This paper assesses the magnitude of the non-indigenous/indigenous test-score gap for third-year and fourth-year primary school pupils in Peru, in relation to the main family, school and peer inputs contributing to the test-score gap using the estimation method of feasible generalized least squares. The article then decomposes the gap into its constituent components using the traditional Oaxaca-Blinder decomposition method, as well as a modified decomposition method based on the estimation of a cognitive achievement production function. The decomposition results from both decomposition methods suggest that almost all of the test-score gap is explained by various peer, student, family and school characteristics. The peer characteristics used in the regression are the main contributors to the gross test-score gap, comprising between 58% and 71% of the language test gap and 45-62% of the mathematics test-score gap, depending on the decomposition method used. (Contains 4 figures, 5 tables, and 11 notes.)