Chapter 2 introduces some of the methodological issues in the research design of this study. Greater detail, the remainder of the issues, and additional descriptive information on the bureaus appear in this appendix.
Many extensive, quantitative approaches can be, and have been, taken to evaluate whether the outputs of a bureau change systematically with changes in control in the White House (see Chapter 1). An extensive quantitative approach, however, is ill suited to the purposes of this study for a number of reasons. The first and foremost reason is the exploratory nature of the study: it was initially unclear what the mechanisms were by which departmental bureaus responded to the preferences of presidential administrations, how readily a bureau's careerists could adjust their activities (even when willing to do so), and how longer-term policy change occurred within bureaus. Despite the limited initial knowledge of these factors, the study needed to assess them to evaluate whether careerists acted in support of presidential policy preferences. 1 These points made an intensive method appropriate and a quantitative approach problematic.
A quantitative approach was problematic both because of the difficulty in specifying a priori the items to count and because of the qualitative nature of many presidential preferences themselves. Quantitative analysis requires that relatively distinct dimensions of performance be identified. Yet many of an administration's expressions of preference require innovation and creativity on the part of the bureau's careerists to find or create appropriate policy alternatives and then to support them. An administration's preferences for policy change often begin as only general conceptions which then undergo a process of detailed specification and evaluation against the technical possibilities, available resources, and cost of achievement. Exploring this process of evolutionary refinement, adjustment, and evaluation of preferences is in many ways more suited to qualitative than quantitative measures.