Copolymer reactivity ratios are often estimated from experimental data by outdated and statistically incorrect methods which are usually based on a linearization of the Mayo-Lewis model. This is true even though a solution using nonlinear regression analysis was published in the literature thirty years ago. Here the uses of nonlinear least squares and the error-in-variables model, which accomodates errors in all measured variables used in a model, are demonstrated and compared. The comparison is carried out by simulation using seven representative copolymer systems. For the Mayo-Lewis model there are circumstances where very little difference exists between the results obtained by the two methods in particular when the error in the monomer feed mole fraction is less than 1%. In this case, although the point estimates will be similar, an approach based on nonlinear least squares may underestimate the size of the joint confidence regions. Under the assumption of model adequacy, experimental plans based on D-optimality yield precise parameter estimates when compared to the more empirical approach of distributing experiments evenly over the monomer feed composition range. Furthermore, the use of data from designed experiments leads to a very simple closed-form solution for estimating reactivity ratios. This study confirms that EVM is the statistically correct state-of-the-art approach to estimating reactivity ratios and should be the preferred technique. Techniques such as the Fineman-Ross approach are suitable only for obtaining starting values for nonlinear estimation procedures.