Theses Doctoral

The Determinants of Post-Compulsory Education Decision in Rural China: With an Analysis of a Grassroots NGO Intervention

Yao, Haogen

In rural China, when approaching the end of nine-year compulsory schooling, students face four equally popular post-compulsory education decisions (PCED): dropout, work after graduation, vocational high school, and academic high school. The literature tends to simply treat PCED as dichotomous (continue vs. leave school), and there is a geographical research imbalance favoring inner China. An increasing volume of studies also suggest that traditionally recognized factors like socioeconomic status and academic performance are not as influential as before in advancing the schooling. People have started to look at socio-emotional support, such as the promotion of self-discipline and confidence. At present, it is grassroots NGOs (GNGO) who take the major responsibility for providing this type of support in rural China, and there is rare discussion of achievements, let alone evaluation of practical impact.
Given the existing problems, the key research questions of this study are: (1) What are the current PCED determinants for China’s rural students? More specifically, what are the PCED determinants for lower secondary students in rural Guangdong, a coastal province? (2) How can GNGO intervention affect PCED by boosting certain subjective factor(s)? The tested treatment is the Lighthouse program, whose one-month summer camp aims to improve student attitudes towards their life, such as making them more confident, organized, and social.
The key to answering the first question is to explore a comprehensive list of variables applying to local populations, which cannot be achieved simply through a literature review. When answering the second question, since Lighthouse participation is voluntary, it is important to deal with selection bias, to ensure that any identified Lighthouse impact results from its activities rather than the student characteristics that lead to their participation.
To overcome these methodological challenges, I first employed the Delphi approach. Delphi is an iterative process used to collect and distill the judgments of experts using a series of questionnaires interspersed with feedback. It is used to identify possible PCED determinants that are missing in the literature, to determine factors that lead to Lighthouse participation, and to collect discussions about both PCED determinants and GNGO intervention. Based on the Delphi results and literature, I then designed five questionnaires for students, households, teachers, principals, and Lighthouse volunteers. In Jun-Oct 2012, I led seven research assistants in conducting two waves of surveys in eight towns, building a firsthand dataset of 6298 valid observations with imputations. Multinomial logit was used to investigate PCED determinants. It predicted the PCED probabilities, given nine groups of independent variables. Propensity score matching was used to evaluate the program impact. It calculates the treatment propensity for each student based on their characteristics, so the Lighthouse impact can be compared between treated and untreated students of similar treatment propensity. Tests of robustness and heterogeneity were conducted after both methods. Qualitative materials collected from Delphi and on-site interviews were used to explore the causal mechanism.
I use relative risk ratios to report the findings of PCED determinants. The findings challenge the existing literature regarding the roles of gender and parental background, further extend knowledge of monetary reward/cost and subjective factors, and confirm new possible determinants that have seldom been investigated in literature. The main model passes the robustness check, and there exist explainable heterogeneity effects. It is notable that education aspiration stands out as a strong PCED determinant, ceteris paribus.
Propensity score matching shows that the Lighthouse program mainly affects PCED by boosting educational aspiration for students with high academic performance, although that impact fades gradually if there is no follow-up service. The novelty of the program to local people, volunteer team morale, and volunteer acceptance of Lighthouse training could help explain why increases in aspiration varied across sites. The role-model effect might explain why the increase in aspiration exists, as there are signs that the students tried to copy the volunteer’s schooling decision once trust was built.
This study makes three major contributions. It can be translated into comprehensive advocacy for education policies related to PCED, such as dropout prevention and the promotion of VHS. It may also suggest the value of, or at least the required improvement to, China’s educational GNGOs, which are young and remain confined by governmental regulations. Last but not least, this is a unique showcase of how qualitative-quantitative sequential mixed-method works better in exploratory analyses. The study has limitations in timing, missing data, external validity, implementation of research methods, and heavy rely on self-reported questionnaires, but they can be largely eliminated by conducting proper further studies.

Geographic Areas


  • thumnail for Yao_columbia_0054D_13123.pdf Yao_columbia_0054D_13123.pdf application/pdf 7.03 MB Download File

More About This Work

Academic Units
Economics and Education
Thesis Advisors
Tsang, Mun C.
Ph.D., Columbia University
Published Here
February 1, 2016