學費的制定攸關教育資源配置的合理性，教育的外部性應會影響學費政策；政府如何選擇預算，及其對最適學費的影響如何呢？對此問題，文獻常只探討某特定教育階段，鮮有分析外部性與教育資源配置的關係，且視政府預算為外生，而忽略其社會成本。本文以當事人 – 代理人模型，探討政府如何制定非線型最適學費，且不同能力的個人選擇其最適教育量，以達到其終生所得最大。突破雙重積分的困境後，得到最適學費的一般明顯解，找出學費制定的一般法則，使各階段教育的學費政策具一致與合理性。重要的政策結論有(1)當教育具總外部性時，最高階段教育的學費恆小於其成本；若為總和外部性，則每階段教育皆額外補貼同一數額，此數額為該外部性對最高教育者總所得之邊際貢獻平減預算之社會邊際收益。(2)假設教育成本非遞減且工資函數與能力分佈為合理特定函數，則當預算的社會迖際收益（預算可選擇時，應等於其成本）小於一時，政府可補貼較高等教育；越高階段的教育，學費越高；且補貼越低。

　　　　　The efficiency and fairness of educational expenditure is regularly challenged in society. Using a principal-agent model to derive a nonlinear tuition policy generally applied to every level of schooling, we analyze cases with/without an externality of education, with/without a fixed government budget constraint. The analytical solution is explicitly derived. Some policy implications are: (1) in the case of aggregate-sum externality, the government should "extra" subsidies every level by an equal amount, which is the marginal contribution of the externality to the income of the highest level deflated by the financing cost of the budget; (2) assume that the marginal cost of producing education is increasing and the wages and individual characteristics are some specific forms. Then, when the marginal cost of financing is less than 1, the tuition policy is increasing and the proportion of the production cost paid directly by the higher education people is higher.