Several experimental and cross-sectional studies have established the stress-buffering effect of social identification, yet few longitudinal studies have been conducted within this area of research. This study is the first to make use of a multilevel approach to disaggregate between- and within-person effects of social identification on subjective and endocrinological stress reactions. Specifically, we conducted a study with 85 prospective students during their 1-day aptitude test for a university sports program. Ad hoc groups were formed, in which students completed several tests in various disciplines together. At four points in time, salivary cortisol, subjective strain, and identification with their group were measured. Results of multilevel analyses show a significant within-person effect of social identification: The more students identified with their group, the less stress they experienced and the lower their cortisol response was. Between-person effects were not significant. Advantages of using multilevel approaches within this field of research are discussed.