[Excerpt] In this paper, we go inside the “black box” of graduate education to investigate what characteristics of graduate programs in the humanities and related social sciences actually influence PhD students’ attrition and graduation probabilities. We make use of data from the Graduate Education Survey; a retrospective survey of all graduate students who entered PhD programs in the treatment and control departments during the 1982-1997 period that was conducted by Mathematica Policy Research Inc. for the Mellon Foundation. After briefly describing the survey in the next section, section III outlines our methodological approach. Section IV presents our empirical findings and brief concluding remarks appear in section V. While our focus is on evaluating the effects of the GEI, we believe the methodological approach that we use can be profitably employed in a wide range of program evaluation studies.