varied regions; Carter et al (2003) study the overall impact of USAID DG programs in six countries; and de Zeeuw and Kumar (2006) look at media, human rights, and election programs in nine postconflict states.
While these studies have generated valuable insights into how programs were carried out, how they were received, and how participants and donors perceived their effects, they are not ideal either for “determining program effectiveness” or to “inform strategic planning.” This is because such studies focused almost entirely on specific DG projects, rather than on the broader context of democratization in the countries being studied. They did not systematically compare cases of varying levels of DG assistance or compare the effects of DG projects with comparison groups that did not receive assistance.
The basic tool of case study analysis is process tracing (George and Bennett 2005). In this method, researchers track the unfolding of strings of events, testing hypotheses regarding the causal relationships among them by considering multiple hypotheses that could account for the strings of events and searching for confirming and disconfirming evidence. The process is not unlike a detective’s efforts to solve a murder mystery by reconstructing a timeline of events, examining all possible suspects and their alibis, assessing plausible motives and opportunities for the observed actions and events, and building a case in favor of one causal chain as having determined the ultimate outcome rather than others.
Like solving any mystery, process tracing can be painstaking and time-consuming work, and the results often depend on an analyst’s skill in recognizing how specific social conditions, motivations, events, and opportunities link to form a coherent explanatory chain. Also like any criminal case, the persuasiveness of pointing out any one factor or event as causal depends on the analyst’s imagination and skill in identifying and considering alternative causal pathways and gathering evidence as to how likely or unlikely they were.1
Case studies to demonstrate the effectiveness of aid programs thus face the same challenge as formal statistical evaluations—they must try to determine what would have happened in the absence of the aid program, whether by including studies of both groups receiving aid and those not receiving aid in their case studies (a comparative case study design) or by trying to trace and account for historical trends and confounding fac-