Performance budgeting is a crucial way to promote responsible fiscal management and ensure that taxpayer funds are used effectively to achieve desired policy outcomes. Scholars have assessed its practical impact, yet few examine whether performance measures yield actual incentives in theory. This study thus introduces a frontier-based approach for performance budgeting, incorporating incentive considerations. The framework considers a scenario where performance budgeting spans a budgeting period and a subsequent post-assessment period. The budgeting period involves designing incentive budget allocation plans based on performance evaluation results, using technical efficiency and marginal contribution measures. In the post-assessment period, the effectiveness of the proposed incentive allocation plans is examined through inferring subordinates’ best strategies. Subordinate expected utility is modelled as the difference between allocated budget and the cost of effort for performance improvement. Subordinates’ strategies are derived from solving an expected utility maximization problem, indicating that optimal strategies involve improving performance to specific levels with the proposed allocation plans. To implement the allocation plans, a multiplicative directional distance function framework with piecewise Cobb-Douglas technology is introduced for measuring subordinates’ performance. The proposed approach is empirically applied to allocate budget for Chinese “985 Project” universities.