Budget revenue forecasts should be best estimates of expected receipts. Often they are not. This paper analyzes the rationale for overstated revenue forecasts and derives conditions for intentional biases. A theoretical model demonstrates that overstated revenue forecasts can be the result of the government's attempt to boost unobserved revenue collection effort. If positive forecast errors are costly and undermine public credibility of budget expenditure plans, the reverse outcome is possible and governments may understate revenue forecasts. A case study for Azerbaijan is presented in support of the former incentive motive.