Context: The lack of representation of people from low socio-economic and socio-educational backgrounds in the medical profession is of growing concern and yet research investigating the problem typically studies recruitment and selection in isolation. This study examines the impacts of home and school socio-economic status (SES) from application to selection in an undergraduate medical degree. Socio-cognitive career theory and stereotype bias are used to explain why those from backgrounds of low SES may be disadvantaged, especially if they are female. Methods: Home and high school SES information for 2955 applicants and 202 medical students at one Australian medical school was related to application rates and performance on three selection tests (high school matriculation, the Undergraduate Medical and Health Sciences Admissions Test [UMAT] cognitive ability test, a multiple mini-interview) and academic performance in medical school. Interactions between gender and SES were assessed using moderated regression analyses. Results: Applicants from backgrounds of low SES were under-represented. They were further disadvantaged at selection by the use of high school matriculation and cognitive ability tests, but not by the interview. They did not perform more poorly in medical school. Although females applied in greater numbers, a significant interaction between SES and gender indicated that female applicants of low SES were the most disadvantaged by the use of cognitive ability testing at selection. A targeted allowance of applicants from regions of low SES overcame this adverse impact to some extent. Conclusions: Efforts to widen participation that focus on recruitment are insufficient when selection tests have adverse impacts on people from backgrounds of low SES. The addressing of low self-efficacy that arises from socio-cultural factors, together with reductions in stereotype threat, may reduce the current disadvantages imposed by SES in the medical profession. Discuss ideas arising from the article at www.mededuc.com discuss.