Global solution methods for dynamic stochastic general equilibrium models are accurate but computationally expensive. In particular, evaluating conditional expectations at numerous points in the state-space leads to significant computational complexity. In the paper at hands, I show how to remove the majority of calculations required for the evaluation of conditional expectations. Therefore, I replace the approximated integrals obtained by e.g. quadrature rules with an analytic expression. I provide a general framework and carry out the approach in detail using Chebyshev basis functions. Subsequently, I adapt the exact expectations technique to the neoclassical model with recursive utility, labor choice and student-t shocks to log-productivity.