Purpose – Education-job mismatches, especially overeducation or vertical mismatch, are generally found to lower the worker’s job satisfaction, which may generate the counter-productive behaviors, such as high rates of absenteeism and turnover in developed countries. The purpose of this paper is to examine the impacts of educational mismatches from their both forms and dimensions (match, overeducation, horizontal mismatch and double mismatch) on the job satisfaction among university graduates in Cambodia. Design/methodology/approach – To deal with the sample selection bias owing to the unobserved job satisfaction of unemployed graduates, this paper applies the Heckman probit model on a survey conducted with 19 higher education institutions in Cambodia. Findings – Results indicate that a half of graduates suffer at least one type of educational mismatch and the both forms of mismatches adversely affect the job satisfaction with the strongest impact from the double mismatch case. Research limitations/implications – The authors take into account the sample selection bias, but are not able to deal with the unobserved heterogeneity, such as individual competences and preferences. With the panel data, it would be possible to isolate those individual fixed effects. Practical implications – The findings underline the importance of improvement in the quality of higher education in Cambodia that seems to play a main role in this education-job mismatch problem. Creating the occupational counseling for the high school students would be also helpful to orientate students to the majors strongly needed by the labor market. Originality/value – This paper focuses on all forms and dimensions of mismatches and takes into account the sample selection bias in the context of a low-income country where the increasing rate of enrollment in higher education seems to be accompanied by an increasing rate of education-job mismatches. Previous research works focused mostly on overeducation and in developed countries. Keywords Higher education, Job satisfaction, Heckman probit regression, Sample selection bias, Vertical and horizontal educational mismatches