In this paper, we present a theoretical analysis of the error with three basic Monte Carlo radiosity algorithms, based on continuous collision shooting random walks, discrete collision shooting random walks and stochastic Jacobi iterations. We show that the variance of these three algorithms is in practice to very good approximation identical, explaining empirical results obtained before. The derivation in this paper can easily be adapted in order to calculate the variance of other Monte Carlo rendering algorithms, such as stochastic ray tracing and density estimation as well.