Background The Intercollegiate Membership of the Royal College of Surgeons (MRCS) examination is undertaken by large numbers of trainees in the UK and internationally as a mandatory step within surgical training. Unlike some high‐stakes medical examinations, the MRCS is yet to be validated. A quantitative study was undertaken to assess its predictive validity by investigating the relationship between MRCS (Parts A and B) and national selection interview scores for general and vascular surgery in the UK. Methods Pearson correlation coefficients were used to examine the linear relationship between each assessment, and linear regression analyses were employed to identify potential independent predictors of the national selection score. All UK medical graduates who attempted the interview in 2011–2015 were included. Results Some 84·4 per cent of the candidates (1231 of 1458) were matched with MRCS data. There was a significant positive correlation between the first attempt score at Part B of the MRCS examination and the national selection score (r = 0·38, P < 0·001). In multivariable analysis, 17 per cent of variance in the national selection first attempt score was explained by the Part B MRCS score and number of attempts (change in R2 value of 0·10 and 0·07 respectively; P < 0·001). Candidates who required more than two attempts at Part B were predicted to score 8·1 per cent less than equally matched candidates who passed at their first attempt. Conclusion This study supports validity of the MRCS examination, and indicates its predictive value regarding entry into specialist training.