This paper discusses some reasons why system control theory has had only a limited impact on economic analysis, especially on government policy¬making. First, in engineering targets are well formulated and known to the controller; in economics, government's targets are often unknown or misperceived by the public. Second, in engineering the distinction between manipulated and non-manipulated variables is given, and the effectiveness of each manipulated variable established. In economics, it is often unclear which manipulated variable is actually used for policy. To this purpose, Granger causality tests -which in fact relate to causality in prediction and not to causality in control -seems inadequate. The paper also discusses some possible uses of the concept of cointegration for inference about unknown targets.