Law and Social Inquiry
Volume 21 (Summer 1996), 761-804
Experienced social scientists know that textbook descriptions of the research process are at best a sanitized description of the messy reality of what happens when researcher meets data. The neat progression of research question to research hypotheses to research design to data collection to analysis and finally to reporting is not unheard of. But most large, complex research efforts lack this linearity, reflecting the uncertainties, ambiguities, and complexities of the social and political world. The course of research is also tied very closely to the background and experiences the researchers and analysts bring to a project, both on an individual level and on a collective level for group projects or projects rooted in strong institutional settings.
Understanding the realities of the research and analysis process itself requires some framework or perspective. Textbooks typically portray research in terms defined by the more positivist approaches to philosophy of science.(1) In this essay I turn instead to literatures on persuasion and interpretation. *762 Specifically, I draw on recent writing on the rhetoric of inquiry and on writing dealing with the problems of interpretation of texts.
Substantively, I focus on one research product from a large research project. In 1993, Harvard University Press published The Hollow Core: Private Interests in National Policy Making by John Heinz, Edward Laumann, Robert Nelson, and Robert Salisbury (whom I will collectively refer to as HLNS(2)). As indicated by the subtitle, The Hollow Core is a study of interest groups; more specifically, it focuses on persons who represent interest groups in Washington. HLNS base their analysis on a large dataset constructed from structured interviews with interest groups, individuals who represent those groups, and government officials who are the targets of influence efforts. HLNS focused their attention on four distinct policy "domains"--energy, labor, health, and agriculture.
The analysis in The Hollow Core is extensive and rich. In an essay of this type, I cannot possibly do justice to the full range of issues HLNS consider in the book. For this reason, my discussion focuses on three key findings: patterns of recruitment and careers of representatives, including the role of lawyers in the interest representation process; the social structure of interest representation;(3) and the impact of interest representation on policy outcomes. In the discussion that follows, I will summarize HLNS's analyses related to these three topics and then present a reconstruction of the analytic journey that culminated in The Hollow Core. My goal is to use the complexities of that journey to develop an understanding of how various factors shape and direct research.
In developing this essay, I have relied on the text of The Hollow Core, a long group interview with the four authors, and some early design documents to which the authors generously gave me access. My discussion is in three sections: an overview of the process of developing and executing research, a summary of HLNS's analyses on the three key points mentioned above, and a reconstruction of HLNS's processes of project development, data analysis, and results interpretation.
As I have noted, the textbook image of research sees a more or less linear process:
Some, perhaps even a lot of, research does flow in this way.(4) However, Shively acknowledges the more frequent reality:
[O]ne of the better-kept secrets in political science [and I would extend this to the social sciences more generally] is that good political scientists generally do not draw up research designs before they start to work on a topic. Nor do they usually "frame hypotheses" in any formal sense before they start to work, though they may have some operational hunches about what they expect to find. . . . Their procedure is much less formal than the one they prescribe for students. They play with data, immerse themselves in what other people have written, argue with colleagues, and think.(5)
There is one clear commonality in the research process (at least in the research we come to know about), and that is the role of the "last" phase, writing. That phase forces the researcher to communicate, and the questions that arise during the writing can often lead back to more analysis, or to additional questions, or even, where possible, to additional data collection. In part, this is because data themselves are often extremely messy. The authors of The Hollow Core take note of the inconveniences of data and the ability of data to conflict with one's neat theories (p. xiii). Similarly, Laumann and Knoke in a related study of interest organizations, acknowledge the "numerous colleagues [who] read, commented upon, criticized, or suffered through our various efforts to wring a coherent account from the data."(6)
*764 But as suggested by Shively, the complexities and inconveniences of research go well beyond the data themselves. While one might describe the research process in many different ways, I would liken a research endeavor to an intellectual journey or safari.(7) Research is not a simple journey to the shopping mall of research results, involving a straight shoot down a super- highway with wide lanes, well-marked exits, and findings to be picked off the store shelves or racks. Rather, it is more akin to setting out on a faint trail through the jungle. There is usually some planned destination or quarry, but the researcher may or may not reach that destination or find that quarry. There may be clearings along the way where surroundings can be seen and evaluation of progress can be made, but overgrowth obscures most of the path, and many distractions appear as the researcher travels along in search of his or her quarry.(8)
Furthermore, the quarry itself may be ill defined. We may think we know what we are looking for, but we may be looking for the wrong thing, or we may not see something that is right in front of our eyes because we are not looking for it. It would be nice if the quarry could stop us and say, "Whoa, stop, you idiot . . . I'm what you're looking for." In fact, researchers may stare at a set of materials for days, weeks, or longer, and then suddenly recognize a pattern or result that significantly clarifies or explains the phenomenon of interest. This failure to see something right before our eyes which in retrospect may seem obvious is not unique to social science: in 1995, biologists discovered, residing on the mouth of the common lobster, the Cycliophora, an entirely new phylum.(9)
While research can be carried out as a mechanical process, high-quality research involves much more: intellectual curiosity, technical skill, an ability to improvise, creativity, and luck (e.g., accidents of juxtaposition). Some of these elements (e.g., technical skill) can be, and are, taught. Others (e.g., *765 knowing how to improvise) can be gained through experience. Some (e.g., curiosity, creativity) are probably largely a function of individual personality.(10) But even with the requisite skill, experience, and curiosity, there are always those elements that are matters of luck: being in the right place at the right time, knowing someone who knows someone who knows the right person. Research, to draw on the title of Shively's book, is a craft.(11) How might we understand how various elements of the research craft combine to result in first-class research products? For this task, I draw on two sets of literature. The first is the recently developed literature on the "rhetoric of inquiry."(12) That literature is a joinder of rhetoric of presentation with the logic of inquiry. It posits that underlying inquiry is a process of argument and persuasion: Before an analyst can make an argument to an audience, the analyst must persuade him or herself as to the results and the meanings of those results; only then can the analyst communicate findings to others. The process of self-persuasion is a form of rhetoric (in the classic as opposed to the pejorative sense).
One prominent figure in the rhetoric of inquiry "movement" is economist Donald McCloskey. He has described what he calls the rhetorical tetrad consisting of logic, fact, story, and metaphor. Each of these has its own process of reasoning: logic through deduction, fact from induction, story from understanding, and metaphor through the process called abduction (or retroduction).(13) The elements of empirical research--theory, data, analysis, and interpretation--draw on the various types of reasoning but not necessarily in ways that map to the simplified view of the research process one finds in most texts on methods. Thus, rhetoric of inquiry casts doubt both on the standard positivist image of research with its emphasis on the deductive logic of theory to hypotheses to data(14) and on the "grounded theory" approach with its emphasis on induction from data to understanding.(15) The *766 rhetorical tetrad focuses our attention on the interrelations among disparate elements of the research endeavor and on the constant moving back and forth among the various logics underlying research practice.
I also draw on a second literature to recognize the centrality of the process of interpretation. In empirical analysis, a key (if not the key) problem is arriving at an understanding of the meaning of the data and analysis. To do this, the analyst must usually interpret data and results because the empirical information seldom presents itself in a clear, unambiguous fashion. In the literature on interpretation as a process, the object of interpretation is most often some type of text: "a group of entities, used as signs, which are selected, arranged, and intended by an author in a certain context to convey some specific meaning to an audience."(16) This definition of text is not limited to written words; it draws on the broader concept of the meaning laden "sign" developed by semiology.(17)
Much, probably most, of the data used by empirical social science falls within this definition of text but with added complexities.(18) The first complexity is that the social scientist typically has a multiplicity of individual ("zero-order") texts (i.e., sets of information or responses from multiple observations or respondents); thus there is no individual "author" and the researcher must extract some common meanings from what is effectively an aggregated ("first-order") text. This first-order text typically constitutes the "fact base" for the analysis, and that fact base actually combines, in an inductive kind of way, the researcher's intentions (by defining what is sought) and the original sources' intentions (because the sources provide the information in response to some type of request).
What probably most differentiates social science from other problems in analyzing text (broadly defined) is that the analyst does not directly interpret either the zero-order or first-order text, but goes on to manipulate the first-order text to generate a set of results that constitute a "second- order" text. The second-order text is both a tool of interpretation (by creating summary information about the lower order texts) and a subject of interpretation (by generating new "signs" that themselves require interpretation). The creation of this second-order text involves a combination of deduction (to identify possible avenues of analysis given the research *767 questions and hypotheses) and induction (to link unexpected results to alternative hypotheses).
A second important difference between traditional textual interpretation and the interpretive efforts of social scientists is that social scientists seek to move well beyond understanding the intentions of the creator of the text. The interpretive goal of most social scientists is to arrive at understandings of their sets of materials that inform their theories or research questions. However, even with this difference of purpose in the interpretation process, many of the tools of interpretation used by social scientists are much like those used by textual analysts.
One common method of interpretation that empirical social scientists use is to propose narratives or stories that bring coherence to the data. The goal here is to understand the data and the data's implications for the analyst's research question and/or theory. In developing these narratives, the analyst works primarily from the second-order text, but also regularly refers back to the first-order text, and possibly even the zero-order text. Once comfortable with a story line, the researcher identifies possible metaphors that link those stories to theoretical structures and to broader societal concerns. These metaphors may lead the analyst to want to revise the stories, which may require additional analyses, which lead the analyst to want to collect additional data.
This sketch raises several points. First, by thinking of data as text, we can see that the data establish boundaries for an empirical analysis.(19) Data need not (seldom?) dictate results; they do normally limit results. The same is true of traditional text: a text does not dictate a meaning, but it does impose constraints on possible interpretations.(20) A key difference between the interpretive task of the empirical social scientist and the literary scholar (or the art historian or the musicologist), however, is that the social scientist virtually always creates the second-order text and often has a great deal of control over the creation of the zero-order and first-order texts.
Second, empirical analysis involves making sense of data both in terms of identifying possible patterns in the data and verifying the existence of those patterns. Our processes of "making sense" revolve around narratives and tropes such as metaphor. A quite common phrase I have heard in presentations of quantitative, empirical research (by economists, political scientists, and sociologists) is "what is the story here?" Recognizing and making sense of patterns as the core of interpretation is by no means limited to social science. For example, art historians have longed puzzled over Raphael's last painting, "The Transfiguration," which pictures two apparently*768 unrelated scenes: the transfiguration of Christ and a New Testament scene depicting the disciples' failure to cure a boy of epilepsy. Why does Raphael appear to show the triumphant assent of Christ with what appears to be a defeat of his powers? Only recently did a professor of medicine observe that the boy appeared to be in the final phase of a seizure rather than in the onset or climax. "In short," the doctor noted, "Raphael showed the boy as cured . . . [and] the two scenes are . . . bound together by showing mutually compatible scenes of the divinity of Christ and His miraculous power to heal." To the recognition of the boy's medical state, the professor added the observation that Raphael had probably been informed by a Vatican official that "Raphael" meant "God hath healed" in Hebrew.(21) Thus, here we see first the recognition of a possible pattern and then the discovery of another piece of evidence that provides a measure of confirmation of that pattern.
When one tries to understand the interpretation process, one quickly realizes that the literature on interpretation has developed largely in the context of textual interpretation, both in literary studies and in qualitative social science. Elsewhere, I have elaborated the argument about why one might want to look to this literature to understand the interpretation process in quantitative analysis.(22) The core of that argument is that quantitative data possess many of the features of traditional text (e.g., complexity, ambiguity, uncertainty, multiple meanings, etc.) and that there are many common elements in how analysts interpret the traditional text and the quantitative "text." In particular, both types of interpretation involve viewing the text in context, identifying or developing narratives, and applying tropological conventions.
Some might argue that a true "science" makes sense using mathematics. However, much (perhaps most) of the natural sciences rely not on mathematical formulations but narrative and tropological formulations. Much of biology (e.g., evolutionary theory), medical science (e.g., etiology of disease), and geology (e.g., formation of geological structures) is largely discussed in the form of narratives. In the more mathematical of sciences, such as physics, theories have both a mathematical and a narrative focus (e.g., the "big bang" theory).(23) And even the most mathematical areas, such as quantum mechanics, are amenable to narratives that aid in understanding.(24)
*769 I now turn to a brief summary of the core lines of analysis and argument in The Hollow Core.
Again let me remind you that my discussion focuses on only three key findings reported in The Hollow Core. The first I consider is patterns of recruitment and careers of representatives of interest groups.
Interest group representatives bring to their work their family and social backgrounds, formal training, substantive nongovernmental experience, and governmental experience. HLNS map these characteristics and then look at these factors as possible influences on the substance of the representatives' day to day work. Those with the strongest political involvement are those whose backgrounds are laden with political experience. In contrast to the common image of lawyers as having some special involvement in politics, the lawyers in the Washington study are more likely to be doing "lawyerly" things than "political" things.
Social background plays a role primarily in terms of who someone represents, with those working on behalf of business and other conservative interests more likely to have backgrounds that conform to stereotypes (come from upper middle class, are Protestant, identify with the Republican Party) and those on behalf of liberal groups reflecting the demographics of the traditional Democratic coalition. Similarly, and not surprisingly, the ideology and outlook of the representatives vary across organizational and party affiliations; not surprisingly, those with ties to traditionally conservative interests have conservative outlooks and those with ties to liberal interests align ideologically with those interests. HLNS find that these linkages reflect patterns of recruitment by interest organizations (p. 172).
One of the long-standing issues in interest representation is the "revolving door" question: do individuals move frequently between government service and lobbying activities, and how much does prior government service affect a representative's ability to exercise influence in agencies where she/he had previously worked? HLNS's data includes detailed career histories that allowed them to follow the revolving door as it turns. Their analysis shows that while a door is there, it usually is an exit, not a revolving door, and when it does revolve, it turns very slowly; only 8% of the representatives interviewed had moved between government service and interest representation more than one time, and fewer than half of those held government *770 positions dealing with matters of immediate concern to their subsequent employers (pp. 119-20). Furthermore, HLNS show that while government experience (particularly congressional experience) is very helpful in a variety of ways (ranging from personal contacts to knowing processes), the value of that experience decays over time.
When HLNS examine the nitty-gritty of the Washington representatives' work, they find three distinct patterns associated with different roles:
The most distinctive group is the latter, who tend to focus their efforts on formally contested matters. The distinctiveness of lawyers is reinforced when HLNS look at whether representatives tend to be substantively specialized within their policy domains; the general answer is no, with the exception of external lawyers (i.e., those who are in private practice as distinct from those directly employed by interest group organizations), who tend to have relatively specific sets of policy concerns. That is, the typical external lawyer working on policy-related matters is not the "generalist, available for hire on a wide range of issues" (p. 92) but someone who specializes in a small number of substantive fields.
The distinctiveness of lawyers and law-related activities is reinforced by a factor analysis of the importance of 18 specific tasks related to policy work. That analysis shows four dimensions underlying the representatives' work: relating directly to government officials and actions ("government relations"), relating to like-minded and opposing interest groups ("interest group networks"), relating to the public ("public presentation"), and employing the legal system for policy advancement ("litigation"). When they look at variations in placement on these dimensions, the most distinctive group is lawyers (both internal and external), which is the only group that stands out on the "litigation" factor.
A final way in which lawyers stand out, particularly external lawyers, is in terms of the autonomy they have in completing their tasks. At least part of this discussion takes place in the context of HLNS's consideration of three potential bases for the social organization of representatives: substantive expertise or training, access to targets of influence, and nature of the client organizations. HLNS find that relatively few representatives (perhaps 20%) work in settings such as law firms or consulting companies where some autonomy is likely to occur; interest organizations directly employ the *771 vast majority of representatives.(25) HLNS sought to assess degree of autonomy directly by asking respondents, "Have you ever had occasion to refuse a potential client or work assignment, not because of a formal conflict of interest, but because of your personal values?" The pattern of responses shows that those whose organizational ties are weaker do in fact have greater autonomy. Outside lawyers and consultants are much more likely (60% or so) to report having refused such assignments than is the sample as a whole; in contrast, only 14% of internal counsel reported having refused such an assignment for this reason.(26) HLNS note that the likelihood of firm lawyers in Washington having refused assignments was much higher (by a factor of about 3) than Nelson had found in his study of four large firms in Chicago (p. 187).
The work of Washington representatives centers on communicating with government officials involved with the policies of concern to the representatives' employers or clients. HLNS explored in some detail exactly whom the representatives contacted, and how often those contacts occurred. The average representative saw 16-20 government targets at least once during the year (p. 194). Some types of targets were common across domains (the White House, congressional leadership), while others were more specific to policy domains (specific committee leaders, committee staffers, and officials in executive agencies). HLNS characterize the structure of contacting as following the "center/periphery principle" (which is consistent with other types of social networks) modified by sectorial differentiation defined by the policy domains (pp. 196-200). Geometrically, this translates into a set of concentric circles segmented into pie slices. The segmentation leads HLNS to apply spatial analyses to assess the nature of the patterns within each domain. They find that the structures do in fact vary substantially across the domains, ranging from "bipolar" in the labor area to "niches" in agriculture. They conclude that conflict within the domains tends to focus on the periphery with the core agencies in the center performing a more neutral coordination or facilitation function (p. 217)
After discussing the interaction between the representatives and their targets, HLNS turn to interaction among interest organizations and among the elite of the Washington representative community (whom HLNS refer *772 to as "notables").(27) Looking first at the organizations, HLNS asked respondents to identify other groups that served as allies on policy conflicts and groups that served as adversaries. There was some difference in the overall ally-to-adversary ratios across domains (1.28 to 1 in labor compared to 2.04 to 1 in health), suggesting differences in the level of conflict in the various domains. They then classified both respondent and nominated groups into domain specific categories (e.g., labor peak associations, unions, citizens' groups, trade associations, business peak associations for the labor domain; farm peak associations, commodity groups, trade associations, and externality groups for the agriculture domain), and cross-tabulated the nominator's category with the nominee's category. They conclude that their results, with the possible exception of the labor area, show the weaknesses of peak associations of the type associated with the corporatist model. In the United States, "the structures of conflict and cooperation are built from the particular configurations of interest organizations that participate in domain politics" (p. 261).
In their analysis of the notables network,(28) HLNS rely both on reports from the notables themselves and on reports from the representatives outside the notable groups. For each notable, the respondents were asked whether they were personally acquainted with the notable and, for those with whom they were acquainted, whether they knew the notable "well enough to be confident that they would take the trouble to assist you briefly (and without a fee) if you requested" (p. 266). Among the representatives outside the notable group, the median representative was acquainted personally with six of the notables and felt able to call on a median of three for a brief bit of assistance (p. 266).
The central part of the notable analysis was a series of smallest space analyses, some relying on the reports of the larger group of representatives and some relying on the notables' own reports. HLNS constructed measures of the degree of acquaintance of each notable with every other notable; these measures served as proximity indicators for the smallest space analysis. The results of these analyses placed each notable into a two- or three-dimensional space. HLNS report these results both for all notables and separately for notables within each domain.
The overall results provide the picture (p. 272) that led HLNS to describe the pattern as having a "hollow core." Dividing the space into quadrants, one finds that the notables from each domain cluster in a particular quadrant, with relatively few notables appearing outside his or her domain's quadrant (notables from the health domain are the most scattered--see p. *773 273). However, looking also at the individual domains, HLNS find cleavage--very striking cleavage in the labor domain--in the placement of notables along representative characteristics such as liberalism, partisanship, income, and interest represented. Similarly, when HLNS look at the acquaintances among the notables themselves in each domain, they find substantial separation based on the specific interest the notable represents (e.g., in agriculture, producers and processors vs. distributors, labelers, and safety; or in health, care providers and researchers vs. pharmaceuticals and consumers). HLNS note that these kinds of cleavages seem to run counter to conventional wisdom that "members of [the] permanent Washington establishment seek to maintain contact with both allies and adversaries" (p. 278); they conclude that their analysis shows that "the four policy domains lack core actors" (p. 298). The notables, whom one might expect to play a mediational role, "appear to function as advocates" (p. 300), as "organizational advocates and mobilizers rather than as mediators whose contacts and connections facilitate the resolution of issues" (p. 301). HLNS conclude that "autonomous brokers who have the capacity to bridge the four areas either do not exist or we [HLNS] were not able to find them" (ibid.).
To assess the nature of conflict over policy as well as the results the interests are able to achieve, HLNS focus on a set of 20 policy-related actions in each domain. All of these actions occurred during the period 1977-82 (covering most of the Carter administration with a Democratic Congress and the early years of the Reagan administration with a Democratic House and a Republican Senate). Representatives answered several questions regarding the policy actions in their domain: whether it was conflictual, nature of interest activity regarding the action, their employer's or client's position on the action, whether their interest in the action was regarding the central issue or more narrow, technical questions, and the degree to which their principal achieved its objectives on the issue.
HLNS find that representatives vary substantially in their centrality to policy actions. Many are involved only in fairly narrow, technical questions; interestingly, this tended to be true of lawyers. Other representatives are more central in the policy process, particularly those with more connections to the notables, those working for organizations with substantial resources, and/or those who specialize in a substantive area. Generally, who is and is not active on specific issues tend to group within each domain, indicating the relatively parochial nature of many policy questions. For example, in the health area, there are food and drug issues, finance issues, issues affecting universities, and issues affecting special client populations.
*774 When HLNS turn to the conflict over specific issues,(29) they found clearly recurring cleavages among interests in each domain. However, the nature of the cleavage differs. Only in the labor domain, is the cleavage regularly along the broad ideological division between labor and management. In the other domains, while there are recurring lines of division, the issues producing those divisions tend to be "technical rather than ideological," reflecting the fact that interest group positions are "highly specific to the issues" rather than reflecting broad positions (pp. 343-44).
HLNS analyze the representatives' self-report of success in the five policy actions in which they were most involved.(30) They apply multiple regression and find that some factors, such as number of notables known, number of years in government, and who employed the respondent, do predict success; however, they conclude that the most important result of their analysis is how much of success they cannot explain (89% in the overall analysis and 78-84% within each domain). They suggest that their inability to account for more of success is because "the determinants of success are usually situation specific" rather than broad factors such as partisan affiliation, organizational resources, lawyering skills, and the like. It also may be the case, they observe, that "meaning of success is uncertain or ambiguous"--it may be a matter of perception; they come to this conclusion because 21% of the instances when they have more than one report from an organization on a particular policy action, the representatives differ substantially in their assessment of the outcome (pp. 351-52).
Regarding the influence of individuals, HLNS conclude that, at least reputationally, influence is not well explained by broad characteristics such as career experience, political activity, work performed, or organizational ties. The variables that best predict a reputation for influence, a track record of success and connections to the notables, might themselves be seen as measures of influence rather than explanations of influence. Does a reputation for success lead to success, or does success lead to a reputation for success, or do they really measure the same thing? While one can talk about "Washington heavyweights" (as HLNS do in the opening pages of the book--more about that below) and one can construct lists of those recognized as prominent players on the Washington scene, HLNS find that the ability of these eminences to achieve actual results for their clients varies by time, by domain, by political situation, and by nature of the issue.
*775 Many political scientists would say that winning and losing is usually a matter of having the votes. Having data spanning two administrations (and differences in congressional control) allow HLNS to look at the question of whether liberals were more successful under Carter and conservatives more successful under Reagan. To some degree this proves to be the case, but there are substantial inconsistencies (e.g., conservatives' success went down sharply in the health domain under Reagan compared with Carter, while the liberal score went up slightly--see p. 353). Part of the inconsistency arises from playing offense versus playing defense: it is easier to win when you are trying to maintain the status quo than when you are trying to accomplish change. More important is the general "uncertainty both in the nature of the decision process and in the outcome of issue events," combined with the fact that most issues "do not significantly involve broad principles of economic ideology, except in the labor policy domain" (p. 358).
One last explanation for success that HLNS examine is what I would term the "meta-issue" addressed by the event. Drawing on work by Kingdon and by Polsby on agenda setting, HLNS categorize events as significantly identified with the ideology and programmatic priorities of the president (administration politics), as arising from systemic demands or crises, or as reflecting ongoing relatively routine issues within a domain (domain politics). They find that the mix of issues in their events vary over time (administration politics was a bigger share of the events during the early years of the Reagan administration than in the later years of the Carter administration) and by domain (reflecting administration priorities and external events). Those proposals stimulated by systemic demands are most likely to succeed, followed by proposals stimulated by domain politics (particularly in domains that are dominated by narrow specific interests, as in the agricultural domain where 11 of 13 proposals succeeded). The proposals least likely to succeed fell into the administration politics category, even in the early years of the Reagan administration when it at least appeared that the administration was getting through a lot of major proposals.
In the end, HLNS emphasize the uncertainty inherent in the policy process. This uncertainty flows from a variety of sources ranging from such things as the expanded significance of federal policies to the pervasiveness of countervailing power (pp. 370-93). All this takes place, according to HLNS (pp. 395-96), in a political context where neither private nor government power is highly concentrated--in a pluralist system.(31) This lack of power concentration leads to the potential for sudden change, "even in *776 highly institutionalized arenas"--the Tax Reform Act of 1986 being an excellent example.(32)
How do social scientists come to create the zero-order texts that both are the focus of the analysis and at the same time create the boundaries for the possibilities for analysis? This is an important question if we are to understand the analyses and interpretations that constitute the published literature of empirical social science. It is almost trite to observe that research projects and the analyses they lead to do not appear out of thin air. Any large project (and most small projects) needs to be understood in the context of how it came to be. Projects can reflect an ongoing development in thinking from a prior project, accidents of conversation or reading (i.e., the juxtaposition of several disparate articles or books might strike a chord that suggests questions, answers, or sources of information/data), personal relations and the intellectual orientations the participants in those relationships bring to the research, and/or the exigencies of funding opportunities. Interestingly, most of these elements are important for understanding the Washington representatives study.
Understanding the context in which HLNS developed and executed the Washington representatives study is particularly important in arriving at a broader understanding of the analyses and conclusions described above. The research reported in The Hollow Core was a project of a group of scholars centered at the American Bar Foundation (ABF), and the ABF provided much of the funding for the research. Furthermore, before embarking on this project, these and other ABF scholars had conducted a wide range of work on the legal profession, much of that work appearing as books published during the 1980s: Chicago Lawyers (John Heinz and Edward Laumann),(33) Partners with Power (Robert Nelson),(34) Beyond Monopoly (Terence Halliday),(35) and From Patrician to Professional Elite (Michael Powell).(36) The work on these (and other) research projects influenced the questions asked, the data collected, and the analyses performed by HLNS.
*777 In this section I present a reconstruction of how The Hollow Core (particularly the analyses I summarized in the previous section) came to be. I start first with the evolution of the project and how that came to define the data set out of which HLNS drew their analyses. I next set out how the data themselves constrained both the analyses and the possible findings. Finally, I turn to my reconstruction of how HLNS arrived at their conclusions, focusing on the role of narratives and metaphors, and including what I believe are some underlying and unstated metaphors that shaped HLNS's conclusions.
To understand the conceptualization of the research project, one must start with Heinz and Laumann's earlier work on the Chicago bar. As the authors of The Hollow Core note, the idea of the Washington study arose from a conversation that they summarize as "now that the Chicago lawyers have been dissected, wouldn't it be interesting to look at Washington, where by all accounts many lawyers are real movers and shakers who get the federal government to adjust public policy in behalf of private interests?" (p. xiv). Aspects of the dissection of the Chicago bar can be found in The Hollow Core (even though HLNS early on came to realize that a study of policy advocacy in Washington could not be limited to lawyers, and eventually that the role of lawyers in policy advocacy was actually relatively small):
Another starting point is the variety of backgrounds, and combinations of backgrounds, among the group of scholars who undertook the Washington project: sociology (Laumann and Powell), sociology and law (Nelson); law plus some political science (Heinz), and political science (Salisbury). They were brought together both by their intellectual interest in lawyers and/or interest groups,(39) and by the interpersonal connections of prior collaboration, and of mentor and student.(40) The starting point for HLNS(P) was the juxtaposition of interest representation and lawyers. The big question was where to go from there.(41) The preferences of the collaborators reflected their prior interests and backgrounds:
Throughout the early discussions, a central question was whether the focus of the research should be on Washington lawyers and their role in the policy process or on Washington representatives and the role of lawyers within that broader group.(47)
As discussions progressed, HLNS(P) agreed that both lawyers and nonlawyers would have to be included, although the relative balance between the two groups was unclear. They realized that roles of lawyers and nonlawyers could be quite diverse in the policy process, including advocacy, advice, monitoring, and influencing, and that these roles could be organized around substance or process and/or around clients or agencies. They also agreed that the research would have to focus on several specific policy domains.
To finance some early fieldwork needed to clarify both substantive and methodological issues, the collaborators obtained a small internal grant *780 from the American Bar Foundation (ABF).(48) The proposal to the ABF emphasized lawyers and their impact as interest representatives on the formulation and implementation of policy.(49) HLNS(P) explained their interests in terms of the long-standing recognition of the importance of lawyers in the American political system (citing classic statements by writers such as Alexis de Tocqueville and Harold Laski) along with developments such as the tripling in the size of the D.C. bar between 1973 and 1981, the sharp growth in D.C. branch offices of major law firms, the growth in the number of lawyers employed by the federal government, and the increase in regulatory activities by the federal government.(50)
HLNS(P) were already advancing tentative hypotheses or questions:
The goal of the first phase was not to test these, or other, hypotheses; rather, as one member of the research team described it, this was a "soaking and poking" phase.(51)
The early trips involved interviews at which most or all of the team gathered in some person's office to discuss both the substance and process of *781 research on interest representation.(52) These sessions were a kind of reverse focus group: instead of a facilitator interviewing a group of respondents, it was a group of researchers interviewing a single respondent, "picking their [sic] brains" in the words of one member of the team.(53) In the initial trips, all of the researchers who were in Washington went to the each of the interviews; as time passed, and the project began to take focus, the researchers split up to increase the number of exploratory interviews they could complete. During these sessions, HLNS(P) tried specific questions of the type they planned to include in later interviews, and they explored whether it made sense to pursue issues from the earlier Chicago study, such as the role of social stratification both among the lawyers and among the nonlawyers, and whether the degree of stratification differed between these two groups.(54)
Despite the recognition of the importance and role of nonlawyers, the emphasis in the project remained on lawyers. This was reflected questions posed in several memos from this period:(55)
Refined hypotheses began to emerge, and these began to focus attention on comparisons between lawyers and nonlawyers:
During this period, the collaborators realized that underlying these questions and issues was a set of three dimensions that could be seen as underlying the nature of Washington representation:
HLNS developed a two-by-two table to capture the link among these dimensions and comparisons between lawyers and nonlawyers (which I would describe as a fourth dimension):(56)
| EPISODIC | RECURRENT/ROUTINE | |
| FACILITATIVE | Nonlawyers Policy Development |
Nonlawyers Bureaucratic |
| ADVERSARIAL | Lawyer Generalists Policy development |
Lawyer
speccialists Routine |
From the preliminary trips and other exploratory efforts, the principals realized that some sequencing would be important in the research process. While it was relatively easy to identify groups with interests in federal government policy decisions, it was less easy to identify who represented those groups in the policy process. Sometimes representation was by members of the groups while at other times it was by hired specialists. Existing lists of lobbyists were simply not a satisfactory sampling base. The sampling issue became central to the later analyses because it would define who and what was actually the subject of the study.
Throughout the period during which the substantive issues were being clarified, HLNS kept thinking about a research design that could produce the kind of data needed to address those issues.(57)
After much debate, they arrived at a research design which included interviews with:*783
With the design finalized, HLNS were ready to undertake the primary data collection.
By the time funding for data collection was in hand, HLNS had come to recognize that the study would have to go beyond Washington lawyers. Nonetheless, the interest in lawyers still was the driving force of the research. HLNS wanted to understand Washington lawyers as representatives rather than on Washington representatives more generally;(58) the nonlawyers were to be included primarily as a group to contrast with lawyers. The proposal to the ABF board requesting funding for the primary data collection described the principal research tasks as "to examine the diversity of lawyers' roles in the Washington context, to determine whether there is a differentiation by role or by institutional setting, and to establish the ways lawyers and non-lawyers differ in the performance of the roles that are not peculiar to lawyers." The sample design made the emphasis on lawyers clear: in each of the four policy domains (by now determined to be agriculture, *784 energy, health, and labor), the researchers would draw random samples (from lists they would build from a variety of sources) of 60 government lawyers active in the policy domain, 150 lawyers in private practice active in the policy domain, and 60 nonlawyer representatives active in each policy domain. In terms of the number of final respondents, HLNS(P) expected to have data from 700 lawyers and 200 nonlawyers.
However, even before the actual data collection started, the shift from a focus on Washington lawyers to a focus on Washington representatives began. The problem of working out final details of sampling procedures pushed HLNS to the recognition that "the emphasis of the project is on Washington representation, as opposed to a study of the D.C. bar."(59) On the target side of the interest representation process, the subject shifted from government lawyers to government officials. This, of course, did not mean that the authors were no longer interested in lawyers.(60) However, as the authors dug into the data, and the relatively limited role of lawyers became increasingly clear, the emphasis shifted still more. The realization of the relatively small role of lawyers first explicitly surfaced in discussions indicating that it might be necessary to supplement the original samples to obtain enough lawyer respondents to permit the comparisons between lawyers and nonlawyers.(61) No supplementation was done, but in the end, only one third of the representatives in the sample were lawyers (p. 76) and only 34% of the government official targets of these representatives were lawyers (computed from table on p. 223).(62)
A brief descriptive essay published by Salisbury in an edited collection on interest groups first reported the relatively small proportion of the sample who had law degrees (34%) and went on to note that the "fraction is not much different from what Milbrath found in the 1950s, but there is some reason to think that lawyers do not hold quite the dominant position they did in the past."(63) By the time of the first extensive published report of *785 the data in 1987, "Private Representation in Washington,"(64) the lawyer element had become secondary to the more generic Washington representative focus. This article clearly reflected the authors' surprise that lawyers were not more dominant among the representatives they identified: "Given the rapidly increasing number of lawyers in the District of Columbia, we might expect lawyers to be the predominant group among representatives. While lawyers are by far the largest professional group, however, only a third of the sample hold law degrees."(65) Milbrath's 1950s study of registered lobbyists found that 43% held law degrees; among the registered lobbyists in HLNS's sample, 41% hold law degrees.(66)
The shift in emphasis from Washington lawyers to Washington representatives was clear in the later articles published by HLNS. One article did take a "lawyer focus," considering careers, tasks, etc.; probably the key finding of that article is that "lawyers occupy a relatively specialized niche in the system of interest representation . . . [which] limits their influence in policy formation."(67) The other articles emphasizing more traditional political science concerns (interest group alliances,(68) the use of experience by lobbyists,(69) and elites in the policy process(70) appeared in major political science journals rather than in law journals.
This takes us to what is customarily thought of as analytic phase of a research project. The project has been conceptualized, and the data have been collected. However, as the history above makes clear, the analytic element is crucial in getting to this stage. Even as the design was being developed, *786 HLNS were realizing that many of their suppositions about the world of representatives were wrong or irrelevant. These realizations reflected informal analyses of the data they were collecting, either from the "exploratory" phase or from sample characteristic data or from minimal tabulations from completed interviews.
In an previous essay on analysis in qualitative research, I observed that the demarcation between data collection and data analysis did not hold up in that context because analysis started as soon as the researcher had any data in hand (i.e., from the time the first interview or observation or document became available). I argued that rather than dividing the research/analysis process between data collection and data analysis, it made more sense to distinguish between data acquisition and data review, recognizing that data analysis occurs as soon as there is any contact between the researcher and data. In a quantitative project involving original data collection, the distinction might be labeled data acquisition and data sifting, again recognizing that researchers will be analyzing what's happening even as the data come in (e.g., what is the implication of the number responding, who is responding, etc.). The type of analysis that occurs in the data acquisition phase of quantitative research probably differs more from the analysis in the data-sifting phase than the qualitative researcher's analysis differs between the data-acquisition and data-review phases.(71)
At one level, to ask what the role of data is might seem absurd. Empirical analysis requires empirical data, and those data are the center of the analysis. On the other hand, one might posit that empirical data are simply the starting point for analysis. The reality is that the data are more than the starting point and less than the central focus of the analysis. For these reasons, I prefer to think of the data as framing or limiting the analysis. That is, data limit the range of what the analyst can look at while at the same time suggesting points of analysis or conclusions.
At one extreme, data can stand up and hit the analyst over the head. The clearest example of this in The Hollow Core is the issue of the role of lawyers in the interest representation process. HLNS began their research with a central focus on lawyers. They quickly realized that nonlawyers would also have to be considered, but through the design phase, the primary focus on lawyers persisted--I say this based on their original image of the samples as being dominated by lawyers and on some of the comments HLNS made to me in discussing their analysis ("we were very surprised by the low percentage of lawyers in the sample"; "we had a hell of a time finding *787 lawyers"). Only when they conducted their first survey of client organizations and began to assemble the sampling frame for representatives did they realize the relatively small role of lawyers.(72) As I noted previously, at one point HLNS became sufficiently concerned about the number of lawyers in their sample that they developed contingency plans for oversampling lawyers if that became necessary to ensure that at least 30% of the representative respondents were lawyers.
One might be tempted to describe the above as an example of letting the data speak. Yes, but data speak only when asked:(73) If the analysts had not asked about lawyers, then the relatively small number of lawyers in the data set would not have been necessarily remarkable.(74) In fact, one could just as easily ask why so many of the representatives are lawyers? The conclusion that 30% is a relatively small proportion is driven by the authors' original expectation, not by something inherent in the data themselves. To continue the assault metaphor, for the data to hit the analyst over the head, the analyst must be in the room where the weapon is located--the weapon does not seek out the analyst, the analyst seeks out the weapon.
As the "fact base" for the analysis, data constrain the analyst's options, both analytically and substantively. At the crudest level, if the collectors of the data did not ask a question (or asked it in an ambiguous or confusing way), the analyst is limited by the data available. A very good example of this limitation in HLNS's analysis revolves around the "hollow core" itself. The data that served as the key base for that analysis were responses by the representatives about their relationship with the notables--would the notable "take the trouble to assist [the representative] briefly without a fee if requested?" Fundamentally, the smallest space analysis looked at the relationships between representatives and notables and used those relationships to place the notables in a three-dimensional space which HLNS found to resemble a sphere with a hollow center. They conclude from this analysis that "the four policy domains lack core actors" (p. 298).
*788 All that HLNS can in fact say from this analysis is that there is no group of interest advocates playing a core or mediational role. That is, no interest actors appear to be power "brokers" in the sense that the term "broker" refers to someone working to bring disparate or competing interests together. HLNS's expectation that they might find such a group flowed from finding such a core group in their analysis of Chicago Bar's notables network.(75) Theoretically, they were led in this direction by their interest in professionalism and the idea that the "autonomous" professional might rise above the conflict of the moment to play such a brokerage role.(76) If such autonomy did exist for Washington lawyers, this would be in contrast to what they found in their study of the Chicago bar, the autonomy envisioned by the sociology of the professions does not necessarily work in the way the theory suggests, and may not exist even at the top of the professional ranks.(77)
In fact, HLNS's data in one sense do support the idea that independent professionals enjoy a level of autonomy, but at the same time the data show that this does not lead to the mediational function suggested by the professions literature. Respondents were asked, "Have you ever had occasion to refuse a potential client or work assignment, not because of a formal conflict of interest, but because of your personal values?" Only 14% of internal counsel (inside lawyers) reported having refused an assignment for this reason; this figure is very close to what Nelson found when he put the same question to lawyers in four large Chicago lawyer firms.(78) However, outside lawyers and consultants are much more likely (over 60%) to report having refused such assignments (p. 187). However, as HLNS discovered, the outside lawyers' and consultants' autonomy really was not all that important in the policy representation process simply because they in fact played a relatively small role in the overall scheme of things.
*789 While there may be no core or mediational actors among those representing interests in the policy process, this does not mean that there are no such actors in the overall process. However, HLNS refer to the policy domains lacking core actors, not just core representatives, and the authors' discussion fails to make clear whether their reference to "actors" is meant to apply actors generally in the policy or simply intended as a synonym for representatives.(79) The possible confusion created by not making this clear is complicated by other analyses they report which did show core actors in terms of the targets of influence (see pp. 206-15). In fact, HLNS note that it may be the case that "government officials act as mediators who bind the system together." While their data preclude assessing that possibility (p. 302), HLNS conclude that "American political elites have only relatively narrow constituencies and do not act cohesively." Furthermore, there is little evidence of elite autonomy because of the strong links to interest networks and client organizations (pp. 304-5).
If the analysis had extended to policy elites, such as members of Congress, senior committee staff, senior staffers within agencies, etc., HLNS might have found that in at least some of the domains those actors do function as a core. HLNS note that this might be the case, and acknowledge that "our data do not permit us to address this directly since government officials were not included on the list of notables." The problem is not just that HLNS included no government officials on the list of notables; the structure of the question used to generate the data for the analysis was entirely inappropriate for use with government officials. One could have imagined a different questioning strategy--for example, "Which of these people would return your phone call if you tried to contact them to discuss a pending policy decision?"--that would have permitted the inclusion of government officials as well as representatives on the list of notables.(80) HLNS did not do this, and the data they had limited the analyses and conclusions that were possible.
*790 The framing or limiting role of data is also evident in HLNS's analysis of success. The data they use for this analysis consist of the representatives' reports of whether their client or organization's objectives were reached on each of five issue events in which they had the greatest interest (p. 345). The focus was on individual events rather than on achieving a broader set of policy objectives (although the notes of a meeting from 1987 make it clear that they realized that event success is "relative to the scope and particulars of a client's goals related to an event."(81) There is nothing wrong with focusing on specific events, but doing so offers only one type of approach to understanding when and why interest groups succeed in the policy arena. HLNS could have assessed success differently, focusing on some broader or alternative notion. For example, rather than identifying events, the respondents could have been asked to identify key policy goals, both in terms of changes desired and existing policies to be maintained; from these, they could have been asked to rate their success in terms of the goals. HLNS's definition of their data in terms of events rather than in more general policy goals limits the analysis of success that they present in chapter 11.(82)
While my discussion of specific examples of the limiting or framing role of data has been in terms of suggesting alternative questions, analyses, and/or measures, I do not intend this as a critique of what HLNS did. My point, rather, is that data always limit the analysis in important ways. In fact, as soon as the analyst begins to define the data for analysis, the analyst has framed or limited the scope of what will or can be looked at. There are a variety of relationships the data can have with the analyst, most of which reflect the limitations the data impose.
Every analyst hopes that the data will be his or her friend and produce the results that confirm the expectations that the analyst brings to the research. When the data act as a friend, we find the expected results with relative ease. For HLNS, the data as friend is clearly evident in the analysis of the representatives' careers. This does not mean that we can change the question (or the rules of the game) whenever we want and the friendly data will simply go along with us. As in a friendly game of baseball, where the kid who owns the bat can take that bat and go home, ending the game, the data will let us push them around only so far before refusing to cooperate further. Furthermore, even when the data are friendly, we may come to realize that there are other things we want to consider, but the data do not permit us to do so.
*791 More often, the data turn out to be a tease. By this I mean that the analyst can find results that are supportive of the hypotheses or expectations. In The Hollow Core, data as tease was evident in the factor analysis of the various tasks the representatives performed; as I will describe in the next section, the authors struggled with the factor analysis results for some time before they felt they understood what those results meant. When the data play the role of a tease, we have to work hard to extract the desired results. We may ultimately be able to argue that our theory is correct, or our results are clear, but it may take a lot of effort to get to that point, or we may have to make our arguments in a muted fashion because the patterns are weak or obscured by other things going on in the data. Alternately, data as tease may entice us to look deeper at the data in ways that we had not originally anticipated.
The data can stand as a gatekeeper. In this role the data do not tell us whether we are right or wrong. Instead, the data kind of yawn at us; the question we are asking is irrelevant as far as the data are concerned. After we think about it for a while, we may come to realize that we have committed what some call a "Type III" error: we have asked the wrong question.(83) This role is not typically evident in the written product of the research, because analyses in which the data play the role of gatekeeper seldom see the light of day. An example from the Washington representatives study which becomes clear only by looking at the original data collection instrument is the revolving- door image of movement between government service and work as a representative. HLNS designed their questionnaire to get detailed information on career patterns, including up to five prior positions in government service, five prior positions in the current organization, and five prior positions in other public or private organizations where the respondent worked in a representative capacity.(84) When the data were in, HLNS discovered that most representatives had very stable careers, with few of the career moves they had anticipated in designing their survey instrument.
The data can play the role of the enforcer. Here they limit us because of the decisions and choices we have made in collecting the data. When HLNS made the choice to ask about their respondents' contacts among only the group of notables, they precluded the possibility of looking at the potential of a mediational role for other players in the policy process. The data in turn enforce this limitation on the analysis. The authors may respond that their interest was only with regard to the roles played by the private interests, *792 and the data enforce this stricture, even when the reader asks, What about the role of the government officials?
Lastly the data can play the role of the opponent. They can tell us things we were not looking for, and perhaps things we did not want to find. In the case of The Hollow Core, the data told HLNS that lawyers did not have the central role in the policy process that HLNS had expected. The result of this very early finding was that HLNS had to recast their interests away from the role of lawyers as advocates in the policy process to broader questions about that process.
What is interesting about the roles I have delineated is that they reflect the interrelationship between the data and the analyst's theory. The data come to play the role they do in large part as a function of the fit between theory and data, both in terms of the correctness of the theory and in terms of the relevance of the theory. When the theory is wrong, the data will tend to play the role of opponent and/or enforcer; when the theory is of no or marginal relevance, the data will tend to play the role of tease of gatekeeper.
In published work, we see only the final iteration of the data analyses that the authors carried out. All experienced analysts know that many, many other analyses underlie the final presentation.(85) Most of the preliminary analyses are simply discarded because of errors, or uninterpretable results, or some other problem.(86) Others help focus the analyst's attention more toward alternative specifications or different types of analysis.
Retracing these analyses is difficult, if not impossible. Discarded results typically disappear into the electronic ether (once upon a time, such results typically had at least some paper life because analysis results could be examined only as computer printout). Intermediate results may exist as either electronic files that the analyst visually examined and retained or as printed results filed away in some manner. In group projects, some of these intermediate analyses may be documented in the form of internal memos, either from analysis staff to senior investigator or from one investigator to another. Later versions of results may appear in convention papers that serve as drafts of eventual articles or book chapters.
As HLNS developed their analysis from their data set, there was little in the way of memoing, but there were a number of interim papers and *793 drafts of papers ("there were some memos . . . but not so much memos as drafts. There were drafts of papers or pieces of papers that were exchanged"), both presented at meetings and published. In some large projects, the senior participants are relatively far removed from the day-to-day work of data analysis. Such was not the case in this project. HLNS did employ a number of staffpersons as data analysts, and while the degree of "hands-on" involvement with the data varied (with Nelson probably being the closest), all the principals were closely involved in working with the data in one way or another, making key decisions about coding, analytic methods, and statistical indicators.
One of the sets of results reported by HLNS is the factor analysis of the respondents' reports of the importance of each of 18 tasks in their day-to- day work (pp. 98-100). While the presentation of these results is matter-of- fact ("[t]he four factors are readily interpretable"), arriving at those results was not easy. Nelson oversaw this analysis and he reports that he and the data analyst "made several attempts" and that "the first set of attempts did not work."(87) A close examination of the results of the analysis indicates its problematic nature. The factor analysis accounted for 46% of the total variation in the 18 items (p. 416 note 6), but the fit of individual items to the factors varied from a high of 70% ("preparing official testimony") to a low of 22% ("resolving conflicts within organizations"); 5 items were in the 20-30% range, 1 in the 30-40% range, 4 in the 40-50% range, 5 in the 50-60% range, and 3 in the 60-70% range.
How did the researchers make sense of these results? They needed to identify and describe some pattern, and the key to the pattern identification appears to have been associating each item with a single factor (i.e., creating mutually exclusive item clusters). Interestingly, in a discussion in an earlier chapter of the types of representatives that organizations reported using, HLNS used 14 similar items to describe the tasks the organization needed done; there they spoke of "activities most clearly linked to lawyers' skills," which included not only litigation (which is the lawyers' dimension in the factor analysis of the representatives' reports of their activities) but also "drafting legislation or regulations" (p. 64). In the factor analysis of the representatives' reports of activities, they identified factors and associated clusters of activities as "government regulations, interest group networks, public presentation, and litigation"; in the earlier discussion of organizational tasks, they had four different clusters (congressional activities, executive branch and regulatory activities, activities in court, and organizational *794 activities). In both analyses, the key tool of pattern identification appears to have been identifying mutually exclusive categories.
Relying on mutually exclusive categories was a type of simplification of complex and ambiguous information. Some method of simplification is often central to identifying patterns. However, simplification can also distort. This is most clear in the item HLNS labeled "drafting legislation or rules." While the authors associated it with the "government relations" factor and cluster, the coefficients they report indicate that "drafting" was also clearly associated with "public presentation" and with the factor they labeled "litigation" but which might have been labeled "legal." In fact, its loading on the public presentation factor was marginally higher than one of the items placed in the public presentation factor.
This factor analysis represents one of the points where we can readily see the importance of interpretation. The early project documents make it clear that the authors did not set out to find that lawyers played a distinctively "legal" role in the policy advocacy process. Rather, their early thinking was that "lawyers aren't special . . . there's nothing special about lawyers." As a hypothesis, HLNS entertained the possibility that lawyers might bring key professional values to their policy work that could have major implications for how politics played out in the Washington arena. The early analysis quickly revealed that their original expectations did not hold up: lawyers "were distinctively lawyerly" and "there were few institutional [in the policy sense] arenas where they dominated." In retrospect, it is not surprising that the two court-oriented activities stand out as relatively distinct because the courts erect barriers to exclude those who are not lawyers. For none of the other activities do equivalent barriers exist.
The analytic consequence that resulted in this negative finding was a search for patterns in the data that had theoretical coherence, and efforts to make sense of (i.e., interpret) the patterns that did appear in the data. The approach HLNS used in their pattern search was to apply a variety of different methods of analyze their data.(88) The most dominant method, not so much because of its frequency as because of the visual impact of the graphics produced from it, is smallest space analysis.
An alternative (or complementary) approach the authors might have employed (and did at least to some degree--perhaps to a greater degree than is evident in the book) would be to "perform the data," a phrase I adapt from the performing arts to suggest that interpretation might be aided by pushing the data intentionally beyond the limits of what is normally deemed to be appropriate. For example, from all indications, the factor analytic*795 methods used by the authors were the standard exploratory methods of extracting factor solutions and rotating the initial loadings according to some standard criteria (in their case, at least in the published report, the criteria applied was "varimax" which seeks to distinguish sharply among factors that are held to be independent of one another). A more radical approach might have been to apply confirmatory techniques that explicitly varied the nature of the factor structure being imposed; this is somewhat more labor-intensive because the imposition of structure requires detailed specifications. However, seeing how the results varied under forced specifications might have served to clarify the structure of relationships and lead to alternative interpretations.
In a sense, this would involve letting the data "perform" through something akin to improvisation (a technique often applied by performers in their development of interpretations). Much as a director might suggest to an actor an improvisation exercise to assist the actor in finding an appropriate interpretation of his or her role, the data analyst can artificially impose structure on some data to see how the data perform under that structure. The goal here is not to use that structure as the reported analysis but rather to find nuances in the data that might otherwise escape the analyst's attention (just as improvisation can assist an actor to grasp the nuances of a dramatic text).
The interpretation of the spatial results did involve at least an element playing the data. Using a computer graphics program, some of the authors looked at several of the three-dimensional smallest space solutions visually using a computer program.(89) This program made it possible to view solutions from a variety of perspectives (visual not theoretical), and it was as they looked at this that they began to recognize the characteristics of some of the structures. This technology became available after the authors had completed much if not most of their analyses, but it did aid in interpreting some of the results of their analyses of the outcomes of policy conflicts (the material presented in ch. 11).
A comment of one of the authors about playing with the smallest space solutions is worth noting. In discussing the interpretation of the smallest space analysis using the graphics rotations, that author noted:
I got worried to some degree about [the] extent we were picking and choosing by doing this. It's one thing to use the dimensions that you are given by the program and it's another thing to say if you happen to look at it from right over here it looks likes this. I [became] concerned about the level of choice that we were exercising over the presentation because if you looked at it from this direction, you might be able to *796 make this argument, but looking at it from another direction might lead to a different argument.
This concern partly reflected the need to limit the number of data points included in the graphic presentations--including too many points would create clutter that would obscure the pattern. The authors wanted to ensure that their choice of which to include or exclude resulted in a fair representation of the full set of points.
One of the most interesting aspects of The Hollow Core is that the end result of the interpretation process is easy to see: the analyst seeks to arrive at stories (narratives) and metaphors that create a coherent picture both for the analyst and as something to communicate to an audience. HLNS open the book with a specific story, "The Lawyer and the Heavyweight," a story that captures a good bit of what they eventually want to convey but also provides something for them to eventually reject. The thrust of this story is as follows:
An oil company was seeking a revision in a regulation. The company retained a lawyer who knew his way around the agency which had jurisdiction; the lawyer found the right person (the "bureaucrat"), painstakingly made the case, and was awaking a decision. Another company, also desiring this regulatory change, got impatient and retained a heavyweight Washington insider. Mr. Heavyweight made a few phone calls for which he collected a big fee. Unfortunately for the oil companies, the intervention of the heavyweight raised all sorts of warning flags, and rather than completing the change, the bureaucrat decided he had better look at the matter closer if it was something of interest to Mr. Heavyweight. In the end, the change went through, but Heavyweight's impact was to slow things down rather than to facilitate the change.
This story sets up several major points of analysis: the role of stature, the role of lawyers, the role of career experience, and the uncertainty of the policy process. At the same time the story challenges conventional wisdom about power and influence.
While the story is interesting in its own right, it also provides an important insight into the link between metaphor and narrative. Much of power of the narrative is created by the metaphorical use of "heavy-*797 weight."(90) In fact, this linkage of metaphor and narrative is a common interpretive tool. Specifically, as I will show below, analysts often start with a metaphor that captures central aspects of the phenomenon of interest, develop a story to amplify and explicate the narrative, discover that the story casts doubt on the starting metaphor, and conclude by developing a new metaphor to encapsulate the thrust of the analysis. In the discussion that follows, I use questions from each of the three areas of analysis I previously summarized to illustrate the interconnection of metaphor and narrative: the career paths of representatives, the interrelationships among representatives, and the impact of representatives on policy outcomes.
The analysis of the role of government service in the careers of representatives starts with a standard metaphor, progresses through a narrative of the career pattern, and concludes with what I would describe as a new, although not clearly stated, metaphor. The standard metaphor in discussing interest group representatives' careers is that of the "revolving door": representatives revolve (move back and forth) between government employment and working on behalf of interest groups. While others have cast doubt on the usefulness of the metaphor as a basis for understanding policy decisions,(91) it has continued to be an important metaphor in the image of lobbyists. HLNS show clearly that only a slight majority (55%) of representatives have an experience in government service and a sizable chunk (9%) have experience only at the state or local level (p. 117), and that the vast majority (more than 80%) of those with federal government experience moved between government and private representative positions only a single time; they didn't "revolve [rather they] walk[ed] through the door once" (p. 119). Among the representatives in the survey, only 51 of 776 "moved between private representative positions and federal government jobs more than once" (ibid.). Further analyses show that these 51 do not stand out as particularly influential (only two could be characterized as generalist heavy-weights) or even as drawing heavily on their past government experience in their current positions (less than half had held government positions that dealt with matters affecting organizations for whom they worked either before or after their spells of government service).
Once HLNS reject the revolving-door metaphor, their next step is to construct a story about how and why government service might be of use to *798 a representative. Here they examined the "who you know versus what you know" issue by asking the representatives who had previous government experience and who believed that government experience had "been significantly helpful" whether each of the following had been helpful: familiarity they had gained with the issues, contacts they had in the administration, contacts they had in Congress, knowledge of the decision-making process, and contacts with other representatives. Not surprisingly, the location of helpful personal contacts depended on where the respondent's government service had been (i.e., with Congress or in an administrative agency). However, knowledge of the decision- making process itself was the standout in importance. The key to their story, however, is that the usefulness of government experience, both people knowledge and process knowledge, is directly related the recency of that experience. The value of experience decays over time, and the nature of that decay varies with the way in which the experience was useful:
HLNS conclude from their analysis of the role of government experience in the work of representatives that such experience is clearly both frequent and useful but far from universal. What you know is more important than who you know at least in part because issues and processes change less rapidly than to who holds particular jobs. Still, even the issues and processes change over a long enough time frame. Congressional experience is more important than executive experience because it provides broader knowledge and provides it more quickly. I would speculate regarding this last point that congressional experience might have been more valuable during the period of HLNS's research at least in part because of greater stability in Congress than in the executive branch; that is, the experience gained on the "Hill" may have decayed more slowly than did executive branch experience.
HLNS's analysis suggests a replacement for the revolving-door metaphor: the "half-life of government experience." The half-life metaphor captures three of the important factors in the importance of government experience. First, all types of government experience tend to decay over time. Second, the speed of the decay depends on the nature of the experience (in the same way the rate of radioactive decay varies across types of *799 atoms). And third, the decay differs by individual, with some individuals experiencing faster decay than others for more or less random reasons (just as some atoms of a given isotope decay before others for no apparent reason).
The metaphor that was the foundation of HLNS's original expectation about the nature of representative networks was that of a wheel and spokes with a group of highly influential heavyweights standing in the middle to mediate among various interests. More generally, they expected to find that outside lawyers (and possibly other outside professionals) could serve something of a mediating role, reflecting professionalism which in turn leads to a commitment to the public interest and autonomous decision making (p. 157). The effect of such mediating groups would be to smooth over conflicts that arise from sharply conflicting interests and ideologies.
The story line that the authors present as a starting point is descended from that of the "power elite" theorists such as C. Wright Mills;(92) many of the more popular books on Washington lawyers also reflect this view.(93) According to this story, there is a core group of heavyweights constituting an "inner core" (p. 8-9).(94) Over the course of the book, HLNS abandon that story and shift to what might be called a "birds of a feather" story: Representatives tend to know and talk to others of like outlook and like interests. Groups tend to recruit representatives from among those groups who have traditionally been allied with the interest (e.g., business interests tend to favor upper status Protestants while labor interests draw heavily from among Catholics and Jews--see pp. 147-48). The birds of the feather story extends to patterns of interaction among representatives: there is no central elite through which representatives interact. Collegial ties form most often through the work context, and that work context tends to be focused on a single outlook rather than a broad outlook across a range of perspectives on issues central to the representatives work (pp. 174-76). Of course there are variations in degree, with some organizational settings showing stronger patterns of socioethnic "inbreeding" (p. 178), but HLNS describe the organizational*800 factors structuring networks of representatives as "more consistent and pervasive . . . than party affiliation, ethnicity, and religion" (p. 180).
While there is some evidence that outside professionals (lawyers and consultants) fall ideologically in the middle (p. 157), this does not appear to provide them with the type of autonomy that the professionalism literature (Parsons, Horsky) imagined. Rather, just as Heinz and Laumann found in the corporate hemisphere of the Chicago bar, the dependence on a relatively narrow, stable set of clients limits the ability of these professionals to mediate (pp. 186-88). Furthermore, the partisanship in the system serves to type lawyers as representing one side or another in a given policy setting (i.e., lawyers do not regularly move back and forth, e.g., representing energy companies on one issue and then consumer interests on another).
The partisan shape of the representatives' networks eventually led HLNS to the "hollow core" metaphor. The purpose of this metaphor is to capture the image of the world of representatives. Contacts occur around the periphery of the network surface (which is not so much a basketball-shaped sphere as sphere that has been squashed down a good bit so that the third dimension has much less variation than the other two). Proximity along the surface is the basis of interaction rather than through any type of central hub. The geometric metaphor captures what HLNS see as the fundamental nature of interaction among representatives in the policy process.
As I discussed in earlier sections, this metaphor may not extend to the policy process as a whole, in large part because of the restricted nature of the data on which the analysis that led to it was based. The spherical object includes only representatives, and leaves out the possibility that nonrepresentative actors provide a hub function. The old power elite image extends beyond the heavyweight lobbyist to include both interest actors and government decision makers; HLNS includes only the former. Nonetheless, the hollow core image stands in sharp contrast to popular images of interest representation and the influence of lobbyists.
While one can argue about definitions of politics, in the policy advocacy setting Harold Lasswell's classic description "who gets what, when, and how" constitutes one of the fundamental questions for analysis (pp. 344-46). The standard metaphor for understanding how policy outcomes come to be is that of the game (p. 359); in fact, much of the sophisticated analysis of game theory can be applied to policy conflict.(95) The key elements *801 of the standard policy game are votes (i.e., number of constituents who can be influenced to vote for or against a policymaker or for someone with direct control over a policymaker), money, and access.(96) The policy game is akin to war or football: resources in terms of money and votes count a lot; they provide access to decision makers, and can sometime directly sway decision makers. The image of this game is that there is a fair degree of predictability (which does not mean there is absence of uncertainty).
The story that HLNS develop about policy outcomes builds on the game metaphor but moves off in a direction that suggests very different types of games than the ones governed by power and resources. The statistical analysis of success reported in the book emphasizes how little of success the authors can account for. In a concluding section titled "The Uncertainty of Influence," HLNS observe (p. 359):
[Al]though the outcomes of individual events may sometimes be foreseen, the respondents' overall rates of success in achieving their objectives cannot be adequately explained by categorical variables. We cannot account for much of the variance in those rates even though we have good measures of several variables that one might expect to enhance influence.
The narrative that HLNS build focuses on the uncertainty of the policy process, within a context of substantial stability in the nature and degree of conflict, the coalitional structures, and the role of broader principles (i.e., ideology). That is, the system itself evidences substantial stability, but this stability does not lead to predictability in policy results based on measurable factors such as resources or contacts.
As a result of the narrative they create, HLNS reject the relatively simple game-theoretic models that might be applied to policy. The central element in the system as they see it is uncertainty that arises from a wide variety of factors (pp. 370-93). Taking these factors, and differences in patterns among the four policy domains in their study, HLNS conclude that "policy-making systems must be seen as historically situated social structures" (p. 403). There are "multiple arenas of politics and multiple forms or methods of interest representation" (p. 404).
What, then, is the underlying metaphor for explaining policy outcomes? The authors largely reject the strategic game metaphor with its emphasis on resources and power. The narrative they have developed actually fits very nicely with the game metaphor, but rather than a game governed by strategy and underlying resources, it is a game governed by uncertainty with rapidly shifting resources. Thus, rather than the metaphor of the football *802 game or the war game (where resources are external to the play of the game itself), the metaphor is more that of a card game such as poker or bridge in which new resources are dealt with each hand of the game. Experience and skill are important, but chance still plays a large (and sometimes a dominant) role in the outcome. Furthermore, things like strategies, what counts as resources, etc., vary across different games (poker vs. bridge vs. blackjack) as they vary across different policy domains (and within sub-domains--think of the possible variations on poker in terms of wild cards, face up versus face down, etc.). Furthermore, the rules can shift depending on who is in control of various positions in the game (again, think of dealer's choice in poker), and this can shift advantages in important ways. But regardless of rules and the specific game being played, the dominant element in card games is the role of chance in the way the cards are dealt in a given round, subject to the possibility of cheating in the deal through marked cards or dealing from the bottom (which is not unlike the possibility of bribery or other types of "cheating" in the policy process).
In all three of the examples I have discussed, one can see a movement along a trail that starts with a metaphor, constructs a narrative, and ends with a new or revised metaphor. Each of these analytic/interpretive trails uses the logic of various statistical methodologies to extract from the data suggestions about direction of the trail. These trails are not like the yellow brick road, obvious for all to see who take the time to look. Rather, they are more akin to an overgrown jungle trail with multiple twists, turns, and branches, many of which lead nowhere, some of which lead off in very different (and often interesting) directions, and a few of which lead the analyst generally in the theoretical and intellectual direction that motivated the research originally. Even in the latter case, what the analyst finds both along the trail and at the stopping point on the jungle trail--note that I do not say the "end of the trail"--often bears only a vague resemblance to what the analyst was expected to find when he or she embarked on the trek.
The progress down these analytic trails can be relatively quick or relatively slow. For example, did HLNS come quickly and easily to the realization that the role of outside lawyers was relatively narrow? As I noted previously, the small number of lawyers in the sample was something that leapt out at the authors even as the data collection was ongoing. One might think that this pointed clearly down the interpretive path that lawyers were simply not as important as originally expected. In fact, this realization was not "a Eureka kind of thing," but something that HLNS came to see "over a period of time" after looking at a variety of elements of the data. Similarly, realizing that the representatives tended to be highly specialized rather than *803 there being groups of generalists did not come immediately, but flowed from "a hell of a lot of analysis" and had to deal with "the glass being half full or half empty and what you use as the base." One of the authors commented, "I remember Bob kept commenting on how little time ministers spend reading the Bible or scripture even though they're specialists in religion. And so trying to make sense of these people only spend[ing] 15% of that time on this subfield--what does that mean?"
What does that mean? This in fact is the constant question in most empirical research. In simple experiments it might be possible for researchers to draw clear meanings from statistical results.(97) However, most social, political, and economic research is not based on simple experiments. The type of clearcut controls and unambiguous results such experiments might permit simply are not practical. Furthermore, the problem of interpretation is not one that will go away with better theories or agreed upon paradigms. Important questions of the type social scientists ask and are asked by others lead to data and then to analyses which themselves are complex and ambiguous. In the end, empirical social scientists (whether doing qualitative or quantitative research) must interpret their data and their analyses if they are going to answer those questions.
The consumers of quantitative research pick up a published study and read the authors' report of their question, their data, their findings, and their conclusion. However, as every researcher knows, the path from question to conclusion is not the linear path of the published research report. Seldom do we see the gestation of a research project, or its infancy, or its adolescence.(98) Moreover, the final product that we do see tends to show little of the ambiguity or uncertainty in the analyses. Rather, that product tends to look (and be presented as) relatively clear and unambiguous.
In this essay, I have described how a major research project evolved from a first casual dinner-table conversation, to the development of a research design, through extensive and varied analyses of a rich, complex dataset, and ultimately to a major book on policy advocates and policy advocacy. The authors started their research safari with their sights set on the Washington lawyer. However, even as they planned their research adventure, *804 they realized that there was other game to be found in the world of interest representation in Washington. Once they arrived in their data jungle, they discovered that lawyers were much less numerous, and of much less importance, than they had originally expected. Rather than abandoning the hunt, they shifted to what was actually there and produced a rich analysis of many aspects of Washington representation.
A combination of theoretical interests and data realities drove HLNS's analysis. Their interest in social structure derived from earlier research they had conducted; their interest in lawyers reflected both earlier research and their institutional setting at the American Bar Foundation. These interests led them to design data collection instruments that in turn constrained both the analyses they could conduct and the results they could arrive at. Note here that I do not say that the data determined their results, only that it constrained the results. A combination of data, theory, analytic technique, and interpretative orientations determine results.
Combining the elements of the rhetorical tetrad (logic, fact, story, and metaphor) with literatures on interpretation helps bring the research and analysis process into clearer focus. The interrelations among the various elements is far from the textbook descriptions of the research process. One of the more striking elements of the account I have presented is the ongoing role of interpretation, broadly defined. Theory is extremely important, but it is not static; it evolves and changes as the researcher confronts empirical issues ranging from research design through data collection and data analysis and ultimately to the need to put down in writing the research findings. Throughout this process the researcher engages in interpretation: What does the theory mean for designing research? What do the difficulties in getting the data mean for the theory? How do the limitations of the data constrain what the researcher can say? What coherent patterns can the researcher identify and verify in the data? All of these issues (and others I have not listed) require as a core element interpretation.
In this essay, I have tried to show how narratives and metaphors serve as a major tool in the interpretation process. While I have not tried to show explicitly the commonalities in use of narrative and metaphor between quantitative and qualitative empirical research, those commonalities are a major element in the reality of high-quality empirical social science. This does not mean that every piece of empirical analysis necessarily employs the interpretive tools I have described. However, research that links back to theory in major and meaningful ways is very likely to require a major element of interpretation. There are undoubtedly other approaches to interpretation. However, regardless of the interpretive methods a researcher employs, those methods are a central element of the researcher's craft.
1. Karl Popper, The Logic of Scientific Discovery (New York: Basic Books, 1961; originally published in 1935 as Logik der Forschung); Ernest Nagel, The Structure of Science: Problems in the Logic of Scientific Explanation (New York: Harcourt, Brace & World, 1961); Carl G. Hempel, Philosophy of Natural Science (Englewood Cliffs, N.J.: Prentice-Hall, 1966).
2. Michael Powell was originally a member of the research team, but he dropped out in August 1982 when he moved to Chapel Hill, N.C., to take a teaching position; in some places, I recognize his participation by using the abbreviation HLNS(P).
3. It is the metaphoric representation of HLNS's conclusion regarding their findings on this issue that serves as the book's title: the structure of interest representation is built around a "hollow core"--there are no central players serving a brokerage role among competing interests.
4. Policies of some scholarly journals sometimes presume a process of this type; for example, the American Journal of Political Science requires authors of data-based articles to prepare an abstract "divided into four sections: Theory, Hypotheses, Methods, and Results."
5. W. Phillips Shively, The Craft of Political Research 24-25 (3d ed. Englewood Cliffs, N.J.: Prentice-Hall, 1990).
6. Emphasis added. Edward O. Laumann & David Knoke, The Organizational State: Social Choice in National Policy Domains (Madison: University of Wisconsin Press, 1987) ("Laumann & Knoke, Organizational State").
7. An alternative metaphor for the research process might be a board game (one does regularly hear references to the "research game" or the "grant game"). One could conceptualize it as combining elements of Monopoly and Scrabble. It's like Monopoly in the sense of having a mostly unidirectional path but with occasional short cuts or side trips to jail or the Reading Railroad (Is an NSF grant like getting an unexpected inheritance or more like slowly building up property?). Alternatively, it is like Scrabble in that it involves putting things together (words in the case of Scrabble, information or data in the case of research). In a competitive research setting (i.e., where multiple researchers are seeking the same goal), it might be like Sorry or Parchesi. One might also conceptualize it a bit like the child's game of Chutes and Ladders where one progresses toward a goal but regularly encounters bad luck (sliding down a chute and having to work your way back up) or ladders (taking advantage of short cuts along the way).
8. Sometimes the effort to fund a project is in itself safari-like. In one large project I was involved with, I recall clearly the reaction when we heard that we had won the contract: We felt like we had succeeded in shooting the elephant, and the carcass had just been delivered to our front yard; we now had to figure out what we actually would do with it.
9. Natalie Angier, "Flyspeck on a Lobster Lip Turns Biology on Its Ear," N.Y. Times, 14 Dec. 1995, at 1. The lowest order of classification is the species, followed by the genus, followed by the family, followed by the phylum.
10. One common frustration in training graduate students is the student who reaches the dissertation stage and is at a complete loss over what to do for his or her dissertation research. To be a successful research scholar, one must be driven by intellectual curiosity. Students should be frustrated at the dissertation stage, but by a surplus of ideas rather than a dearth of ideas.
11. One might be tempted to question the role of luck in a craft; however, for most craftspeople, the really successful product is often a matter of luck because of the uncertainties associated with much of the work that a craftsperson does.
12. For useful collections, see John S. Nelson, Allen Megill, & Donald N. McCloskey, eds., The Rhetoric of the Human Sciences (Madison: University of Wisconsin Press, 1987), and other books in the UW Press series; Herbert W. Simons, ed., Rhetoric in the Human Sciences (Newbury Park, Cal.: Sage Publications, 1989).
13. Donald N. McCloskey, Knowledge and Persuasion in Economics 61-63 (New York: Cambridge University Press, 1994).
14. See references to Popper, Nagel, and Hempel cited in note 1.
15. Barney G. Glaser & Anselm L. Strauss, The Discovery of Grounded Theory: Strategies for Qualitative Research (Chicago: Aldine Publishing Co., 1967); Anselm L. Strauss, Qualitative Analysis for Social Scientists (New York: Cambridge University Press, 1987); Jack D. Douglas, Investigative Social Research: Individual and Team Field Research (Beverly Hills, Cal.: Sage Publications, 1976).
16. Jorge J. E. Gracia, A Theory of Textuality: The Logic and Epistemology 4 (Albany: State University of New York Press, 1995) ("Gracia, Theory of Textuality").
17. See, e.g., Roland Barthes, Elements of Semiology (New York: Hill & Wang, 1977); or Umberto Eco, A Theory of Semiotics (Bloomington: Indiana University Press, 1976).
18. Some types of demographic characteristic data might be arguably "nontextual" in that they are naturally occurring, but then someone had some intention in defining the categories (e.g., race, gender, age, etc.) which are used in collecting the data. Economic indicators (e.g., money supply, unemployment rate, GNP, price index, etc.) clearly are constructed by an "author" to convey some "meaning" in a given "context," and the choices made by those who construct the index have important implications in understanding the meanings of that index.
19. In a sense, data serve to establish a frame for the analysis; see Erving Goffman, Frame Analysis: An Essay on the Organization of Experience (Cambridge: Harvard University Press, 1974).
20. See Gracia, Theory of Textuality 119-23; I am tempted to refer to this limitation as the quality of "textuality," but this does not appear to be a standard usage of that term.
21. William H. Honan, "Cardiologist Answers a Raphael Question," N.Y. Times, 16 Dec. 1995, at 15. The Times article translates "Raphael" as "God heals"; my College Edition of Webster's New World Dictionary of the American Language 1205 (Cleveland: World Publishing, 1964), translates it as I have shown, "God hath healed," which seems to me to fit a bit better.
22. "The Data Puzzle: The Nature of Interpretation in Quantitative Research," 40 Am. J. Pol. Sci. 1 (1996).
23. See Stephen W. Hawking, A Brief History of Time: From the Big Bang to Black Holes (New York: Bantam Books, 1988).
24. See, e.g., Robert Gilmore, Alice in Quantumland (New York: Springer- Verlag, 1995).
25. The largest groups of representatives are divided between executive positions in the client organizations (45%) and positions specifically devoted to government affairs (22%); the remaining representatives are internal lawyers or internal research staffers.
26. Also, as a group, those who work as outside representatives, both lawyers and nonlawyers, tend to be middle of the road (p. 157).
27. HLNS briefly examine the stability of issues and issue concerns here as well (pp. 249-52).
28. Interestingly, more of the notables, 45 of 72, were lawyers (18 were selected for each of the four policy domains), a higher proportion than was true of the representatives as a whole (pp. 264-65).
29. HLNS show and discuss smallest space results for specific cleavages in each of two issues from each domain.
30. They also provide a minimal assessment of the validity of the self- reports by comparing the level of success reported by those representing the "winners" as perceived by the government officials HLNS interviewed in each domain with the level of success reported by the losers. HLNS find that the mean success reported by the winners was significantly higher than the mean success reported by the losers.
31. The other end of the spectrum, where both government and private power is highly concentrated, would be a corporatist system.
32. Just before this reform was introduced and then enacted, one of my colleagues published a book which argued that major reform was extremely unlikely because of the entrenched interests in the then current system; see John F. Witte, The Politics and Development of the Federal Income Tax 379-86 (Madison: University of Wisconsin Press, 1985).
33. Chicago Lawyers: The Social Structure of the Bar (New York: Russell Sage Foundation, 1982) ("Heinz & Laumann, Chicago Lawyers").
34. Partners with Power: The Social Transformation of the Large Law Firm (Berkeley: University of California Press, 1988).
35. Beyond Monopoly: Lawyers, State Crises, and Professional Empowerment (Chicago: University of Chicago Press, 1987).
36. From Patrician to Professional Elite: The Transformation of the New York City Bar Association (New York: Russell Sage Foundation, 1988).
37. The former relies on a small set of clients who are able to demand and get deference to their demands and needs while the latter has a broad client base so that no one client is particular significant in the overall scheme of things.
38. This contrasts sharply with some of the more popular presentations of the Washington insider-lawyer-lobbyist who some authors portray as "insiders," "fixers," or "power brokers"; see Joseph Goulden, The Super-Lawyers: The Small and Powerful World of the Great Washington Law Firms (New York: Weybright & Talley, 1972) ("Goulden, Super-Lawyers"); Charles Horsky, The Washington Lawyer (Boston: Little, Brown, 1952) ("Horsky, Washington Lawyer"); Mark Green, The Other Government: The Unseen Power of Washington Lawyers (New York: Grossman Publishers, 1975) ("Green, Other Government").
39. All the more senior members of the team had or were doing work on interest groups. Laumann was in midstream of a project on the role of organizations in policy change that would eventually be published as Edward O. Laumann & David Knoke, The Organizational State (cited fully in note 6). Salisbury had built his reputation on the study of interest groups, and Heinz had worked with Salisbury on interest groups during his year of graduate study and published several early pieces on interest group issues. Laumann and Heinz were in many ways the central pivot because they were the ones who had a foot in both research areas (i.e., legal profession and interest groups), Laumann perhaps more than Heinz because of his then current research.
40. Heinz had briefly been a political science graduate student of Salisbury, and Nelson was a student of Heinz's; Powell was a student of Laumann's. Thus, the project brought together three generations of scholars.
41. Progress in developing the design was slow because of the other commitments each of the participants had (e.g., completing books, dissertations, existing research projects, etc.) and because of the vagueness of the initial idea. The initial discussions that began the research the participants came to refer to as the "Washington Study" occurred in May 1980; the design of the basic project was not worked out until early 1982, and the primary data collection did not occur until 1983 and 1984.
42. Parsons's writings on the legal profession had also included a functionalist analysis of interest representation; see Talcott Parsons, "A Sociologist Looks at the Legal Profession," in Essays in Sociological Theory (New York: Free Press, 1954); see also Talcott Parsons, "The Law and Social Control," pp. 65-78 in William M. Evan, eds., Law and Sociology 65-78 (New York: Free Press, 1962)
43. Salisbury memo ("Further Thoughts Apropos Study of Washington Lawyers"), no date ("early 1981").
44. Salisbury letter of 16 Dec. 1980.
45. Memo (author unknown) entitled "Proposed Washington Study: Heinz, Laumann, Nelson, Powell Discussions," 3 Feb. 1981.
46. Nelson memo ("Suggestions about Research Design") of 5 Jan. 1981.
47. This has many implications, the most immediate of which was the practical problem of defining the population to be sampled: Should it be based on standard directories, either of the bar (e.g., Martindale-Hubbell) or of "Washington representatives" (using annual compilations published by Columbia Books in Washington Representatives)?
48. At this stage HLNS(P) envisioned the larger project as ultimately involving five separate data collection steps: a random mail survey of 1,000 nongovernment lawyers resident or practicing in the D.C. area; a followup survey of those in the initial sample who were engaged in a "distinctively Washington practice" (expected to be about two-thirds of the original sample); a random survey of 500 government-employed lawyers in the D.C. area; a "small, random sample of 'Washington Representatives' (N = 250)" to determine the relative significance of lawyers in policy representation and to determine in what ways lawyers differ from nonlawyers in their performance of this role; and a survey of lawyers active in each of the issue domains selected for analysis (50 to 70 lawyers in each domain).
49. The proposal was entitled "Washington Lawyers and National Policy Making."
50. HLNS(P) were concerned that the ABF board recognize the central role of the legal profession-related aspects of the project; see Powell memo of 3 April 1981.
51. See Robert H. Salisbury, with John P. Heinz, Edward O. Laumann, & Robert L. Nelson, "Soaking and Poking among the Movers and Shakers: Quantitative Enthnography along the K Street Corridor" (presented at 1984 Annual Meeting of American Political Science Association, Washington Hilton, 30 Aug.-2 Sept. 1984. The phrase "soaking and poking" is from the work of Richard F. Fenno, Jr.; see "The Political Scientist as Participant Observer," in Watching Politicians: Essays on Participant Observation (Berkeley, Cal.: IGS Press, 1990).
52. To a significant degree, the subjects of those interviews reflected preexisting contacts in Washington (from personal relationships or prior research).
53. Otherwise unattributed quotations are taken from my group interview with the authors (13 Feb. 1995).
54. This concern reflected an early hypothesis that stratification would be structured less by law school than was true of the Chicago bar because of the importance of government service, which would tend to wash out earlier factors.
55. Nelson memo, "Suggestions about Research Design," 5 Jan. 1982; "Salisbury Memorandum re Washington Lawyers Project" (undated, but estimated as Jan. 1982).
56. Heinz, "Notes from Washington Study Conference--Friday Feb. 19, 1982"); Laumann memo (as revised by Heinz), "A Framework for Analyzing the Differing Roles of Washington Representatives in Issue Resolution," 26 March 1982.
57. They identified (untitled memo of 18 Feb. 1982) the variables the data would need to inc