Theory testng and evaluation research are similar, and although the two processes are typically carried out by different researchers with little cross fertilization, evaluation entails testing of some theory, even if implied. This relationship is widely ignored in the field of criminal justice, with the result that much evaluation is done with results never being applied to the cumulation of theoretical knowledge. Two prominent economic approaches to evaluation, cost-effectiveness evaluation and the theoretical approach of economics identified most prominently with Becker and Erlich, provide a good example of the problems in relating evaluation to theory building. The first type of analysis contributes little to knowledge, rather it contributes facts to which theories must fit. Results which indicate the price of a program in relation to measurable benefits are descriptive and useful as far as the input data are reliable and valid. They do not empirically verify theoretical principles, and results of similar analyses often differ dramatically, depending on the population, setting, etc. The economic theory of behavior applied to crime, as outlined by Erlich, is deceptively simple. First it necessitates converting some subjective phenomena into objective indexes for purposes of analysis and ignoring others. For instance, Erlich's first assumption asserts that offenders behave 'as if they seek to maximize their expected utility.' No consideration is given to the issue of the extent to which such utility is a function of structurally determined condition or of emerging situations. A second assumption presumes that the distribution of individual preferences for crime is relatively stable across different communities at a given point in time. The approach does, however, have promise when used in conjunction with more traditional psychological and sociological approaches. What is needed overall is for research strategies to go beyond data collection and beyond formalizing theory that is potentially falsifiable and to designing test cases in terms of the theory so that measurement error can be specified. The problems of evaluation as a contribution to knowledge are demonstrated in a study of delinquent gangs and detached worker projects done in Chicago and another series of experiments in educational settings based on status characteristics or expectation states theory.