It is the purpose of this article to suggest that the technique of the analysis of covariance dominates, in a statistical sense, the use of dummy variables. We shall first show how dummy variable regressions and the analysis of covariance are related. This comparison will show why we think the analysis of covariance is a more useful technique than the use of dummy variables in a regression analysis. Then in an attempt to make our prescription of the analysis of covariance somewhat easier to apply we shall develop a set of computational short cuts. In our view these result in computational difficulties of such small extra cost that they seem clearly to suggest that the analysis of covariance is a technique much more useful than simply inserting dummy variables into regressions.