Carbon Emission Reduction Effects of Government Talent Attraction Policies: Evidence from Fujian Province, China

Abstract

Fujian Province launched a talent recruitment policy in 2012 to integrate top university graduates into grassroots government roles, aiming to support green development. This study investigates the impact of recruiting “three-high” talents—those who are highly educated, skilled, and specialized—on reducing county-level carbon emissions. Using panel data from 134 counties between 2007 and 2021, we apply a time-varying difference-in-differences model. Robustness checks, including propensity score matching estimation, placebo tests, and fixed-effect controls, confirm the reliability of our results. We find that the policy significantly reduces carbon emission intensity, primarily by enhancing green technological innovation. The effect is more pronounced in urban, economically developed, and non-resource-based regions, especially where public awareness of green practices is higher. These findings suggest that localized talent policies can play a critical role in advancing low-carbon development. Our results offer new evidence for integrating human capital strategies into environmental policy design and highlight the importance of aligning recruitment efforts with regional development needs to support China’s carbon neutrality goals.

1. Introduction

General Secretary Xi Jinping emphasized at the National Organizational Work Conference “China Communist Party has always placed great importance on selecting and appointing talents, treating the selection and use of talents as a key and fundamental issue related to the Party and the people’s cause. A good cadre must be firm in belief, serve the people, be diligent and practical in governance, be willing to take responsibility, and remain clean and honest”. This underscores the party’s strategic emphasis on talent as a foundational pillar of effective governance and the country’s sustainable development.

To identify and cultivate high-quality young government cadres, Fujian Province launched an innovative recruitment policy in 2012, targeting doctoral and master’s graduates from top-tier universities such as Tsinghua University, Peking University, and Renmin University of China. These graduates were directly appointed to roles such as Deputy County (District) Head or Deputy Township (Town) Head for Science and Technology in 48 counties and districts across the province.

Since the policy’s implementation, these young officials have facilitated the introduction and advancement of over 500 projects, attracting total investments exceeding 80 billion yuan. Previous studies have demonstrated that local leaders play a vital role in fostering regional economic growth, driving structural transformation, and promoting industrial upgrading [1,2,3]. Nevertheless, the effectiveness of leadership often depends on individual backgrounds and professional experiences. For example, leaders with business experience tend to have a significantly positive impact on regional development, and younger leaders with longer expected tenures can exert greater long-term influence [4,5]. These findings suggest that leadership quality—especially educational background and professional competence—may influence not only economic growth but also environmental outcomes.

Despite this, existing literature largely focuses on provincial or municipal leaders, with limited empirical evidence on how county-level leadership—especially young, high-educated officials—affects green governance. Moreover, prior studies emphasize firm-level or policy-level instruments for emission reduction, while the role of human capital within government structures remains underexplored. This constitutes a key research gap that our study aims to address.

Therefore, the 2012 recruitment policy in Fujian—which introduced highly educated, professionally competent, and specialized individuals (commonly referred to as “three-high” talents)—raises a critical question: Can such talents meaningfully contribute to green development? And if so, what mechanisms explain their potential role in promoting environmentally sustainable economic transformation?

This study treats the government talent attraction policies as a quasi-natural experiment. We compile panel data for 134 counties from 2007 to 2021 and employ a multi-period difference-in-differences (DID) model to identify the policy’s effect on county-level carbon emissions. We conduct extensive robustness checks, including propensity score matching (PSM), fixed effects, placebo tests, and variable replacements to ensure the credibility of the empirical results.

Our findings indicate that the policy significantly reduces carbon emission intensity, mainly through enhanced green technological innovation. By assigning top-tier graduate students to grassroots leadership roles, the policy promotes innovation in low-carbon technologies. Heterogeneity analysis further reveals that the carbon reduction effects are most evident in economically developed, non-resource-based, and urban counties, especially where public awareness of green practices is higher.

This study makes several key contributions to the literature on green development and public governance. First, it shifts the analytical lens from traditional emission governance approaches—typically focused on firms or environmental regulations—to the role of human capital within government institutions. By examining the impact of highly educated grassroots officials, it provides new insights into how leadership quality can shape local environmental outcomes.

Second, the study uncovers the mechanism through which government talent attraction policies reduce carbon emissions, namely by enhancing green technological innovation. This “green innovation effect” enriches the theoretical understanding of how talent policies can contribute to sustainable development goals.

Third, from an empirical perspective, the study addresses the problem of endogeneity by treating the policy as a quasi-natural experiment. It employs a multi-period difference-in-differences (DID) approach, supplemented with propensity score matching (PSM), placebo tests, and the inclusion of province-year fixed effects to ensure the robustness and credibility of the causal identification strategy.

Finally, by using county-level data—a governance level that is critical for policy implementation yet often overlooked in existing research—this study provides valuable micro-level evidence for the design of talent-driven environmental policies in developing regions.

The subsequent contents of this study are organized as follows: Section 2 reviews the relevant literature on leadership traits, talent attraction policies, and their links to carbon emissions, identifies key research gaps, and develops the theoretical framework and research hypotheses. Section 3 outlines the empirical strategy, including the multi-period difference-in-differences (DID) model, variable construction, data sources, and sample selection. Section 4 presents the empirical results, covering baseline regressions, robustness checks, mechanism testing, and heterogeneity analysis. Section 5 offers a detailed discussion of the findings, evaluates robustness, explores mediating mechanisms, and provides policy recommendations for enhancing the impact of government-led talent programs on low-carbon development. Section 6 concludes the study by summarizing key insights, acknowledging limitations, and suggesting directions for future research.

2. Literature Review

Lin (2018) argues that a “proactive government” can foster economic growth and industrial upgrading by selecting industries that align with its resources. Government leaders serve as the embodiment of this proactive governance, playing a crucial role in enhancing governance capacity and modernizing governance structures, particularly in areas such as economic development, social governance, and environmental protection. Despite their significance, existing studies have certain limitations in fully capturing or understanding this role [6].

Most notably, prior research tends to focus on provincial or municipal leaders [7,8,9,10], while the actions and influence of county-level leaders—those directly responsible for policy execution—remain underexplored. Furthermore, the literature often emphasizes variables such as tenure, turnover, and cadre rotation, rather than examining leaders’ educational backgrounds or their recruitment pathways [11,12,13,14].

In addition, although the impact of leadership on economic growth and institutional reform has been widely studied, the relationship between leadership characteristics and environmental outcomes, such as carbon emissions or green governance, is still understudied [3,4,15,16,17]. The key references cited in this section are summarized in

Table 1. Literature Review on Local Officials and Regional Development in China.

Year	Authors	Brief Description of Research Content	Summary of Research Conclusions
2023	Xu et al. [3]	Studied whether the industrial structure experience of officials from their previous posts affects the product structure adjustment paths of enterprises in their current jurisdictions after cross-regional transfer.	Official transfers lead to convergence in the product structure of enterprises in the current jurisdiction with that of their previous jurisdiction. This trend is significantly influenced by characteristics such as the economic development level, tenure length, and typicality of the officials’ previous posts, reflecting a path dependence in industrial policies.
2016	Xu & Wang [4]	Investigated whether local officials with corporate executive experience (“political CEOs”) can more effectively promote economic growth, and explored their functional pathways.	Officials with market-oriented enterprise experience significantly boost local economic growth, with longer tenures enhancing this effect. The mechanisms involve management capabilities, knowledge structure, and social capital.
2007	Xu et al. [7]	Constructed a panel dataset on cross-regional exchanges of provincial officials and empirically tested its impact on economic growth in the receiving regions.	Cross-regional exchange of provincial governors can significantly increase the GDP growth rate in the receiving regions by approximately 1 percentage point, with the mechanism primarily achieved through promoting the development of the secondary industry and driving industrial structural adjustment.
2012	Du et al. [8]	Developed an “official experience index” to explore the relative roles of provincial officials’ career resumes and economic growth in political promotion.	Official promotion is influenced by both their economic performance and experience background, with experiential qualifications (especially central government or enterprise backgrounds) showing stronger explanatory power in promotion.
2014	Fan & Li [9]	Based on a quasi-natural experiment of ministerial replacements, used the difference-in-differences method to study the impact of political connections on fiscal transfers to the ministers’ home regions.	The appointment of new ministers significantly increases special transfer payments to their home regions, with this effect exhibiting a non-linear “inverted U-shaped” pattern.
2019	Xu & Ma [10]	Using enterprise capacity utilization data, examined whether changes in municipal party secretaries lead to excess capacity behavior driven by performance evaluation.	Official turnover exacerbates excess capacity among local enterprises, particularly during non-routine leadership transitions, with land and financial resource allocation identified as key intervention tools.
2014	Eaton & Kostka [11]	Based on field research in multiple Chinese regions, analyzed the impact of high-frequency local official turnover on the effectiveness of environmental policy implementation.	Short official tenures weaken their motivation for environmental protection, leading to low-quality, low-cost policy implementation methods, thereby reducing the overall effectiveness of environmental governance.
2014	Piotroski & Zhang [12]	Explored the influence of local officials’ promotion incentives on the timing of enterprises’ initial public offerings (IPOs).	Officials’ promotion expectations significantly drive an early surge in IPO activities within their jurisdictions, but such IPO enterprises generally perform poorly in financial metrics and long-term returns, reflecting market distortions caused by performance pressure.
2015	Gan et al. [13]	Using enterprise data, analyzed how officials incentivize enterprise expansion through resource allocation (e.g., land and financing) during their term-end periods.	To pursue political performance, officials at the end of their terms tend to provide preferential resources to enterprises, leading to a decline in enterprise capacity utilization—with state-owned enterprises benefiting most notably.
2017	Qian & Cao [14]	Based on payment system data, examined whether capital flows during local officials’ tenures exhibit regional preferences.	A distinct “money follows officials” effect exists, with capital flowing more to officials’ birthplaces and original regions during their tenures.
2007	Zhang & Gao [15]	Using 1978–2004 provincial panel data, analyzed the mechanisms through which official tenure length and cross-regional exchanges affect economic growth.	Official tenure and economic growth show an “inverted U-shaped” relationship, with cross-regional exchanges exerting a stronger positive effect on economic growth in eastern regions.
2019	Xu & Li [16]	Aggregated manufacturing enterprise capital data to the county level to study whether officials prefer allocating resources to their hometowns.	During officials’ tenures, manufacturing capital and enterprise numbers in their hometown counties grow significantly, alongside notable regional economic growth, reflecting hometown preference behavior.
2022	Zhao & Luo [17]	Used a multi-period difference-in-differences method to evaluate the impact of central performance evaluation mechanism adjustments on local governments’ expenditures in people’s livelihood areas.	Reforms to performance evaluation mechanisms effectively enhance local officials’ emphasis on education, healthcare, and other livelihood investments, weakening the “tournament-style” behavior of purely pursuing GDP.

for ease of reference. This forms a notable gap in the literature that this study seeks to address.

This study addresses a gap in the literature by extending the perspective on how the background and capabilities of grassroots officials influence regional green and low-carbon development. Carbon reduction policies typically operate through two main mechanisms: administrative regulation and market-based approaches. Previous research has shown that carbon trading markets significantly reduce emission intensity [18], and that government intervention can enhance this effect [19].

Fujian Province has adopted a distinctive talent recruitment strategy by appointing outstanding master’s and doctoral graduates from prestigious universities as Deputy County Heads for Science and Technology. This approach institutionalizes the flow of talent from academia to government, facilitating a smooth transition from knowledge generation to policy implementation. Such a shift has been shown to foster regional innovation, patent output, and R&D activities [20,21].

This policy has contributed to improvements in green technological innovation and energy efficiency, generating what scholars refer to as a “technological dividend” [22]. Technological progress enables more efficient resource utilization, lowers urban carbon intensity, and supports sustainable economic growth [22].

At the same time, the talent attraction initiative has created a sound institutional framework for the recruitment, training, evaluation, and motivation of young officials. With strong professional expertise and extensive academic networks, these officials are better equipped to formulate strategies tailored to local needs and resource constraints. By integrating ecological priorities with economic objectives, they have facilitated coordinated development in environmentally sensitive counties.

Based on the above review and theoretical logic, this study proposes the following hypotheses:

Hypothesis 1.

Government talent attraction policies, by introducing highly educated, skilled, and specialized officials, significantly reduce county-level carbon emissions and thereby enhance the effectiveness of green governance.

By recruiting high-level talent from academic institutions into local leadership roles, the Fujian provincial government has brought a forward-looking perspective rooted in elite education into county governance. These young cadres are well-versed in local ecological conditions and adept at aligning national policies with local needs, enabling the development of targeted green strategies. They leverage their alma maters’ research resources to promote collaboration between enterprises and research institutions, accelerating the adoption of advanced clean technologies [23]. Through such industry–university–research partnerships, enterprises are able to continually refine their production processes and enhance resource efficiency [24]. As a result, while advancing energy conservation and emission reduction goals, firms can simultaneously gain competitive advantages through green technological innovation [25].

Additionally, talent recruits apply their research expertise to assess the green potential of investment projects, helping attract low-carbon enterprises and promote clean production. By advancing green technology innovation, they improve energy efficiency and reduce carbon emissions at the county level. This innovation further drives industrial upgrading, encourages low-carbon inputs, and supports lean production. As outdated, high-emission industries are phased out, enterprise resource efficiency improves, contributing to lower CO₂ emissions [26].

Based on the existing literature and realistic logic, this study proposes the following hypothesis:

Hypothesis 2.

Government talent attraction policies reduce carbon emissions by strengthening local green technological innovation, which serves as the key mediating mechanism.

3. Research Design

3.1. Econometric Model

As a top-down initiative, government talent attraction policies can be viewed as an exogenous shock event, enabling the design of a quasi-natural experiment that offers a unique empirical opportunity to assess the policy’s impact on carbon emission reduction. By considering the fact that the implementation of these policies began at different times across regions, the difference-in-differences (DID) method using panel data is employed to capture the effects and variations resulting from changes in policy implementation. This approach aims to provide more accurate causal evidence regarding the impact of government talent attraction policies on county-level carbon emissions. The model follows the standard DID specification commonly used in policy evaluation studies [27], accounting for time and regional fixed effects and allowing for staggered treatment adoption across counties. The multi-period DID model is constructed as follows:

$${\left({\mathrm{C}\mathrm{O}}_2\right)}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{k}}{\sum }_{\mathrm{k}}\left({\mathrm{X}}_{\mathrm{k}\mathrm{i}\mathrm{t}}\right)+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(1)

where the explanatory variable ${\left({\mathrm{C}\mathrm{O}}_2\right)}_{\mathrm{i}\mathrm{t}}$ represents the level of carbon emissions at the county level, and the core explanatory variable ${\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}$ denotes the dummy variable indicating whether county i implements the government talent attraction policies in year t, taking a value of 1 if the policy is implemented and 0 otherwise. ${\mathrm{X}}_{\mathrm{k}\mathrm{i}\mathrm{t}}$ represents the set of control variables, with k indicating the number of control variables. ${\mathsf{\mu }}_{\mathrm{i}}$ denotes regional-fixed effects, ${\mathsf{\gamma }}_{\mathrm{t}}$ denotes time-fixed effects, and ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$ represents the random error term. The regression coefficients ${\mathsf{\beta }}_1$ associated with ${\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}$ measure the net effect of government talent attraction policies on the level of carbon emissions at county level, which is the primary focus of this paper.

3.2. Variable Construction

3.2.1. Dependent Variable: County-Level Carbon Emissions

Drawing on the approach of Shao et al. (2019) [28], the carbon emissions at the county level have been standardized. Considering that the size of the county’s economy can have a certain impact on the level of carbon emissions, carbon intensity is used to characterize the carbon emissions of the counties. Specifically, carbon intensity is defined as the ratio of the total carbon dioxide emissions of the county to the regional GDP, representing the amount of carbon dioxide produced per unit of GDP.

3.2.2. Independent Variable: Government Talent Attraction Policies

The core explanatory variable in this study is the government talent attraction policies. From 2012 to 2021, a total of 48 districts and counties in Fujian Province have successively implemented these policies. This study designates the time when these 48 districts and counties began implementing the policy as the policy shock point. A dummy variable is created to measure this effect, where districts and counties that have implemented the policy are assigned a value of 1, while those that have not are assigned a value of 0.

3.2.3. Control Variables

To account for the potential influence of other factors on the level of carbon emissions at the county level, a series of control variables are incorporated, following the approach of Zhang and Zhong (2022) [29]. These control variables include the production level of enterprises, regional population density, population size, industrial structure, level of industrial development, urban–rural income gap, and the degree of internal openness.

3.3. Data Sources and Sample Selection

This study uses panel data from 134 counties (including districts and county-level cities) spanning 2007 to 2021. The sample includes 84 counties in Fujian Province and 50 counties from neighboring provinces (Guangdong, Jiangxi, and Zhejiang), yielding an initial 2055 area–year observations. After removing three regions with missing data—Kinmen County, Meilie District, and Longgang City—the final dataset contains 2010 observations.

The period 2007–2021 was selected for two reasons. First, since the talent attraction policy was launched in 2012, starting from 2007 allows for adequate pre-policy observations to meet the parallel trend assumption in DID analysis. Second, 2021 is the latest year with complete and reliable data on key variables such as carbon emissions, green patents, and socioeconomic indicators. Data beyond 2021 were incomplete or unavailable at the time of the study.

In the data processing phase, the data are adjusted with 2007 as the base year. We use the regional GDP index of each prefecture-level city to deflate the values, resulting in the calculation of real regional GDP. Additionally, all continuous variables undergo a two-way winsorization at the 1% level to mitigate the influence of outliers.

The data used to compute county-level carbon emissions is sourced from EDGAR (The Emissions Database for Global Atmospheric Research), which provides global estimates of anthropogenic greenhouse gas and air pollutant emissions based on national and spatial grids.

The list of counties in Fujian Province that have implemented government talent attraction policies is obtained through web scraping and organizing data from the talent introduction lists published by the Student Career Guidance Service Center at Peking University, the Student Career Development Guidance Center at Tsinghua University, and the Employment Information Website of Renmin University of China.

The remaining research data are derived from the “China County Statistical Yearbook” and the statistical yearbooks of each province for the corresponding years. For any missing data in districts or counties, we utilize the CSMAR (China Stock Market & Accounting Research) and EPS (Economy Prediction System) databases for supplementation, and individual missing values are filled in using interpolation methods. Descriptive statistics of the sample data are presented in

Table 2. Descriptive Statistics of Main Variables.

Variables	Symbol	Measurement	Average Value	Standard Deviation
Carbon intensity	CO2	total CO2 emissions/real regional GDP	8.900	1.132
Government talent attraction policies	TALENT	districts and counties that have implemented the policy are assigned a value of 1, while those that have not are assigned a value of 0.	0.133	0.339
Production level of enterprises	PE	number of industrial enterprises with annual revenue of more than 20 million yuan	5.231	1.012
Regional population density	PD	total population at the end of the year/area of the administrative region	6.033	1.220
Population size	POP	total regional population at the end of the year	3.889	0.701
Industrial structure	IS	value added of the secondary industry/value added of the tertiary industry at county level	0.173	0.554
Industrial development	ID	value added of the secondary industry/nominal regional GDP	−0.808	0.331
Urban–rural income gap	GAP	urban disposable income per capita/rural disposable income per capita	1.100	0.369
Internal openness	OPEN	total retail sales of consumer goods/real regional GDP	−1.076	0.385

Note: Except for the variable representing government talent attraction policies, all other variables are log-transformed.

.

4. Empirical Analysis

4.1. Benchmark Regression Results

Table 3. Results of the Benchmark Regression.

Variables	CO2
	(1)	(2)	(3)	(4)
TALENT	−0.1469 *** (0.0300)	−0.1576 *** (0.0304)	−0.1253 *** (0.0310)	−0.1225 *** (0.0306)
PE		−0.1333 *** (0.0244)	−0.1253 *** (0.0310)	−0.1003 *** (0.0249)
PD		−0.0745 (0.0817)	−0.0608 (0.0807)	−0.0517 (0.0797)
POP		0.0058 (0.1071)	−0.0265 (0.1059)	−0.0490 (0.1047)
IS			−0.3982 *** (0.0609)	−0.2422 *** (0.0640)
ID			0.3803 *** (0.0951)	0.2053 ** (0.0975)
GAP				0.0921 (0.0600)
OPEN				0.2668 *** (0.0370)
Constants	10.0428 *** (0.0854)	11.1508 *** (0.3821)	11.5470 *** (0.3895)	11.5018 *** (0.4067)
County FE	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes
Observations	2010	2010	2010	2010
R2	0.9286	0.9298	0.9316	0.9335

Note: ** and *** indicate statistical significance at the levels of 5%, and 1%, respectively. The data in brackets are t-value. It is the same with following tables.

presents the results of the benchmark regression, showing that the estimated coefficients of the policy dummy variable Talent are significantly negative at the 1% level, both with and without control variables. This finding suggests that government talent attraction policies have a substantial negative effect on carbon emission levels, supporting Hypothesis 1. In other words, these policies effectively reduce county-level carbon emissions, contributing to carbon peaking and carbon neutrality goals. For instance, in Column (4), the Talent coefficient is −0.1225 (significant at the 1% level), indicating that, on average, pilot counties experience a 12.25% reduction in carbon emission intensity compared to non-pilot counties. This strongly confirms the policy’s positive impact on promoting carbon emission reduction at the county level.

4.2. Validity Tests for Time-Varying DID Model

4.2.1. Parallel Trend Test

The application of DID relies on the assumption that the experimental and control groups satisfy the parallel trend assumption, meaning that, in the absence of policy intervention, the outcome variables for both groups would follow a similar trend. To verify this assumption, the approach of Beck et al. (2010) [30] is adopted, utilizing an event analysis method. This involves setting the relative timing of the government talent attraction policies as a dummy variable for each experimental group and including it in the benchmark model. This approach tests whether the trend is indeed parallel before the treatment and examines the dynamic effect of the policy on county-level carbon emissions.

As shown in Figure 1, prior to policy implementation, the government talent attraction policies had no significant effect on county-level carbon emissions, indicating that the experimental and control groups were not significantly different before the policy’s introduction. This suggests that the DID model satisfies the parallel trend assumption. After the policy’s implementation, a significant difference emerges between the experimental and control groups, confirming that the government talent attraction policies have a significant negative impact on county-level carbon emissions.

4.2.2. Placebo Test

To thoroughly assess the impact of government talent attraction policies on county-level carbon emission reduction and to eliminate potential bias, this study also employs a placebo test. This involves randomly selecting counties as a simulated treatment group, constructing a pseudo-treatment group, and conducting 500 regression iterations. Figure 2 shows the distribution of regression coefficients and p-values for the pseudo-treatment group. Most of the estimated coefficients cluster around 0 and are substantially distant from the real sample’s estimated coefficient of −0.1225. This finding indicates that the observed carbon emission reduction effect of the talent attraction policies is unlikely due to random factors or omitted variables, thus validating the reliability of the regression results.

4.2.3. Goodman–Bacon Decomposition

When interpreting causal effect estimates in DID under two-way fixed effects (TWFE), it is crucial to ensure not only that the parallel trend assumption holds, but also that the control group remains unaffected by the policy shock. However, in practice, this condition is challenging to achieve, and the heterogeneity of treatment effects or the presence of negative weights can introduce estimation bias into DID coefficients, potentially compromising result accuracy [31]. To address this, the Goodman–Bacon decomposition is applied in this study, breaking down the multi-period DID coefficients into multiple 2 × 2 DID coefficients to assess the extent of bias in time-varying DID estimates under TWFE. The analysis reveals a negative weight of 5.5% in the estimation results, while appropriate treatment effects account for a substantial proportion. Therefore, the potential bias is minimal, affirming the reliability of the study’s conclusions.

4.3. Robustness Checks

4.3.1. Propensity SCORE Matching Estimation (PSM-DID)

If counties implementing government talent attraction policies are selected based on specific criteria rather than randomly, significant initial differences may occur, leading to sample self-selection bias that affects the accuracy of the findings. To mitigate this issue, the study employs propensity score matching (PSM-DID) to refine the sample before re-evaluating the impact of the government talent attraction policies on county-level carbon emissions.

Initially, a matching test is conducted to ensure high similarity between the treatment group (counties implementing the policy) and the control group (counties not implementing the policy) across multiple key control variables. Figure 3 shows that the standardized bias for most variables is significantly reduced and generally within 10% after matching, indicating effective balance.

Various matching strategies, including 2-nearest neighbor matching with a caliper, kernel matching, and radius matching, are then used to further analyze the carbon emission reduction effects. The regression results in

Table 4. PSM-DID estimation results.

Variables	CO2
	Nearest Neighbor Matching with a Caliper	Kernel Matching	Radius Matching
TALENT	−0.1219 *** (0.0435)	−0.1252 *** (0.0309)	−0.1227 *** (0.0306)
Control variables	Yes	Yes	Yes
Constant	12.1022 *** (0.7133)	11.6618 *** (0.4727)	11.5227 *** (0.4074)
County FE	Yes	Yes	Yes
Time FE	Yes	Yes	Yes
Observation	1171	1965	2008
R2	0.9196	0.9321	0.9336

Note: *** indicate statistical significance at the levels of 1%.

show that the estimated effects remain consistent in significance and direction across all methods, reinforcing the robustness of the benchmark results.

4.3.2. Exclusion of Other Policy Impacts

The carbon emission reduction effects of government talent attraction policies may be influenced by related policies, such as environmental protection and energy conservation, potentially affecting existing estimates. This study covers the period from 2007 to 2021, during which the “low-carbon city pilots” initiated in 2010 and the “pilot scheme on comprehensive ecological protection compensation” implemented between 2018 and 2020 may have impacted local carbon emissions. Therefore, the inhibitory effect observed in the benchmark regression may be attributed to these policies.

To account for the potential influence of these exogenous policy shocks, a robustness test is conducted by incorporating dummy variables for both policies into the benchmark regression model. The results in Column (1) of

Table 5. Robustness test.

Variables	CO2
	Exclude Parallel Policies	Joint Fixed Effects	Replace Explained Variables		Change Policy Time
	(1)	(2)	PCO2 (3)	TCO2 (4)	AD3 (5)	AD4 (6)
TALENT	−0.1033 *** (0.0309)	−0.0672 ** (0.0329)	−0.4411 ** (0.1774)	−0.2628 *** (0.0954)	−0.0599 (0.0710)	−0.0442 (0.0662)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
Constant	11.9772 *** (0.4271)	11.4475 *** (0.4155)	9.3421 *** (2.3603)	2.2522 * (1.2699)	11.3322 *** (0.5791)	11.2904 *** (0.5654)
County FE	Yes	Yes	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes	Yes	Yes
Province × year joint fixed effects	No	Yes	No	No	No	No
Observation	2010	2010	2010	2010	2010	2010
R2	0.9343	0.9353	0.9633	0.9703	0.9331	0.9330

Note: *, ** and *** indicate statistical significance at the levels of 10%, 5% and 1%, respectively.

show that even after controlling for these policy shocks, government talent attraction policies still significantly reduce county-level carbon emissions, confirming the strong validity of the benchmark results.

4.3.3. Joint Fixed Effects Control

Different provinces implement varying carbon emission reduction policies in specific years, leading to corresponding changes in carbon emission trends. To account for this variation, following the approach outlined by Liu et al. (2018) [32], the joint fixed effects of provinces and years are controlled in the benchmark regression model to capture the time-varying policy effects specific to each province. The regression results, presented in Column (2) of Table 4, demonstrate that controlling for these joint fixed effects does not fundamentally alter the regression outcomes. The significant carbon emission reduction effect of government talent attraction policies remains intact, further reinforcing the robustness of the research conclusions.

4.3.4. Substitution of Dependent Variables

To capture dynamic changes in carbon emissions, this study uses per capita carbon emissions (PCO₂) and total carbon emissions (TCO₂) as new explanatory variables for robustness analysis. Employing a time-varying DID approach, we conduct regressions on the benchmark model to assess the validity of our conclusions after replacing the explanatory variables. The results in Columns (3) and (4) of Table 4 reveal that the core variable, Talent, remains significantly negative for both total and per capita carbon emissions. This confirms that government talent attraction policies effectively reduce carbon emissions at the county level, supporting carbon peaking and carbon neutrality goals, thus validating the benchmark regression results.

4.3.5. Alteration of POLICY Implementation Timeline

To validate the robustness of the regression results, this study conducts a counterfactual robustness test on the carbon emission reduction effect of government talent attraction policies by advancing the implementation time by 3 years (AD3) and 4 years (AD4) for a placebo test. The estimation results in Columns (5) and (6) of Table 4 show that when considering 2009 and 2008 as the policy implementation years, the randomly constructed policy variables exhibit no significant inhibitory effect on carbon emissions compared to the 2012 policy shock time. This confirms the robustness and credibility of the research findings.

4.4. Mechanism Analysis

The theoretical analysis suggests that government talent attraction policies play a significant role in promoting green technological innovation. To empirically test this mechanism, we draw on the approach of Jiang (2022) [33] and Shao et al. (2022) [22], using the number of green patents per capita at the county level as a proxy for green innovation capacity. The patent data are sourced from the China National Intellectual Property Administration (CNIPA) and filtered based on the International Patent Classification (IPC) Green Inventory. The specific model is constructed as shown in Equation (2).

$${\mathrm{G}\mathrm{P}}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{k}}{\sum }_{\mathrm{k}}\left({\mathrm{X}}_{\mathrm{k}\mathrm{i}\mathrm{t}}\right)+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(2)

where ${\mathrm{G}\mathrm{P}}_{\mathrm{i}\mathrm{t}}$ is the log of the number of green patents per capita in county i and year t, used to proxy green innovation capacity. Other variables are defined as in Equation (1).

Column (1) of

Table 6. Mechanism regression results.

Variables	(1)	(2)
TALENT	0.7345 *** (0.1376)	0.5289 *** (0.1426)
Control variables	No	Yes
Constant	4.9067 *** (0.3914)	9.7793 *** (1.8973)
County FE	Yes	Yes
Time FE	Yes	Yes
Observation	2010	2010
R2	0.6932	0.7042

Note: *** indicate statistical significance at the levels of 1%.

examines the impact of government talent attraction policies on regional green technology innovation. Building on Column (1), additional control variables are included in Column (2). The results show that the core explanatory variable, Talent, is significantly positive at the 1% level, regardless of whether other influencing factors are controlled, indicating that government talent attraction policies indeed help elevate regional green technology innovation.

Government talents bring not only strong professional expertise and foresight but also resources from their alma maters and valuable connections with scientific research teams, which foster green technology innovation based on regional resources and accelerate the application of scientific research achievements. Green technology innovation, in turn, transcends existing resource allocation boundaries, enhancing resource efficiency to further control and reduce carbon emissions. Consequently, the policy of attracting government talents can facilitate carbon emission reductions at the county level by advancing green technology innovation. These findings confirm Hypothesis 2.

4.5. Heterogeneity Analysis

4.5.1. Heterogeneity by County Type

Resource-based cities are those in which the extraction and processing of local natural resources, such as minerals and forests, serve as the primary industries. This category includes both prefecture-level administrative regions (such as prefecture-level cities and regions) and county-level administrative regions (such as county-level cities and counties).

According to the Sustainable Development Plan of National Resource-based Cities (2013–2020), there are currently 262 resource-based cities in China. These cities are a crucial support for the sustained and healthy development of the national economy; however, their economic development is accompanied by substantial carbon emissions.

Therefore, promoting carbon emission reductions in resource-based cities is a key element in advancing China’s “dual carbon” goals (carbon peaking and carbon neutrality). For this purpose, the classification of cities into resource-based and non-resource -based cities is applied for sub-sample regression analysis, with the regression results presented in Columns (1) and (2) of

Table 7. Heterogeneity test results.

Variables	County Type		Economic Development
	(1) Resource-Based	(2) Non-Resource-Based	(3) Low	(4) High
TALENT	0.0073 (0.0346)	−0.2088 *** (0.0442)	−0.0666 (0.0558)	−0.1392 *** (0.0376)
Control variables	Yes	Yes	Yes	Yes
Constant	12.9823 *** (0.8350)	11.4829 *** (0.5297)	11.5490 *** (0.8496)	12.4205 *** (0.5103)
County FE	Yes	Yes	Yes	Yes
Time FE	Yes	Yes	Yes	Yes
Observation	660	1350	1166	844
R2	0.9592	0.9203	0.9048	0.9690
p value tested by Talent coefficient group difference	0.026 **		0.074 *

Note: *, ** and *** indicate statistical significance at the levels of 10%, 5% and 1%, respectively.

.

The analysis of sub-sample regression results suggests that the policy of attracting government talents significantly promotes carbon emission reduction in non-resource-based counties, while showing no substantial impact in resource-based counties. This indicates that the policy effectively lowers carbon emission levels in non-resource-based counties. However, in resource-based counties, existing industries remain heavily dependent on natural resources, such as minerals. Consequently, the green technology innovation introduced by government talents does not significantly advance the green transformation of local industries, thereby diminishing its effect on regional carbon emission reduction.

4.5.2. Heterogeneity by Economic Development Level

Differences in economic development levels among counties may influence the carbon emission reduction effects of government talent attraction policies. A higher level of economic development typically indicates a well-established market order, industrial structure, and institutional environment, which is more conducive to enabling government talents to leverage their scientific expertise, carry out green technology innovation suited to local conditions, and promote the greening of industrial chains.

This study measures the economic development level of counties using per capita real GDP. The sample is further divided into two groups for separate testing: specifically, the average per capita real GDP of counties in each year is taken as the classification criterion. Counties with values above the average are classified as the “high group”, while those with values below the average are classified as the “low group”.

The results in Columns (3) and (4) of Table 6 indicate that the carbon emission reduction effect of government talent attraction policies is less significant in counties with lower economic development. In contrast, in counties with higher economic development, the policy’s effect on reducing carbon emissions is both significant and strongly negative.

This finding suggests that robust financial support and a solid economic foundation provide the necessary resources and support for government talents to advance green technology innovation. Furthermore, this economic foundation facilitates the innovation-driven upgrading and optimization of the industrial structure, aiding in the achievement of carbon peaking and carbon neutrality targets.

4.6. Extended Analysis

Achieving carbon peaking and carbon neutrality goals is a long-term endeavor. During the green and low-carbon transformation of the economy and society, both administrative units and the public play vital roles in realizing these objectives.

First, municipal districts are typically the core areas of high-quality economic development, characterized by environmentally friendly consumption patterns, modern industrial structures, and clean energy frameworks. Additionally, municipal districts often house provincial or municipal governments. As local officials, the introduced talents, under the pressure of performance appraisals and environmental protection assessments, are likely to enhance the supervision of environmental performance, thereby contributing to the dual goals of economic development and environmental protection.

Second, when the public in a region possesses a strong awareness of green and low-carbon practices, they are more likely to adopt sustainable lifestyles. This awareness can influence various sectors, including production, transportation, and energy consumption. The public can also actively monitor the carbon emission behaviors of enterprises, thereby playing a crucial role in advancing carbon peaking and carbon neutrality strategies.

Based on this analysis, this study establishes two variables: the administrative unit dummy variable (RANK) and public green and low-carbon awareness (PUBLIC). In the RANK variable, regions classified as municipal districts are assigned a value of 1, while those classified as county-level cities are assigned a value of 0. The measurement of public green and low-carbon awareness (PUBLIC) follows the approach of Tan (2023) [34], using the Baidu index of five keywords: “carbon peak”, “carbon neutrality”, “dual carbon”, “carbon emissions”, and “low carbon”. The regression model is expressed as Equation (3):

$${\left({\mathrm{C}\mathrm{O}}_2\right)}_{\mathrm{i}\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1{\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_3{{\mathrm{T}\mathrm{A}\mathrm{L}\mathrm{E}\mathrm{N}\mathrm{T}}_{\mathrm{i}\mathrm{t}}\times \mathrm{Z}}_{\mathrm{i}\mathrm{t}}+{\mathsf{\beta }}_{\mathrm{k}}{\sum }_{\mathrm{k}}\left({\mathrm{X}}_{\mathrm{k}\mathrm{i}\mathrm{t}}\right)+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\gamma }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{t}}$$

(3)

where Z_it represents either RANK_it or PUBLIC_it, and the interaction term TALENT_it*Z_it captures how the policy effect varies across regions with different administrative statuses or public awareness levels. All other variables are defined as in Equation (1).

The regression results for administrative units are presented in Column (1) of

Table 8. Influence of administrative units and public green and low-carbon awareness on the carbon emission reduction effects of government talent attraction policies.

Variables	(1)	(2)
TALENT	0.0966 (0.0911)	−0.1234 *** (0.0305)
RANK	−1.9826 *** (0.3008)
TALENT × RANK	−0.3889 *** (0.1098)
PUBLIC		0.0911 (0.3638)
TALENT × PUBLIC		−0.1166 *** (0.0377)
Control variables	Yes	Yes
Constants	12.2586 *** (0.6381)	2.3410 *** (0.4953)
County FE	Yes	Yes
Time FE	Yes	Yes
Observations	1365	2010
R2	0.9268	0.9339

Note: *** indicate statistical significance at the levels of 1%.

. The regression coefficient for the interaction term between government talent attraction policies and the administrative unit (Talent × Rank) is significantly negative at the 1% level, indicating that the carbon emission reduction effect of these policies is more pronounced in municipal districts.

Column (2) of Table 7 presents the regression results for public green and low-carbon awareness (PUBLIC). The regression coefficient for the interaction term between government talent attraction policies and public green and low-carbon awareness (TALENT × PUBLIC) is also significantly negative at the 1% level. This finding suggests that stronger public awareness of green and low-carbon practices enhances the carbon emission reduction effect associated with the government talent attraction policies. In other words, maintaining a strong connection with the public remains an effective approach to achieving China’s carbon emission goals.

5. Discussion

5.1. Benchmark Regression Interpretation

The benchmark regression results (Table 3) demonstrate that government talent attraction policies have a statistically significant and robust negative effect on county-level carbon emissions. The coefficient for the TALENT variable remains significantly negative across all specifications, with the most comprehensive model (Column 4) showing a 12.25% reduction in carbon emission intensity. This confirms Hypothesis 1 and provides compelling evidence that the policy contributes effectively to carbon peaking and carbon neutrality goals through improved local governance capacity.

5.2. Robustness and Validity Assessment

Multiple robustness tests validate the consistency of the benchmark results. The parallel trend test (Figure 1) confirms the suitability of the DID approach, while the placebo test (Figure 2) shows that the observed effect is unlikely due to random allocation or omitted variables. The Goodman–Bacon decomposition further supports the credibility of the estimated effects by showing only 5.5% negative weighting, indicating minimal distortion from treatment heterogeneity.

Robustness is also reinforced through PSM-DID estimation (Table 4), which addresses potential sample selection bias. Additionally, Table 5 demonstrates that the main findings hold after controlling for other concurrent policies, province-year fixed effects, alternative dependent variables, and counterfactual policy timing—thus confirming the reliability of the estimated causal effect.

5.3. Mechanism Analysis: Green Technological Innovation

Table 6 provides evidence supporting Hypothesis 2, namely that the policy effect is mediated by green innovation. The positive and significant relationship between TALENT and green patent output suggests that the presence of high-quality government recruits enhances the county’s capacity for green technological innovation. This innovation facilitates more efficient resource use and cleaner production processes, thereby serving as a critical mechanism through which the policy reduces emissions.

5.4. Heterogeneity of Policy Impact

The sub-sample regression results presented in Table 7 reveal notable heterogeneity in the effectiveness of government talent attraction policies across different regional contexts. Specifically, the policy exhibits a significant carbon emission reduction effect in non-resource-based counties, while the effect is statistically insignificant in resource-based counties. The findings indicate that regions heavily reliant on natural resource extraction and processing may encounter structural constraints that impede the full potential of talent-driven green innovation. Moreover, policy effects are more pronounced in economically advanced areas, where robust financial resources, sophisticated industrial infrastructure, and well-developed institutional frameworks create favorable conditions for high-level talent to apply their expertise, implement innovative initiatives, and facilitate effective green transitions. These insights underscore the importance of tailoring policy implementation to regional development contexts in order to optimize environmental outcomes.

5.5. Interaction Effects: Administrative Units and Public Awareness

Table 8 examines the interaction effects of administrative units and public green and low-carbon awareness on the effectiveness of government talent attraction policies. The results show that the interaction term between TALENT and RANK is significantly negative, indicating that the policy has a more pronounced carbon reduction effect in municipal districts. This may be attributed to the fact that such areas typically possess better infrastructure, stronger regulatory capacity, and more mature innovation ecosystems, which enhance the impact of introduced talent.

Similarly, the interaction term between TALENT and PUBLIC is also significantly negative, suggesting that the higher the public’s awareness of low-carbon practices, the more effective the policy becomes. This finding underscores the important role of public participation and civic engagement in strengthening policy outcomes and promoting effective green governance.

5.6. Policy Implications

Based on the above findings, we propose the following policy recommendations:

First, taking Fujian Province as an example, the feasibility and benefits of government talent attraction policies are demonstrated, providing a successful model for other counties. This approach encourages provinces, prefecture-level cities, and counties across the country to adopt targeted policies for attracting government talent. By learning from the experiences of regions that have already implemented such policies, local governments can continuously refine and improve their strategies to meet the needs of regional development. Tailoring policies to local conditions not only helps cultivate a high-quality cohort of young officials but also enhances the carbon reduction effectiveness of these policies, thus contributing to China’s achievement of its dual carbon goals in the new development stage.

Second, while emphasizing top-level policy design, it is necessary to expand the role of higher education institutions in the process of government talent recruitment. Greater attention should be paid to the management, training, and deployment of these talents to ensure their effective attraction, retention, and utilization. Relevant departments should proactively formulate comprehensive support policies, including multi-dimensional talent development plans, opportunities for continuous professional growth, follow-up assessments, and family support mechanisms for the recruited talents. It is important to design targeted development programs to support the growth of young grassroots cadres. In addition, leveraging unique regional resources to offer financial subsidies and credit support can further attract and retain talent while enhancing their sense of belonging to the local community.

Third, special assessments should be established for the performance of recruited government talent, with an increased emphasis on environmental outcomes. Upon completion of the two-year temporary assignment, evaluations should focus on the county’s environmental performance, including carbon emission levels. Key evaluation areas should include green and low-carbon industrial development, green technology R&D, and public participation in low-carbon consumption. These measures aim to incentivize recruited talent to lead their counties toward a balanced path of economic development and environmental protection while also preventing superficial or short-term carbon reduction strategies.

Finally, the long-term nature of China’s carbon peaking and carbon neutrality goals may conflict with the short-term nature of such temporary assignments. Therefore, it is essential to develop a targeted and scientifically grounded evaluation system that takes into account regional differences in natural resource endowment, economic development level, and industrial structure. Such an approach is vital for supporting counties in sustainably achieving their carbon reduction targets over the long term.

6. Conclusions

This study investigates the impact of government talent attraction policies on county-level carbon emissions using a multi-period difference-in-differences (DID) approach with panel data from Fujian Province and its neighboring regions between 2007 and 2021. The empirical results demonstrate that these policies significantly reduce carbon emission intensity, primarily through the enhancement of green technological innovation. The effect is more pronounced in non-resource-based and economically developed counties, particularly in municipal districts and areas with stronger public awareness of environmental issues. These findings highlight the importance of tailoring talent policies to regional characteristics and promoting public participation to improve the effectiveness of green governance.

Despite the valuable insights provided, this study is subject to several limitations. First, the analysis may suffer from potential omitted variable bias or the confounding effects of concurrent environmental policies, which could affect the precision of the estimated treatment effects. Second, the proxy used for green innovation—green patent applications—may not fully capture the quality, relevance, or actual environmental impact of the innovations. Third, as the empirical analysis is limited to a specific region (Fujian and neighboring provinces), the generalizability of the findings to other parts of China or other countries with different administrative, economic, or environmental contexts may be constrained.

Future studies could build upon this work by exploring the long-term impacts of talent attraction policies on environmental outcomes and evaluating their effectiveness in different institutional and regional settings. Researchers may also adopt more granular data, alternative indicators for innovation quality, and advanced causal inference techniques (e.g., synthetic control or event study methods) to enhance identification robustness. In addition, comparative studies across provinces or countries could help uncover contextual factors that shape the success of talent-driven green development initiatives, offering richer insights for policy design.