The First Edition was popular for its integration of multiple regression with evaluation desig, and for offering systematic ways to select the proper goals for single- and multiple-outcome evaluations. This edition has been revised to cover new issues and to clarify further the concepts used in impact analysis. New features include: expanded coverage and explanation of quasi-experiments; a section on impact analysis theory; a section on the use of qualitative research for impact analysis; expanded coverage of significance testing for programme evaluation; an explanation of why the comparative-change design is better than an ex-post-facto design from the standpoint of causal inference; a clarification of the effects of volunteering or self-selection; and a new simplified appendix on regression artefacts.