This paper examines whether educational production in secondary school involves joint production among teachers across subjects. In doing so, it also provides insights into the reliability of value-added modeling. Teacher value-added to reading test scores is estimated for four different teacher types: English, math, science and social-studies. The initial results indicate that reading output is jointly produced by math and English teachers. However, while falsification tests confirm the English-teacher effects, they cast some doubt about whether the math-teacher effects are free from sorting bias. The results offer a mixed review of the value-added methodology, suggesting that it can be useful but should be implemented cautiously. (Contains 10 tables.)