How Can Vocational Education Drive Poverty Alleviation and Career Growth in Rural China? Evidence from Low-Income Areas

Impact of Vocational Education on Poverty Alleviation

The Uniqueness of Vocational Education in China

In China, vocational education exhibits strong regional disparities. Eastern coastal areas enjoy better resources and coverage, while rural communities in central and western regions face significant disadvantages. Moreover, curricula often fail to match labor market demand, undermining graduates’ competitiveness (Wang et al., 2023). Human capital theory underscores the importance of aligning education with employment needs (Wu & Fang, 2024). Evidence shows that well-matched programs increase employment and income, whereas mismatched training leads to inefficiency (Forsler et al., 2024). In rural China, challenges remain in identifying labor market needs, designing flexible curricula, and fostering school–enterprise collaboration (Lv et al., 2022; Zhang et al., 2024).

Literature Gaps and Purpose of the Study

Despite broad recognition of the potential of vocational education, three key gaps remain. First, most studies rely on data from developed or urban regions, overlooking rural and low-income communities. Second, prior research focuses mainly on short-term economic outcomes, with little exploration of long-term social and cultural effects. Third, heterogeneity across gender, region, and curriculum match has not been systematically examined.

Building on these gaps, this study investigates the role of vocational education in alleviating poverty in rural China. Specifically, it examines impacts on employment and income, explores differences across subgroups, and analyzes the mechanisms through which vocational education contributes to poverty alleviation. The findings aim to enrich academic understanding, inform policy optimization, and provide lessons for other developing countries.

Data Sources and Sample Representativeness

The data for this study come from the China Rural Vocational Education Survey (CRVED), two rounds of national longitudinal surveys conducted by the authors’ team between 2021 and 2024. A total of 10,000 questionnaires were initially distributed. After excluding 323 participants with missing baseline key variables and 693 participants who did not participate in the endline survey or had missing endline data, 8,984 valid samples were included in the final analysis, with an overall effective response rate of 89.84%. This response rate is consistent with the standards of large-scale longitudinal surveys in rural China and ensures sufficient statistical power for subsequent analyses.

Inclusion criteria were:

aged 16–25 years;

Exclusion criteria were:

missing key baseline variables;

During the two-wave follow-up, 1,016 participants were lost, with a sample attrition rate of 10.16%. The main reasons for attrition included out-migration for work, refusal to participate again, and invalid questionnaires. Balance tests confirmed no significant differences in gender, age, parental education, or family economic status between lost and retained samples (p > 0.05), suggesting no serious selection bias and good sample representativeness. Detailed results of the balance tests (including standardized mean differences for all covariates) are available upon request to enhance replicability. Regarding missing data, we first examined the missing pattern of all core variables (employment, income, social capital, and family background) and found that the overall missing rate was 2.3%, which is below the 5% threshold for negligible missingness. Given the low missing rate and random missing pattern, listwise deletion was employed to handle incomplete responses, which is a robust and appropriate strategy for small missing rates in longitudinal surveys. The data cover six provinces in central and western China—Henan, Anhui, Hunan, Sichuan, Guizhou, and Gansu—and include 30 representative national-level poverty-designated counties. Sample stratification criteria considered the level of regional socioeconomic development (central vs. western), the gender ratio (male vs. female), and the type of school (vocational school vs. general high school). Based on population statistics from the six provinces, proportional allocation was applied to ensure that the sampling ratios reflected the underlying population structure. For example, provinces in the western region contributed a larger share of the sample in line with their rural population size, and the male–female ratio in the sample (51:49) closely matched census distributions. Similarly, the proportions of students from vocational schools and general high schools were determined to mirror their actual enrollment ratios in the target counties. In addition, a multistage random sampling method was employed to ensure the representativeness of the sample. To further clarify the study design, this research adopts a two-wave longitudinal design, with baseline data collected in 2021 (Time 1) and endline data collected in 2024 (Time 2), which allows for the examination of changes in employment, income, and social capital over time and avoids the ambiguity of cross-sectional analyses. (Figure 1)

OPEN IN VIEWER

Figure 1.

Stratified sampling flowchart.

The design of the questionnaire focused on the mechanisms of poverty alleviation and heterogeneity analysis, covering the following key variables. The dependent variables included the employment rate (whether full-time employment was achieved), income level (measured as the natural logarithm of monthly income), and social capital score. Social capital was measured as a composite index based on community participation and social trust, adapted primarily from the World Bank Social Capital Assessment Tool (Theduy, n.d.) and widely used social capital measures (Kang & Mok, 2022; Putnam, 2000). The measure included six items in total: three items captured the frequency of participation in village-level collective activities (e.g., “In the past 12 months, have you taken part in community meetings or cooperative events?”), and three items assessed generalized trust toward others in the community (e.g., “Most people in this community can be trusted”). All items were rated on a five-point Likert scale (1 = strongly disagree, 5 = strongly agree), standardized, and then rescaled to a 0–10 index to ensure comparability with other dependent variables. To assess the measurement quality of the social capital scale, we examined its internal consistency reliability using Cronbach's α coefficient, a widely accepted indicator of scale reliability in social science research. The six-item composite demonstrated good internal consistency (Cronbach’ s α = 0.81), and an exploratory factor analysis confirmed the uni-dimensionality of the construct, supporting the meaningful combination of community participation and social trust into a single latent indicator of social capital. Specifically, the Cronbach's α coefficient of 0.81 exceeds the 0.70 threshold for acceptable internal consistency, confirming that the six items reliably measure the same latent construct of social capital. The results of the exploratory factor analysis (EFA) further supported the construct validity of the scale, with all six items loading on a single factor (factor loading range: 0.68–0.79) and explaining 67.3% of the total variance. The key independent variable was vocational education participation, defined as whether respondents were enrolled in formal secondary vocational schools or certified technical training programs providing skill-based curricula. The comparison group consisted of students enrolled in general high schools without vocational training. Control variables included gender, age, parental education, and family socioeconomic status (SES). SES was measured using a single self-reported item on household asset status (e.g., “How would you rate your household's economic condition relative to others in your village?”) on a five-point scale. Although a single-item indicator, this measure has been commonly employed in rural studies of China and shows good face validity, with strong correlations to objective indicators of household wealth.

To ensure data quality, strict quality control measures, including questionnaire pretesting, enumerator training, double data entry and outlier removal, were implemented during the survey process.

Propensity Score Matching (PSM) and Sample Balance

To control for sample selection bias, this study employed the propensity score matching (PSM) method to match the vocational education participation group (treatment group) with the nonparticipation group (control group). Specifically, students’ propensity scores for choosing vocational education were estimated via a logit model, with independent variables including gender, age, parents’ education level, and family economic status.

Quasi-Experimental Methodology and Group Construction

The data for this study are well represented, encompassing vocational schools and general high schools across 30 impoverished counties and involving approximately 10,000 students. The data collected include information on students’ gender, age, educational choices, family income levels, parents’ education levels, asset status, and details about the type of program, teaching quality, and employment support provided by the schools.

We used a quasi-experimental design to divide the students into a treatment group and a control group. First, we excluded samples with missing baseline key variables (e.g., employment status, income level, family economic status) from the raw data, resulting in the exclusion of 323 students. Second, at the end-line stage, we excluded students with missing key variables and those who did not participate in the final survey, leading to the exclusion of an additional 693 samples (478 + 215). We ultimately obtained a total of 8,984 valid samples for the subsequent analyses, with 5,031 students in the treatment group and 3,953 in the control group. To reduce selection bias, we used the propensity score matching (PSM) method to further divide the treatment group into a ‘high match group’ (2,616 students) and a ‘low match group’ (2,415 students). (Figure 2)

OPEN IN VIEWER

Figure 2.

Sample screening and grouping flowchart.

While this study employed a quasi-experimental design with a non-equivalent control group, we acknowledge that vocational school students and general high school students may follow distinct career trajectories. Vocational education students typically enter the labor market earlier, while general high school students may delay labor market entry due to higher education aspirations. Our comparison is therefore intended to capture the short-term employment and income effects of vocational education, rather than long-term career returns. To mitigate potential bias, the propensity score matching (PSM) process balanced observed baseline characteristics such as gender, parental education, and family economic status. Age was included as one of the covariates, but it was not treated as the sole proxy for career stage. In this way, the analysis focuses on the relative advantages of vocational education at comparable ages in the early career stage, consistent with previous empirical studies in developing country contexts.

Data Analysis Methods

The data analysis methods used in this study include the following. First, descriptive statistical analysis was used to compare the mean differences between the treatment and control groups in terms of key indicators such as the employment rate and income level. Second, multiple regression analysis was conducted to estimate the predictive relationship of vocational education on poverty alleviation by constructing a model that controls for confounding variables such as gender, age, region, and family economic status, thereby improving the reliability of causal inference. Given the nested structure of the data with students clustered within schools and regions, cluster-robust standard errors were used in all regression models to account for intra-cluster correlation and ensure valid statistical inference. Third, heterogeneity analyses were performed to explore differences in the effects of vocational education across gender, region, and course-matching groups by introducing interaction terms, with a particular focus on assessing the moderating effect of aligning course design with market demand on poverty alleviation outcomes. To examine the moderating effects, we adopted a hierarchical regression approach: Model 1 included control variables and the treatment indicator, Model 2 added the moderator variable, and Model 3 further introduced the interaction term. The incremental variance explained (ΔR²) between models was used to evaluate the significance of the moderating effects. Fourth, mechanistic analyses were conducted through mediation analysis to examine how vocational education influences the employment rate and income level via mechanisms such as skill upgrading, social capital enhancement, and employment opportunity expansion. Regarding temporal ordering, vocational education at baseline (Time 1) was used to predict mediators, with outcomes measured at follow-up (Time 2); thus, the identified mechanisms represent statistical mediation rather than strict causal mediation given some concurrent mediator measurement. Direct, indirect, and total effects were estimated, and the significance of indirect effects was assessed using bootstrap confidence intervals with 5,000 resamples. Finally, all analyses were performed using Stata 16.0, the significance level was set at 0.05, and robustness tests were carried out to verify the reliability of the model results.

Definition of Variables and Matching Methods

Definition and Description of the Main Variables

The core dependent variables of this study include the employment status (measured by whether monthly income reaches the standard of basic livelihood security), income level (log-transformed to reduce skewness in the distribution), and social capital score (a continuous variable ranging from 0 to 10 based on community participation and social trust). The key independent variable is vocational education participation status, which categorizes students into a treatment group and a control group. The control variables include individual characteristics such as gender, age, and household economic background (e.g., family economic status and parents’ education), which are used to account for confounding factors that could influence the dependent mechanisms. Additionally, the mechanism variables include skill enhancement (measured by the acquisition of vocational qualification certificates) and employment opportunity expansion (assessed by participation in school-enterprise cooperation programs). Variable specifications follow standardized indicators provided in the questionnaire. (Table 1)

OPEN IN VIEWER

Table 1.

Core Variable Definitions and Measurement Methods.

Variable	Name	Define	Measurement Methods	Range of Values
Dependent variables	Employment rate	Whether students achieve full-time employment after vocational education.	Derived from the question, “Do you currently have a full-time job?” Recorded as a binary variable: 1 = Yes, 0 = No.	Binary variable (0 or 1)
Dependent variables	Income level	Students’ current monthly income.	Recorded in RMB and log-transformed to minimize distributional bias.	Continuous variables
Dependent variables	Social capital score	Reflects an individual's level of social participation and social trust.	Calculated as a weighted average of responses to questions on “social participation” and “social trust.” Scores range from 0 to 10.	Continuous variable (0–10)
Independent variable	Vocational education participation status	Whether students are involved in vocational education.	Participation was scored as 1 and nonparticipation as 0, based on whether the student was enrolled in a vocational school and had completed at least one skills training module.	Binary variable (0 or 1)
Control variables	Gender	Gender of the student.	Recorded as a binary variable: Male = 1, Female = 0.	Binary variable (0 or 1)
Control variables	Age	Gender of the student.	Recorded as the students’ current age.	Continuous variables
Control variables	Parental education	Educational level of the student's parents.	Categorized as: “Elementary school and below” “Middle school” “High school and above” Dummy variables were created for each category.	categorical variable
Control variables	Family economic status	Asset status of the student's family.	Calculated using standardized responses to the questionnaire item on “main assets of the family.”	Continuous variables (positive)

The selection of variables for this study is grounded in the theoretical framework of vocational education for poverty alleviation. Employment and income serve as direct indicators of poverty reduction, representing the core outcomes affected by vocational education (Wen et al., 2021). Social capital score serves as an indirect indicator, reflecting how vocational education can strengthen social participation and trust to improve individual economic and social standing (Fischer, 2022). Control variables were included to account for confounding influences: gender and age shape both program participation and labor market outcomes, while family economic status and parental education capture background-related disparities in educational selection.

Propensity Score Matching (PSM) Method

To mitigate selection bias associated with nonrandom vocational education participation, this study applied propensity score matching. Propensity scores were first estimated using a logit model that included gender, age, family economic status, and parental education. These baseline factors significantly predicted program enrollment. Nearest neighbor matching was then implemented with a caliper of 0.05. Post-matching balance checks confirmed that standardized mean differences for all covariates were sufficiently small, indicating effective balancing between the treatment and control groups.

In summary, this study employed propensity score matching (PSM) to effectively address sample selection bias between the treatment and control groups. As shown in Table 2, the logit regression model indicates that key individual and family characteristics significantly shape students’ vocational education enrollment decisions, providing a sound basis for propensity score estimation. As presented in Table 3, post-matching balance diagnostics confirm that covariate imbalance was substantially reduced, supporting the credibility of the matched sample. Post-matching comparisons further demonstrate that vocational education is associated with a significantly higher employment rate (see Table 4), underscoring its potential to improve labor market outcomes for rural youth. These findings validate the utility of PSM in reducing selection bias and establish a reliable foundation for subsequent analyses of causal effects, poverty alleviation mechanisms, and heterogeneous impacts. Further analyses using the matched sample will examine the specific effects of vocational education on employment, income, and social capital, as well as variations across key subgroups.

OPEN IN VIEWER

Table 2.

Logit Model Predicting Students’ Propensity Scores to Choose Vocational Education.

Variable	Coefficient(β)	Standard Error(SE)
Gender(Male = 1)	0.154	0.043
Age (Years)	0.087	0.021
Parental education(Middle school)	0.272	0.058
Family economic status(Index)	0.185	0.045

OPEN IN VIEWER

Table 3.

Standardized Mean Differences (SMD) in Baseline Characteristics of Samples Before and After Matching.

Variable	Before Matching (SMD)	After Matching (SMD)
Gender (Male = 1)	0.15	0.03
Age (Years)	0.12	0.04
Parental education (Middle school)	0.21	0.05
Family economic status (Index)	0.25	0.06

OPEN IN VIEWER

Table 4.

Comparison of Employment Rates in the Matched Treatment and Control Groups.

Employment Rate	Treatment Group(%)	Control Group(%)
Employed	75.8	63.4
Unemployed	24.2	36.6

Descriptive Analysis

Figure 3 presents standardized comparisons of the treatment and control groups across the three outcome variables. The results indicate that, after standardizing all indicators into z-scores, vocational education students exhibited higher standardized values in employment rate, income, and social capital compared with the control group. Although the magnitudes differ, the overall pattern suggests that vocational education is associated with consistently better outcomes across multiple dimensions.

OPEN IN VIEWER

Figure 3.

Standardized comparisons of vocational education effects on employment, income, and social capital. Notes: All outcomes were transformed into z-scores prior to plotting; the y-axis denotes standardized values (standard deviation units).

Table 6 presents balance checks between the treatment and control groups on baseline characteristics. Most covariates, including gender and age, were similar across groups. However, significant differences emerged in parental education and family economic status, suggesting that family background may shape selection into vocational education. Further analyses integrating evidence from Table 5 and Figure 3 indicate that the vocational education group performed better across all key outcomes: employment rate, income level, and social capital.

OPEN IN VIEWER

Table 5.

Data Statistics of Treatment and Control Groups in Terms of Dependent Variables.

Variable	Treatment Group Mean（SD）	Control Group Mean（SD）	P-Value
Employment rate (%)	75.8%(0.43)	63.4%	0.000***
Income level (RMB)	3580(1250)	2980(1100)	0.000***
Social capital score	7.2(1.9)	6.3(2.1)	0.000***

p-value indicates statistical significance at the 0.1% level (p < 0.001).

OPEN IN VIEWER

Table 6.

Baseline Characteristics of Treatment and Control Groups.

Variable	Treatment Group Mean (SD)	Control Group Mean（SD）	P-Value
Gender (Male = 1)	54.8% (0.49)	53.6% (0.5)	0.120
Age (Years)	18.5 (1.8)	18.4 (1.7)	0.145
Parental education (Middle school)	38.6% (0.49)	34.2% (0.47)	0.015*
Family economic status (Index)	2.9 (1.2)	2.6 (1.3)	0.002**

Note: *p < 0.05; **p < 0.01.

Vocational education was associated with a substantially and significantly higher full-time employment rate, supporting its role in strengthening labor market participation among rural disadvantaged students. Regarding economic outcomes, the treatment group also achieved markedly higher monthly income levels, confirming the direct poverty-alleviating benefits of skill-focused training. In addition, participants in vocational education displayed significantly higher social capital, suggesting that such programs can indirectly support vulnerable groups by strengthening social engagement and trust, thereby reinforcing broader poverty reduction effects.

Regression Analysis and Exploration of Heterogeneity

This study employs hierarchical regression models by sequentially adding variables to examine the impact of vocational education on poverty alleviation outcomes. Model 1 includes only control variables and treatment indicators, Model 2 adds the moderator variables, and Model 3 further introduces interaction terms to explore the moderating effects. This hierarchical approach allows us to analyze the incremental impact of adding each variable in the model.

Regression Analysis

To test the core research questions, we compared the effects of vocational education against general high school education on employment, income, and social capital using hierarchical regression (see Table 7). Results are presented in three panels corresponding to each outcome variable, with consistent model specification across all analyses.

OPEN IN VIEWER

Table 7.

Hierarchical Regression Results of Vocational Education on Key Outcomes.

Panel A. Employment Rate
Variables	Model 1 (Reduced)	Model 2 (+ Moderator)	Model 3 (+ Interaction)
Treatment (Vocational = 1)	0.34**	0.31**	0.29**
Moderator (Family asset status)	/	0.18*	0.15*
Treatment × Moderator	/	/	0.21**
Gender (Male = 1)	0.12*	0.11*	0.10*
Age	0.04	0.04	0.04
Parental education	0.07	0.06	0.06
Region (Urban = 1)	0.22**	0.21**	0.21**
Constant	1.85**	1.77**	1.70**
R2	0.24	0.26	0.29
ΔR2	/	0.02*	0.03*
N	8984	8984	8984

Panel B. Income Level
Variables	Model 1 (Reduced)	Model 2 (+ Moderator)	Model 3 (+ Interaction)
Treatment (Vocational = 1)	0.28**	0.25**	0.22**
Moderator (Course-market match)	/	0.19*	0.17*
Treatment × Moderator	/	/	0.20*
Gender (Male = 1)	0.10	0.09	0.09
Age	0.05	0.05	0.05
Family asset status	0.12*	0.11**	0.11*
Parental education	0.08	0.07	0.07
Region (Urban = 1)	0.19**	0.18**	0.18**
Constant	2.10**	2.00**	1.95**
R2	0.27	0.29	0.32
ΔR2	/	0.02*	0.03*
N	8984	8984	8984

Panel C. Social Capital
Variables	Model 1 (Reduced)	Model 2 (+ Moderator)	Model 3 (+ Interaction)
Treatment (Vocational = 1)	0.22**	0.20**	0.18**
Moderator (Community trust)	/	0.16*	0.14*
Treatment × Moderator	/	/	0.19*
Gender (Male = 1)	0.08	0.08	0.07
Age	0.03	0.03	0.03
Family asset status	0.10*	0.09	0.09
Parental education	0.06	0.05	0.05
Region (Urban = 1)	0.15**	0.15**	0.15**
Constant	1.65**	1.58**	1.55**
R2	0.21	0.23	0.26
ΔR2	/	0.02*	0.03*
N	8984	8984	8984

Notes: *p < 0.05, **p < 0.01. Control variables include gender, age, family asset status, parental education, and region.

Panel A (Employment Rate)

Vocational education showed a significant positive association with employment in the baseline model. Introducing family asset status as a moderator improved overall model fit, and the interaction term was also statistically significant. These results indicate that family economic background shapes the strength of employment gains associated with vocational education.

Panel B (Income Level)

Vocational education similarly predicted higher income levels in the baseline model. Adding the course-market matching moderator increased explanatory power, and the corresponding interaction term remained significant. This pattern implies that greater alignment between vocational curricula and labor market demand strengthens the income-improving effects of skill training.

Panel C (Social Capital)

A comparable pattern emerged for social capital: vocational education was significantly and positively associated with social capital in the baseline model. Including community trust as a moderator enhanced model fit, and the interaction between vocational education and community trust was significant. This suggests that vocational education operates more effectively to improve social capital in contexts with higher levels of community trust.

Overall, results indicate that vocational education exerts meaningful positive effects on employment, income, and social capital. Furthermore, hierarchical regression models confirm that these effects are moderated by family economic background, course-market alignment, and community trust. The consistent improvements in model explanatory power further support the robustness of these moderating roles.

Heterogeneity Analysis

In line with the third research aim, we examined whether the effects of vocational education on poverty alleviation varied across gender and regions. This section first analyzes gender heterogeneity by focusing on female employment, income, and social capital outcomes, and then compares the differential impacts of vocational education between the central and western regions of China.

Gender Heterogeneity Analysis

The term ‘gender heterogeneity’ refers to differences in how vocational education impacts employment, income, and social capital outcomes between male and female students. This analysis focuses on how gender influences the effects of vocational education, not on variance heterogeneity.

Table 8 indicates that participation in vocational education has a more significant effect on the employment rate and income level of females, resulting in a 12.4% increase in the employment rate and a 10.2% increase in income. This finding suggests that allocating more resources to vocational education for females, particularly in the central and western regions, could lead to greater poverty alleviation outcomes.

OPEN IN VIEWER

Table 8.

Impact of Vocational Education Treatment Groups on Female Work Rates, Income Levels, and Social Capital.

Variable	Coefficient(β)	Standard Error(SE)	P-Value
Employment rate (%)	0.124	0.015	p < 0.001*
Logarithm of income level	0.102	0.022	p = 0.001

Notes: Male is the reference group, coded as 0. The data in Table 8 reports the female work rate, and male data is implicit as the baseline.

*p < 0.01.

Regional Heterogeneity Analysis

We then analyzed regional heterogeneity by comparing the central and western regions, in order to assess whether differences in resource allocation and institutional support shape the poverty alleviation effects of vocational education.

The results in Table 9 reveal a significant difference in the poverty alleviation effect of vocational education between the central and western regions. In the western region, vocational education led to a 24.5% increase in the employment rate and an 18.6% increase in income, highlighting that the poverty alleviation effect of vocational education is more pronounced in resource-poor areas.

OPEN IN VIEWER

Table 9.

Heterogeneity Analysis of the Impact of Vocational Education Treatment Groups on Work Rates and Income Levels in the Central and Western Regions.

Variable	Central Region Coefficient(β)	Western Region Coefficient(β)	Standard Error(SE)	P-Value
Employment rate (%)	0.178	0.245	0.245	p < 0.001*
Logarithm of income level	0.152	0.186	0.186	p < 0.001*

Note: *p < 0.01.

Figure 4 shows that, in terms of course matching, the employment rate of the high match group in the western region is significantly higher than that in the central region, with an increase of more than 28%. Conversely, Figure 5 shows that the vocational certificate acquisition rate among students participating in vocational education is 7% lower in the western region than in the central region. This suggests a particular need to enhance vocational certificate training in the western region.

OPEN IN VIEWER

Figure 4.

Moderating effect of the vocational course match on work rates in the central and western regions.

OPEN IN VIEWER

Figure 5.

Vocational certificate acquisition rates for central and western regions among vocational education participants.

Through regression and heterogeneity analyses, vocational education is shown to have a significant poverty alleviation effect in both the central and western regions, particularly in the western region. However, female participation and course matching stand out as critical factors, highlighting the importance of strengthening vocational certificate training for females and increasing investment in course match programs.

Building on these heterogeneous effects, we further investigated the mechanisms through which vocational education influences employment, income, and social capital.

Mediation Analysis

To further clarify the mechanisms through which vocational education exerts its poverty alleviation effects, mediation analyses were conducted, with skill upgrading, social capital enhancement, and employment opportunity expansion as potential mediating variables. Direct, indirect, and total effects for employment and income outcomes are reported in Table 10.

OPEN IN VIEWER

Table 10.

Mediation Analysis Results of Vocational Education on Employment and Income.

Panel A. Employment Rate
Effect Type	Estimate (β)	95% CI
Total Effect	0.36***	[0.25, 0.47]
Direct Effect	0.14*	[0.03, 0.25]
Indirect Effect	0.22***	[0.13, 0.32]
Skill upgrading	0.11***	[0.05, 0.18]
Social capital	0.07**	[0.02, 0.12]
Employment opportunity	0.04*	[0.01, 0.08]

Panel B. Income Level
Effect Type	Estimate (β)	95% CI
Total Effect	0.42***	[0.31, 0.53]
Direct Effect	0.13*	[0.02, 0.24]
Indirect Effect	0.29***	[0.18, 0.40]
Skill upgrading	0.15***	[0.08, 0.23]
Social capital	0.09**	[0.04, 0.16]
Employment opportunity	0.05*	[0.01, 0.10]

Notes: Bootstrap resampling with 5,000 iterations; *p < 0.05, **p < 0.01, ***p < 0.001.

Robustness Test

To examine the stability of the results obtained through the propensity score matching (PSM) method, this study employed alternative matching methods, including radius matching and kernel matching, as substitutes for nearest neighbor matching (NNM). Additionally, the impact of vocational education participation on employment rates and income levels was re-estimated. To further validate the robustness of the model, the study also reassessed the effect of vocational education on poverty alleviation by using the growth rate of household income as an alternative dependent variable.

Table 11 and Figure 6 demonstrate that the estimation results from the different matching methods are highly consistent numerically and all indicate a significant positive effect of vocational education participation on employment rates and income levels, highlighting the strong robustness of the results. Table 12 and Figure 7 illustrate that vocational education participation significantly increases the growth rate of household income (12.8%), which aligns with the results of the analysis of the logarithm of income, further validating the positive effect of vocational education on household economic improvement.

OPEN IN VIEWER

Figure 6.

Comparison of regression coefficients for employment rates and income levels under different matching methods.

OPEN IN VIEWER

Figure 7.

Comparison of regression coefficients using the logarithm of income and the income growth rate as dependent variables.

OPEN IN VIEWER

Table 11.

Impact of Vocational Education Participation on Employment Rates and Income Levels Under Different Matching Methods.

Variable	Nearest Neighbor Matching Coefficient(β)	Radius Matching Coefficient(β)	Kernel Matching Coefficient(β)	Standard Error(SE)	P-Value
Employment rate (%)	0.212	0.205	0.218	0.035	p < 0.001*
Logarithm of income level	0.152	0.145	0.150	0.025	p < 0.001*

Note: *p < 0.01.

OPEN IN VIEWER

Table 12.

Robustness Test Results with Household Income Growth Rate as Dependent Variable.

Variable	Vocational Education Participation Coefficient (β)	Standard Error (SE)	P-Value
Growth rate of household income (%)	0.128	0.018	p < 0.001*

Using the above analytical framework and empirical results, this study confirms the significant role of vocational education in improving employment rates, income levels, and social capital in low-income rural communities while also revealing its heterogeneous characteristics and robustness. These analytical results provide strong empirical support for optimizing vocational education policies and achieving the goal of precise poverty alleviation.

Vocational Education Significantly Enhances Employment Rates and Income Levels

The study's data showed that vocational education significantly increased the employment rate and income levels of individuals living in low-income rural communities. After matching the sample via propensity scores, the employment rate of students who participated in vocational education reached 75.8%, which was significantly higher than the 63.4% employment rate of the control group that did not participate in vocational education (p < 0.001). In terms of income, the average monthly income of the vocational education participants was RMB 3,580, whereas the average monthly income of the nonparticipants was RMB 2,980, with the difference being statistically significant (p < 0.001). These findings suggest that vocational education provides substantial economic opportunities for rural low-income populations by directly enhancing individual skills and employment ability. This finding is consistent with those of previous studies (Lee et al., 2021). Additionally, vocational education participants had a significantly higher rate of acquiring vocational qualifications than nonparticipants did, further strengthening their competitiveness in the labor market.

Heterogeneous Effects of Vocational Education Across Groups

Vocational education has significant heterogeneity in its poverty alleviation effects on the basis of gender, region, and course match with labor market demands. This finding is consistent with the study by Tadesse et al. (2022). In terms of gender differences, female vocational education participants experienced a more significant increase in employment rates, which rose by 12.4% (p < 0.001) compared with females who did not participate in vocational education. This result suggests that vocational education has a stronger effect on improving the economic status of women (Chongcharoentanawat et al., 2022). Compared with those in the central region, students in the western region demonstrated significantly greater increases in employment rates and income levels, highlighting the greater potential of vocational education to alleviate poverty in resource-poor areas. However, despite the central region's higher starting point, the increase in employment rates and income resulting from vocational education remains significant. This regional disparity suggests that the effectiveness of vocational education is heavily influenced by the local level of economic development and the availability of educational resources (Heinesen & Lange, 2024). The analysis of course match further emphasizes the importance of a strong match between course content and labor market demand. The data show that the employment rate for the high match group is as high as 82.3%, whereas it is only 68.5% for the low match group. This finding underscores the critical role of optimizing course design to better align with market needs, thereby enhancing the overall effectiveness of vocational education (Dalby & Noyes, 2022).

Analysis of the Poverty Alleviation Mechanism of Vocational Education

This study further examines the specific mechanisms through which vocational education contributes to poverty alleviation. The results indicate that vocational education achieves positive poverty alleviation effects primarily through three pathways: skill enhancement, employment opportunity expansion, and social capital enhancement. In terms of skill enhancement, 64.7% of the students who participated in vocational education obtained vocational qualification certificates, whereas only 21.5% of the nonparticipants did (p < 0.001). This significant difference demonstrates that vocational education empowers participants by providing systematic skills training, thereby improving their employment ability. The expansion of employment opportunities is another critical pathway through which vocational education alleviates poverty. The findings reveal that students who participated in school-enterprise cooperation programs experienced an average monthly income growth rate of 18.6%, which was significantly greater than the income growth of those who did not participate in such programs. This result suggests that school-enterprise cooperation not only enhances students’ employment opportunities but also equips them with more competitive vocational skills through practical, hands-on training.

In addition, vocational education indirectly contributes to poverty alleviation by enhancing students’ community involvement and social trust. The data revealed that the social capital score of the vocational education participants was 7.2, which was significantly higher than the score of the nonparticipants, which was 6.3 (p < 0.001). This finding suggests that vocational education not only helps economically disadvantaged groups escape poverty but also supports the overall social development of poor areas by promoting community participation and fostering social cohesion (Putnam, 2000; Zhou et al., 2022).

Theoretical and Practical Implications of the Poverty Alleviation Effect of Vocational Education

This study confirms the significant role of vocational education in promoting employment, increasing income, and enhancing social capital, particularly in low-income rural communities. These findings align with the theoretical perspectives in the literature (Yi et al., 2023), which suggest that vocational education contributes to economic improvement by directly upgrading skill levels (Wang et al., 2024). Moreover, this study provides targeted empirical evidence for central and western rural low-income areas, revealing the multiple mechanisms through which vocational education alleviates poverty, including skill enhancement, employment opportunity expansion, and social capital enhancement (Luimes, 2023; Tian et al., 2023). Furthermore, the results demonstrate that vocational education not only significantly improves the economic status of participants in the short term but also indirectly fosters social cohesion in impoverished areas by promoting community participation and strengthening social trust. These findings offer an important theoretical foundation and practical insights for policymakers aiming to optimize vocational education policies and achieve the goal of precise poverty alleviation.

Implications of Heterogeneous Effects for Policy Optimization

This study highlights the significant heterogeneous effects of vocational education on the basis of gender, region, and course match, offering valuable insights for policy design. First, the female cohort demonstrated greater employment and income growth after they participated in vocational education. This finding suggests that female-targeted vocational education programs have greater potential for poverty alleviation. It is recommended that rural areas prioritize female-specific resource support, such as gender-focused skills training programs and incentives for vocational certification tailored to females (Chongcharoentanawat et al., 2022; Salas-Velasco, 2024). Second, the analysis of regional differences reveals that while vocational education results in greater employment and income growth in the western region, the baseline effect remains lower than that in the central region. This outcome reflects the unequal distribution of educational resources and underscores the need for greater investment in vocational education in resource-poor regions (Heinesen & Lange, 2024). Suggested actions include expanding school-enterprise cooperation programs and improving the quality of teaching resources in these areas. Finally, the significant impact of course match underscores the importance of dynamically adjusting curriculum design (Dalby & Noyes, 2022; Yang & Chia, 2024). Policymakers should conduct in-depth analyses of labor market demands and optimize vocational education courses to better align with market needs. This approach would enhance the adaptability of vocational training, enabling them to better meet the specific requirements of local economic development.

Research Limitations and Future Directions

While this study provides strong empirical evidence for the poverty alleviation effects of vocational education, several limitations point to opportunities for future research.

First, the study's data are primarily from low-income rural areas in central and western China, which excludes eastern or more remote regions. Given that the study sample is limited to central and western rural low-income areas, the research findings are primarily applicable to these regions rather than the entire rural areas of China. Specifically, the conclusions cannot be generalized to eastern China's developed rural areas or other economically advanced rural regions. This limitation may affect the generalizability of the findings. Future research should broaden the scope of data collection to include regions with diverse geographic and economic contexts, enabling a more comprehensive understanding of the poverty alleviation effects of vocational education.

Second, this study focuses on the short-term economic impacts of vocational education but does not delve into its long-term benefits, such as mitigating inter-generational poverty transmission and enhancing social mobility. Although the present study employs a two-wave longitudinal design, the follow-up period remains relatively short. Future studies may adopt longer-term, multi-wave longitudinal designs to examine the long-term and intergenerational impacts of vocational education. In addition, because the two groups differ in their typical career trajectories—vocational students tend to enter the labor market earlier, while general high school students often delay employment to pursue higher education—the observed advantages in income, employment, and social capital should be interpreted primarily as early-career outcomes rather than evidence of long-term superiority. In the long run, higher education students may catch up or even surpass vocational students, depending on their degree attainment and field of study. To address this limitation, future studies should rely on panel or cohort data to track students’ career development over time and provide more comprehensive insights into both short- and long-term effects.

Additionally, the mechanisms behind social capital enhancement were not fully explored in this study. Future research could combine qualitative methods to investigate how vocational education fosters community participation and social cohesion, further uncovering potential pathways for social capital formation. Furthermore, future research may further explore the development of targeted career guidance and counseling interventions for rural youth to support sustainable career construction and long-term labor market adaptation. Given the evolving landscape of career development in the twenty-first century, future studies could integrate innovative frameworks of sustainable career development that emphasize decent work, lifelong adaptability, and alignment with long-term societal and ecological challenges (Guichard, 2022; Hartung & Di Fabio, 2024). Practical interventions could include personalized career planning services, skills matching guidance, job-search support, and postgraduation follow-up counseling to strengthen the link between vocational education and sustainable career growth. These efforts would help translate vocational skill training into stable, long-term career success and further enhance the sustainability of poverty-alleviation effects. Collectively, these research directions will contribute to a more comprehensive understanding of the social and economic benefits of vocational education and provide policymakers with a stronger scientific basis for optimizing vocational education policies.