We prove an upper bound on the convergence rate of Markov Chain Monte Carlo (MCMC) algorithms for the important special case when the state space can be aggregated into a smaller space, such that the aggregated chain approximately preserves the Markov property.