The Impact of Financial Literacy on Clean Cooking Fuel Choice Among Rural Households: Evidence from China

Abstract

Promoting clean cooking fuel adoption among rural households is critical for China’s rural energy transition. However, existing studies have largely overlooked the role of endogenous cognitive capabilities, particularly financial literacy, in shaping household energy choices. To fill this gap, this study uses 2018 China Family Panel Studies (CFPS) data and employs Probit and IV methods to investigate the association between financial literacy and clean cooking fuel choice among rural households, as well as the potential channels underlying this relationship. The results show that financial literacy is significantly and positively associated with clean cooking fuel adoption. Exploratory heterogeneity analysis further reveals that this positive association is more pronounced among households with lower educational attainment, and that the effect of financial literacy diminishes as the number of children in the household increases. Channel analysis suggests that financial literacy may operate through easing credit constraints, promoting non-agricultural employment and entrepreneurship, and enhancing social trust. These findings suggest that strengthening financial education and improving rural credit systems may help alleviate borrowing constraints and optimize livelihood strategies, thereby promoting the adoption of clean energy in rural households.

1. Introduction

Energy is not only the material foundation of human survival but also the core driving force for sustainable economic and social development [1]. Ensuring universal access to affordable, reliable, and sustainable modern energy is one of the core components of the United Nations Sustainable Development Goals (SDGs), notably SDG 7. However, constrained by inadequate infrastructure and low income levels, a large proportion of the global population still faces difficulties accessing clean energy. Data from the International Energy Agency shows that as of 2020, approximately 2.4 billion people worldwide still lacked access to clean cooking fuels; according to the World Health Organization, the resulting indoor air pollution causes 3.2 million deaths annually. More recent estimates, based on the 2024 edition of the Tracking SDG 7: The Energy Progress Report (jointly published by IEA, IRENA, UNSD, the World Bank, and WHO), indicate that approximately 2.1 billion people still lacked access to clean cooking fuels and technologies as of 2022, with the number remaining largely stagnant. As the world’s largest developing country and energy consumer, China faces a particularly acute challenge in its vast rural areas, where the clean energy transition is both a national policy priority and a critical pillar of the broader rural revitalization strategy. Despite notable progress—World Bank data indicates that the proportion of rural households using clean cooking fuels rose from 54.6% in 2016 to 77.4% in 2023—a considerable gap remains. These statistics reveal that while advancements have been made, a substantial share of rural households, particularly in less developed central and western regions, still rely on traditional solid fuels, signaling an incomplete transition. This persistent challenge highlights the need for continued research into the drivers of clean energy adoption, such as financial literacy.

To promote clean rural energy transition, the Chinese government has successively introduced a series of policy measures. The Rural Revitalization Strategic Plan (2018–2022) issued by the State Council in 2018 explicitly identified guiding the transition of rural energy consumption toward high-efficiency and clean patterns as a key development orientation; the 2025 Central Document No. 1 further emphasized the need to stimulate the endogenous development momentum of rural areas and improve farmers’ development capacity through enhanced development-oriented assistance, providing solid support for alleviating the difficulties in accessing clean energy in rural areas. The successful advancement of rural clean cooking fuel choice requires not only government “effective blood transfusion”—improving energy infrastructure construction and policy subsidy mechanisms—but also the enhancement of farmers’ “self-generating” capacity. Among these, financial literacy, as a key component of farmers’ endogenous development capacity, has drawn increasing attention from academia regarding its role in energy consumption decision-making [2]. Financial literacy refers to the comprehensive ability of individuals to master financial knowledge and apply financial skills to make sound financial decisions [3]. It not only affects household financial market participation behavior but may also influence the adoption of clean energy—a decision with dual economic and environmental attributes—by optimizing resource allocation efficiency and enhancing risk assessment capabilities. Against this backdrop, systematically investigating the impact mechanism of financial literacy on rural households’ clean energy adoption holds significant theoretical value and practical implications for activating the endogenous momentum of rural energy transition and improving the rural revitalization policy system.

The existing literature has accumulated a rich body of research on the factors influencing clean energy adoption, primarily focusing on two dimensions. The first is the external conditions dimension, encompassing factors such as energy infrastructure accessibility [4], environmental regulations and policies [5], and energy prices [6]. For example, Yang and Wan (2022) [7] found that the closer the proximity to clean energy supply points, the stronger farmers’ willingness to adopt. The second is the household internal characteristics dimension, including income levels [8], environmental perception [9], and household size [10]. He et al. (2023) [8] confirmed that rising income levels can significantly drive farmers’ demand for clean energy sources such as natural gas and electricity.

In research on the association between financial factors and clean energy adoption, scholars initially focused on financial inclusion. Acheampong et al. (2025) [11] demonstrated that financial inclusion has a significant positive effect on clean energy adoption. As research has deepened, some scholars have begun to focus on financial literacy as a core micro-level variable. Ankrah Twumasi et al. (2022) [12] used Ghana as the research subject to verify the promoting role of financial literacy in residents’ renewable energy adoption. Koomson et al. (2025) [13] examined financial literacy and clean energy adoption in the South African context, identifying entrepreneurship, financial inclusion, and generalized trust as transmission channels. Wang et al. (2025) [2] used China Household Finance Survey (CHFS) data to examine the link between farmers’ financial literacy and energy poverty, and employed CFPS data to examine the mechanisms driving financial literacy’s impact on energy structure. Vargas-Hernández (2026) [14] comprehensively examined the relationship between financial literacy and renewable energy adoption in Mexico, finding that while Mexico has tremendous potential for renewable energy, its low financial literacy levels constitute a significant barrier to the widespread application of clean energy technologies. Evidence from developed countries further supports this relationship. For instance, Blasch et al. (2021) [15] proposed the concept of energy-related financial literacy and found that it is positively associated with the adoption of energy-efficient durable goods in a large household survey across three European countries (Italy, Netherlands, and Switzerland). Kalmi et al. (2021) [16] similarly found that higher energy-related financial literacy is associated with lower electricity consumption in Finnish households. Boogen et al. (2021) [17] further documented that stronger energy-related financial literacy is linked to higher electricity efficiency in the European residential sector.

However, the existing research still has notable deficiencies. First, empirical research targeting China’s rural context is relatively scarce. Most of the literature focuses on foreign contexts, and the unique energy structure, financial market environment, and policy system in rural China make it difficult to directly apply research findings from other countries to Chinese practice. Second, mechanism analysis lacks systematicity. Existing studies have largely failed to clarify the specific transmission pathways through which financial literacy affects clean energy adoption, and the exploration of heterogeneous characteristics remains insufficiently in-depth. Third, some studies suffer from data limitations—either small sample sizes or a lack of core indicators directly measuring financial literacy, which undermines the reliability and generalizability of research conclusions.

Against this backdrop, this paper adopts data from the 2018 CFPS to examine the relationship between financial literacy and rural households’ clean energy adoption, and to investigate the mechanisms through which any such association may operate. We also explore whether the effects vary systematically across different household characteristics. Through this analysis, we aim to provide empirical evidence that can inform the design of more effective policies for promoting rural clean cooking fuel choice.

Relative to the existing literature summarized above, this paper offers three marginal contributions. First, we advance the core research focus by examining the transition to clean cooking fuels among rural households in China. While a few studies have explored the link between financial literacy and clean energy adoption, existing evidence predominantly stems from contexts such as Ghana, South Africa, and Mexico. By systematically evaluating how financial literacy drives this transition within the specific setting of rural China, our study fills a critical gap in the literature regarding the determinants of clean cooking fuel adoption. Second, we broaden the theoretical understanding of the mechanisms through which financial literacy affects clean energy adoption. Distinct from previous studies that primarily emphasized trust, financial inclusion, or entrepreneurship, this paper—grounded in the socioeconomic context of rural China—identifies and empirically validates two key transmission channels: promoting non-agricultural employment and alleviating credit constraints. These channels reveal how financial literacy drives clean energy adoption by diversifying household economic structures and relaxing binding budget constraints, offering a theoretically grounded framework for understanding the finance–energy nexus in rural developing economies undergoing structural transformation. Third, we provide granular heterogeneous effect analysis that yields actionable micro-level evidence for targeted policy design. This study systematically examines how the impact of financial literacy varies across dimensions of human capital (e.g., household head’s education) and demographic structures (e.g., household dependency burden) within rural China. These heterogeneity findings identify which household characteristics condition the effectiveness of financial literacy in promoting clean energy adoption, thereby informing the design of differentiated rural energy and financial education policies.

2. Theoretical Analysis and Research Hypotheses

To establish a rigorous theoretical foundation for the formulation of our hypotheses, this study integrates two complementary theoretical perspectives. First, Human Capital Theory conceptualizes financial literacy as a critical form of human capital that enhances individuals’ capacity to process information, evaluate intertemporal trade-offs, and make economically efficient decisions [18]. Within the context of household energy consumption, clean energy technologies are typically characterized by high upfront costs but yield significant long-term economic and health benefits. Individuals with higher financial literacy are better positioned to accurately assess these intertemporal returns, thereby effectively overcoming the cognitive biases that often result in underinvestment in clean energy.

Second, the Sustainable Livelihoods Framework offers a structural lens for understanding the mechanisms through which financial literacy operates. This framework identifies financial, human, social, and physical capital as interacting assets that shape household livelihood strategies [19]. In this context, financial literacy functions as an enabling capability—a meta-competency that optimizes the mobilization and combination of these assets. Specifically, it: (1) facilitates access to financial capital by mitigating information asymmetry and enhancing creditworthiness; (2) strengthens human capital-based strategies by facilitating non-agricultural employment and entrepreneurship; and (3) bolsters social capital by improving the understanding of market rules and contract enforcement, thereby fostering generalized trust. Synthesizing these theoretical insights, we propose four hypotheses that capture both the direct effect of financial literacy (H1) and its indirect effects mediated through credit constraints (H2), non-agricultural employment and entrepreneurship (H3), and social trust (H4). Each hypothesis is explicitly derived from and anchored in the theoretical perspectives articulated above.

2.1. The Internal Logic of Financial Literacy Affecting Rural Households’ Clean Cooking Fuel Choice

With the popularization of environmental protection concepts and the advancement of energy structure transition, household clean energy adoption has become an important micro-level pillar for promoting sustainable development and green transition, and the role of financial decision-making ability in this process has gradually drawn scholarly attention [13]. Financial literacy is not merely simple numerical ability; rather, it represents an individual’s comprehensive ability to acquire and process financial information and make intertemporal optimization decisions. The existing literature has extensively explored the importance of financial literacy for various economic behaviors [20]. In the context of household cooking fuel choice, transitioning from traditional solid biomass (such as firewood and crop residues) to modern clean cooking fuels (such as LPG, natural gas, and electricity) typically involves an intertemporal trade-off. Traditional solid fuels are often “free” to collect but are highly inefficient and generate severe indoor air pollution. Conversely, adopting clean cooking fuels requires households to make an initial investment in new stoves and to incur recurring expenditures to purchase commercial fuels. Evaluating this transition requires decision-makers to weigh current and ongoing financial costs against future benefits, including time savings, improved health outcomes, and enhanced environmental comfort. Individuals with higher financial literacy generally possess stronger intertemporal decision-making abilities and can more accurately assess the ongoing expenditures and long-term net benefits associated with clean cooking fuel use [3]. Additionally, traditional cost–benefit analyses often overlook non-monetized utilities, while financially knowledgeable decision-makers tend to have broader horizons and can better understand the long-term potential benefits of choosing clean cooking fuels, thereby incorporating these factors into their decision function [21]. In other words, financial literacy can help households eliminate cognitive biases caused by short-sightedness or insufficient calculation ability, enabling them to recognize that even though upfront stove investments and recurring fuel purchases impose continuous financial burdens, the long-term cost-saving potential, health dividends, and time freed from biomass collection can outweigh these immediate financial constraints. Based on this, financial literacy can effectively alleviate the cognitive barriers and behavioral biases that hinder households from transitioning to clean cooking fuels. Thus, the following research hypothesis is proposed:

H1. 

Financial literacy can increase the probability of households using clean energy.

2.2. The Mechanism of Financial Literacy Affecting Rural Households’ Clean Cooking Fuel Choice

2.2.1. Credit Constraint Mechanism

Credit constraints are a key factor constraining households’ large durable goods consumption. The first possible channel through which financial literacy affects households’ clean cooking fuel choice is by alleviating the credit constraints households face. First, financial literacy can reduce the information asymmetry households face in obtaining credit funds. Households with low financial literacy are often unable to effectively identify information from formal financial channels due to their lack of understanding of financial product terms, thereby being excluded from formal financial services [22]. Conversely, households with higher financial literacy can more proficiently master financial knowledge and convey effective signals to financial institutions, thereby increasing households’ demand for and access to formal credit [23]. Second, financial literacy helps households optimize credit costs and improve debt management ability. Consumers with low financial literacy often struggle to identify cost differences across credit products and are prone to paying higher interest or falling into debt traps [24,25]. Individuals with high financial literacy can identify credit products that match their economic capacity, such as selecting lower-interest mortgage loans or avoiding high-cost internet financial products, thereby reducing household debt costs [26]. Finally, the alleviation of credit constraints directly translates into households’ ability to pay for clean energy. Clean energy facilities often require substantial upfront acquisition costs, and for low-income or liquidity-constrained households, access to external financing is crucial. Research suggests that households with access to financial services are more likely to smooth consumption through credit support, thereby improving household clean energy consumption patterns [27,28]. Thus, the following research hypothesis is proposed:

H2. 

Financial literacy increases the probability of households using clean energy by alleviating credit constraints, lowering the barriers and costs of accessing funds.

2.2.2. Non-Agricultural Employment Mechanism

The second possible channel through which financial literacy affects households’ clean cooking fuel choice is by promoting farmers’ non-agricultural employment and entrepreneurship activities, thereby driving clean cooking fuel choice through income and conceptual effects. On the one hand, financial literacy can stimulate households’ entrepreneurial motivation and willingness to engage in non-agricultural employment by enhancing risk management capability. Existing research indicates that individuals with higher financial literacy often possess more rational risk preferences and stronger risk tolerance, which increases the probability of opportunity-driven entrepreneurship [29]. Additionally, financial literacy can enhance individuals’ sensitivity to market information and resource allocation capability, helping them more precisely identify non-agricultural employment opportunities or business projects [30]. For example, farmers with higher financial literacy can transition from being risk-averse to becoming rational risk-takers, thereby tending to seek higher returns through entrepreneurship or non-agricultural employment. On the other hand, the successful development of non-agricultural employment and entrepreneurship activities significantly improves households’ economic performance and modernization level. Financial literacy is an important component of entrepreneurs’ enterprise management capability and can significantly improve business outcomes and household income [31]. As household income increases and wealth accumulates, households’ pursuit of higher-quality living standards will prompt them to change traditional energy consumption habits [10]. Furthermore, the process of entrepreneurship and non-agricultural employment itself is a process of information acquisition and concept renewal, which facilitates the promotion of clean energy technologies and products [32]. For instance, entrepreneurship contributes to the market-based dissemination of innovative clean energy technologies, influencing their accessibility and affordability [33]. Therefore, the following research hypothesis is proposed:

H3. 

Financial literacy increases the probability of households using clean energy by promoting farmers’ non-agricultural employment and entrepreneurship, raising household income and modernized perspectives.

2.2.3. Social Trust Mechanism

The third possible channel through which financial literacy affects households’ clean cooking fuel choice is by enhancing households’ level of social trust. Clean energy equipment is, to some extent, a “trust good” or “experience good,” and farmers’ degree of trust in technology providers and the market environment directly influences their adoption decisions. First, financial literacy can enhance individuals’ understanding of market rules and the spirit of contracts, thereby improving generalized trust. The improvement in financial literacy means that individuals possess stronger financial knowledge and skills, which help them better evaluate and interpret social interactions and economic contracts [34]. When individuals can understand the logic of market operations, their defensive psychology toward others decreases, thereby raising their level of trust in economic and social actors (including policymakers and clean energy suppliers). Second, social trust can reduce perceived risks and transaction costs in clean energy trading processes. Due to information asymmetry in clean energy technology, individuals with higher generalized trust are less likely to believe they will be deceived or misled in the market [35]. This sense of trust can reduce the perceived risks associated with high upfront costs [36], making households more willing to accept recommendations from early adopters or policymakers. Finally, the promotion of clean energy often requires collective action and social capital support. Community peer demonstration effects are particularly critical in the adoption of facilities such as solar photovoltaic systems, and this collective action largely depends on trust [37,38]. Individuals with high financial literacy are more adept at leveraging social capital and enhancing collaborative capacity through trust-building, thereby removing psychological barriers to clean energy entering households. Therefore, the following research hypothesis is proposed:

H4. 

Financial literacy increases the probability of households using clean energy by enhancing households’ social trust level, reducing transaction costs and perceived risks.

3. Data Sources, Variable Selection, and Descriptive Statistics

3.1. Data Sources

The data used in the empirical analysis of this paper comes from the 2018 China Family Panel Studies (CFPS) published by the Institute of Social Science Survey of Peking University. As a nationally representative comprehensive longitudinal social survey project, CFPS covers multiple dimensions including society, economy, population, education, and health. Since the launch of its baseline survey in 2010, it has adopted a three-stage unequal probability cluster sampling method and conducted follow-up surveys every two years. The main reasons for selecting the 2018 data are as follows. First, compared with the 2016 and 2020 data, the CFPS 2018 specifically includes financial literacy and migration modules. Through three classic objective questions covering compound interest calculation, inflation perception, and risk cognition, it provides key support for quantifying micro-household financial capacity. Moreover, compared with other domestic micro databases, the CFPS 2018 data also contains detailed household energy consumption information, which can meet the analytical needs of household energy use behavior. Based on data availability and the suitability of variable settings, this paper ultimately selects this dataset for analysis. The initial dataset contained 14,218 household-level and 37,354 individual-level observations. After merging the two databases, we restricted the sample to household heads aged 16 to 85 years. We then excluded those with an urban hukou to focus on the rural population. Finally, after removing observations with missing values for key variables, the final estimation sample comprised 3960 observations distributed across 946 villages and 29 provinces. This sample size aligns with related research using the same data source; for instance, Nan et al. (2023) [39] utilized a final regression sample of 4292 observations.

3.2. Variable Definitions

3.2.1. Dependent Variable

This paper takes household clean cooking fuel choice behavior as the dependent variable. This variable is defined based on respondents’ answers to the question “What type of fuel does your household primarily use for cooking?” in the CFPS household questionnaire. Drawing on the studies by Ma et al. (2022) [40] and Zhu et al. (2023) [41], if a rural household’s primary cooking fuel is clean energy (including canned gas/liquefied petroleum gas, natural gas/piped gas, solar energy/biogas, and electricity), the household clean cooking fuel choice behavior is assigned a value of 1; if the primary fuel is non-clean energy (including straw, firewood, or coal), it is assigned a value of 0.

3.2.2. Explanatory Variable

The core explanatory variable of this paper is financial literacy. Currently, there are two main approaches to measuring this variable in academia. The first is from a subjective perspective, measuring subjective financial literacy based on respondents’ self-assessment of their financial ability or their familiarity with various financial products and services. The second is from an objective cognitive dimension, quantifying objective financial literacy based on correct answer rates on core financial literacy questions [29]. The resident financial literacy assessment system proposed by Lusardi and Mitchell (2014) [3] contains three core questions corresponding to cognitive understanding of the time value of money (compound interest calculation), inflation perception (comparison of purchasing power across different periods), and risk diversification (comparison of differences between stock and fund risks). This method has been widely adopted and applied in academia [42]. The CFPS 2018 questionnaire covers these three questions, which can be used to measure objective financial literacy. In this paper, answer results are assigned values of 0–1 (correct = 1, incorrect = 0), generating three dummy variables; drawing on the approaches of Liao et al. (2017) [43] and Wu et al. (2018) [44], factor analysis is used for dimensionality reduction to obtain a factor-based financial literacy indicator.

Table 1. Factor analysis results.

Variable	(1)	(2)	(3)
	Factor 1	Factor 2	Factor 3
Eigenvalue	1.31134	0.90150	0.78716
Proportion	0.4371	0.3005	0.2624
Cumulative	0.4371	0.7376	1.0000
Bartlett test of sphericity Chi-square (p-value)	284.799 (0.000)

reports the results of the factor analysis. Bartlett’s test of sphericity yields a p-value of 0.000 (i.e., p < 0.001), well below the 0.05 threshold, thereby confirming the suitability of the data for factor analysis. Based on the eigenvalue-greater-than-one criterion, a single factor is retained, which represents financial literacy.

Table 2. Factor loading results.

Variable	(1)	(2)
	Factor 1	Uniqueness
Compound interest calculation	0.5797	0.6640
Inflation perception	0.6715	0.5490
Risk diversification	0.7241	0.4756

presents the factor loadings, based on which we construct our financial literacy index. Meanwhile, following the study of Abreu and Mendes (2010) [45], a summative financial literacy indicator is constructed based on the total number of correct answers, with 1 point per correct answer and a maximum of 3 points. Given that this indicator does not reflect the differential importance of different financial questions, it is only used for subsequent robustness checks.

3.2.3. Control Variables

To the greatest extent possible, to exclude the influence of observable variables on the estimation results, drawing on the approaches of Lu et al. (2025) [46] and Koomson et al. (2025) [13], this paper primarily controls for the demographic variables of rural residents, household variables, and provincial fixed effects. The demographic variables of rural residents include head of household gender, head of household age, head of household age squared, head of household education level, head of household marital status, head of household hukou health status, and head of household political status. Household variables include family size, number of minors under 16 in the household and number of elderly persons over 65 in the household.

3.3. Descriptive Definitions

Table 3. Variable measures and descriptive statistics.

Variable	Variable Description	Variable Types	Mean	SD
Clean cooking fuel choice	Primary cooking fuel is electricity/natural gas/LPG/solar power = 1; Primary cooking fuel is straw/firewood/coal = 0	Binary	0.823	0.376
Financial literacy	Respondents’ cognition of compound interest, inflation, and risk. Correct answer = 1, otherwise = 0. Factor analysis used for dimensionality reduction.	Continuous	−0.153	0.540
Head gender	Male = 1, Female = 0		0.504	0.500
Head age	-	Continuous	47.69	13.56
Head age squared	Quadratic term of household head age	Continuous	2458	1345
Head education	Senior high school and above = 1, others = 0	Binary	0.227	0.419
Head marital status	Married with spouse = 1, Unmarried/widowed/divorced = 0	Binary	0.884	0.320
CPC party membership	CPC member = 1, others = 0	Binary	0.005	0.071
Self-reported health	Self-rated health: relatively healthy, very healthy, extremely healthy = 1; unhealthy, fair = 0	Binary	0.694	0.461
Family size	-	Continuous	3.947	1.831
Number of children under 16	-	Continuous	0.349	0.648
Number of seniors over 65	-	Continuous	0.232	0.489
Whether a household uses the Internet	Household uses the Internet = 1; household does not use the Internet = 0	Binary	0.216	0.411
The total amount of household financial assets	The logarithm of household financial assets	Continuous	8.265	4.339
N			3960	3960

Notes: The mean values and standard deviations of all continuous variables are reported.

reports the definitions, assignments, and descriptive statistics of the above variables. From the perspective of the dependent variable, the mean of “household clean cooking fuel choice” is 0.823, indicating that 82.3% of the sample households have adopted clean energy for cooking, demonstrating a relatively high penetration rate of clean energy in rural China. From the perspective of control variables, regarding head of household characteristics, the average age of heads of household is 47.69 years, approximately 50.4% are male, 22.7% have senior high school education or above, 88.4% have a spouse, 0.5% are CPC members, and 69.4% consider themselves to be in good health. Regarding household characteristics, the average family size is 3.947 persons.

4. Identification Strategy and Empirical Analysis

4.1. Identification Strategy

To identify the relationship between financial literacy and rural households’ clean cooking fuel choice, this paper first employs the following OLS model (linear probability model) as the benchmark model:

$${Fuel}_{ij}={\beta }_0+{\beta }_1{FL}_{ij}+{\beta }_2{X}_{ij}+{\omega }_j+{\epsilon }_{ij}$$

(1)

where $i$ and $j$ represent rural households and provinces, respectively; ${Fuel}_{ij}$ is the binary dummy variable indicating whether rural household $i$ uses modern energy, taking the value 1 if clean energy is used and 0 otherwise;${FL}_{ij}$ represents the household’s financial literacy; $X_{ij}$ represents demographic characteristics such as head of household age, head of household gender, family size, etc. (see Table 1 for details); ${\omega }_j$ represents provincial fixed effects, primarily used to control for unobservable provincial infrastructure and other factors affecting household clean cooking fuel choice; and ${\mathrm{\epsilon }}_{\mathrm{i}\mathrm{j}}$ represents the random error term.

Since household clean cooking fuel choice is a binary dummy variable, this paper further employs the following Probit model to identify the effect of financial literacy on rural households’ clean cooking fuel choice:

$${Pr(Fuel}_{ij}=1|X_{ij})=\varphi ({\alpha }_0+{\alpha }_1{FL}_{ij}+{\alpha }_2{X}_{ij}+{\pi }_j)$$

(2)

where $i$ and $j$ represent rural households and provinces, respectively; ${Fuel}_{ij}$ is the binary dummy variable indicating whether rural household i uses modern energy, taking the value 1 if clean energy is used and 0 otherwise;${FL}_{ij}$ represents the household’s financial literacy; $X_{ij}$ represents demographic characteristics such as head of household age, head of household gender, family size, etc. (see Table 1 for details); and ${\pi }_j$ represents provincial fixed effects, primarily used to control for unobservable provincial infrastructure and other factors.

4.2. Results

Table 4. Regression Results of Financial Literacy on Household Clean Cooking Fuel Choice.

Variable	(1)	(2)	(3)	(4)
	OLS		Probit
Financial literacy	0.061 *** (0.012)	0.034 *** (0.012)	0.270 *** (0.052)	0.165 *** (0.053)
Head age		−0.001 (0.004)		−0.017 (0.017)
Head age squared		−0.000 (0.000)		0.000 (0.000)
Head gender		−0.030 * (0.016)		−0.137 ** (0.070)
Head education		0.051 *** (0.016)		0.296 *** (0.096)
Head marital status		0.060 ** (0.030)		0.276 ** (0.116)
CPC party membership		−0.138 (0.122)		−0.682 (0.425)
Self-reported health		0.012 (0.017)		0.048 (0.070)
Family size		−0.006 (0.005)		−0.033 (0.022)
Number of children under 16		−0.005 (0.020)		−0.021 (0.068)
Number of seniors over 65		−0.040 ** (0.016)		−0.154 *** (0.059)
Provincial fixed effects	Yes	Yes	Yes	Yes
N	3960	3960	3938	3938
R2	0.094	0.128	—	—

Notes: Standard errors are clustered at the village level. The marginal probability effects are 0.063 (0.012) in Column (3) and 0.037 (0.012) in Column (4). The reduction in sample size from OLS (3960) to Probit (3938) is due to the perfect prediction of success by certain province dummy variables in the Probit model. *** p < 0.01, ** p < 0.05, * p < 0.1.

presents the regression analysis results of the impact of financial literacy on rural households’ clean cooking fuel choice behavior. To ensure the reliability of conclusions, this paper employs both the OLS linear regression model and the Probit probability model for comparative testing, where columns (1) and (2) are OLS model regression results, and columns (3) and (4) are Probit model regression results. All models control for provincial fixed effects, and standard errors are clustered at the village level to address heteroscedasticity. Column (1) presents the OLS benchmark regression results. Without incorporating other control variables, only regional differences are controlled through provincial fixed effects. The results show that the regression coefficient of financial literacy is 0.061 and is significantly positive at the 1% level, implying that, controlling for regional differences, each unit increase in financial literacy is positively associated with the probability of rural households using clean energy by 6.1 percentage points. This preliminarily verifies the positive effect of financial literacy on rural households’ clean cooking fuel choice.

To exclude the interference of head of household individual characteristics and household-level factors, column (2) adds a series of control variables including head of household gender, head of household age, head of household education level, head of household marital status, head of household health status, head of household political status, family size, number of family members under 16 and over 65. The results show that after adding control variables, the regression coefficient of financial literacy decreases from 0.061 to 0.034 but remains significantly positive at the 1% level, confirming that financial literacy is positively related to the probability of rural households using clean energy. The Probit model regression results in columns (3) and (4) are broadly consistent with the OLS results, verifying the robustness of the core conclusions. Hypothesis H1 is thus verified.

4.3. Robustness Checks

To ensure the reliability of the benchmark conclusions, this paper conducts robustness checks from four dimensions.

4.3.1. Alternative Measure of Financial Literacy

To verify whether the measurement of financial literacy affects the research conclusions, this paper replaces the factor-based financial literacy indicator with a sum-score financial literacy indicator [39]. Specifically, three core questions on compound interest calculation, inflation cognition, and risk cognition are scored based on the number of correct answers (1 point per correct answer, maximum 3 points). The regression results in column (1) of

Table 5. Robustness Check Results 1.

Variable	(1)	(2)	(3)	(4)	(5)
	Based on Summed Financial Literacy Indicator	Adding Control Variables	Excluding Households with Financial Investment Experience	Control Function Approach
Based on summed Financial literacy indicator	0.018 *** (0.007)	-	-	-	-
Financial literacy		0.023 * (0.012)	0.029 ** (0.012)	0.378 *** (0.094)	-
Village average financial literacy	-	-	-	-	0.800 *** (0.123)
Residual	-	-	-	−0.308 *** (0.094)	-
Control variables	Yes	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes	Yes
F-statistic	-	-	-	-	11.587 (0.001)
N	3960	3924	3814	3960	3960
R2	0.128	0.135	0.129	0.130	0.101

Notes: Standard errors are clustered at the village level. *** p < 0.01, ** p < 0.05, * p < 0.1.

show that the coefficient of the sum-score financial literacy indicator is significantly positive at the 1% level, confirming the reliability of the benchmark conclusions.

4.3.2. Incorporating Additional Control Variables

Since whether a household uses the Internet and the total amount of household financial assets may affect the estimation results of this paper, we include these two control variables here to conduct a robustness test. Column (2) of Table 5 reports the regression results after adding these variables. The results show that after incorporating additional controls, financial literacy still has a significantly positive effect on rural households’ clean cooking fuel choice at the 5% level, indicating that the core conclusions are robust.

4.3.3. Excluding Households with Financial Market Investment Experience

In real economic scenarios, some households participate in financial market investment activities. Such consumers often accumulate richer investment experience and professional knowledge through “learning by doing” in financial decision-making, and their financial literacy levels are correspondingly relatively high. Including such consumer groups in the sample may easily lead to overestimation of the policy effect of financial literacy. Based on the CFPS 2018 survey question “Does your household currently hold financial products such as stocks, funds, government bonds, trust products, or foreign exchange products?”, households holding any of the above financial products are defined as having financial market investment experience. To verify the reliability of the benchmark conclusions, this section conducts robustness checks by excluding such households from the sample. Column (3) of Table 5 reports the relevant results, showing that the regression results after excluding such samples are consistent with the benchmark model conclusions, indicating that financial literacy improvement can significantly raise the probability of household clean cooking fuel choice, and the core estimation results possess robustness and reliability.

4.3.4. Control Function Approach

Following Wooldridge (2015) [47], we employ the Control Function approach to mitigate potential endogeneity issues. The core principle of this method involves introducing the residual term from the first-stage regression directly into the original equation as an additional regressor, without substituting the original explanatory variables. Notably, the estimated coefficient is substantially larger than the baseline estimates. This difference, mirroring the IV results, can be attributed to the correction of attenuation bias arising from measurement error in the latent variable of financial literacy. As reported in Column (4) of Table 5, the coefficient on financial literacy remains significantly positive, confirming the robustness of our baseline regression results.

4.3.5. Instrumental Variable Method

In the baseline regression, the potential endogeneity of financial literacy was not addressed. Following the approach of Wang et al. [2], we employ the average financial literacy of other rural residents within the same village (excluding the subject household) as an instrumental variable for household financial literacy to conduct a robustness analysis. The validity of this IV rests on a credible exclusion restriction. In rural China, household cooking fuel choice is fundamentally constrained by local infrastructure, such as the availability of piped gas and stable electricity grids. This infrastructure is relatively uniform within a village and is determined by broader government planning, making it exogenous to individual villagers’ financial literacy. Consequently, while the average financial literacy of other villagers may influence a given household’s own financial literacy through peer effects and network spillovers, it is unlikely to directly determine that household’s specific cooking fuel choice, except through its impact on the household’s own financial literacy (which subsequently shapes their awareness of available fuels, ability to navigate subsidy programs, and willingness to invest in fuel-switching). Column (1) of

Table 6. Robustness Check Results 2.

Variable	(1)	(2)	(3)
	First-Stage	IV	IV-Probit
Financial literacy	-	0.338 *** (0.097)	1.705 *** (0.144)
Village average financial literacy	0.800 *** (0.123)	-	-
Control variables	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes
Kleibergen–Paap rk LM statistic	-	33.198 (0.000)
Kleibergen–Paap rk Wald F statistic	-	42.603 (0.000)
N	3960	3960	3938
R2	0.101	-	-

Notes: Standard errors are clustered at the village level. The marginal probability effects are 0.683 (0.208) in Column (3). These marginal effects are reported for completeness and comparability, and they are not used to quantify the effect of financial literacy. *** p < 0.01.

reports the first-stage regression results, demonstrating that the instrument possesses strong predictive power for financial literacy. We further conduct tests for weak instruments; as shown in Column (2), the Kleibergen–Paap rk Wald F statistic is 42.603, which substantially exceeds the critical value of 16.38 at the 10% maximal IV size based on the Stock-Yogo test, indicating that our instrument is not weak. Additionally, we perform an underidentification test. Column (2) shows that the Kleiberge–-Paap rk LM statistic is 33.198 (p-value = 0.000), which allows us to reject the null hypothesis of underidentification. This confirms that the model is correctly identified. Collectively, these tests confirm the validity of our instrumental variable. Column (2) and (3) of Table 6 respectively present the estimation results of IV and IV-probit. The coefficients obtained via the IV approach are substantially larger than the baseline estimates in Table 4. This amplification, which aligns with findings in the similar top-tier literature, can be attributed to the correction of attenuation bias caused by measurement error and the identification of the Local Average Treatment Effect (LATE). The significantly positive coefficients on financial literacy further corroborate the robustness of the baseline regression results. However, given the near-ceiling adoption rate, the large magnitudes observed in the linear IV and IV-Probit models may be unreliable for precise quantitative interpretation.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity by Educational Attainment

As an important proxy variable for human capital, educational level is directly related to farmers’ information processing ability and cognitive horizon. Therefore, the impact of financial literacy on rural households’ clean cooking energy adoption may exhibit heterogeneity at the head-of-household education level. To test this hypothesis, we introduce an interaction term between financial literacy and household head education into the baseline regression. The estimation results in Column (1) of

Table 7. Heterogeneity Analysis Results.

Variable	Clean Cooking Fuel Choice
	(1)	(2)
Financial literacy	0.042 *** (0.011)	0.043 *** (0.013)
Financial literacy * head education	−0.030 * (0.015)	-
Financial literacy * number of children under 16	-	−0.036 * (0.021)
Control variables	Yes	Yes
Provincial fixed effects	Yes	Yes
N	3960	3960
R2	0.129	0.129

Notes: Standard errors are clustered at the village level. *** p < 0.01, * p < 0.1.

suggest that the effect of financial literacy on clean cooking energy adoption is more pronounced for households where the head has an education level below high school; specifically, the coefficient on the interaction term is negative and statistically significant. A plausible theoretical hypothesis is that a “substitution effect” may exist between financial literacy and formal education in influencing farmers’ decisions. For households with heads having senior high school education or above, they themselves possess relatively high general human capital and strong information acquisition and cognitive abilities, enabling them to increase income to cover clean energy acquisition costs through non-agricultural employment and entrepreneurship. As a result, the marginal contribution of financial literacy is “diluted” by the high education level, making its impact relatively small. In contrast, heads of household with below-senior-high-school education have relatively limited cognitive abilities. Financial literacy can enhance their risk assessment capability, helping them break through agricultural path dependence and more precisely grasp non-agricultural employment and entrepreneurship opportunities. At the same time, financial literacy effectively lowers the barriers and costs of households accessing credit funds, alleviating borrowing constraints. Through the synergistic empowerment of these two channels, the household income of low-education farmers increases, effectively breaking through the capital constraints for clean energy acquisition. Therefore, the impact effect of financial literacy might be more significant in these households. At this stage, we present this substitution effect as a possible theoretical interpretation rather than a definitive empirical conclusion.

4.4.2. Heterogeneity by Number of Children Under 16

A higher number of children in a household implies a greater demand for time and resources dedicated to care and education. This directly reduces the effective labor supply and increases the economic burden, leading to relatively lower household disposable income. Since income levels directly influence household consumption budgets and the capacity to purchase clean energy equipment, the impact of financial literacy on rural household clean energy adoption may exhibit heterogeneity based on the number of children. To test this hypothesis, we introduce an interaction term between financial literacy and the number of children under 16 into the baseline regression. The estimation results in Column (2) of Table 7 show that the positive impact of financial literacy on clean cooking energy adoption significantly diminishes as the number of children in the household increases; specifically, the coefficient on the interaction term is negative and statistically significant. A plausible theoretical hypothesis for this attenuation effect is that households with more children face more pressing daily consumption expenditures, such as education, healthcare, and basic living costs. These rigid expenditures crowd out the limited savings and credit resources available for clean energy equipment investment. In such households, even if the head possesses higher financial literacy and can identify non-agricultural employment opportunities or access credit effectively, the marginal increase in income or financing is more likely to be prioritized for immediate living needs rather than investment in durable goods like clean energy. Furthermore, the time constraints induced by caring for multiple children limit the flexibility of household heads to engage in migrant work or entrepreneurship, thereby dampening the transmission efficiency of financial literacy through non-agricultural employment and entrepreneurship channels. Consequently, the marginal effect of financial literacy on clean energy adoption declines monotonically as the number of children in a household increases, which is consistent with the observed diminishing promotional effect on clean cooking fuel choice, though further empirical testing is needed to fully verify these specific channels.

4.5. Mechanism Analysis

4.5.1. Credit Constraint Mechanism

Financial literacy not only helps households accurately understand credit product terms and reasonably evaluate their own debt servicing capacity but also improves their credit records and trust in financial institutions, thereby effectively lowering the barriers and costs of households accessing credit funds and alleviating borrowing constraints [29]. Existing research suggests that the alleviation of borrowing constraints can promote households’ adoption of clean energy by providing funding support for household clean energy equipment acquisition, reducing upfront investment pressure and liquidity risks [40]. To test the mediating role of borrowing constraints between financial literacy and clean cooking fuel choice, drawing on the approach of Nan et al. (2023) [39], this paper selects “whether the household has outstanding bank loans” as a proxy variable for credit accessibility (Despite its measurement limitations, this variable is constructed following established practices in the literature). Having an outstanding bank loan indicates that a household has successfully navigated the borrowing threshold, thereby facing fewer liquidity constraints compared to credit-rationed households. The results in column (1) of

Table 8. Mechanism Analysis Results.

Variable	(1)	(2)	(3)	(4)
	Credit Constraint	Non-Agricultural Employment	Entrepreneurship	Trust
Financial literacy	−0.061 *** (0.014)	0.134 *** (0.016)	0.021 * (0.013)	0.421 *** (0.070)
Control variables	Yes	Yes	Yes	Yes
Provincial fixed effects	Yes	Yes	Yes	Yes
N	3401	3357	3960	3954
R2	0.132	0.193	0.059	0.048

Notes: Standard errors are clustered at the village level. *** p < 0.01, * p < 0.1.

show that financial literacy has a significantly negative impact on households’ borrowing constraints. This suggests that households with higher financial literacy face lower capital borrowing thresholds and have more unblocked financing channels. The above results indicate that financial literacy promotes households’ clean cooking fuel choice possibility indirectly by alleviating the borrowing constraints they face, i.e., borrowing constraints play a significant mediating role between the two. Thus, the results provide supporting evidence for the proposed theoretical transmission channels, and the findings are consistent with the theoretical expectations regarding the mediating role of these variables as stated in hypothesis H2.

4.5.2. Non-Agricultural Employment Mechanism

Farmers with relatively high financial literacy typically possess better information processing and risk-coping abilities, making them more likely to seize non-agricultural employment opportunities and prompting household labor to transfer from agriculture to the non-agricultural sector [48]. A large body of literature has confirmed that non-agricultural employment not only raises farmers’ income but also affects their lifestyle and energy choices, thereby promoting clean energy adoption [49,50]. To test this mechanism, this paper uses whether the head of household’s main job belongs to non-agricultural employment and entrepreneurship to measure the level of farmers’ non-agricultural employment. Among these, this paper uses whether any household member is engaged in self-employment or has established a private enterprise as an indicator of entrepreneurship behavior. The regression results in columns (2) and (3) of Table 8 show that financial literacy has a significantly positive impact on both the head of household’s non-agricultural employment level and entrepreneurship, which is statistically significant. This suggests that financial literacy promotes farmers’ transition from agriculture to non-agricultural employment and entrepreneurship, thereby improving household economic conditions and energy use preferences, ultimately raising the probability of rural households’ clean cooking fuel choice. Thus, the results provide supporting evidence for the proposed channels and are consistent with the theoretical expectations of these variables playing a mediating role, as hypothesized in H3.

4.5.3. Social Trust Mechanism

Financial literacy is not only an embodiment of human capital but also an important foundation for building social capital. Farmers with relatively high financial literacy typically possess stronger information processing ability and risk assessment capability, which can effectively reduce defensive psychology arising from information asymmetry, thereby raising their level of generalized trust in others [13]. The existing literature suggests that social trust can promote clean cooking energy adoption [13]. To test the mediating role of social trust between financial literacy and clean cooking fuel choice, this paper selects the average level of households’ trust in strangers as a proxy variable. The regression results in column (4) of Table 8 suggest that financial literacy significantly raises the trust level of rural households. This means that financial literacy exerts a significant “social capital building effect,” reducing households’ search costs and monitoring costs for obtaining clean energy services by enhancing farmers’ trust in strangers. Financial literacy can raise the probability of rural households’ clean cooking fuel choice through the mediating pathway of enhancing households’ social trust level. Thus, the empirical results provide supporting evidence for the proposed transmission channels, aligning with hypothesis H4 and the theoretical expectations that these variables play a mediating role.

5. Conclusions and Suggestions

Based on the 2018 CFPS survey data, this study finds across various model specifications that financial literacy is significantly and positively associated with rural households’ clean cooking fuel choice. Exploratory heterogeneity analysis further reveals that this positive association is more pronounced among households with lower educational attainment, and that the effect of financial literacy diminishes as the number of children in the household increases. The channel analysis provides suggestive evidence for three potential transmission channels: easing credit constraints, promoting non-agricultural employment and entrepreneurship, and enhancing social trust. These findings carry several implications. Theoretically, they underscore the potential importance of cognitive capabilities—specifically financial literacy—as an enabling factor in household energy decisions, complementing the traditional focus on income and infrastructure constraints. The heterogeneous patterns, while exploratory, also suggest that financial literacy may function as a partial substitute for formal education, a proposition that warrants further investigation in other contexts. Methodologically, the study illustrates the value of using micro-level survey data combined with instrumental variable approaches to better understand the drivers of household clean cooking fuel choice, while also acknowledging the limitations of the current identification strategy and the need for stronger empirical designs in future research.

Nevertheless, this study has several limitations. First, this study relies on a cross-sectional dataset from 2018, which precludes a definitive assessment of the temporal sequencing of the mechanisms. We hope that future studies will leverage updated panel data to further investigate this relationship. Second, the instrumental variable strategy relies on village-average financial literacy, which may not fully satisfy the exclusion restriction. Third, the study focuses on a single-country context (China), and the generalizability of the findings to other settings warrants caution. Future research using panel data, quasi-experimental designs, or experimental interventions would be valuable to further validate these findings. Comparative studies across countries at different development stages would also help clarify the contextual boundaries of the relationship between financial literacy and clean energy adoption.

Before presenting the policy implications, it is important to qualify the evidentiary basis of our recommendations. Our empirical strategy, while employing instrumental variable estimation to address endogeneity, relies on cross-sectional data from a single survey wave. The IV approach addresses certain sources of endogeneity but may not fully eliminate all threats to causal identification. Moreover, the mediation analyses are associational rather than experimental. We therefore frame the following policy recommendations as suggestive implications derived from our empirical findings, rather than as definitive policy prescriptions. Where the evidence is more robust (e.g., the core association and the heterogeneity patterns), we express recommendations with greater confidence; where the evidence is more tentative (e.g., specific mechanism pathways), we express recommendations with appropriate caution. Readers should interpret these recommendations as evidence-informed directions for policy consideration, not as proven interventions.

Based on the aforementioned research findings, the following three policy implications are proposed to leverage the positive role of financial literacy, promote the popularization of rural clean energy, and support rural revitalization and ecological environmental protection.

First, strengthen the popularization of financial knowledge in rural areas, with a specific focus on households with lower education levels and specific family structures. The empirical results suggest a significant positive correlation between financial literacy and clean energy usage, with the marginal effect being more pronounced for households with lower educational attainment and fewer children under 16 years of age. Therefore, government departments could consider incorporating financial literacy training into the rural public service system and conducting targeted financial knowledge dissemination activities. Enhancing farmers’ information processing and intertemporal decision-making capabilities contributes to partially compensating for cognitive limitations caused by a lack of formal education, guiding them to rationally evaluate the long-term returns of clean energy, which may facilitate the diffusion and adoption of clean energy technologies in rural areas.

Second, improve the rural financial credit service system to alleviate household borrowing constraints. The mechanism analysis reveals that alleviating borrowing constraints is a crucial pathway through which financial literacy drives clean cooking fuel choice. Therefore, policymakers could consider deepening inclusive financial reforms and innovating green credit products tailored for clean cooking fuel choice. By lowering financing thresholds and optimizing credit approval processes to enhance farmers’ financial accessibility, these measures may help alleviate the funding barriers faced by farmers during the energy transition, enabling financial literacy to be more effectively translated into actual clean energy consumption capabilities.

Third, expand the diversified pathways for financial literacy to empower farmers’ livelihood development and stimulate endogenous dynamics. The study finds that promoting non-agricultural employment and entrepreneurship, as well as enhancing social trust, are key mechanisms through which financial literacy operates. Consequently, relevant policies could consider integrating financial education components into vocational training, entrepreneurship support, and rural social governance. Strengthening farmers’ non-agricultural skills and entrepreneurial willingness contributes to improving household economic status; meanwhile, enhancing farmers’ understanding of and trust in social rules may create a more favorable social environment for the adoption of clean energy, thereby achieving the synergistic development of improved household livelihoods and upgraded energy consumption structures.