In the Dutch Educational Assessment Program, the students' language proficiency is measured in grade 9, at age 15. Writing performance is measured through several performance-based writing tasks, rated on numerous aspects such as content, style, organization, punctuation, spelling and grammar. As a consequence, national performance levels are reported in a rather detailed fashion. This paper focuses on the efficiency of the measurement and reporting of writing performance using a sample of 1,451 students during 1987-88. Multilevel factor analysis (MLFA) was used to determine dimensionality and reliability. At the school level, a distinction between two factors was warranted: composition versus technical conventions. At the student level, the composition factor remained intact, but the technical conventions factor had to be divided into the subskill factors of punctuation, spelling, and a factor interpreted as technical quality at the sentence level. Factor structure was not found to be independent of the writing task. Indications of the unreliability of measurement of functional writing proficiency suggest that the usefulness of the national assessment data for school effectiveness studies is limited. Recommendations are made for improvement of assessment. (Contains 5 tables and 57 references.) (Author/SLD)