and narrowing peak width, indicating declining emission intensities. Nevertheless, similar to the counties showing weak rural population shrinkage, counties showing strong rural population shrinkage displayed small secondary peaks in the later period, resulting in polarization.

4.2.2. Spatial Changes in Carbon Emission Intensity in the Yellow River Basin

In 2010, the overall spatial pattern of the county carbon emission intensities exhibited a polycentric, differentiated distribution, with high-value zones concentrated in specific areas and low-value zones scattered across the basin. By 2015, the high-value zones had shrunk, the number of counties with a higher emission intensity decreased, and the proportion of zones with low values increased. This eased the spatial disparities and shifted the pattern towards a balance. By 2020, the spatial pattern was dominated by low-value zones, which expanded significantly, while high-value zones became sporadic; most counties fell into the low-value categories. Compared to 2010, the spatial heterogeneity in the basin-wide emission intensity markedly weakened, with the overall pattern becoming more coordinated and balanced. This shift from pronounced differences and concentrated clusters of counties with high values to moderate disparities and a growing number of counties with low values and finally to a balanced distribution with scattered counties with high values reflects the gradual optimization of the county-level carbon emission intensity spatial patterns in the basin (Figure 7).

4.3. Spatial Correlation Between Rural Population Shrinkage and Carbon Emission Intensity

To examine whether a spatial correlation exists between population growth/shrinkage and carbon emission intensity and to map the spatial association between rural population shrinkage and carbon emission intensity, Geo Da software (1.20.0.36) was utilized to conduct a bivariate spatial autocorrelation analysis. This analysis incorporated the degree of rural population growth and shrinkage alongside carbon emission intensity during the 2010–2020 period. Overall, the resulting Moran’s I statistic was 0.126, indicating a significant positive spatial autocorrelation between the two variables.

The regions with high rural population shrinkage and a high carbon emission intensity (H-H) were mainly concentrated in the upper reaches of the Yellow River, such as in Gansu and Inner Mongolia, and showed a contiguous distribution pattern within the upper reaches. The regions with high rural population shrinkage and a low carbon emission intensity (H-L) were mostly distributed in the border zone between the middle and upper reaches of the Yellow River Basin, such as in northern Shaanxi and eastern Gansu, and exhibited a clustering characteristic in the transitional zone between the middle and upper reaches. The regions with low rural population shrinkage and a high carbon emission intensity (L-H) were scattered in the lower reaches of the Yellow River Basin, such as northern Shandong. They are mainly economically developed counties with weak rural population shrinkage or slight population growth. The regions with low rural population shrinkage and a low carbon emission intensity (L-L) were primarily concentrated in the core optimization zone of the Yellow River Basin’s lower reaches, such as in Shandong and Henan, which are the leaders in industrial upgrading. These regions clustered into a low-carbon demonstration zone in the lower reaches (Figure 8).

4.4. Rural Population Shrinkage Impacts on Basin Carbon Emission Intensity

4.4.1. Benchmark Regression Analysis

When analyzing the whole sample, -tpsr was found to exert a significant positive effect on cei, with a coefficient of 1.850 (

Table 3. Total population benchmark regression results.

Variable	(1) All Counties	(2) Population Growth	(3) Weak Shrinkage	(4) Strong Shrinkage
-tpsr	1.850 *** (0.630)	1.038 (1.206)	−0.783 (1.617)	2.310 ** (0.876)
ins	0.051 (0.072)	−0.128 (0.195)	−0.073 (0.073)	0.164 (0.104)
gov	−0.039 *** (0.009)	−0.008 (0.013)	−0.043 *** (0.012)	−0.146 ** (0.058)
med	−0.032 *** (0.008)	−0.035 *** (0.011)	−0.026 (0.016)	0.022 (0.032)
edu	0.033 *** (0.005)	0.022 ** (0.009)	0.023 *** (0.007)	0.063 *** (0.016)
loan	−0.001 (0.001)	−0.002 (0.002)	−0.000 (0.001)	0.002 (0.005)
_cons	2.922 *** (0.262)	3.234 *** (0.490)	2.527 *** (0.447)	2.623 *** (0.797)
N	960	960	960	960

Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs re-ported per norms, and **, *** denote 5%, 1% significance (p < 0.05, 0.01).

), indicating that total population shrinkage suppressed carbon emission intensity. Among the control variables, government financial capacity and medical resource allocation showed negative coefficients (−0.039 and −0.032, respectively) suggesting that an enhanced government financial capacity and optimized medical resources can reduce cei. In contrast, basic education level had a positive coefficient (0.033), implying that expanded basic education moderately contributed to higher carbon emissions. In the counties with strong population shrinkage, the -tpsr coefficient increased to 2.310, reflecting a more pronounced inhibitory effect on carbon emissions. In the full sample, -rpsr also had a significant positive effect on cei, with a coefficient of 0.629 (

Table 4. Rural population benchmarking regression results.

Variable	(1) All Counties	(2) Population Growth	(3) Weak Shrinkage	(4) Strong Shrinkage
-rpsr	0.629 ** (0.263)	−0.312 (0.446)	−0.406 (0.381)	1.201 *** (0.365)
ins	0.051 (0.071)	0.995 (0.696)	−0.046 (0.141)	0.117 (0.080)
gov	−0.039 *** (0.009)	−0.034 (0.062)	0.013 (0.019)	−0.041 *** (0.012)
med	−0.030 *** (0.008)	0.026 (0.024)	−0.032 *** (0.010)	−0.018 (0.013)
edu	0.035 *** (0.005)	−0.000 (0.058)	0.029 *** (0.006)	0.042 *** (0.009)
loan	−0.000 (0.001)	−0.007 (0.006)	−0.003 (0.003)	0.001 (0.001)
_cons	2.767 *** (0.273)	2.975 (1.677)	2.468 *** (0.388)	2.861 *** (0.489)
N	960	960	960	960

Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs re-ported per norms, and **, *** denote 5%, 1% significance (p < 0.05, 0.01).

), indicating that rural population shrinkage suppressed cei. Among the controls, government financial expenditure reduced cei through initiatives such as clean energy subsidies and rural grid upgrades, resulting in a coefficient of −0.039. Medical resource allocation lowered rural energy consumption-related cei by reducing residents’ travel for urban medical services, resulting in a coefficient of −0.030. Education level played a positive regulatory role with a coefficient of 0.035. Improved education may expose rural residents to urban consumption patterns, increasing automobile and appliance ownership and directly boosting energy use. In the counties with strong rural population shrinkage, the -rpsr coefficient rose to 1.201, mirroring the total population trend, indicating that stronger shrinkage correlated with a more significant emission-inhibiting effect.

4.4.2. Regional Heterogeneity Analysis

Due to significant differences in natural endowments, industrial structures, and policy orientations among the upper, middle, and lower reaches of the Yellow River Basin, the impact of the population shrinkage rate on the carbon emission intensity exhibited spatial heterogeneity. Based on the basin’s geographical location and administrative division, the study area was divided into the upper reaches (Gansu, Inner Mongolia, Ningxia, and Qinghai), middle reaches (Shaanxi and Shanxi), and lower reaches (Shandong and Henan). Regional differences in how the different populations affect carbon emissions were explored using the benchmark regression results.

Geographic heterogeneity tests confirmed significant regional differences in how population shrinkage affects the carbon emission intensity, validating the regional heterogeneity hypothesis (

Table 5. Heterogeneity regression results for the upper, middle, and lower reaches of the Yellow River Basin.

	-tpsr/cei			-rpsr/cei
Variable	(1) Upper	(2) Middle	(3) Lower	(1) Upper	(2) Middle	(3) Lower
-tpsr	−2.364 *** (0.779)	5.015 *** (1.216)	0.268 (0.981)
-rpsr				−1.403 *** (0.372)	2.521 *** (0.475)	0.332 * (0.178)
ins	−0.077 (0.124)	0.149 (0.105)	−0.134 (0.112)	−0.103 (0.116)	0.155 (0.105)	−0.230 * (0.121)
gov	−0.038 ** (0.018)	−0.060 (0.037)	−0.034 *** (0.008)	−0.024 (0.020)	−0.043 (0.035)	−0.034 *** (0.007)
med	−0.020 (0.022)	0.017 (0.023)	−0.029 *** (0.005)	−0.034 (0.024)	0.019 (0.025)	−0.029 *** (0.006)
edu	0.053 *** (0.016)	0.044 *** (0.010)	0.010 *** (0.003)	0.047 *** (0.0165)	0.044 *** (0.011)	0.010 *** (0.002)
loan	−0.002 (0.002)	−0.010 ** (0.004)	0.001 ** (0.001)	−0.003 (0.002)	−0.006 (0.004)	0.001 ** (0.001)
_cons	3.178 *** (0.564)	3.572 *** (0.547)	2.217 *** (0.291)	3.347 *** (0.536)	3.586 *** (0.513)	2.135 *** (0.287)
N	960	960	960	960	960	960

Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs re-ported per norms, and *, **, *** denote 10%, 5%, 1% significance (p < 0.1, 0.05, 0.01).

). In the upper reaches, total population shrinkage exerted a significant negative impact on carbon emission intensity, with a coefficient of −2.364. Rural population shrinkage also showed a significant negative effect, with a coefficient of −1.403, indicating that population shrinkage in the upper reaches promoted carbon emissions. In the middle reaches, total population shrinkage had a significant positive impact on carbon emission intensity, with a coefficient of 5.015. Rural population shrinkage similarly exhibited a positive effect, with a coefficient of 2.521. This implies that population shrinkage in the middle reaches inhibited carbon emissions. The impact in the lower reaches was relatively complex. Total population shrinkage did not have an effect on carbon emission intensity, while rural population shrinkage showed a weak positive effect with a coefficient of 0.332, suggesting that rural population shrinkage weakly inhibited carbon emissions in this region. Overall, the impact of population shrinkage on emission intensity exhibited regional heterogeneity across the basin, with the influence magnitude showing a trend of middle reaches > upper reaches > lower reaches.

For the control variables, in the upper reaches, local fiscal general budget revenue (gov) exhibited a negative effect while education level (edu) showed a significant positive effect on cei. In the middle reaches, education level (edu) had a significant positive effect and loan exhibited a significant negative impact. In the lower reaches, industrial upgrading (ins) and the number of beds in hospitals and health centers had significant negative effects, whereas education level (edu) and loans (loan) displayed significant positive effects.

4.4.3. Robustness Test

To verify the endogeneity robustness of the core conclusions, this study introduced a lag into the total population shrinkage rate (tpsr) and rural population shrinkage rate (rpsr) of one period (2009–2019). The regression results show that the coefficients for total population shrinkage for the overall sample and the strong shrinkage group were 2.870 and 4.805, respectively, while the coefficients for rural population shrinkage for the overall sample and the strong shrinkage group were 1.015 and 1.501, respectively. These results are completely consistent with the benchmark regression results in terms of coefficient sign (positive vs. negative) and significance, with only slight increases in values due to the accumulation of long-term effects, proving that the core conclusions were not influenced by reverse causality. To ensure consistency in the endogeneity treatment for the heterogeneity analysis, this study simultaneously applied lagged term (2009–2019) regression to the sub-samples of the upper, middle, and lower reaches. The results indicate that the inhibitory effect of population shrinkage in the upper reaches, the promoting effect in the middle reaches, and the weak effect in the lower reaches were all consistent with the benchmark heterogeneity results. This proves that the regional differences are real structural characteristics rather than biases caused by endogeneity (

Table 6. Robustness test for whole sample.

Variable	(1) All Counties	(2) Population Growth	(3) Weak Shrinkage	(4) Strong Shrinkage
-tpsr	2.870 *** (0.630)	−3.052 *** (1.136)	1.052 (2.051)	4.805 *** (1.468)
ins	0.037 (0.072)	−0.229 *** (0.55)	0.540 (0.324)	0.108 (0.075)
gov	−0.040 *** (0.009)	−0.008 (0.008)	−0.068 *** (0.023)	−0.198 *** (0.061)
med	−0.0316 *** (0.008)	−0.027 *** (0.007)	−0.030 * (0.018)	0.068 * (0.040)
edu	0.032 *** (0.005)	0.030 *** (0.007)	0.023 *** (0.007)	0.052 *** (0.014)
loan	−0.001 (0.001)	−0.002 (0.002)	−0.003 (0.005)	0.001 (0.006)
_cons	2.839 *** (0.267)	3.234 *** (0.490)	3.196 *** (0.447)	2.377 *** (0.789)
N	960	960	960	960

Note: Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs reported per norms, and *, *** denote 10%, 1% significance (p < 0.1, 0.01).

,

Table 7. Robustness test for rural population.

Variable	(1) All Counties	(2) Population Growth	(3) Weak Shrinkage	(4) Strong Shrinkage
-rpsr	1.015 *** (0.308)	−2.749 (1.954)	0.517 (0.441)	1.501 *** (0.410)
ins	0.057 (0.074)	−0.266 *** (0.082)	0.183 * (0.104)	0.145 (0.112)
gov	−0.026 *** (0.009)	−0.035 (0.062)	−0.055 *** (0.013)	−0.016 (0.014)
med	−0.024 *** (0.008)	0.021 (0.024)	−0.019 (0.015)	−0.017 (0.012)
edu	0.032 *** (0.005)	−0.025 ** (0.029)	0.032 *** (0.007)	0.035 *** (0.008)
loan	−0.001 (0.001)	−0.004 (0.002)	−0.002 (0.001)	0.001 (0.001)
_cons	2.713 *** (0.268)	3.380 (0.731)	2.569 *** (0.425)	2.305 *** (0.443)
N	960	960	960	960

Note: Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs reported per norms, and *, **, *** denote 10%, 5%, 1% significance (p < 0.1, 0.05, 0.01).

and

Table 8. Robustness test for regional heterogeneity.

	-tpsr/cei			-rpsr/cei
Variable	(1) Upper	(2) Middle	(3) Lower	(1) Upper	(2) Middle	(3) Lower
-tpsr	−5.085 *** (1.387)	5.840 *** (1.532)	0.667 (0.775)
-rpsr				−1.274 *** (0.316)	2.540 *** (0.476)	0.002 * (0.195)
ins	−0.041(0.1111)	0.150 (0.103)	−0.143 (0.115)	−0.076 (0.020)	0.152 (0.111)	−0.128(0.123)
gov	−0.035 * (0.020)	−0.054 (0.033)	−0.035 *** (0.008)	−0.043 ** (0.021)	−0.027(0.033)	−0.033 *** (0.008)
med	−0.011 (0.022)	0.007 (0.023)	−0.028 *** (0.006)	−0.025 (0.024)	0.014 (0.026)	−0.028 *** (0.006)
edu	0.048 *** (0.015)	0.040 *** (0.010)	0.009 *** (0.002)	0.048 *** (0.016)	0.035 *** (0.011)	0.010 *** (0.002)
loan	−0.002 (0.002)	−0.007 * (0.003)	0.001 ** (0.001)	−0.003 (0.002)	−0.002 (0.003)	0.001 ** (0.001)
_cons	3.540 *** (0.571)	3.381 *** (0.507)	2.221 *** (0.278)	3.542 *** (0.568)	3.530 *** (0.511)	2.166 *** (0.281)
N	960	960	960	960	960	960

Note: standard errors in parentheses; two robust SEs yield consistent results, clustered SEs reported per norms, and *, **, *** denote 10%, 5%, 1% significance (p < 0.1, 0.05, 0.01).

).

5. Discussion

5.1. The Spatial Association Between Rural Population Shrinkage and Carbon Emission Intensity in the Yellow River Basin

5.1.1. The Spatial Distribution of Counties with Different Degrees of Rural Population Shrinkage

A total of 206 counties exhibited total population shrinkage and a pattern of urban growth and rural shrinkage. Among these, 55 were in the upper reaches, 93 in the middle reaches, and 58 in the lower reaches, exhibiting a pattern of multi-regional distribution with discontinuous patches across the basin. They were primarily concentrated in the middle and lower reaches of Shaanxi, Henan, and other regions, showing that despite overall population shrinkage in the Yellow River Basin, urban population agglomeration and rural population outflow remained prominent. Additionally, 83 counties mainly distributed in Shaanxi, Qinghai, Gansu, and other regions showed total population growth alongside urban growth and rural shrinkage; their urban–rural population flows also showed a pattern of urban absorption and rural outflow. Notably, there were no counties with urban shrinkage and rural growth, highlighting the stability of the urban–rural differentiation pattern in population changes across the Yellow River Basin from 2010 to 2020. Nationally, most population shrinkage areas in China are characterized by rural population shrinkage and urban population growth. Although the number of counties with urban population shrinkage increased slightly between 2010 and 2020, they were mainly concentrated in Northeast China. The absence of the reverse urban–rural population exchange (urban shrinkage and rural growth) in the Yellow River Basin aligns with the national population evolution pattern noted by Liu Zhen [39], confirming that the urban–rural population changes in the basin are consistent with China’s overall population change trends.

The number of counties with total population shrinkage, urban growth, and rural shrinkage was twice that of the counties with total population growth, urban growth, and rural shrinkage (Figure 9). Among the counties showing population shrinkage, 312 (97.5%) exhibited rural population shrinkage, with only 8 showing rural population growth. Spatially, counties with strong rural population shrinkage were concentrated in the upper and middle reaches in Gansu, Inner Mongolia, Shaanxi, and Shanxi, while the counties with weak rural population shrinkage were distributed in the lower reaches in Shandong and Henan. This pattern aligns closely with the spatial characteristics of the counties showing total population shrinkage, which were clustered in the middle and lower reaches and scattered in the upper reaches (Figure 4).

5.1.2. The Spatiotemporal Link Between Rural Population Shrinkage and Carbon Emission Intensity

The carbon emission intensity across the Yellow River Basin exhibited a steady annual decline (Figure 6). However, in counties with weak and strong rural population shrinkage, side peaks in the carbon emission intensity emerged, indicating that some counties with an extremely high or low emission intensity clustered at these side peaks, which were distinct from the main group. This reveals a polarized distribution of counties with strong population shrinkage, which manifested as high- and low-carbon emission counties (Figure 5). High-carbon emission areas expanded from their traditional agglomeration in the middle reaches to the upper reaches and surrounding counties, with reduced spatial uniformity (Figure 4). This suggests that the spatial distribution of counties with a high total population carbon intensity and rural population carbon intensity across the Yellow River Basin counties is gradually shifting from stable agglomeration toward dispersion.

In terms of the spatiotemporal patterns of changes in population and carbon emission intensity, the Yellow River Basin and China’s three northeastern provinces exhibited distinct characteristics. As traditional industrial bases, the three northeastern provinces have an industrial structure dominated by heavy industries, with sectors such as iron and steel, chemicals, and machinery manufacturing accounting for a relatively large share [60]. This structure means that their spatial distribution of carbon emission intensity is more strongly influenced by the concentrated layout of large industrial enterprises, forming relatively clustered, high-intensity industrial emission zones. In contrast to the Yellow River Basin, there are fewer instances of high-carbon emission counties and low-carbon-emission counties in this region.

5.2. The Impact of Population Shrinkage on Carbon Emissions

At the whole basin scale, both total and rural population shrinkage exhibited significant inhibitory effects on carbon emission intensity, with these inhibitory effects being more pronounced in the counties showing strong shrinkage (Table 3 and Table 4). However, the overall inhibitory effect of total population shrinkage was stronger than that of rural population shrinkage. This may be related to cities and towns being the core areas of carbon emissions, as changes in urban populations are more directly and intensively linked to energy consumption in production and living sectors, thereby amplifying the emission reduction effect due to population shrinkage [61]. The control variables showed a similar regulatory mechanism in the two population shrinkage scenarios. Improved government fiscal capacity helps reduce the carbon emission intensity, an effect that may be achieved through policy tools such as clean energy subsidies and strengthened environmental regulations [62]. Meanwhile, optimized allocation of medical resources also lowers carbon emission intensity, indicating that improved medical services can reduce residents’ transportation demand in order to access medical resources, thereby indirectly inhibiting energy consumption and carbon emissions [63]. However, basic education level showed a significant positive impact. This may be because while the popularization of education improves human capital, it also promotes the upgrading of consumption patterns, leading to an energy rebound effect that partially offsets the effectiveness of energy-saving technologies [64].

In the analysis of regression results on regional heterogeneity, due to differences in natural resources, economic development, and other aspects among the upper, middle, and lower reaches, population shrinkage has different impacts on carbon emission intensity in the Yellow River Basin (Table 5). As an ecological barrier with inherent low-carbon characteristics, the upper reaches experienced accelerated rural-to-urban population concentration driven by shrinkage, increasing urban energy consumption and straining urban infrastructure. In the short term, this heightened demand outpaces the efficiency improvements from clean energy adoption or low-carbon technologies, thereby intensifying carbon emissions [65]. As a high-energy-consuming region, massive rural population outflow drives urban migration, reducing the industrial labor supply. Energy-intensive enterprises thus curtail production or adjust industries due to labor shortages, directly lowering fossil energy use and emission intensity [66]. In the lower reaches, the weak positive effect of rural population shrinkage stemmed from the partial balance between rural emission reductions and newly added urban emissions. However, the insignificant effect of total population shrinkage results from the combined impact of a stable total population and an optimized economic structure. This highlights the downstream region’s buffering capacity against environmental impacts amid urban–rural transformation and industrial upgrading [67]. The control variables also show distinct regional influences. Industrial upgrading (ins) had the strongest emission-inhibiting effect in the lower reaches, highlighting tertiary industry growth and manufacturing upgrading as core drivers of this region’s low-carbon transition [68]. Education (edu) showed positive effects throughout the basin. In the upper reaches, it reflects the human capital’s role in eco-technology adoption while in the middle reaches, it reflects industrial technology’s influence on energy consumption [69]. Local finance (gov) reduced emissions in both the upper and lower reaches [70], while lower-reach financial loans (loan) weakly increased emissions by supporting capital-intensive industries (Figure 10).

5.3. Regional Identification and Policy Implications for Rural Population Shrinkage and Carbon Emission Intensity

The significance of this study lies in its identification of problem areas centered on H-H regions and categorization of H-H, H-L, L-H, and L-L regions that require different interventions, which can guide precise policymaking for zonal management in the Yellow River Basin. This study found that from 2010 to 2020, the rural population in the Yellow River Basin showed a continuous declining trend. The counties with population shrinkage accounted for 71.25%, among which there were 156 counties with strong rural population shrinkage (Figure 4). The counties with rural population shrinkage in the upper reaches accounted for 28%, while the middle reaches emerged as the core area of strong rural population shrinkage with 62% of its counties showing this pattern; 22% of the counties in the lower reaches showed strong shrinkage (Figure 4). The regions with an annual change in carbon emission intensity exceeding the median during 2010–2020 were defined as high-carbon emission areas; 86.3% were located in the upper reaches, 10.2% in the middle reaches, and 3.5% in the lower reaches. These results are consistent with those from the bivariate LISA cluster map (Figure 9), and through mutual verification, the H-H, H-L, L-H, and L-L regions were confirmed (Figure 11).

The H-H regions were mainly concentrated in the upper reaches of the Yellow River Basin. Dominated by ecological barrier functions, rural population shrinkage in the upper reaches is primarily driven by factors such as ecological migration and weak industrial absorption capacity [71]. However, traditional energy development and energy-intensive industries in the region have not completed the low-carbon transformation. Population loss has not effectively driven industrial emission reductions; instead, the weakening of scale effects has led to a lock-in effect characterized by rural population loss alongside a persistently high carbon emission intensity [72]. The H-L regions were mostly distributed in the border zone between the middle and upper reaches of the basin. This border zone is subject to both the adjustment of energy-intensive industries in the middle reaches and ecological constraints in the upper reaches. Strong rural population shrinkage has driven passive industrial emission reductions [73], while ecological policies have inhibited carbon emissions [74], forming a transitional zone effect where population shrinkage is accompanied by a reduced carbon emission intensity. The L-H regions were scattered in the lower reaches. Due to their high industrial density, the low rural population shrinkage rate in these areas indicates a sufficient labor supply, and thus the industrial production scale has not decreased due to population loss. Traditional industries still exhibit a high carbon emission intensity, and this is coupled with growing residential energy consumption driven by a stable population, resulting in weak population shrinkage alongside a high carbon emission intensity. The L-L regions were mainly concentrated in the core optimization zone of the lower reaches. Synergy between industrial upgrading and a stable rural population, combined with low-carbon policies [75], has formed a low-carbon pattern characterized by a stable population and green industries, making it a benchmark zone for low-carbon transformation in the basin.

Targeted policies are required for these four types of regions. For H-H regions, efforts should be made to promote ecological migration and the coordinated transformation of industries. Ecological compensation funds should support new energy and ecological industries to replace traditional energy-intensive industries, thereby breaking the lock-in effect [76]. In H-L regions, policy coordination in the border zone between the middle and upper reaches should be strengthened, and platforms for ecological industry cooperation should be established to consolidate the achievements of the low-carbon transformation [77]. For L-H regions, low-carbon technological transformation of traditional industries in downstream counties should be promoted, and sufficient labor should be utilized to develop a circular economy, reducing carbon emissions from industrial and residential energy consumption [78]. In L-L regions, it is recommended to refine the low-carbon transformation model of the core downstream zone, build demonstration models featuring green industries and stable population, and disseminate such experiences throughout the basin.

5.4. Limitations

While this study covers the 2010–2020 period, a critical decade for urbanization and demographic transition, its relatively short timeframe limits the observation of long-term trends. For instance, the lagged effects of the industrial path dependence in the middle reaches and the maturation of clean energy infrastructure in the upper reaches may require a longer period to fully manifest. Given that official census data is only available up to 2020, future research should incorporate the latest census data to further deepen our understanding of the impact of population decline on carbon emission intensity in the Yellow River Basin at the county level. In addition, future research could focus on carbon emission indices in small-scale geographical areas such as townships based on nighttime lighting data in order to formulate more targeted carbon reduction policies. The year 2020 was when the COVID-19 pandemic broke out. Although this study has verified, through the Seventh National Population Census bulletin and the trend extrapolation method of nighttime remote sensing carbon emission data, that the pandemic did not change the long-term evolution of rural population shrinkage and carbon emission intensity, the pandemic’s disturbances to short-term population mobility and production activities in some industries in 2020 may still have had a slight impact on the micro-details of the cross-sectional data. This potential limitation needs to be further verified in subsequent studies with longer time-series data.

6. Conclusions

Our analysis of 320 Yellow River Basin counties during 2010–2020 demonstrated that rural population shrinkage dominated the overall changes in population. From 2010 to 2020, 97.5% of the counties experiencing population shrinkage saw a reduction in their rural population, with both the scale and speed of rural population shrinkage being significantly higher than those of the total population and urban population. This validates Hypothesis 1 which states that there is a predominant role of rural population shrinkage, by virtue of its substantial proportion, in driving the overall population decrease in the Yellow River Basin, despite urban growth. The benchmark regression results indicated that both -tpsr and -rpsr had significant positive impacts on carbon emission intensity, meaning that stronger shrinkage leads to greater inhibitory effects on carbon emissions; these results support Hypothesis 2. Regionally, this effect exhibited a distinct gradient: strongest in the middle reaches, followed by the upper reaches, and weakest in the lower reaches. The upper reaches experienced a vicious cycle of concurrent population shrinkage and rising emissions, whereas the middle reaches showed a positive effect due to shrinkage. Industrial upgrading in the lower reaches further amplified emission reductions, collectively supporting Hypothesis 3 regarding regional heterogeneity. This study clarifies the correlation law between population shrinkage and carbon emission intensity in the Yellow River Basin as well as the logic of regional differentiation. The identification of four types of regions can provide precise spatial guidance for formulating targeted policies to promote sustainable development under the “dual-carbon” goals.