In this paper Monte Carlo simulation is used in order to compare the performance of five different methods to estimate quarterly models with partly missing quarterly observations. The methods are compared on the basis of the parameter estimates they produce. The first three methods solve the estimation problem in two steps: first the yearly series is disaggregated into a quarterly one and then the quarterly model is estimated. The fourth method considers disaggregation of the yearly series within the context of the model to be estimated and arrives simultaneously at estimates of the missing quarterly observations and of the parameters of the model. The last method simply consists of maximum likelihood estimation of the yearly model. The conclusions from this simulation study are twofold: (i) none of the methods that are developed for the purpose of estimating quarterly models with partly missing observations performs significantly better than maximum likelihood estimation of the yearly model; (ii) the standard errors that result from application of the first three methods are deceptive.