Contrary to their significance for the adjustment of econometric models, dummy variables are seldom addressed in the literature. They serve as a tool for improving the fit of the model equations to the data and with it the prognostic performance of a model, at least lastly. Referring to the RWI-business cycle model (version 59) as an example, it is shown in this article that all the dummy variables taken as a whole are in fact reducing the error measures of the most of the endogenous variables; on average, the mean absolute percentage error of a dynamic solution of the example model over the same sample period its parameters were calculated is reduced to about 45 percent (yearly) or, respectively, to about 61 percent (quarterly). As it is well known, the price one has to pay for this improvement is a narrowing of the theoretical basis of a model. In the case of a simulation of political measures with short-term effects (for instance, additional investments to the infrastructure made by the government) the differences between a model with and the same model without dummies are rarely of practical significance – but this depends on the variable analyzed and on the point in time the impulse was set in the previous year. One of the consequences of this study consists in the question, whether the cost caused by the scrutiny for shifts and breaks in macroeconomic time series and caused by the implementation of dummy variables which take account of those breaks is counterbalanced by the slightly better fit and performance of the models. Investing more time and deliberation in the theoretical backing of the equations that made up an econometric model seems to be a reasonable alternative.