"Policy Research," The American Sociologist, Vol. 6, Supplementary Issue (June, 1971), pp. 8-12. Translated into French in Synopsis (September-October, 1971), pp. 1-12. Reprinted in Robert B. Smith (ed.) An Introduction to Social Research, Vol. I of the Handbook of Social Science Methods (Cambridge, Mass.: Ballinger Publishing Co., 1983), pp. 77-92.

Amitai Etzioni is director of the Center for Policy Research in New York City and chairman of the Department of Sociology at Columbia University. He has been a consultant for numerous domestic government agencies and has testified before congressional committees. His many articles on public policy matters published in professional and popular journals range from peace plans for the Middle East to schemes to make television participatory. His book that deals most directly with public-policy on a theoretical level is The Active Society: A Theory of Societal and Political Processes (1968).

Policy research is concerned with mapping alternative approaches and with specifying potential differences in the intention, effect, and cost of various programs. It differs from applied research in much the same way that strategy differs from tactics; it is more encompassing, longer-run in its perspectives, and more concerned with the goals of the unit for which it is undertaken; that is, it is more critical.

Policy research differs from basic research as strategy differs from theory. It is much less abstract, much more closely tied to particular actions to be undertaken or avoided. While basic research aims chiefly to uncover truth, policy research seeks to aid in the solution of fundamental problems and in the advancement of major programs.

Social scientists have been conducting all three kinds of research for decades; therefore, lengthy conceptual exercises in explaining the differences between the three kinds of research seem gratuitous. There are, though, a few key characteristics of policy research, probably the most undernourished of the three types, that deserve brief commentary.

I discuss first the critical stance of policy research and the structural implications of the specific intellectual position involved. The problems resulting from the need of policy researchers to communicate with policy makers are explored next, as well as the difficulties intrinsic to the policy research "product."

The Critical Stance

One reason social scientists in the United States are conducting more policy research than they did a decade or two ago is their increased interest in becoming involved or "active" in their professional research as well as in their "after hours" citizen activity. Policy research is a legitimate opportunity for scholars to do "relevant" research.

The debate about whether or not social science can be neutral is pertinent only for basic research. Applied and policy research cannot claim neutrality. Applied research accepts specific assignments from clients and tries to serve their needs largely in their terms. It may develop, for example, a psychological test that can discriminate between employable and unemployable persons, or a statistical projection to forecast the number of inmates released from a state's prison system who will be back in jail within twelve months, or a projection to determine what actions are most likely to mobilize opposition to the war in Vietnam.

Policy research ought not to take on specific assignments but to concern itself with the problems of the social unit (or units) to which it relates. Thus while applied researchers concern themselves with such things as de-designing instruments to distinguish employable from unemployable persons, policy researchers address themselves to such questions as which are the best methods of reducing unemployment in a given society in a given period. The latter will also consider the underlying assumptions on which the research is based-in this case, that unemployment rather than the labor force be reduced. They may raise such questions not in terms of their personal values, not, for example, whether researchers view a life of reflection and leisure without material goods as superior to a life of affluence, but in terms of the needs and basic values of those afflicted with unemployment as well as the needs and values of members of the society in which unemployment occurs. A policy researcher might conclude, for instance, that the goal of reducing welfare costs and taxes might be better served not by generating more jobs but by keeping more people out of the labor market-occupied in, let us say, studying. Similarly, the goal of providing socially useful lives for all citizens might be realized through alternatives other than the reduction of unemployment.

Society, or at least important segments of it, is, however, committed to another relevant value, the value of work as an end in itself. Here, policy research may help to highlight the tensions and conflicts among the three societal commitments-work, leisure, and low unemployment-and specify the economic and social usefulness of alternatives heretofore only dimly perceived. Thus, work in traditional terms, reduced tax burdens, and social usefulness are typically contrasted with loafing, higher taxes, and social usefulness. But the citizens and legislatures may support such options, options that are socially useful but not dependent on work in the traditional sense. It is not enough, of course, to indicate that such options are logically possible; specific programs must -be outlined (for example, workers taking an educational sabbatical and, thus, spreading employment more widely), relative costs must be estimated, and effects on other values of the social unit must be considered. In terms of the theory of action, applied research deals with means, taking the goals for granted. Policy research deals with values and seeks to clarify goals and the relations among them, as well as among goals and sets of means. Applied research is, by definition, instrumental. Policy research is inevitably critical.'(1)

The societal function of such a critical orientation should be noted. Policy makers, whether leaders of government or of a social movement, tend to encapsulate themselves in sets of assumptions they come to take for granted and are reluctant to examine. For example, behind most programs to help unemployed persons rests an assumption about the relative ease with which people can be educated and trained. The rationale of programs and their curricula, attitudes toward the trainees (which ones, ask the policy makers, are "good material"?), and the hiring of educationalists and psychologists are all built around this assumption. When a social scientist suggests that changing a person is a very difficult (and maybe normatively dubious) process, that jobs should be changed to accommodate people (which is very difficult too, but not as difficult as changing people), and that people and jobs could be much better matched, he encounters severe resistance to even examining the idea because he challenges a core assumption.

Frequently we find that it is considered unnatural as well as damaging to the interests and psychic comfort of those involved to reexamine openly the assumptions that underlie their policy, especially after a broad consensus about the assumptions has been reached. Such reexamination has real costs in terms of time, budget appropriations, psychic tensions, etc.; it is hence justified only when the assumptions clearly and significantly deviate from reality. The reason policy research is so useful for enhancing the reality-testing of a system is that even under these conditions, and especially when reality leaves the policy assumptions far behind, there is a considerable tendency to adhere to the old assumptions; hence, there is a great need for institutionalization of the responsibility to prepare alternative rationales and to pry the policy maker loose from his antiquated assumptions.

In contrast, the applied researcher tends to focus on "bits," on improving the implementation of existing policies; he tends to accept uncritically-consciously or unwittingly-the basic underlying assumptions of the policy makers.

Reference is made here to roles and not persons, to the modes of conduct that the roles typically legitimate and that seem most prevalent. Most policy researchers occasionally conduct some applied research and most applied researchers are sometimes critical. However, there are specific mechanisms that keep the applied researcher from becoming too policy-oriented (and, to a lesser extent, the policy researcher from becoming too applied), namely, the mechanisms that provide funds for and supervise research.

Division of Labor and Structural Separation

The need for policy research lies not in the fact that policy makers are personally inept, unable to conduct research, or unable to review their assumptions critically-although these deficiencies may exist-but in the fact that policy makers do not have the prerequisites for conducting research. Time, training, books, and computers are all lacking. While the staff directly subordinate to most policy makers could be assigned to conduct some studies, they tend to suffer from similar limitations. For example, the average length of time the Council of Economic Advisors is given to explore a specific topic before it goes to the President for a decision is reported to be one week. Of course, the CEA draws on background material that may have been amassed over a long period, and members may have worked in the general area prior to joining the CEA (Wilensky, 1967:94-109), but their "backgrounds" and the material available frequently are insufficient to allow them to deal directly and specifically with the researchable questions the CEA must face at a particular moment. Using available data under rush conditions requires considerable "stretching" by way of conjecture, extrapolation, and interpretation, which weakens the empirical power of the CEA's recommendations.

Policy research is thus best conducted by organizational units that specialize in such research, units that are close to the policy makers (attentive to their problems and aware of their constraints) but at the same time sufficiently independent to allow critical analyses. The RAND Corporation, whose main source of funds is the Air Force, provides a case in point (Smith, 1966). The two centers for policy research set up by the United States Office of Education in Syracuse, New York, and Palo Alto, California, seem too removed from OE to have a meaningful interaction with the agency; the institute supported by the Office of Economic Opportunity at Madison, Wisconsin, seems too tightly controlled by the policy makers to have the required capacity for fully independent policy research.

Some significant work is conducted by such radical centers as the Center for Democratic Institutions in Santa Barbara and the Institute for Policy Studies in Washington, D.C. These facilities tend to nourish the policy making of the opposition rather than the policy making of those in power. Since those in power usually can gain research inputs more easily than can those out of power, these radical centers have a balancing value. Their main weakness is their tendency to quickly produce pamphlets instead of careful conceptualizations and empirical research.(2)

Policy studies are now most commonly conducted by small ad hoc task forces that service an agency or the White House (for example, the Rivilin task force on higher education) or by a commission (for example, the Kerner commission). These task forces or commissions because of their ad hoc nature frequently lack the facilities, experience and continuity-in short the structural prerequisites-for conducting high-quality policy research. What seems necessary is more RANDs in which policy research is institutionalized. As more than enough RANDs are working for the armed services, the main lack of such facilities is on the domestic front.

There are several structural features that are also relevant to policy research; their value can be derived from the basic division of labor among (a) insiders (to the policy making organs) who do chiefly staff work, (b) unattached outsiders best suited for basic and applied research, and (c) autonomous yet related policy researchers (related in terms of commitment to problems of the policy makers and easy access to them). Other structural needs are for a permanent staff of sufficient stature and size (a critical mass seems unattainable if the policy research unit is too small), location of the unit within easy reach of the policy makers, and regular flow of relevant data between the units involved. (The last prerequisite may sound trivial but is in fact important and often difficult to achieve.)


Policy researchers must communicate their "products" to the policy makers, whether they are the President and his staff or a local chapter of the League of Women Voters. While this may seem obvious, it is in fact not fully understood, and the need for communication is often overlooked.

The typical and traditional way research findings are communicated is through publication. Papers read at professional conferences, articles, books, and even stenciled reports are the most widely used modes of communication. Experience suggests, however, that while these may be the most effective channels for communicating basic and applied research, they do not serve policy research nearly as well.

The policy researcher must expect some resistance to his conclusions, resistance on the part of the policy maker that is emotional (due to his commitments to other policies), cognitive (due to his lack of information and training), and self-interested (the recommended policies may serve some groups less well than the obsolete ones d(3)o).3 Formal means of communications, such as those listed above, are particularly ineffective in overcoming such resistance. Frequently repeated, face-to-face exchanges seem very necessary. Presocialization (preparing for the report) and follow-up (after it is handed in)-are essential. The report itself often is not really necessary; it serves more to fulfill contractual obligations and to provide an opportunity for and legitimization of interaction with the policy maker than to communicate with him.

Put differently, policy researchers, unlike basic researchers, must be willing to invest a significant amount of their time and energy in communicating.(4) They must learn how the world looks to the policy makers, how best the policy makers will absorb the new ideas, which means of documentation and presentation will be most successful, and in what sequences the findings should be offered. One of the secrets of RAND's success is the hundreds of briefings RAND sets up for people in the Air Force, Department of Defense, Congress, the press, and groups of citizens (Smith 1966: chap. 6).

A successful policy researcher seems to require a psychic profile quite different from that of a basic researcher. Basic researchers can be "loners," people who get along best with libraries, computers, test tubes, and research associates. A policy researcher must, be able to interact effectively with politicians, bureaucrats, housewives, and minority leaders.

It may be argued that the task of communicating might be delegated to go-betweens, leaving the researchers in their cubicles, but this is inefficient. Direct contact with policy makers, with the questions they raise, is useful and necessary to the policy researcher; it clarifies in his mind the constraints policy makers face and must live with and must be shown how to overcome. Frequently the policy researcher alone has the authority and knowledge to deal with the policy maker's questions."(5)

The Product

Probably the least useful tool for communication between policy researcher and policy maker is the one most commonly used-the hundred-odd-page stenciled report packed with statistics, footnotes, and technical terms. The shortcomings of this mode of communication are more than stylistic and cannot be overcome simply by providing an abstract or a "rewrite" in idiomatic English, sans tables and references.

A major difficulty is the level of abstraction (or analysis) at which the problem is conceptualized. The more abstract the conceptualization, the more difficult it is to provide concrete recommendations. Social science disciplines differ considerably in the extent to which they are abstract (or analytic). History and much of social anthropology are relatively concrete in the concepts most widely used. Parsonian sociology is extremely analytic. Institutional economics is much less analytic than econometrics.

The more analytic sciences use an esoteric language that must be translated before it has any meaning for policy makers, and the technical terms often contain assumptions that are typically not specified when such translation takes place. This is the case because the assumptions are universal and hence not more relevant to one report than to another and because they are numerous and hence to recite them in a report dealing with a specific question would make the communication ponderous. Many policy researchers are not even conscious that they subscribe to these assumptions (for example, the assumption that man is not rational) but without sharing these assumptions, their recommendations are not fully intelligible to the policy maker. Even if the assumptions were spelled out with greater regularity, most policy makers would not have the long training necessary to take them into account systematically in evaluating the reports.

Probably the most important assumption shared by all analytic scientists is that certain factors remain constant. Thus, an economist dealing with the relation between prices and unemployment rates takes for granted that unless otherwise specified all other economic and noneconomic variables that affect this analytic relation remain unchanged. However, reality, or that part of it with which the policy maker is concerned, is affected by a great many factors simultaneously; almost none of them remain constant. Theoretically, this problem can be handled by combining the inputs of various analytic studies into a synthesized recommendation. In practice, however, such piecing together is not practicable. Few researchers and, in effect, no policy maker can accomplish it. Without such compensation for the analytic nature of the research input a responsible transition to policy cannot be made. How the jump is made when policy researchers use pure analytic studies is not known because few studies have been made of the transition from research to decision on the policy level. That the results are frequently not productive is common knowledge.

So as not to leave the policy maker to his own devices, a policy researcher may mix what his model tells him with his own "common sense" (thus, to curb inflation he may recommend a tax increase that is smaller than one suggested by a Keynesian economic model because the researcher "knows" that the Keynesian rate of increase would be politically unacceptable). The policy researcher may also downgrade some of his conclusions and build up others on the basis of his experience rather than on his theory or research. Or the analytic recommendation may be implemented in its "pure" form and lead to "unanticipated consequences," which then need adjustment; the policy research may prepare the policy maker for this need. None of these are very satisfactory solutions.

An alternative approach is for policy research to be based on a less formalized, less analytic discipline that falls between the analytic sciences and policy recommendations, the way medicine comes between physiology and pharmacy and the way international relations come between economics, sociology, anthropology, and political science on the one hand and policy making on the other hand.

This in-between discipline, called policy science, furnishes nonabstract concepts that are understandable to the policy maker, it furnishes practical knowledge that has no analytic foundation but is useful for moderating or complementing analytic knowledge,, and it is, in general, not, far removed from the observed world and that experienced by the policy maker. It thus reduces the need for transition and translation. For example, the person who attempts to run a prison system on the basis of a psychological or sociological theory will soon find that practical experience is a needed element in criminology. One who seeks to run a welfare program on the basis of an analytic science will soon find it necessary to know the bureaucratic and legal constraints and the subcultural habits of the recipients.

Frequently policy researchers have tried to move too directly from highly analytic social science theories to policy recommendations. The necessary intermediary discipline that records and evaluates the programs and options available has been slow in developing and is now a major area on which policy research as a discipline must focus.

For studies that are to serve policy making on the national (or societal) level, a macro-action conceptual frame of reference is essential. Data about individual members (persons, firms, or other member units) are useful, but unless there is a clear conception of how these members are combined into larger units and what the attributes and dynamics of the units are, policy studies will not be relevant to the treatment of societal problems. Thus, in addition to such notions as emotions, perceptions, and beliefs-all individual attributes-and in addition to such high-level abstractions as the Parsonian pattern variables that apply to any and all units, we need to build our analyses around such concepts as class, race, region, strategy, mobilization, and societal investment in production of knowledge, concepts that are shared by the policy maker and the policy researcher and common to the English language and to scholarship. This is not to reiterate the frequent complaint against jargon but to point up the need to supplement abstract and microvariables with more concrete and macrofactors.

Moreover, there is a distinction as vital to the policy researcher and policy maker as it is irrelevant to the basic researcher, namely the degree to which a variable is "moveable," that is, the degree to which the phenomenon it characterizes is malleable.(6) Thus, sociologists regularly break down social data into categories of sex, education, income, class, and race since from a basic research viewpoint they all have a similar (or "independent") status. From a policy viewpoint, however, some of these variables are "given" or extremely difficult to change (sex), while others are relatively more changeable (income).

A reviewer of this paper for The American Sociologist commented here: "So, if women are up in arms about discrimination by sex we mustn't tell policy makers about it because sex can't be changed?" As I see it, the source of discrimination is not the sex difference but the attitude toward sex; the fact that sex is extremely difficult to change is not relevant to a particular policy, and if it should be established that the sex difference or some other social phenomenon (or some parts of it, as, most phenomena are affected by more than one factor) is not changeable, this is important information we should share with policy makers. The waste of considerable energy on challenging windmills rather than fueling change is one of the key characteristics of liberal elites and radical movements.

The ranking of factors in terms of their malleability is, of course, important in itself. Malleability is not of less status for basic research or general theory than, say, the distinctions flagged by Parsons's pattern-variables. However, while for general theory malleability is but one of many fruitful bases for theory building and research, it is central to policy science. Policy science as a conceptual discipline must alert the researcher and the policy maker to differences in malleability; it must focus its attention and research efforts on the more moveable variables and on the conditions under which the less moveable ones can become more open to modification.

From a consideration of differential malleability, the next step is to consider who can move what. On the macrosocietal level, I have found it fruitful to apply a conceptual scheme called societal guidance. I distinguish between controlled and uncontrolled processes, and I specify the attributes of the controlling agents in terms of their knowledge, commitments, decision-making strategies, and amounts and kinds of power; I also explore the capacities of those controlled and the conditions under which they are mobilized to support, as against oppose, the controlling units (Etzioni, 1968).

Policy research also differs from basic research in its central methodological considerations, considerations that apply to a lesser extent to applied research. For the basic researcher, science is an open-ended enterprise. There are no intrinsic reasons for the completion of a study at any particular deadline, and the dictum "until proven otherwise" is always at least implied. For the policy maker, there are specific times when fundamental decisions will be made and the decisions made then will become the base for more detailed decisions. The policy researcher must schedule his research so as to produce conclusions by that point (unless, of course, he can delay the decision until better data and analyses are available). For the basic researcher to conclude that the data at hand are too thin to warrant conclusions is both fully legitimate and in line with self-interest (he protects himself from any backlash from conclusions based on insufficient data and increases his chances for obtaining additional funds for his work). For the policy researcher to reach such a conclusion, unless the data are extremely poor, is an abrogation of his responsibility; the policy maker is likely to make a decision anyhow, and he probably will make it less well if the policy researcher has not shared whatever data and analyses are available, of course highlighting their limitations.

The basic researcher is not constrained to encompass the unknown so as to gain at least crude estimates of the segments of the world for which there are no studies. He shares no responsibility to act and can follow other research procedures-for instance, proceeding step by step and exploring each step fully before he takes the next one. The policy researcher must often cover his chosen field by first approximations before he can work in detail on one segment.

One way that such approximations are accomplished is by the use of qualitative techniques that are then followed by more formal studies. Sargent Shriver, for instance, as the head of the Peace Corps, is reported to have frequently used journalists who visited Peace Corps units in the field for a few weeks. They wrote informal evaluative reports at a small fraction of the time and cost of a formal (and, typically, much more narrowly conceived) evaluation study.

The basic researcher takes pride in generating new data. The policy researcher takes pride in finding existing data that, through secondary analyses, are usable so that he need not be delayed by the costly and long procedure of generating new data. It might be said here that policy researchers "live off" basic studies, a statement that is quite true. Policy researchers ought to acknowledge their intellectual and data debts to basic research; they ought not, however, to stop using these data and conceptions, as the potential use by policy research is one justification for the investment in basic research.

In toto, little should be made of the distinct attributes of policy research. The demarcation lines are not sharp, and many studies cross them freely. There seems, though, to be a body of work that more directly serves the making of fundamental decisions, and there is little doubt that we could benefit from research specifically aimed at improving these decisions. The policy makers have not done too well with the assistance they have had so far. While the sources of their ineptness are many, lack of relevant knowledge is not the least of them.


Barnett, Richard 1968 Intervention and Revolution. New York: World.

Dror, Y. 1967 "Public administration: four cases from Israel and the Netherlands." Pp. 418-426 in P. F. Lazarsfeld, W. H. Sewell, and H. L. Wilensky (eds.), The Uses of Sociology. New York: Basic Books.

Etzioni, Amitai 1968 The Active Society: A Theory of Societal and Political Processes. New York: Free Press.

Gouldner, A. 1965 "Explorations in applied social science." Pp. 5-22 in A. Gouldner and S. M. Miller (eds.), Applied Sociology. New York: Free Press.

Smith, Bruce L. R. 1966 The RAND Corporation. Cambridge, Mass.: Harvard University Press.

Wilensky, Harold L. 1967 Organizational Intelligence: Knowledge and Policy in Government and Industry. New York: Basic Books.

Wirth, L 1947 "Responsibility of social science." The Annals 249 (January): 143-51.

1. In his discussion of applied social science, Gouldner (1965: 11-15) makes a similar distinction between what he terms "engineering" and "clinician" approaches to applied social science. The latter orientation corresponds to what we claim is a different overall discipline, rather than just a different class of the same phenomenon.

2. There are individuals in these centers whose work is atypically "hard," for example, Richard Barnett at the Institute for Policy Studies. See, for instance, his Intervention and Revolution (1968).

3. Further insights into the problem of resistance to accepting or using the social scientist's conclusions are presented by Gouldner (1965:15-21). An earlier but not less insightful statement of the problems confronting the application of social science is given by Wirth (1947:146-147).

4. This is true also for applied researchers, but to a lesser degree because they may expect less resistance and seek less-encompassing changes.

5. Some brief illustrations of the problems dealt with here can be found in Dror (1967).

6. For a discussion of the issues involved in this concept, see Etzioni (1968).

The Communitarian Network
2130 H Street, NW, Suite 703
Washington, DC 20052