...... POLICY IMPLEMENTATION: AN EMPIRICAL INVESTIGATION OF THE HANDICAPPED EFFORT IN HEAD START Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY SELGUK OZGEDIZ 1976 IIIII IIIIIIIIII II III I III IIIII IIIII “O'NETKAL. ... 4;. 3.? ~ 4-,; .. ”C . I . ‘ {This 1's to certify that the thesis entitled POLICY IMPLEMENTATION: AN EMPIRICAL INVESTIGATION OF THE HANDICAPPED EFFORT IN HEAD START presented by Selguk Ozgediz has been accepted towards fulfillment of the requirements for Ph.D. degree in Political Science c/ ‘ "\ I 3/ /' / 73%4‘41 ' fif’?"-'-’*>{ 1" V 7 V Major professor Date A (4-13.; 6 1’ / ’7” 71 0-7639 rr'? —---—.— I I I CID, IJI'\ ABSTRACT POLICY IMPLEMENTATION: AN EMPIRICAL INVESTIGATION OF THE HANDICAPPED EFFORT IN HEAD START BY Selguk Ozgediz This study is on empirical explanation of the level of success achieved in implementing a new public policy. The specific policy examined is the so-called handicapped policy in Project Head Start which stems from a 1972 congressional requirement that at least ten per cent of the total number of enrollment opportunities in Head Start be made available to handicapped children. The fieldwork was conducted in April-May 1974. The sample consisted of thirty-five pro- grams. The data were collected through on-site, person-to- person interviews with the directors of the sampled programs. The research design used for collecting the data was primar- ily a cross-sectional survey design . The limited set of literature on implementation is reviewed, definitions for the terms "policy" and "implemen- tation" are offered, and the place of policy implementation within the overall policy process is discussed. The conceptual model of the study has its origins in the expectancy theory of organizational psychology. The ’1‘?! (I) I (n s-— #4- Selcuk Ozgediz primary independent variable of the study is level of imple- mentation success (LIS). The two variables used for explain- ing LIS are implementation potential (IPOT) and level of policy-specific effort (EFRT). The secondary dependent vari- able of the model is EFRT and the independent variables used for explaining it are environmental forces for implementation (ENVR), policy support (PSUP), and self-evaluation of capa- bility (CAPE). All of these variables are measured through indices based on the survey data. The effects of IPOT and EFRT on LIS and those of ENVR, PSUP, and CAPE on EFRT are studied through regression anal- ysis. Two models, an additive and an interactive one, are formulated and tested for each relationship. Implementation of the policy in the sampled programs is described and discussed. Observed differences between successful and unsuccessful programs are identified and interpreted and a profile of the implementation in a typical program is presented. It is shown that there is strong empirical support for the overall conceptual model. Hypotheses confirmed by the data include the following: (1) controling for IPOT, LIS increases as EFRT increases, (2) controling for EFRT, LIS increases as IPOT increases, (3) controling for ENVR or CAPB or both, EFRT increases as PSUP increases, and (4) con- troling for ENVR, EFRT increases as PSUP x CAPB increases. The hypothesis: "controling for PSUP, EFRT increases as either ENVR or CAPB (or both) increase" is disconfirmed by Selcuk Odediz the data. It is also found that there is strong empirical sup- port for both the additive and the interactive models pro- posed for the two relationships. However, the criteria used in comparing the alternative models favor the two additive models over the interactive models. Implications of the analytical findings for the hand- icapped policy in Head Start and for the study of policy implementation are discussed and a dynamic version of the model is recommended for future empirical testing. POLICY IMPLEMENTATION: AN EMPIRICAL INVESTIGATION OF THE HANDICAPPED EFFORT IN HEAD START BY Selguk Ozgediz A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1976 © Copyright by SELQUK OZGEDIZ 1976 ii ACKNOWLEDGEMENTS I owe a great debt to a number of individuals for their support and assistance during different stages of the study. My committee chairman, Frank A. Pinner provided invaluable guidance and support throughout the study. The other members of the committee, Charles Cnudde and Paul Conn helped shape the dissertation with their constructive sug— gestions. I also owe them a great deal for their timely response to my inquiries. The fieldwork portion of the study was conducted as a part of the Systems Research Incorporated-Syracuse Univer- sity project entitled “Evaluation of the Handicapped Effort in Project Head Start" under a contract with the 0.8. Department of Health, Education and Welfare. While co- managing this project I enjoyed the continuous guidance and support of Alan F. Bogatay, to whom I am also indebted per- sonally and professionally. Several members of the Systems Research Incorporated project team provided valuable assis- tance during the conduct of the project. I am grateful to William Baucom, Gerard Fiala, Walter Sowles, Sandra Filion, Deborah Grether, Willard Walker, William Maas, and Sunni Harvey for their cooperation and expert assistance. It was a distinct pleasure to work with Burton Blatt. iii He was a source of inspiration to everyone affiliated with the project. Gail Ensher's constant energy and dedication kept the project moving at all times. I am grateful to her and the other members of the Syracuse University project team, notably Dan Sage and Robert Bogdan, for their coopera- tion during the conduct of the study. I also owe a great deal of appreciation to the members of the Senior Consultant Group of the project for their suggestions and the graduate students who conducted the fieldwork for their dedicated effort. The directors and the staff of the sampled pro- grams, who c00perated with us fully and took the time to openly discuss the problems faced in implementing the handicapped policy, deserve my sincere gratitude and appre- ciation. There are a few other individuals who served as cata- lysts in the preparation of the dissertation. I am grate- ful to Ray Lanier, Alan Bogatay, and Larry Lewin for their constant support and encouragement. I thank Sevgin Olcay for proofreading the manuscript. Finally, I owe a long standing debt to my wife who personally sacrificed a great deal-for the completion of my graduate studies. In addition to her never-ending moral support and intellectual stimulation, she patiently typed and edited the draft and the final capy. Although I take full responsibility for any errors, the dissertation and the honors that come with it are, in many ways, as much hers as they are mine. iv . . Ll CF PL” ktn:- {‘21 \qJ-L List of List of CHAPTER CHAPTER CHAPTER CHAPTER CHAPTER CHAPTER Tables TABLE OF CONTENTS Figures ......... . ........................... I. II. III. IV. VI. INTRODUCTION . ......................... Purpose and Objectives Sc0pe LITERATURE ON POLICY IMPLEMENTATION ... Literature Search Major Empirical Studies of Implementation Conclusion from Past Work on Implementation CONCEPTS AND THEORY ................... Concepts Theory DATA .0... OOOOOOOOOOOOOOOOOOOOOOOOOOOOO The Setting The Data FINDINGS ......O.‘ .......... O ...... O... A Descriptive Profile of the Implemen- tation Empirical Comparison of the Alternative Models 1 CONCLUSIONS AND RECOMMENDATIONS . ...... Conclusions Recommendations vii 14 37 90 154 209 E” v 1 LI APPENDICES: APPENDIX A. METHODOLOGY AND PROCEDURES APPENDIX B. MEASUREMENT OF SECONDARY VARIABLES APPENDIX C. SUPPLEMENTARY TABLES LIST OF REFERENCES vi 222 249 262 272 ll. 12. 10. 11. 12. 13. 14. 15. .16. LIST OF TABLES Percentage of Severely and Mildly/Moderately Handicapped Children ..... ..... ........ ...... Implementation Success Scores ........ ......... Pearson Correlation Coefficients Between Variables Considered for the EFRT Index ..... Final Item Analyses for the EFRT Index ..... ... Level of Effort Scores ................ ........ Pearson Correlation Coefficients Between Variables Considered for the IPOT Index ..... Initial Item Analysis for the IPOT Index ...... Implementation Potential Scores ............ ... Final Scores for Environmental Forces for Implementation .......... ......... .... ....... Policy Support Scores ........ . ............ .... Self-Evaluation of Capability Scores .......... Selected Characteristics of the Distributions of Indicators Used for Measuring the Variables of the Model .... .......... ........ Lead Time by Implementation ........ ...... ..... Previous Experience in Serving the Handicapped by Implementation Success ................... Agreement with the Policy by Implementation success 0.00.00...0. ......... 0......0. ....... Attitudes Towards Serving the Severely Handicapped in Head Start by Implementation Success ... ................... vii Page 137 138 141 142 143 146 146 147 149 151 152 153 157 160 165 167 170 18. 19. 20. 210 22. 23. 24. 25. Attitudes Towards Serving the Severely Handicapped in Head Start by Agreement With the POlicy .0.....000......000000.0..... Self-Evaluation of Capability by Implementa- tion Success ................................ Primary Disabling Conditions of Handicapped Children Enrolled in Head Start by Severety of the Handicap ........... ..... ............. .Disabling Conditions of Handicapped School-Age and Head Start Children ..................... Correlation Matrix for the LIS Equations . ..... Comparison of the Additive and the Interactive MOdelS--Exp1anation 0f L15 0 o o o o o o o o o o o o o o o o 0 Correlation Matrix for the EFRT Equations ..... Comparison of the Additive and the Interactive Models--Explanation of EFRT ................. Comparison of the Predictive Ability of Alternative Model Combinations .............. Distribution of Number of Head Start Programs by Size and Cluster .0.00.0.0000.0.......00.. Distribution of Number of Head Start Programs which Met the Screening Criteria by Size and Cluster ......0.....0..0..........00..... Distribution of the Sites Selected for the Second Round Visits by 0CD Region, Program Size, and Cluster ......0..........0......00O Knowledge of the Size of the Handicapped Population by Implementation Success ........ Qualifications of Staff by Implementation success .0....0..00....0...0..00.000......0.. Appropriateness of Facilities by Implementation success .......0............. ..... 0.0.0.0.... Reaction from Competing Community Agencies by Implementation Success ................... Reaction from Members of the Target Group by Implementation Success ............ ..... .. viii 169 172 175 177 194 197 199 202 205 236 238 241 262 262 263 263 264 C.11. C.12. C.13. Reaction from Members of the Target Group by Reaction from Competing Community AgenCieS ....00.00....00.....0........ ...... Reaction from Competing Community Agencies by Agreement with the Policy ............... Reaction from Members of the Target Group by Agreement with the Policy .. ........ ..... Previous Experience in Serving the Severely Handicapped by Attitudes towards Serving the Severely Handicapped . .................. Self-Evaluation of Capability by Agreement with the Policy .... ........................ Effort to Organize by Implementation Success.. Planning Effort by Implementation Success .... Coordination Effort by Implementation Success . .............. .. ............. . ..... Effort to Mobilize Resources by Implementation success 0000...... ..... .....0.............00 Effort to Obtain Additional Funds by Implementation Success ..................... Effort to Add New Staff by Implementation success ......O......O..0.... ...... .... ..... Training and Technical Assistance Effort by Implementation Success .... .............. ... Modification of Facilities by Implementation success 0000000000000 0.00.00... ...... . ...... ix 265 265 266 266 267 268 268 269 269 270 270 271 271 LIST OF FIGURES Figure Page 1. The proposed conceptual model .................. S7 2. Scatter of the residuals from the estimated values of LIS--the additive model ....... ..... 196 3. Scatter of the residuals from the estimated values of LIS--the interactive model ......... 196 4. Scatter of residuals from the estimated values of EFRT--the revised additive model ... 201 5- Scatter of residuals from the estimated values of EFRT--the interactive model . ....... 201 6- An extension of the conceptual model ...... ..... 219 CHAPTER I INTRODUCTION This dissertation is about policy implementation. More specifically, it is about how bureaucracies execute legislative mandates. During the last decade political scientists have paid increasing attention to the question of how policies are formulated. Indeed, the start of the policy analysis move- ment in political science has marked the proliferation of research on policy formulation. Yet, virtually no one has paid attention to the question of what happens to a policy after it has been formulated. This is quite surprising. One would ordinarily expect a large body of literature in an area that has such a direct and immediate effect on the well-being of millions of persons. In the United States alone billions of dollars are being spent annually on hundreds of governmental programs; and each of these programs is being implemented daily by one or more agencies. The eventual success of these programs depends upon how well these agencies can translate the original "intent" of the program into "effect.” Policy implementation, in the sense we have used the term here, refers to those tasks and activities that are l 31 to 1311 as th in 2 performed after a policy has been formulated and operation- alized. By operationalization of a policy we are referring to formulation of rules, orders, and procedures and prepara- tion of manuals and guidelines which indicate how the for- mulated policy is to be executed by the implementation agencies. Implementation, therefore, includes all tasks and activities undertaken by the implementing agency in order to execute the policy in the manner it has been Operation- alized. A successfully implemented policy is one which exhibits a high positive correlation between the implementa- tion objectives and the eventual outcomes. Why study implementation? We can think of at least three reasons: scientific, professional, and political. From a scientific vieWpoint we study implementation in order to gain an understanding of how and why some policies are implemented successfully and others are not. Here the emphasis is on scientific explanation of the results of the implementation. Knowledge gained through scientific studies of implementation increases the relevance of and contributes to the theory building efforts within policy analysis and political science. We can also study policy implementation for professional reasons. Scientific and universally generalizable explana- tions of implementation can lead to far-reaching improvements in program execution and help bring the effects of a program closer to its original intent. Here the emphasis is on application of knowledge gained from scientific studies of 3 implementation to the solution of practical problems. Finally, we can study policy implementation for political reasons. Improved understanding of what makes the implementation of some policies more successful can lead to improvements in policy formulation and policy operational- ization. Periodic feedback about the status of implementa- tion of a policy can generate information which the policy- makers can use to change the initial course of the policy or to devise new policies. Needless to say, the same informa- tion could also be used by the advocates and proponents of the policy for furthering their own interests. In addition to its information generation aspects, studies of implementation can lead to a better understand- ing of the institutions and processes of implementation and the "capacity" of the bureaucracy to execute the specific policies formulated by the legislative branch of government. This, in turn, can lead to changes in bureaucratic and orga- nizational design for the betterment of the governmental delivery mechanism. In this dissertation we come as close to formulation and empirical verification of generalizations about imple- mentation as our data permit. Although we are dealing here with a "marginal" policy within a single governmental pro- gram--the handicapped effort in Head Start-~our findings have relevance to the implementation of other policies. Naturally, further verification of our findings by other researchers will increase the credibility of the 1‘? ("f L s“ 4 generalizations upon which these findings are based. We have divided this chapter into two sections. In the first section we outline the purposes of the disserta- tion. The second section includes a description of the sc0pe and limitations of the study. Purpose and Objectives We see as our principal task the formulation and empir- ical verification of at least one heretofore nonobvious generalization regarding policy implementation. In our view, fulfillment of this purpose entails achievement of the following study objectives: 1. To review the existing literature on policy implementation; 2. To define the central concepts, such as "policy", and "imple- mentation", and illustrate the place of implementation within the overall policy process; 3. To formulate an analytical/conceptual scheme for explaining the variations in the primary dependent variable (level of implementation success); 4. To define and discuss the variables included in the concep- tual scheme and their preposed relationships to one another; 5. To examine the theoretical support for the conceptual scheme and the hypothesized relationships; 6. To describe the setting within which the specific policy examined here (the handicapped policy in Project Head Start) was formulated, operationalized, and implemented; 7. To describe and discuss the research design and the methods and procedures used in collecting the data; 8. To describe and discuss measurement of the variables included in the conceptual scheme; 9. To analyze the data, test the alternative hypotheses, and summarize the empirical findings; 10. To discuss the implications of the findings, draw conclusions t“ l’ ‘r 5 regarding policy implementation, and formulate recommenda- tions concerning future work in this area. These ten, then, are the specific objectives of the dissertation. We address objective 1 in Chapter II and objectives 2 through 5 in Chapter III. Objectives 6 through 8 are accomplished in Chapter IV and objectives 9 and 10 are covered in Chapters V and VI of the dissertation, respec- tively. Scope The term "sc0pe" generally refers to those character- istics of a piece of work that define and describe a set of boundaries. In the case of a doctoral dissertation, these include the boundaries of the problem being studied, the limitations of the techniques and procedures employed, and the generalizability of the empirical findings. We will be discussing each of these issues in great detail in the com- ing chapters. Our purpose here is to state our responses to four major boundary-related questions. These four ques- tions are: 1. Should the study of policy implementation be regarded as a proper branch of politics and political science? 2. What types of policies are the generalizations contained here most applicable to? 3. What type of a research design has been used in this study and what are the major dependent and independent variables? 4. What are the major limitations of the study? The first question relates to the broad issue of whether tfliis dissertation should be accepted as a "prOper" political 6 science dissertation. The other three questions all relate to important specific aspects of the work presented here. Policy Implementation Versus Political Science Our thesis with regard to this issue is that so long as policy analysis, the study of causes, processes and con- sequences of governmental activity, is regarded as a branch of political science, policy implementation, the study of the execution of public policy, should also be accepted as one of its sub—branches. We do not believe there is much opposition in the discipline to the view that policy analysis is a proper branch of political science. In fact, some claim that, fol- lowing the traditionalist and the behavioralist movements, political science today is rapidly shifting towards policy analysis.1 This is also evidenced by the increasing number of political scientists who are preoccupied with analytical- conceptual as well as empirical studies in policy analysis. Other indicators of the emergence of policy analysis as a proper branch of political science, if not as a discipline in itself, include publication of new journals in policy 2 analysis, the increase in the number of universities with 1See: Thomas R. Dye, Understanding Public Policy (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1972), Chapter 1. 2In particular: Policy Sciences (Amsterdam: Elsevier Scientific Publishing Company) and Policy Studies Journal (Urbana, Illinois: The Policy Studies Organization). I: ..4. CI c‘e 7 Ph.D. programs in public policy, and the number of under- graduate and graduate courses in public policy offered by departments of political science.1 What about policy implementation? Wouldn't the dra- matic changes in the status of policy analysis within polit- ical science imply the occurence of similar changes in one of its subdisciplines? The answer, as we will demonstrate more fully in Chapter II, is "no." Most policy scientists have limited the sc0pe of policy analysis to the study of the causes and effects of governmental activity. Implementation, or the "activity" itself, has strangely been left out. Dye defines the sc0pe of policy analysis as "the description and explanation of the causes and consequences of governmental activity."2 Sharkansky jumps from "determinants of public policy" to "evaluation of public policy."3 Ranney, although he uses the term "implementation" in passing, views the central concern of public policy analysis as formulation and evaluation of public policies.4 Other examples could be given where, in perhaps a pure 5 Eastonian tradition, emphasis is placed on the inputs, 1See: H. Hugo Heclo, "Policy Analysis", British Journal of Political Science, Volume 2, Part 1, January 1972, p. 83. sze, Understanding Public Policy, p. 3 3Ira Sharkansky, ed., Policy Analysis in Political Science (Chicago: Markham Publishing Company, 1970). 4Austin Ranney, ed., Political Science and Public Policy (Chicago: Markham Publishing Company, 1968), pp.3-21. 5See: David Easton, A Framework for Political Analysis (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1965). 0111 wi1 ;H ‘u( 2"; A."\ 8 outputs and the environment of a public policymaking system with little or no concern for what happens within the "black box." Needless to say, if one were to use the Eastonian paradigm, the place of policy implementation would be inside the black box. One of the few policy scientists to recognize a role for policy implementation within the field of policy analy- sis is Dror. In Public Policymaking Reexamined he offers an "optimal model of public policymaking."1 The Dror model includes three major stages--metapolicymaking, policymaking, and post-policymaking--and eighteen phases. Two of the phases under the post-policymaking stage are what Dror has called "motivating the executing of policy" and "executing the policy.‘ Although Dror's discussion of these two phases falls far short of systematic analytic treatment, his is one of the few non-empirical studies recognizing implementation as a major branch of policy analysis. If policy implementation were a straightforward process, i.e., if prior to implementation one could predict with great certainty what the outcomes would be, there might have been no need to study implementation. However, experi- ence has taught us that implementation does not always pro- ceed as specified during policy formulation or Operational- ization. Policies are sometimes modified to such a degree during implementation that one is not sure if it is the 1Yehezkel Dror, Public Policymaking Reexamined (Scranton, Pennsylvania: Chandler Publishing Company, 1968), Chapters 13 and 14. 9 initially formulated policy that is being implemented. Knowledge of the consequences of a policy allows one to compare its results with the original intent. However, this comparison does not tell one why and how the eventual consequences were attained. Without studying its imple- mentation all we can say about a policy is whether its intended effects were realized. Studying implementation helps us understand and explain why a particular policy succeeds or fails. In conclusion, policy implementation should be viewed as an important branch of policy analysis and political science. It is a much talked about but seldom studied field. Analytic studies of implementation are required not only for making improvements in the formulation and execution of pol- icies but also for explaining their observed consequences. Universality of the Generalizations Implementation of the handicapped policy in Head Start is the only policy studied in this dissertation. Therefore, from a purely statistical viewpoint, the empirical general- izations of this dissertation can be applied only to the Head Start program and its handicapped policy. From another standpoint, however, our findings have applicability beyond Head Start. Leaving aside the issues relating to data and empirical verification, the theoretical- conceptual chapter of the dissertation was prepared somewhat independently of the handicapped policy in Head Start. In 10 preparing this chapter we were more concerned with formula- tion of a conceptual scheme that would be applicable to all federally-formulated but locally-implemented human service programs such as Head Start, Neighborhood Youth Corps, Model. Cities and the Public Employment Program. Our first hand knowledge of the difficulties faced in implementing these programs at the community level had a lot to do with the crystallization of our thoughts and ideas. It is for this reason that we believe most of the arguments made here could be highly relevant to the implementation of other human ser- vice programs. Relevance of our generalizations to a policy area beyond human services, such as foreign policy or trans- portation policy is an issue we have not investigated. Universality of a generalization determines, to a great extent, the credibility of the theory from which the generalization has been derived. As argued by Stinchcombe, the credibility of a theory increases as the number of cases in which different empirical consequences of the theory 1 In this instance, Head Start provides only one increases. example of an empirical consequence. If the generalizations derived here hold true in other instances--as we expect they will in the case of federally-formulated, locally-implemented human service programs--the credibility of our conceptual model will increase accordingly. 1Arthur L. Stinchcombe, Constructing Social Theories (New York: Harcourt, Brace and World, Inc., 1968), Chapter Two. (1' (j ’4 (I) e"? a I J?" C AV‘LI l I (D 11 Research Design and Dependent and Independent Variables This study is based on a cross-sectional survey design with a sample of thirty-five Head Start programs visited in the spring of 1974 to collect data on the implementation of the handicapped policy. The design, therefore, suffers from the usual threats to internal and external validity that are common to most non-experimental designs that are based on small-sample survey research. The analytic/conceptual scheme used includes six vari- ables. Of these, one is the primary dependent variable, one is an intervening variable, and the other four are indepen- dent variables. The principal dependent variable is the level of implementation success and the intervening variable is the level of implementation effort. The independent vari- ables of the model are implementation potential, environmen- tal forces for implementation, policy support, and self- evaluation of capability. Four of the six variables are measured through indices, i.e., through a set of "secondary" variables or indicators, and the remaining two are measured directly. The hypothe- sized relationships concern (1) the effects of implementation potential and level of effort on implementation success and (2) the effects of environmental forces for implementation, policy support, and self-evaluation of capability on level of effort. In the case of the first relationship three and in the case of the second relationship two alternative 12 models are proposed. The explanatory power of each alterna- tive model is assessed using the Head Start data. Limitations of the Study Perhaps the most important limitation of the study is the weakness of the theoretical support for the conceptual model proposed. Expectancy theory in organizational psychol- ogy provides the only major theoretical base for the hypothe- sized relationships. Other than this, the model stands mainly on intuitive and empirical grounds. The second limitation of the study is the size of the sample upon which the empirical generalizations are based. Implementation of the handicapped policy in only thirty-five local Head Start programs is investigated. This alone limits the number of independent variables that can be considered in explaining the dependent variable. In addition, the size of the sample affects the magnitude of errors due to sampl- ing. The third limitation of the study has to do with the internal validity of the cross-sectional design. Although attempts were made to estimate some longitudinal data, the design lacks the desirable features of experimental and quasi-experimental research designs. These features include pre- and post-measurement, availability of a control group, and random assignment of programs to the two groups. These deficiencies of the design have been partially removed by adopting appropriate fieldwork procedures and by controling 13 for some variables through statistical means. Finally, the fourth limitation of the study relates to the validity of the indices used for measuring four of the variables. Without the aid of reference measures or indicators the process of index construction becomes highly judgmental. The four limitations discussed above, along with our earlier comments on generalizability, constitute the major limitations of the dissertation as we see them. Undoubt- edly there are other aspects of the study which could be included here as limitations. However, we believe that they are less major than the ones discussed. In any event they will be reviewed in greater depth in the remainder of the dissertation. This concludes our discussion of the purpose and the scope of the dissertation. In the next chapter we will review the literature pertinent to the study of policy implementation. L) CHAPTER II LITERATURE ON POLICY IMPLEMENTATION There is not much literature on policy implementation. 1 and This was reported previously by Pressman and Wildavsky the supplemental literature search effort we undertook did not lead to conclusions different from theirs. In light of this negative finding, our purposes in this chapter are to describe this literature search effort, to review the most important of the analytic/conceptual and empirical works we examined, and to identify the common variables used or pro- posed by others for explaining implementation success. The organization of the chapter is as follows. In the first section we summarize Pressman and Wildavsky's findings from a major bibliographic work on implementation and we follow this with a description of our additional search for literature relevant to the present work. In the second sec- tion we review what we consider to be the four most impor- tant empirical studies on implementation. The third and final section of the chapter is devoted to recapitulation of and conclusions from our literature review effort. 1Jeffrey L. Pressman and Aaron Wildavsky, Implementation (Berkeley, California: University of California Press, 1973). l4 15 Literature Search The only major bibliographic work on implementation is reported in Appendix 2 of Pressman and Wildavsky's Implementation. Expecting a large body of literature, Press- man and Wildavsky have conducted an extensive review of major works in disciplines closely related to implementation. They report the results of their investigation as follows: "Analytical study (as Opposed to mere mention) of implementation seems so eminently reasonable that few can imagine it does not exist. . . . (E)xcept for the few pieces mentioned in the body of this book, we have been unable to find any significant analytic work dealing with implementation. "How shall we persuade others that it is fruitless to look for a literature that does not exist, or that we cannot connect our work with nonexistent analyses by others? It is a little like searching for the little man who isn't there. It is always more difficult to establish a negative rather than posi- tive relationship. While we cannot hope to provide conclusive evidence (after all, there can always be something we have over- looked) we do hope to be persuasive beyond the mere assertion that there is no analytic literature on implementation."1 Pressman and Wildavsky present their findings from the extensive literature search under five headings: periodical indexes, poverty, civil rights, operations research, and public administration. In all instances they conclude that nothing of importance can be found in the works examined. In the case of public administration, the texts reviewed by Pressman and Wildavsky generally use the term "implementa- tion" in the introductory sections of the work for defining other terms. But implementation itself is never defined or explained. It is mostly used in the sense of "the end 11bid., pp. 166-167. 16 product of administrative efforts." In conclusion, the search conducted by Pressman and Wildavsky reveals that implementation is a much-talked-about but rarely-studied topic. Except for a few empirical works, which we review in the second section of this chapter, they have found no analytic studies of importance on policy implementation. We supplemented the Pressman-Wildavsky search reported above with a search of our own. Our purpose in conducting this investigation was to confirm Pressman and Wildavsky's findings and to carefully scan literature not covered by Pressman and Wildavsky. In the fipgp instance we undertook a library search using the subject catalogues of the Main Library at Michigan State University. We searched for titles under topics such as policy implementation, implementation, and execution. This search did not reveal anything different from the find- ings of Pressman and Wildavsky. Second, we conducted an in-depth review of some of the leading texts in public policy analysis. This review included, principally, the works by Dye, Dror, Rourke, Ranney, Sharkansky, Havemen and Margolis, Jones, Lindblom, Braybrooke 1 and Lindblom, Laswell and Kaplan, and Easton. Although none lThomas R. Dye, Understandinngublic Policy (Englewood Cliffs, New Jersey; Prentice-Hall, Inc., 1972); Yehezkel Dror, Public Policymaking Reexamined (Scranton, Pennsylvania: Chandler Publishing Company, 1968); Francis E. Rourke, Bureaucracy, Politics and Public Policy (Boston, Massachusetts: Little, Brown and Company, 1969); Austin Ranney, ed., Political Science and Public Policy (Chicago, Illinois; Markham Publish- ing Company, 1970); Robert H. Haveman and Julius Margolis, ed., Public Egpenditures and Policy Analysis (Chicago, Illinois: Markham Publishing 17 of these major works used the term "implementation", some referred to it as the "application" and others as the ”exe— cution" of a policy. Except for Dror's inclusion of imple- lementation as a major stage in his "optimal model of policy- making" we could not find an analytic-conceptual discussion of policy implementation in these works. SHEER! we turned our attention to policy evaluation, which, in the recent years has been emerging as an important discipline within policy analysis. The literature we reviewed in this area, such as works by Suchman, Caro, Weiss, I seemed to be more Wholey, et. al., and Rossi and Williams, concerned with questions of definition, measurement and design than the relationship of implementation to program success. Fourth, and finally, we conducted an extensive, article-by-article review of six major periodicals. This included all issues of four journals: Policy_Sciences, Company, 1970); Charles 0. Jones, An Introduction to the Study of Public Policy (Belmont, California: Wadsworth, 1970); Charles E. Lindblom, The Policy-Making Process (Englewood Cliffs, New Jersey: Prentice—Hall, Inc., 1968); David Braybrooke and Charles E. Lindblom, A Strategy of Deci- sion (New York, The Free Press, 1963); Harold Laswell and Abraham Kaplan, Power and Society (New Haven, Connecticut; Yale University Press, 1950); David Easton, A Framework for Political Analysis (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1965). 1Edward A. Suchman, Evaluative Research (New York, Russell Sage Foundation, 1967); Francis G. Caro, ed., Readings in Evaluation Research (New York, Russell Sage Foundation, 1971); Carol H. Weiss, Evaluation Research (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1972); Joseph S. Wholey, et al., Federal Evaluation Policy (Washington, D.C.: The Urban Institute, 1971); Peter H. Rossi and Walter Williams, eds., Evaluating Social Programs (New York: Seminar Press, 1972). 18 Policy Studies Journal, British Journal of Political Science, 1 and Public Choice. In addition, we reviewed the issues of The American Political Science Review published after 1968 and Public Administration Review published after 1969. This investigation revealed a few articles of relevance. We devote the remainder of this section to a review of four of these articles. First, in a recent article in Policy Sciences Thomas B. Smith presents a discussion and a conceptual model of the policy implementation process.2 Smith argues that the assumptions about implementation made in most policy studies (e.g., "once a policy is formulated, its implementation is automatic") is a major reason for the neglect of policy imple- mentation in the models constructed for studying policy- making. In Smith's model policies are viewed as serving a tension generating force in society. Accordingly, the four components of the policy implementation process--the ideal- ized policy, the implementing organization, the target group, and environmental factors--interact to produce "discrepancies and tensions." The tensions result in transactions, trans- actions sometimes generate institutionalization; and, "feed- back in the form of relieved tensions or increased tension is introduced back into the tension generating matrix from 1The first few issues of Public Choice were published under the title : Pcpers on Non-Market Decision Making. 2Thomas B. Smith, "The Policy Implementation Process," Policy Sciences, Volume 4, 1973, pp. 197-209. 19 transactions and institutions."1 The variables proposed by Smith for studying the four components of the policy implementation process include, among others, type of the policy, level of commitment of the government to the implementation of the policy, level of demand or support for the policy, characteristics of the target group, characteristics of the implementing organiza- tion, and intensity and care taken by the implementing agency to organize for the implementation. These and other variables, according to Smith, can be used to examine the within- and inter-component tensions and their consequences. Smith's argument falls far short of systematic formulation and none of his implied hypotheses have been verified empiri- cally. Second, in another study Thomas Smith reports his findings from an empirical analysis of individuals in policy formulation and policy implementation roles in New Zealand.2 Based on a sample of 119 individuals in 27 government depart- ments, the study compares policy formulators and implementors in terms of their social background, career patterns, educa— tion, job satisfaction, career ambition, work load, hierar- chial relations, and political saliency. Smith's principal findings are stated as follows: 1Ibid., p. 203. 2Thomas B. Smith, "Policy Roles: An Analysis of Policy Formula- tors and Policy Implementors," Policy Sciences, Volume 4, 1973, pp. 297-307. 20 "From the evidence of this study, the New Zealand administrative system does not allocate policy formulating roles to individuals different from those who implement policies. Differences between formulators and implementors in the New Zealand system appear to stem from the nature of the work of the two policy roles." The theoretical significance of these findings are questionable. The specific finding for New Zealand is per- haps more significant for elite theorists than for policy scientists. Third, in a recent article in Policy Sciences, Douglas Bunker, arguing for the view that implementation is an inte- gral part of the policymaking process proposes a heuristic model for studying different implementation strategies.2 In Bunker's model, which is based on Gergen's leverage-point concept,3 implementation is viewed as a problem of social persuasion. A three dimensional scheme is presented for "locating actors and selecting directions of preferred change to optimize probabilities of implementation."4 The three dimensions of the scheme refer to variables representing (1) the centrality of the issue for a particular actor, (2) the number and potency of political resources he has lIbid., p. 297. 2Douglas R. Bunker, "Policy Sciences Perspectives on Implemen— tation Processes," Policy Sciences, Volume 3, 1972. 3This concept is based on the argument that the implementation process can be influenced at a finite number of leverage points. Groups and individuals within an organization may also be conceived of as leverage points. See: Ibid., p. 75. Also see: K. Gergen, "Assessing the Leverage Points in the Process of Policy Formulation" in R. A. Bauer and K. H. Gergen, eds., The Study of Policy Formation (New York: The Free Press, 1968). 4Bunker, "Perspectives on Implementation Processes", p. 76. 21 available, and (3) the degree to which he is in agreement with the policy. Bunker also argues that policy implementa- tion in a federal framework requires multi-level participa- tion of four groups of participants: administrative- bureaucratic, political leadership, rational-analytic, and constituent elites. Successful implementation of a federal policy, according to Bunker, demands "both vertical (central office-field) coordination and the integration of multiple functional inputs."l Bunker's analysis, as he himself recog- nizes, has more heuristic than practical value. The fourth article we will cover here is a brief essay by Marshall pointing the need for research on implementation 2 "Political scien- of federal poverty and welfare programs. tists have given more attention to the formulation of federal policy than to its implementation. Thus we know much less about what happens after the legislative drama fades and the bureaucrats go to work to implement the legislation."3 According to Marshall there are three major issues in the study of implementation. The first issue concerns the results of implementing social policies; the second issue is about the characteristics of the implementation process itself; and the third issue relates to the factors 1Ibid., p. 78. 2Dale Rogers Marshall, "Implementation of Federal Poverty and ‘Welfare Policy: A Review Essay," Policy Studies Journal, Volume 2, No. 3, Spring 1974. 31bid., p. 152. 22 explaining variations in the way different cities respond to federal policy. After reviewing the extent of research conducted in these three areas, Marshall concludes his essay as follows: "Just as the pattern of policy implementation becomes familiar, so does the pattern of review essays with their concluding (self—serving?) pleas for more research. . . . But, alas, as is blatantly obvious from the amount of verbiage expended, it is much easier to call for studies than to do studies and stand the gaff of the methodologists attacking lack of rigor and the practitioners bemoaning documentation of the insignificant."l Perhaps Marshall's remarks are as good a point as any for concluding this section of the chapter. In the next section we will review some of the major empirical studies of imple- mentation. Major Empirical Studies of Implementation During the literature search stage of the preparation of this dissertation we were able to identify only four major empirical studies on policy implementation. These four studies, in the order they will be reviewed in this section, are: the Pressman-Wildavsky study of the implemen— tation of an experimental Economic Development Administra- tion project in Oakland, California; the Bailey and Mosher study of the implementation of The Elementary and Secondary Education Act of 1965; the Sudquist and Davis study of pro- gram coordination at the community level; and Martha Derthick's investigation of the implementation of a housing program for the poor. 11bid., p. 156. 23 There is a fifth study which we could have included in this section along with the above four, one which could per- haps be regarded as the "first" empirical study on policy implementation. This is Selznick's renowned work on the Tennessee Valley Authority (TVA).1 Although the Selznick study and the concepts it has introduced, such as cooptation, grass-roots administration, voluntary participation, and commitment, have not lost their relevance, they have been restated by others in forms more appropriate to today's pro- blems. Indeed, some of the studies we review below use a line of argument similar to Selznick's in explaining the variations in successful policy implementation. A Major Study of Implementation The only major work we have come across that carries the title "implementation" is the Pressman and Wildavsky study of the Economic Development Administration's (EDA) experimental project in Oakland, California.2 The Oakland Project, started in 1966, was the first project funded within EDA's major rebuilding program. "To help solve the problems of unemployment and racial unrest, the new program (was designed to) finance public works and business loans that would lead to the Creation of jobs for the unemployed, primarily black."3 Prior to implementation of the project 1Philip Selznick, TVA and the Grass Roots (Berkeley, California: University of California Press, 1949). 2Pressman and Wildavsky, Ipplementation. 31bid., p. 1. 24 it was announced by EDA that more than twenty-three million dollars in federal funds would be offered for public works grants and business loans in Oakland. EDA estimated at that time that the direct employment effect (without spinoffs) of these grants and loans would be about 3,000 additional jobs. Yet, in 1969, three years after the project began, only three million dollars had been spent and the total employment effect of the project was negligible. Pressman and Wildavsky present an insightful analysis of the apparent failure of the Oakland Project. Here we shall summarize the four reasons given by Pressman and Wildavsky. Figpp, implementation of the Oakland Project involved participation of a large number of agencies and individuals and each of the participants had a "distinctive perspective from which it viewed the EDA operation."1 Although none of the participants disagreed strongly with the major goals of the project, they differed in terms of the means of attain- ing them. They each had their own interests and preferences about the sequence of events to be followed. This intro- duced lengthy delays in clearances and agreements necessary to implement the program. Second, the number of "decision points" that had to be passed for the program to continue, the number of partici- pants involved at each major decision point, and the total 11616., p. 98. 25 number of agreements required presented major roadblocks to successful implementation. Third, the two major goals of the program, construc- tion of public works and creation of jobs for the hard-to- employ, required two separate decision paths and series of action. Delays in one decision path stalled the other. Fourth, the project experienced, what Pressman and Wildavsky call, "the emergence of unexpected decisions." As they state: "Not only was it necessary to complete two decision paths in order to achieve the EDA program goals; but within each path, the agency found that the number of decisions and clearances required was constantly growing. . . . The paths of required decisions . . . were soon characterized by more unexpected elements than expected ones: they were anything but straight lines leading directly to goals."1 In concluding the study Pressman and Wildavsky sum- marize their major finding as follows: "What had looked like a relatively simple, urgent, and direct program - involving one federal agency, one city, and a sub- stantial and immediate funding commitment - eventually involved numerous diverse participants and a much longer series of decisions than was planned. None of the participants actually disagreed with the goal of providing jobs for the minority unemployed, but their differing perspectives and senses of urgency made it difficult to translate broad substantive agreement into effective policy implementation."2 The significance of Pressman and Wildavsky's work does 11616., p. 112. 21bid., p. 113. 26 not lie in the methodological soundness of the explanation offered in the study. Instead, it lies in the fact that it is the first major empirical study carrying the title "imple- mentation." The study demonstrates the usefulness of study- ing policy implementation. From a theoretical standpoint, it introduces a few variables which could explain variations in implementation success in other instances. The study also illustrates the potential power of the case-study method in the formulation of hypotheses which could be fur- ther tested in other circumstances. Implementation of a Major Act: The Case of ESEA Another major empirical work on implementation is found in Bailey and Mosher's account of the 0.5. Office of Educa- tion's administration of The Elementary and Secondary Educa- tion Act of 1965 (ESEA).1 In contrast to the conceptual- analytic nature of the Pressman-Wildavsky study the Bailey and Mosher piece is more of a descriptive-historical under- taking. Although a lengthy chapter is devoted to the pro- cess of implementation,2 this section goes little beyond a description of the tasks and activities carried out during the implementation of the six titles of the act. Lack of data on successfulness of implementation seems to be the apparent reason for this emphasis on description of tasks 1Stephen K. Bailey and Edith K. Mosher, ESEA: The Office of Education Administers a Law (Syracuse, N.Y.: Syracuse University Press, 1968). 2Ibid., pp. 98-159. 27 and activities rather than explanation of variations in success. The conclusions reached by Bailey and Mosher about implementation are based on their assessment of the barriers to successful implementation. "Although each one of the Titles of ESEA produced its own particular difficulties and problems, the major issues that developed in the partnership of implementation cut across most, if not all of the titles."1 These major issues were: 0 difficulties in interpreting the legislative intent of the mandate; o lateness of funding; a difficulty of finding specialized staff; 0 paper work, red tape, and "federal control"; 0 difficulties created by interagency clearance requirements; and, o administrative and constitutional issues related to application of ESEA provisions to non—public schools. Bailey and Mosher present considerable qualitative data in describing the problems faced during implementation. Because of lack of data on successfulness, however, Bailey and Mosher were not able to test their hypotheses empirically. A Study on Coordination A recent study undertaken by the Brookings Institution concentrates on an important aspect of implementation: llbid., p. 188. 28 program coordination at the community level.1 The Brookings study does not explicitly deal with the problems of imple- menting a specific program. Instead, it addresses the ques- tion of coordinating the simultaneous implementation of sev- eral federal programs at the local level. The study is based on about 700 interviews conducted during 1967—69 with federal, state, and local officials administering federally sponsored programs, citizen leaders and informed observers in eight states. Sundquist and Davis zero in on problems of coordina- tion in "urban America" and compare them with coordination difficulties faced in implementing programs in "rural America." In the case of urban America, they study the failure of the Community Action Agencies as local coordi- nating structures and contrast it with the relatively greater success level achieved by Model Cities. For resolv- ing the coordination problems in rural America Sundquist and Davis suggest that the model to be adopted by federal, state, and local governments should be a "universal system of multi-county agencies with responsibility both for plan- ning and for facilitating action programs covering the entire range of community activities relating to economic and community development."2 In the Sundquist-Davis study coordination is viewed 1James L. Sundquist and David W. Davis, Making Federalism Work (Washington, D.C.: The Brookings Institution, 1969). 21bid., p. 231. as C81 fee 29 as a prerequisite to successful implementation. Indeed, the central premise of the study is that effective execution of federal programs depends heavily upon the competence of local institutions to plan, initiate and coordinate. Coordi- nation, according to Sundquist and Davis, is best achieved when it is assigned to a "neutral" agency instead of one with implementation responsibilities. The implication of this argument for policy implementation is that, given the abun- dance of federally-Sponsored, locally-implemented programs which are dependent on one another, successful implementation of these programs will require each urban community and rural area to establish a series of "neutral" coordination centers. A Comparative Study of Policy Implementation The final work we will cover in this section is the comparative study on implementation reported in Martha Derthick's New Towns In-Town.l Derthick conducted an in- depth study of a small federal program initiated during the Johnson Administration. The program was based on the idea of building model new communities on surplus, federally- owned land in metropolitan areas. Its aims were to provide housing for the poor and revitalize the nation's cities. Although the program seemed to have a quick and ambitious start, in three years it faced complete failure. In three of the seven local areas selected for program implementation 1Martha Derthick, New Towns In—Town (Washington, D.C.: The Urban Institute, 1972). th pr de (L H OT‘. Su. 30 the projects died outright. In the other four areas the projects faced eventual failure. Three years after Presi- dent Johnson's announcement of the program only three hun- dred housing units were under construction.1 Derthick's study mainly concentrates on the individual projects. The historical-analytic account of each project attempts to explain the reasons for the failure of the pro- jects and the program. Since the projects, taken as a group, do not show much variation in terms of successfulness, explanation of success in implementation is not possible through this study.2 By studying cases of failure, however, one gains insights into what might have made the program more successful. Derthick's analysis yields four reasons for the failure of the New Towns In-Town program:3 0 Failure, on the part of federal officials, to anticipate and counteract resistance to the projects from some local groups; 0 Lack of federal incentives to induce local governments to implement the program; 0 Inability of federal and local officials to reinforce one another and act effectively for generating public support, 1President Johnson announced the program on August 30, 1967. In his announcement he declared that the program would generate modern housing and related services for about 25,000 persons, in Washington, D.C., alone: See: Derthick, p. 7. 2This should not be viewed as a criticism of the Derthick study. Implementation of the program involved only seven projects; and Derthick studied all of them, i.e., the "total population" of projects. Needless to say, the researcher cannot be blamed for lack of variation in the dependent variable. 3Derthick, pp. 83—102. In I‘V. 31 overcoming opposition and building an administrative frame- work; . Statement of objectives that were too ambitious for attain- ment at the local level. In the author's own words: "In summary, the surplus lands program failed both because the federal government had limited influence at the local level and because it set impossibly high objectives. Its goals exceeded by far its capacity to achieve them. . . . Fundamentally, the program failed because of characteristics of the federal govern- ment that are associated with . . . its central position in the governmental system. These characteristics are the scale of its jurisdiction and its separation from the actual execution of domestic programs. Separation, in turn, results from the divi- sion of authority among governments in a federal system and the distance from the 'top' and the 'bottom' level of government hierarchy in a large, complex society."1 This concludes our review of the four major empirical studies on policy implementation. In the next section we will recapitulate what was presented earlier and analyze their implications for the present study. Conclusions from Past Work on Implementation Let us now summarize what we have covered in the two preceding sections and assess where we stand. First of all, we documented, rather laboriously, and demonstrated the fact that policy implementation has not been studied extensively. There are less than a dozen studies that can be claimed as policy implementation studies, some of those by stretching the definition and boundaries of the field and others by lowering the standards for scientific inquiry. 11bid., p. 93. (D L) "f C). (D [II 32 Secondly, we presented summaries/reviews of a few key works on implementation. In terms of their theoretical soundness perhaps none of the studies reviewed can pass a tough methodological test. A few of these were purely heuristic attempts at conceptualization. The others, the ones we have called "empirical,” were historical/analytical accounts of the reasons for the failure of one or more fed- eral programs. None of the studies examined included a "grand theory" of implementation; each followed a line of reasoning that seemed to be most suitable to the Specifics of the case under investigation. Finally, in our review of most of the studies we listed the key explanatory variables proposed by the inves— tigators. We did not use any rigid criteria in choosing the variables to be included in these lists. Instead, we relied heavily on the authors' assessment of the importance of a potentially explanatory factor. In the remainder of this chapter we will attempt to integrate the principal explanatory variables of the eight major works reviewed in the first two sections of this chap- ter. Our purpose in carrying out this exercise is to find out what commonalities exist between these eight, rather diverse studies. The procedure we followed is quite straightforward. First, we listed all the explanatory variables prOposed in each study. This yielded a list of about twenty-five vari- ables. Some of these were "composite variables", i.e., 33 they consisted of a combination of more than one "simple variable."1 In such instances we divided each composite variable into its simple variables. This increased the total number of variables to about thirty. We then examined this last set of variables in order to group them into broader categories. After a few iterations of this, what might be called "conceptual factor analysis," we ended up with five new variables. Despite the trial-and-error nature of the procedure used, we believe that these five categories adequately represent the host of explanatory variables pro— posed by the eight studies. The following is a list of the five new variables. 0 Level of governmental commitment o Straightforwardness of the implementation 0 Level of inter- and intra-agency agreement 0 Capability of the implementing agency 0 Local support for the policy The first variable, level of governmental commitment, refers to commitment to the implementation of the policy by govern- mental agencies, including the legislature, at levels higher than the implementing agency. In most instances this vari- able reflects the commitment of the federal government. It encompasses Derthick's "federal incentives," Smith's "fed- eral commitment,” and Bailey and Mosher's "funding level." 1We are using the term "simple variable" to refer to a variable with which a simple hypothesis can be formulated. A simple hypothesis is of the form "if A then B". 34 In addition, the commitment concept covers variables such as ”establishment of realistic (attainable) objectives," "clarification of policy intent," and "design and distribu- tion of clear guidelines." Straightfomardness of the implementation covers most of the variables proposed by Pressman and Wildavsky. These include "number of decision points to be passed," "number of partic- ipants involved at each decision point," and "total number of agreements required." This variable also covers Bailey and Mosher's "number of interagency clearance requirements." Level of inter- and intra-agency agreement is a variable encompassing a host of factors. In the first instance it refers to the stand taken by the leadership of the implement- ing organization vis-a-vis a positive advocate of the policy (Bunker). In the second instance, it refers to consensus within the implementing organization, i.e., agreement of the subordinates with the leadership. Finally, it refers to "differences in perspectives of the agencies involved in implementation" (Pressman and Wildavsky), "communication between and working together of federal and local agencies" (Derthick), and "existence of 'neutral' coordinating struc- tures" (Sundquist and Davis). The fourth variable, capability of the implementing agency, covers three of Smith's factors: "stability and structure of the implementing organization," ”qualifications of the implementing organization's staff," and "intensity and care taken by the implementing agency to organize for the 35 implementation of the policy." It also encompasses Bunker's notion of "number and potency of the resources available to the implementing agency" and Bailey and Mosher's "staff capability." The final variable, local support for the policy, refers to the level of support the policy receives from the members of the target group and related local agencies. Most of the studies reviewed proposed at least one support-related vari- able for explaining implementation success. Are these five variables independent of one another? We do not think so. Even though we attempted to make them as independent as possible, it is not difficult to construct conceptual schemes and a rationale to illustrate the partial dependence of the five variables and the concepts they rep- resent. Nevertheless, the important point is that during the reconstruction exercise we consciously tried to generate variable categories that are at least as independent of one another as in any other categorization. The significance of this exercise, in our opinion, lies in its heuristic value. By generating a set of vari— ables that covers the full range of concepts used by others to explain policy success we have a clearer idea of what has worked in other instances and what might work in the present undertaking. This concludes our review of the literature related to policy implementation. In the next chapter of the disserta- tion we present a conceptual/theoretical framework which is 36 partly based on the variables summarized above. In addition, we outline the theoretical support for the model and list the specific hypotheses which are later tested in Chapter V. CHAPTER III CONCEPTS AND THEORY In this chapter we introduce, describe, and discuss the major conceptual and theoretical issues which underlie our major purpose: explanation of variations in the level of success achieved in implementing a new public policy. We have divided the chapter into two broad sections. In the first section, entitled "Concepts," we discuss the definitions offered previously for the term "policy" and introduce a definition of our own. We follow this by a discussion of the role of policy implementation within the overall policy process. In the second section of the chapter, entitled "Theory," we concentrate on methodological and theoretical issues. First, we introduce a conceptual model which summarizes our hypotheses regarding explanation of implementation success. Second, we review the theoretical support for the conceptual model. In the third part of the section we define, describe and discuss the six variables of the model. In the fourth part we examine the postulated relationships between these six variables. In the final part of the section we review a set of methodological issues which concern the conceptual model and its theoretical bases. 37 38 Concepts The Term "Policy" The central concept of this dissertation is implementa- tion. However, implementation, in and of itself, has very little meaning. To assign it a meaning, "a verb like 'imple- ment' must be associated with an object like 'policy'."1 Therefore, it is essential that we arrive at a working defini- tion of "policy" before we can talk about implementation. Policy, as it is commonly used, refers to a strategy or a means adopted for achieving some prespecified goals. This coincides with the definition of policy given by Webster--"any governing principle, plan, or course of action"2--and the syn- onyms cited by Roget: "polity, principles; program, procedure, course, line, plan of action; platform."3 In everyday lan- guage, however, the term policy is used quite inconsistently. Sometimes we talk of a policy in the sense of a goal or a desirable outcome ("our policy is to minimize the tension in the Middle East"). In other instances policies refer to generalizations from our past behavior ("our policy is to train all caseworkers." meaning that we already did train them)- 1Pressman and Wildavsky, Implementation, p. xv. 2See: Webster's New World Dictionary, College Edition (1966). Webster includes two other definitions under "policy": "1. political wisdom or cunning; diplomacy; procedure; artfulness. 2. wise, expedient, or crafty conduct of management." The one we have quoted in the text, and the one we are interested in, is the third definition given by Webster. 3See: Roget's International Thesaurus, Third Edition (1962). 39 There are also a wide variety of definitions of policy offered by political and policy scientists. These range from simple definitions such as Sharkansky's ("policies are actions taken by governments")1 to complex and sometimes ambiguous formulations such as the one by Reynolds: "A policy is a move or a series of moves (acts, speeches, etc.) made from a position in a practice which, other things being equal, could be reasonably expected to have at least one of the following results: (1) a modification of that practice or some other practice or (2) a lack of modification of that practice or some other practice when a modification could be reasonably expected to occur were the move not made or (3) the "2 creation of another practice. In between these two extremes lie a variety of definitions. Laswell and Kaplan refer to it as "a projected program of goal values and practices."3 Braybrooke and Lindblom view it as a term encompassing "both decisionmaking and the course that policies take as a result of intercorrelations among decisions and/or in which (policy- making) incorporates certain political processes, in addition to analytic processes, into the determination of courses of action."4 According to James Robinson a policy "refers to 1Ira Sharkansky, "The Political Scientist and Policy Analysis," in Ira Sharkansky, ed., Policy Analysis in Political Science (Chicago: Markham Publishing Companyg'l970), p. 1. James F. Reynolds, "Policy Science: A Conceptual and Method- ological Analysis," Policy Sciences, Volume 6, Number 1, March 1975, p.7. Harold Laswell and Abraham Kaplan, Power and Society (New Haven, Connecticut: Yale University Press, 1950), p. 71. 4 David Braybrooke and Charles E. Lindblom, A Strategy of Decision (New York: The Free Press, 1963), footnote 1, p. 249. .d ,.._.-4 40 goals (objectives, ends) of any social system, the means chosen to effectuate these goals, and the consequences of the means, i.e., the actual distribution of values."1 Dror views policies as guidelines: "Public policymaking is a very complex, dynamic process whose various components make different contributions to it. It decides major guidelines for action directed at the future, mainly by governmental organs. These guidelines (policies) formally aim at achiev- ing what is in the public interest by the best possible means."2 To Easton policies are "decision rules adopted by authorities as a guide to behavior. . . . In this sense, policies would be just a term for a kind of authoritative verbal output."3 Finally, Austin Ranney defines a policy in terms of its five principal components (elements): (1) a particular object or set of objects, (2) a desired course of events, (3) a selected line of action, (4) a declaration of intent, and (5) an imple- mentation of intent.4 The eight definitions of policy quoted above are not all 1James Robinson, Con ress and Foreign Policy-Making (Homewood , Illinois: Dorsey Press, 2), p. 3. 2 Yehezkel Dror, Ppplic Policymaking Reexamined (Scranton, Pennsylvania: Chandler Publishing Company, 1968), p. 12. 3 David Easton, A S stems Analysis of Political Life (New York: John Wiley and Sons, 1965;, p. 358, quoted in‘Rdbert‘H.’Salisbury, "The Analysis of Public Policy: A Search for Theories and Roles," in Austin Ranney, ed., Political Science and Public Policy (Chicago: Markham Publishing Company, 1968), p. 152. 4 Austin Ranney, "The Study of Policy Content: A Framework for Choice," in Austin Ranney, ed., Political Science and Public Policy (Chicago: Markham Publishing Company, 1968), pp. 6-7. 41 contradictory. Perhaps the most common element among these definitions is that policies refer to courses of action, which is precisely how Webster and Roget define the term. The major drawback of the set of definitions considered here is that they are not consistent with each other, i.e., a course of action which is a policy according to one defini- tion may fail to meet the specifications of another. In view of this inconsistency--and in light of the centrality of the concept in our work--we need to formulate a single definition for use throughout the dissertation. The definition we chose to use here reflects our notion of policy. Of the eight definitions quoted earlier it is closest to Ranney's view of policy. Yet there are some major differences between his definition and ours. These differences will become clear in the course of our discus- sion of the definition which follows its formal statement. A policy is a statement, usually written, which has been (or could potentially be) acted upon by an authoritative decision- making body and which at least includes (1) a target object or group of'objects, (2) a set of’goals or objectives regard- ing one or more aspects or characteristics of these objects, and (3) a well-defined strategy or course of’action. We will nOW'elaborate on this definition by parti- tioning it into its components (concepts and phrases) and discussing their significance and implications. Qpc, policies are statements and a great majority of them are in the form of written statements, such as those included in legislative acts or resolutions of decision- making bodies ( a city council, board of commissioners of a 42 county, board of trustees of a university, executive board of a corporation, etc.). Unwritten policies are not uncom- mon and they may relate to significant as well as insig- nificant matters. Thus, a small corporation may have an unwritten "marketing policy" or a social welfare department may have an unwritten "outreach and recruitment policy." However, most of the policies studied by political and policy scientists are in the form of formal, written state- ments. Egg, policies are not always the final outcomes of decision-making processes. Some policies are "considered" during the decision-making process but not adopted at the end. Only one of these alternative policies is eventually implemented. The alternatives considered, if they meet the other specifications of the definition, are also regarded as "policies." The policy selected during the decision- making process is what we call "the adopted policy." 32533, the decision-making body acting on the policy question has to have an authority relationship to the target group of objects defined by the policy. If there is no authority relationship, i.e., if the group of objects do not fall within the legitimate span of control of the decision- making body, the statement adopted by the group is not consid- ered a policy, it only reflects the opinions and desires of the decision-making body in question. Fppp, if the decision-making body is one generally accepted as a "public decision-making body." the adopted fl ,-..-4 43 pol icy--and the alternative policies considered--are referred to as "public policies." Five, policies are designed to affect, i.e., induce changes in one or more characteristics of, a well-defined target object or group of objects. "Well-defined" here means that there are clear rules with which one can deter- mine if a particular object falls within or outside the target group. The target group may consist of the whole nation or any of its subgroups (elderly, youth, unemployed, handicapped, etc.), institutions or organizations (small business organizations, universities, non-profit hospitals, etc - ) , physical or geographic entities (regions, river basins, highways, urban areas, etc.), or, decision-making rules (the filibuster rule in the Senate, equal rights for women, the age of majority, etc.) . Six, policies are purposive, or goal-directed state— ments. The goals or objectives relate to aspects (communi- cation structure, coordination, resource allocation, etc.) or Characteristics (employability, health, education, etc.) Of the target group of objects. The goals and objectives define a desired state of affairs which is to be attained through implementation of the policy. Without goals it becomes quite impossible to judge policy success. Seven, a policy prescribes a particular strategy or course of action which is to be followed during its implemen- tation- "Course of action" refers to a sequence of acts to be PeI‘formed by a prespecified group of actors in a 44 particular manner and within a given time frame. Well- defined policies indicate what (course of action), who (what actors), when (time-frame) , and how (sequence of acts). Most policies also specificy (directly or indirectly) why the particular course of action was chosen from among the alter- natives considered. Strategies are the "backbones" of policies. They contain the principal "if X, then Y"-type cause—effect rela- tionships implicit within the policy statement. In a sense strategies form a bridge between scientific knowledge and its application to the resolution of real-life problems. Without a scientific base, strategies would be little more than "random stabs in the dark." What we have called "strategy" above is what is gener- rally referred to in the policy analysis literature as "P01 icy content"1 and most of the typologies Froman calls "abStract policy categories” are in effect typologies of this construct.2 The so called "traditional policy 1See the essays by Ranney, Van Dyke, Froman, and Salisbury in “183111: Ranney, ed., Political Science and Public Polipy. 2Included among these are: (1) the distinction made by Berelson, Lazarsfeld, and McPhee between "style" and "position" issues; (2) the "mateI‘ial" and "symbolic" satisfaction dichotomy proposed by Edelman; (3) Huntington's distinction between "structural" and "strategic" policy issues; (4) the classification of policy issues as "distributive," "regu- lat°rY," and "redistributive" suggested by Lowi; (5) Salisbury's reclas- 31fication of the Lowi trichotomy (distributive, regulatory, redistribu- give, and self-regulatory); (6) Froman's distinction between "areal" and segmental" city policies; and (7) the distinction made by game theorists between "zero-sum" and "non-zero—sum" policies. For sources of these clas- Sifications see the essays by Lewis A. Froman and Robert H. Salisbury in Austin Ranney, ed., Political Science and Public Polipy. 45 cat-.egories"l in fact refer to the "boundaries" of the policy as opposed to its "content." Policy Implementation and the Policy Process Implementation is but one link in a temporal chain com- monly referred to as the "policy process." Our purpose here is to describe and discuss the place of policy implementa- tion within this process. We are using the term "process" here in the same sense it is used in industry. When we talk of the production process we generally refer to a series of stages the manu- facturing of a good goes through. Such processes usually start with a "design" stage, followed by a sequence of Stages which reflect the conversion of inputs into outputs through the application of a given technology. The process usually ends with a stage such as "quality control" and/or "Packaging." At this point another process, marketing, takes over and carries the product until it reaches the final consumer. 1Included among these are the following: (1) categories defined 1;; 8ubstantive policy areas (health, manpower, ed ucation, etc.); categories defined by the specific decision-making authority 1nV01ved with the policy (presidential, congressional, city-council, etc.) ; (3) categories defined by the level of the decision-making authority within the governmental hierarchy (federal, state, regional, 10°31. etc.); (4) categories defined by the type of target group :2?“ resource, institutional, environmental, etc.); (5) categories (ageged by the subgroups within a particular type of target group 0 b handicapped, poor, youth, etc.); and (6) categories defined d: time periods (Nixon Administration years, post-Vietnam era, Augiission period, space age, etc.). See Froman's essay in n Ranney, ed., Political Science and Public Policy. I | 46 The term process, therefore, usually refers to a set of sequentially-related stages. The stages are also time- related and the outputs of one constitute part of the inputs of another. The process is a transformation process. In- puts, such as managerial, labor, material, and technology are transformed into desirable (and some undesirable) out- puts. The policy process, at least in concept, is not sig- nificantly different from the production process of industry. Much like the production process, the policy process refers to a transformation, conversion of inputs into desirable (and some undesirable) outputs through the machinery of the government. It is the policies that guide this transforma- tion. They specify the desirable outcomes (through the goals) and how these outcomes are to be achieved (through the strategies). However, policies, in and of themselves, are no more than designs for action, much like the designs of production processes. Policies as designs (and as state- ments on pieces of paper) do not produce the desired changes in people's lives. It is the implementation of policies that produce these changes. Where does the policy process start and where does it end? This is not an easy question to answer. In most instances specification of the starting point of the process and the stages that follow it are up to the investigator and the purposes of his investigation. In our case we view the policy process as consisting of the following four major 47 components: 0 formulation 0 operationalization o implementation 0 evaluation Briefly, policy fomulation refers to emergence of a pol- icy issue and development of a strategy which addresses that issue. It covers the period up to the adoption of a partic- ular policy and, in the case of most federal policies, it ends after the legislative branch of government passes an act on the policy issue in question. Policy operationalization bridges the gap between formulation and implementation. It refers to preparation of rules, orders, procedures, plans, and manuals which refine the adopted policy and guide its implementation. Policy implementation, in most general terms, refers to execution of operationalized policies. Finally, policy evaluation involves assessment of the effectiveness of the policy in question. How do we define implementation. Better yet, can we define it any other way than in Webster or Roget? Perhaps not. Synonyms provided by Webster ("to carry into effect; fulfill; accomplish")1 and Roget (carry out, execute, pro— duce, accomplish, perform)2 fully convey the meaning of the verb "to implement." The thing that is executed, 1See: Webster's New World Dictionary, Collcgp Edition. (1966). 2These are selected synonyms from Roget's International Thesaurus, Third Edition. (1962). 48 carried out, or accomplished, of course, is a policy as defined and described during its operationalization. When operationalization is extremely detailed, the degree of flexibility given the implementing agency is usually limited. In such instances an implementor is frequently judged in terms of whether he has "complied with" the instructions outlined in the guidelines. In a sense, therefore, implementation is like carrying out an order. The more specific the order, the less is up to the implemen- tor to decide what actions he should take. In military or pseudo-military settings both the incentives for success- fully implementing an order and the sanctions against not complying are well specified. In these settings the imple- mentor knows, prior to implementation, how he will be judged on his performance. Complying or not complying with the order becomes a matter of individual decision making with known outcomes. In most human service organizations, however, the incentives for compliance and the sanctions against noncom- pliance are not known to the implementors. In fact, in most instances, it is not even known what constitutes successful implementation. This usually reflects a deficiency in the components of the policy process which precede implementation. This brings us to the question of what constitutes successful implementation. Before answering this we must make a distinction between policy evaluation and evaluation of policy implementation. This distinction is necessary 49 because it is quite feasible for an implementor to be extremely successful in implementing a policy which, over- all, does not generate any of the desired effects. This is the case of "bad policy-good implementation". The implemen- tor is not at fault. Failure of the policy is either due to poor policy formulation, improper operationalization, or both. Thus, in evaluating the implementation of a policy success/failure is judged by comparing the immediate out- comes of the implementation with the original implementation objectives. In evaluating the policy overall, however, we compare the effects--usually called "impacts"--of the pol- icy with the overall goals the policy was designed to attain in the first place. Theory The principal aim of this dissertation is explanation of the differences between successful and unsuccessful imple- mentation of a public policy. We want to find out why some implementors are more successful than others and what factors account for the observed variations in their levels of success. In Chapter II we reviewed some of the previous work on implementation and summarized the key factors proposed by others as potential explanator variables. It became apparent during the discussion in Chapter II that most of the explana- tions attempted were nonsystematic and/or nonempirical. The 50 major empirical studies covered concentrated on the implemen- tation of a policy in one or a few cases, and, therefore, did not allow formulation of strongly supported empirical generalizations. Single program case studies or comparative analysis of policy implementation in a few cities, despite their value in helping us understand some of the factors accounting for poor implementation, have not provided us with broad, systematic, and sound conceptual structures which can be used for generating a set of testable hypoth- eses about implementation. Our aim in this section is to attempt to fill this gap and pave the road for empirical confirmation or disconfirmation of some hypotheses about the specific relationships between implementation success and some explanatory variables. We have divided this section into five parts. In the first part we introduce a conceptual model of policy imple- mentation. In the second part we review the main elements of expectancy theory in organizational psychology which provides strong theoretical support for the conceptual model. In the third part of the section we define, describe, and discuss the six variables of the model. In the fourth part we examine each of the two functional relationships and for- mulate the alternative hypotheses that are tested in Chapter V. The final part of the section is devoted to clarifica- tion of a set of methodological issues concerning the con- ceptual model and its theoretical bases. 51 The Conceptual Model The primary concern of policy scientists is explanation of policy success. Understanding the ingredients which make a policy more successful is essential for increasing the responsiveness and the effectiveness of governmental activity. Policy success, when simply defined, refers to the difference between goals and accomplishments. The closer the eventual effects of the policy are to its original goals, the more successful is the policy. The goals and objectives referred to here are the ones identified and defined during policy formulation. They reflect the overall direction of the policy and the specific targets to be attained at the end of its implementation. These goals may be, and sometimes are, different from the goals used in assessing the successfulness of policy implementation. The overall goals of the policy may be stated over a longer time span than the implementation goals. Or, the overall goals may reflect national targets, whereas the implementation goals may vary between the implementing agencies. And, sometimes, the policy goals may be modified during policy Operationalization, resulting in the genera- tion of a different set of goals for implementation. The term policy success, when defined as above, does not speci- fically refer to successful implementation and a sucessfully implemented policy may turn out to be a failure when evalu- ated in terms of the overall policy goals. Thus, when we refer to level of policy success we talk about the overall 52 successfulness of the policy as measured against the policy goals identified during policy formulation. Alternatively, we use the term level of implementation puccess (LIS) when referring to the successfulness of policy implementation as measured against the implementation goals. Our primary emphasis in the dissertation, of course, is on explanation of the variations in LIS. The overall success of the policy depends to a large extent upon what takes place between statement of the goals and the measurement of the results. If the policy formu- lated during the first stage of the policy process is deemed to be effective, i.e., if there is strong theoretical, causal, and other rationale strongly supporting the point that the formulated policy will lead to the desired effects, the chances for the success of the policy will be large, other things being equal. Similarly, if the formu- lated policy is interpreted, Operationalized, and otherwise put into action with minimum deviation from its original specifications and with full understanding of the circum- stances which may potentially affect its implementation, the chances for policy success will similarly be large, other things held constant. Finally, if the policy is implemented exactly as it is operationalized and if the implementation objectives are attained or surpassed, the policy's chances for success will be substantially enhanced . . . again, other things being equal. Successful accomplish- ment of the goals assigned to each stage, therefore, will 53 ensure the attainment of the policy's overall goals. And, because of the strong linkages between the three stages, a failure in any one stage may increase the chances for a sizable discrepancy between the desired goals and the actual results. If the failure takes place during the formulation stage, in all likelihood, an ineffective policy will be imple- mented. If operationalization is done unsuccessfully, what is eventually implemented is likely to be different from the original policy. If however, the first two stages are com- pleted successfully but implementation turns out to be a failure, the original goals of the policy will still remain far ahead of the actual results. Thus, the overall policy success is largely affected by the level of success achieved at each of the three stages of the process. As the preceding discussion indicates, our conception of the policy process--its stages, interrelationships between the stages, and the role played by each stage in the achieve- ment of policy success—-necessitates, at the very least, three different explanations: explanation of the variations in the levels of success reached during policy formulation, policy operationalization, and policy implementation. Here, we are principally concerned with the last of these three explanations. We recognize the importance of policy formula- tion and operationalization in the achievement of policy success. However, our study of these two aspects of the policy process does not go beyond this recognition. How does one go about explaining the overall policy 54 success and/or level of success at each stage empirically? Here we will discuss three alternatives. The first one is what we will call the policy-specific approach. Here the emphasis is on explaining the success- fulness of a single policy. Explanation of success during policy formulation usually takes the form of a case study, since there is not any variation as such to study. Opera- tionalization, if done at a central level, is also studied using the case study method. Of course, decentralized operationalization can be studied by collecting and analyz- ing data on each operationalizing unit. The heaviest emphasis while using the policy-specific approach is placed on implementation since there are usually quite a few imple- menting agencies executing the policy. We have entitled the second approach the pplicy area approach. Here explanation centers on several individual policies within a broad policy area, such as housing, educa- tion, welfare, etc. Because of the presence of several policies, formulation, operationalization and implementation can all be studied using empirical methods that capture the variation between policies and between agencies. The third approach entails empirical investigation of all policies at a given level of government, regardless of the specific policy area, and can be entitled the ppp: policy-specific approach. The emphasis here is to reach generalizations concerning policy formulation, operational- ization, or implementation at the federal, state, or local k="'—' i 55 level. As in the previous approach, variations among poli- cies, policy areas, and agencies allow the analyst to employ a wide variety of empirical methods. In this dissertation our interest lies with the first of these, the policy-specific approach. We are involved with empirical investigation of the implementation of a single policy across. several implementing agencies. Formulation and operationalization were done centrally, by the Congress and the Office of Child Development (HEW), respectively. The fact that formulation and operationalization were done cen- trally implies that the effects of policy formulation and policy operationalization variables on LIS can be assumed to have remained as constants. Each implementing agency can be viewed as having been subjected to the same effects of formulation and oper- ationalization, and, thus, these two variables can be assumed to have been controlled for. This implies that in attempting to explain LIS we need not be concerned, except for descriptive purposes, with what has taken place during the early stages of the policy process. The types of vari- ables we should consider for the explanation should reflect salient aspects of the implementation stage alone. The con- ceptual model we introduce below satisfies this requirement. We contend that level of implementation success, our primary dependent variable, is a function of two key explan- atory variables. These are implementation potential (IPOT) and level of effort (EFRT). Briefly, implementation potential refers to the overall capacity of the agency to implement 56 the policy in the same manner as it is operationalized. It indicates the presence/absence of a set of capacity-related factors which would be required for effective and efficient execution of the policy. Level of effort refers to the amount of policy-specific effort expended during the implementation. Our secondary dependent variable is level of effort and our contention is that it can be explained in terms of three independent variables: environmental forces for implementation (ENVR), policy support (PSUP), and self- evaluation of capability (CAPB). Environmental forces for implementation refers to the net balance of the forces for and against the implementation of the policy. It reflects the net effects of the pressures and influences directed towards the implementing agency by other agencies and inter- ested groups of individuals. Policy support refers to the position taken by the leadership of the implementing agency vis-a-vis a positive advocate of the policy. It reflects the extent to which the leadership of the implementing agency agrees with and supports the policy. Self-evaluation of capability refers to the perception of the leadership of the implementing agency regarding the chances that the agency can implement the policy successfully. The conceptual model which connects the four indepen- dent variables and one intervening variable to LIS, our primary dependent variable, is illustrated in Figure 1. Detailed description and discussion of the variables and of the hypothesized relationships between variables are the . -—-'v 57 subject matters of two later parts in this section. l—ENVR :J’\‘—_#l; IPOT l PSUP EFRT Figure 1. The proposed conceptual model. We now move to a discussion of the theoretical support for this conceptual model. Theoretical Support for the Conceptual Model The conceptual model we just introduced is new for explaining an aSpect of organizational performance. However, it is not altogether new because an analytical scheme simi- lar to this one has long been recognized as one of the most powerful theories explaining variations in employee perfor- mance in organizational settings. This is the so-called expectancy theory in organizational psychology and it deals with motivational determinants of the performance of an indi- vidual. Our conceptual model, therefore, can be viewed as an extension of this theory from the level of an individual to the level of an organization responsible for implementing a public policy. Porter and Lawler and Vroom trace the origins of S8 expectancy theory to Tolman and Lewin, two of the early advocates of cognitive theories of behavior.1 Basic to these theories is the argument that (l) behavior is purpose- ful and goal-directed, (2) individuals have "beliefs" or "expectations" about future events, in particular, about the potential outcomes of the actions they could take, and (3) individuals behave in a manner which pppy think will lead to the attainment of goals they value. "Expectancy" is a cognitive concept and it refers to a person's subjective assessment of the chances that a par- ticular act will yield a particular outcome. The same con- cept has sometimes been called "subjective probability," indicating that expectancies may take on values between zero (for the case where the individual thinks that the specific act will never result in the valued outcome) and one (he is certain that his act will be followed by the outcome). Expectancies enter the theory indirectly, as one of the determinants of "motivation," the key explanatory vari— able of the model. In addition to motivation the theory uses two other predictor variables to explain variations in performance: ability and organizational characteristics. Ability, in general, refers to the individual's capacity to 1See: Lyman W. Porter and Edward E. Lawler, Managcrial Attitudes gpd Performance (New York: Dorsey-Irvin, 1968) p. 9; Victor H. Vroom, lflgrk and Motivation (New York: John Wiley and Sons, Inc., 1964) pp.l3-l4. The Specific works of Tolman and Lewin cited by Porter and Lawler and Vroom are: E.C. Tolman, Purposive Behavior in Animals and Men (New York: Century, 1932) and K. Lewin, The Conccptual Representation and the Mea- surement of Psychological Forces (Durham, N.C.: Duke University Press, 1938). 59 perform a given task or a set of tasks. Defined as such, the term covers a host of individual characteristics. Included among these are intellectual factors (verbal, numeric, sym- bolic, spatial skills and the like), manual factors (such as strength and dexterity), and personality traits. It reflects the capabilities of the individual, i.e., those characteris- tics which enable him to accomplish a specific task. Organizational variables enter the model in the form of goal-setting and extrinsic and intrinsic reward mechanisms. Motivation, the most interesting of the three predictor variables, refers to what we have called "effort," a charac- teristic which reflects how vigorously the individual will use his abilities in performing a given task. Others have given different names to this concept. Vroom reports that it is similar to the Lewinian concept of "force," Tolman's "performance vector," Atkinson's "aroused motivation," Luce's "subjective expected utility," and Rotter's "behavior potential."l Two aspects of motivation, its intensity and its direction, determine how much "force" the individual will exert in performing the given task. Intensity of the individual's motivation reflects his overall state of arousal to engage in any behavior. Direction shows whether his motivation is channeled towards performance of the Specific task in question. Intensity of the individual's motivation is measured 1See Vroom, Work and Motivation, p. 18 for full reference on these concepts. 60 in terms of the attractiveness to the individual of various outcomes that could follow the performance of the task. This amount of attractiveness is referred to in the model as the valence of a given outcome. There are two types of goals or outcomes which might motivate an individual. Extrinsic outcomes refer to factors such as pay and promo- tion. Intrinsic outcomes refer to factors such as feelings of accomplishment and satisfaction. The types of extrinsic and intrinsic outcomes that are perceived as valent may dif- fer from individual to individual. Thus, the theory sug- gests that by focusing on the valence of outcomes one can assess the individual's intensity of motivation to perform the task. The directionality of motivation, or whether the individual will direct his efforts towards high performance, is viewed as a function of two "contingencies" which can be both expressed in probability terms. Cummings and Schwab1 summarize these as follows: . . . First, the employee must feel that if he attempts to be a high performer, he has a good chance of doing so. This con- tingency is referred to as the individual's expectancy percep- tion, or effbrt-perfbrmance probability. A worker might feel that his effort-performance probability is low because he lacks the ability to perform well, because the task itself is poorly structured for successful performance, or for a variety of other reasons. "The second contingency is actually a set of contingencies that refers to the individual's expectations that high performance will result in various valent intrinsic and/or extrinsic out- comes. These contingencies are frequently referred to as the 1L. L. Cummings and Donald P. Schwab, Performance in Orgppizations (Glenview, Illinois: Scott, Foresman and Company, 1973) . ..‘ 61 individual's instrumentality perceptions, or perfbrmance-reward probabilities."1 Thus, according to expectancy theory, the individual first examines the task to be performed and assesses the chances that he will be successful if he attempts to per- form it. He then calculates the chances that his high per- formance will lead to a Specified set of extrinsic and intrinsic outcomes. Finally, he evaluates the level of attractiveness of these outcomes to himself. If either or both probabilities are low or if the individual does not see the outcomes to be positively attractive to him, the "motivation" portion of the theory predicts that the individ- ual would not be a high performer. There is a large body of literature relating to one or more aspects of expectancy theory. Unlike most other theories of performance, expectancy theory is general enough to be applicable to a wide variety of situations. According to Cummings and Schwab, the empirical evidence on the pre- dictive ability of expectancy theory is quite promising: "The research conducted on expectancy theory to date looks very promising for the basic motivational hypotheses of the theory. Clearly, additional research is needed, and limitations in the theory may eventually surface which are not evident at the present time. Nevertheless, the evidence already available strongly suggests that an expectancy theory approach to under- standing aep predicting organizational performance behavior is fruitful." Findings from one of the more recent empirical studies (31 the predictive power of expectancy theory are reported by 1Ibid., p. 30. 2Ibid., p. 33. 62 Porter and Lawler. In this investigation Porter and Lawler tested a set of propositions derived from a conceptual model similar to ours, using survey data collected from close to 500 managers in seven organizations. Three of these organiza- tions were divisions of state governments. Specific variables on which data were collected included: manager's attitudes towards their pay, their perception of the type of role behavior required for success on the job and the degree of need satisfaction provided by the job, self-rating of job per- formance and effort expended, and superior's ranking of job performance and effort expended. Only limited data were col- lected on abilities. Porter and Lawler summarize their empirical findings as follows: " . . . Managers who saw pay closely tied to performance factors received higher performance and effort ratings than managers who did not see such a close relationship (i.e., who had lower effort-reward probabilities). . . . Furthermore, the strongest relationships between perceptions of pay being based on perfor- mance factors and our measures of effort existed for those man- agers who attached the greatest importance to pay as a reward. This latter finding not only supports the model's contention that both value of'reward and perceived effbrt-reward probability are involved in determining a manager's effort, but it also indicates that the form of the combination of these first two variables in the model is interactive rather than additive." " . . . role perception-performance relationships were stronger for managers who were seen as exerting high effort (i.e., were "highly motivated") compared to those seen as exerting relatively low effort. Thus, not only is the notion of a combined effect of effbrt and role perceptions on perfbrmance supported, but also the results indicate the relationship may well be an interactive one. In other words, if extremely high effort were to be combined with extremely inaccurate role perceptions, the prediction would be that the subsequent performance would be evaluated as rela— tively ineffective." 1Porter and Lawler, Managerial Attitudes and Performance, pp.l60—l61. 63 On the effects of ability and motivation on performance Vroom reports: "The evidence which does exist suggests rather strongly that ability and motivation, as typically measured or manipulated, do not have independent effects on performance but rather interact with one another."1 The empirical studies cited by Vroom include the following: 0 Elizabeth French's study of the joint effects of intelli- gence and achievement motivation on the problem solving success of airmen; o Fleishman's experimental study on the effects of ability and motivation on the performance of 400 air force trainees in a complex coordination task; 0 Vroom's study of the effects of participation in decision making on supervisor's satisfaction and performance; . Wyatt's experiments on the effects of economic incentives on the level of performance achieved by fenale workers in a candy factory.2 Clearly, most of these studies are isolated attempts to test one or more aspects of expectancy theory. Although the "universality" of the theory has not been empirically established, the extent of empirical support from different settings indicates that the major propositions of the theory (i.e., those relating (l) ability and motivation to perfor- mance and (2) valance, expectancy, and instrumentality to 'motivation) have held in most test situations. There are quite a few similarities between the struc- ture of expectancy theory and the conceptual model we 1Vroom, Work and Motivation, p. 200. 2Full references for these studies can be found in Vroom's Work and Motivation, pp. 200-204. 64 introduced earlier. In particular, there are four parallels between the components of expectancy theory and the vari- ables encompassed by our conceptual model. First, both structures contain a variable which ref- lects capability. Expectancy theory talks about task-ori- ented abilities of the individual. In our model this is paralleled by the concept of implementation potential which, to a large measure, reflects the agency's capability to implement the given policy. Second, the notion of motivation in expectancy theory is very similar to the concept of effort used in our model. Expectancy theory predicts that highly motivated individuals will be more successful than those with low motivation, other things being equal. we predict that, controling for their implementation potential, agencies with higher levels of policy-specific effort will attain a greater level of implementation success than those with low effort. Third, the concept of valence in expectancy theory resembles our notion of policy support. Expectancy theory refers to the attractiveness to the individual of a given goal or outcome; we are talking about the level of support (bf a given policy by the leaders of the implementing agency. Expectancy theory argues that if the goal or outcome is highly attractive to the individual, he will be highly moti— Vated to perform the task. We are postulating that if the leadership of the agency support the policy strongly they wilhl take apprOpriate actions to put forth the effort 65 necessary to implement it fully. Fourth, and finally, the effort-performance probability of expectancy theory is parallelled by our notion of self- evaluation of capability. Expectancy theory suggests that individuals who feel that they can, with high probability, successfully accomplish the given task are more likely to perform better than those who view this probability as low. According to our conceptualization, agencies which believe that they have the capabilities to implement the policy are more likely to exert the necessary effort than those who have less confidence in their capacity. We realize that similarities between the variables and the structure of expectancy theory and those of our concep- tual model cannot alone be regarded as constituting full the- oretical justification for our explanatory scheme. This is especially true since expectancy theory has been formulated for explaining the performance of an individual, not the success of an agency in implementing a public policy. How- ever, extension of a "micro" theory to a "macro"-level is not uncommon and, if the micro theory happens to have been empirically verified in a number of different situations, such an extension may offer the most promosing starting Point for theory building at the macro-level. This completes our discussion of the theoretical sup- POrt for the conceptual model. We now turn to a detailed desoription and discussion of each of the six variables included in the conceptual model. A 66 Variables of the Model Level of implementation success This variable refers to how much successful goal achievement has been accomplished as a result of the imple- mentation of the policy in question. As indicated earlier in this section, the yardstick used for assessing the level of goal achievement is the set of implementation goals iden- tified and defined during policy operationalization. In many instances, as in the case of the specific policy studied here, these goals coincide with the policy goals identified during policy formulation. Implementation success is a measure of organizational performance concerning the execution of a specific public policy. If the agency's mission does not extend beyond the execution of a single public policy, implementation success and organizational performance become identical. On the other hand, if the agency in question is responsible for simultaneous implementation of two or more public policies, the agency's success level in implementing one policy may differ from its success in executing the other policies. In general, however, the overall organizational performance (Df the agency is strongly correlated with the average suc- cess level it attains in implementing all the policies it is mandated to execute. Assessment of the policy's implementation requires "Omltcome" rather than "impact" indicators. Briefly, the term "iInpacts" refers to lasting changes in the lives of people A 67 which have come about as a result of implementation of the policy in question. Outcomes, on the other hand, refer to the immediate (and more visible) effects of the policy. To give an example, the number of hours a participant has been exposed to skill training may be one of the outcomes of a manpower program whereas the program's impact would most likely be measured in terms of the change in the partici- pant's employability that can be attributed to the program in question. Similarly, the immediate outcome of a social services program might be the number of units of counseling and other services provided. The impact of such a program, however, is usually assessed in terms of the extent to which these services have made the recipients economically self-sufficient or less dependent on the welfare system. The policy goals (or desired impact statements) estab- lished during the policy formulation stage are translated into implementation objectives (or desired outcome statements) during policy operationalization. Thus successfulness of the implementation is purely a matter of level of attainment of the policy's implementation outcomes and not its desired impacts. ,Egvel of effort By the term effort we are referring to the amount of "energy" expended by the implementing agency in the execution Of' the policy in question. In simple terms, with this vari- able we are attempting to find a response to the question, 'Tkrw hard is the agency trying to implement the policy?" A 68 The kind of effort we are talking about is human effort, made up of the individual efforts of persons affiliated with the agency--managers and nonmanagers; professionals, para- plrofessionals, and nonprofessionals; volunteers and nonvolun- t<11 (kindergarten or first grade). . . . These programs are primarily for- «children of age three up to the age the child enters the school system, but may include some younger children." See: U.S. Department Of I‘Iealth, Education and Welfare, Head Start Child Development Programs: M'a-mual Of Policies and Instructions (Washington, D.C.: Office of Chil-(fl Development, U.S. Department Of HEW, 1967), p. 4. 93 areas. The average cost Of serving each child in the Full Year 1972 program was $1,118. Figures for the Summer 1972 program were as follows: There were 434 programs in Operation, enrolling 86,400 chil— dren. The summer program staff consisted Of 5,700 profes- sionals, 9,000 paraprofessionals and 17,738 volunteers. Average cost per child in the Summer 1972 program was $230. In summary, Head Start was a program that started "big" and remained SO throughout its ten-year history. Its imme- diate acceptance and continued funding are indicators Of its popularity. ItS SCOpe is also illustrated by the fact that approximately two per cent Of the present population Of the U.S. has attended Head Start sometime during the last ten years. The origin Of Head Start dates back tO the Title II-A Of the Economic Opportunity Act Of 1964, as amended. Like other programs administered by the Office Of Economic Oppor- tunity (OEO) it was designed in such a way as tO give local communities broad flexibility in program planning and com- munity involvement. The program called for "maximum feasible participation" Of the parents Of children. The laissez-faire attitude of CEO in designing and implementing the program led to wide variations among Head Start centers.1 030 spon- sorship Of the program continued until mid-1969. Then, the responsibility to administer Project Head Start was dele- gated tO The Office Of Child Development (OCD), a 1See: Smith and Bissell, Report Analysis, pp. 56—60. 94 newly-created agency under HEW'S Office Of the Secretary. Initially, OCD'S management Of the program was not different at all from that Of OEO. The same "Manual" was used tO guide local Operations, no major policy changes were made. In short, during the first years Of OCD sponsor- ship local program Operations were not any different from under OEO. During 1970 and in the following years OCD initiated several major efforts tO strengthen Head Start. The first one Of these was a comprehensive monitoring program launched in 1970. This was followed by a significant undertaking, called the "Head Start Improvement and Innovation Program" (I and I), which aimed at generating improvements in local program performance. Attempts were made to clarify Head Start program goals; a set Of "performance standards" were drafted to induce local programs tO evaluate and improve their programs; training and technical assistance activities were-expanded; local programs were Offered several "program n1 Options. several experimental projects were funded in the tradition Of earlier experimental efforts such as Follow 1There were five basic Options Offered: (1) the "Standard" Head Start model (the five—day-per-week, center-based classroom format), (2) the "variations in center attendance" model (serving some or all children on a less than five-day—per-week basis), (3) the "double ses- sions" model (scheduling two classes Of children per day), (4) the "home-based" model (serving children in their home, with the parent as the primary factor in the child's development), and (5) the "locally- designed Options" model (local programs designing and proposing their own "model"). See: U.S. Department of HEW, OCD Notice N-30-334-1 on Program Options for Project Head Start, (Washington, D.C.: U.S. Department Of HEW, Office Of Child Development, 1972). 95 Through, Planned Variations, Health Start, Parent and Child Centers, and Home Start. The initial concept of Head Start, however, remained the same. Concept Head Start, as its name implies, is based upon the simple concept that preschool children from disadvantaged families, like their parents, suffer from various ills Of poverty and that giving them a "head start" may help break the vicious circle these children would most likely find themselves in while growing up and after becoming adults. This head start was to be given the preschoolers through a comprehensive child development program which aims at bring- ing about a greater degree Of social competence in disadvan- taged children. "Social competence. as used here, refers to "the child's everyday effectiveness in dealing with his environment and later responsibilities in school and life."1 The term, according to Head Start literature, "takes into account the interrelatedness of cognitive and intellectual development, physical and mental health, nutritional need, and other factors that enable a child to function Optimally."2 Thus, the rationale underlying the Head Start concept is based on a line Of reasoning Similar to the following: o All children have certain basic needs (such as physical 1U.S. Department Of HEW, OCD Notice N-30-364-1 on Head Start Program Performance Standards, (Washington, D.C.: U.S. Department of HEWJ, Office of Child Development, 1973, p. 6). 2 Ibid., p. 6. in PIO 96 and mental health, balanced nutrition, cognitive and intellectual growth, and social and emotional health). 0 Preschool children whose basic needs are not met are less likely to be socially competent. 0 Preschool children from disadvantaged familities are less likely to have their basic needs met than children from more advantaged families. O Therefore, unless an intervention is made tO meet their basic needs, preschool children from disadvantaged families are less likely tO be socially competent. 0 Therefore, if we want all our preschoolers to be socially competent, there is a need for an intervention program such as Head Start. Our Simplified version Of its concept is also reflected in the program goals for Project Head Start. These goals provide for: "A. The improvement Of the child's health and physical abilities. B. The encouragement Of self-confidence, spontaneity, curiosity, and self discipline which will assist in the development Of the child's social and emotional health. C. The enhancement Of the child's mental processes and skills with particular attention to conceptual and verbal skills. D. The establishment Of patterns and expectations Of success for the child, which will create a climate Of confidence for his present and future learning efforts and overall development. E. An increase in the ability of the child and his family to relate to each other and to others in a loving and support- ing manner. F. The enhancement of the sense of dignity and self-worth within the child and his family." .AS implied by these goals, Head Start's approach is inter- ciisciplinary and comprehensive, allowing for the provision (Df a broad range Of services. In addition, parental and Community involvement are viewed as essential elements Of —; 1Ibid., p. 7; also see: Head Start Manual, pp. 2-3. 97 the Head Start approach. Parental involvement is given high priority because Of the belief that the family is the prin- cipal influence on the child's development. Community involvement is required for the reason, among others, that Head Start, alone, does not have the resources required to provide the broad range Of services a disadvantaged child might need. Organization Head Start's organization and the principal guidelines governing its Operation are fully described in the Head Start Manual and the OCD Transmittal Notices which revise some Of the material in the Manual.1 Here we provide a brief summary Of some of the pertinent aspects Of Operating a local program as background to studying the problem at hand. 1. Sponsorship.-—Loca1 Head Start programs are usually Sponsored by Community Action Agencies established under Section 204 and 205 of the Economic Opportunity Act, school systems, or other non—profit agencies. The applicant agency is referred to as the "grantee agency" and need not neces- sarily Operate a program. A grantee agency may subcontract all or part Of the operation Of the program to one or more "delegate agencies." 2. Funding.--Applications for Head Start grants are made to and reviewed by one of the ten Regional Offices Of the Office of Child Development. Applicants are usually required to finance twenty per cent of the total cost Of the program from non-federal sources. Part or all of the twenty per cent can be met through non-federal in—kind contributions, such as use Of space and facilities in non- federally owned buildings. 1Another useful background document is the following: U.S. Depart- ment of Health, Education and Welfare, Project Head Start 1969-70: A Descriptive Report Of Programs and Participants (Washington, D.C.: Office Of Child Development, U.S. Department of HEW, 1972). 3. 98 Types of programs.--Head Start is primarily for preschool children of age 3 and up. Summer Head Start programs operate mostly during the summer months and are only for children eligible for kindergarten or first grade. The minimum length of a summer program is 120 hours and it must operate at least 15 hours per week. Full Year Head Start programs may Operate for a period of 8 to 12 months for either part of a day or a full day. The minimum length Of the weekly program is also at least 15 hours. Eligibility.--Head Start is primarily for children from poor families. Children from families that are on welfare or that have an annual family income that is less than the fed- erally established poverty line are eligible for Head Start. Up to ten per cent of the children enrolled in each class can be from non-poor families. Parent Participation.--Each Head Start program is required to have an "effective" parent participation program. Parents may be involved in Head Start in one or more Of the following ways: o participation in program-level decision making (in such capacities as a member of the Policy Advisory Group or an elected member of the Parent Committee) 0 participation in classroom activities as paid employees, volunteers or observers; 0 working with their child in cooperation with Head Start staff; 0 participation in educational or other activities for parents which they have helped to develop. Program components.--Head Start has four basic program com- ponents: education, social services, parent involvement, and health services. Health services are further broken into four subcomponents: medical, dental, mental health, and nutrition. Each of these basic services is headed up by a program director. Other activities such as recruit- ment, volunteer activities, and career development are each administered through a coordinator. Other program Staff include teachers (usually one per 15 children), teacher aides (usually two per 15 children, one paid and one volun- tary), nurses, health aides, social workers, community aides, and cooks. 99 The Policy The Legislation On September 19, 1972, President Nixon signed into law the Economic Opportunity Amendments of 1972. With this Sign- ing, the new law, P.L. 92-424, Opened a new chapter in the efforts Of the federal government to serve the handicapped. The act contained the following provision: "The Secretary of Health, Education and Welfare shall establish policies and procedures designed to assure that not less than 10 per centum of the total number of enroll- ment opportunities in the Nation in the Headstart program shall be available for handicapped children (as defined in paragraph (1) of section 602 of the Elementary and Secondary Education Act of 1965, as amended) and that services shall be provided to meet their special needs. The Secretary shall implement his responsibilities under this paragraph in such a manner as not to exclude from any project any child who was participating in the program during the fiscal year ending June 30, 1972. Within six months after the date of enactment of this Act, and at least annually thereafter, the Secretary shall report to the Congress on the status of handicapped children in Headstart programs, including the number of children being served, their handicapping conditions, and the services being provided such children."1 A definition for the term "handicapped children" in the legislation is provided in the Elementary and Secondary Education Act of 1965, as amended: "The term "handicapped children" means mentally retarded, hard of hearing, deaf, speech impaired, visually handi- capped, seriously emotionally disturbed, crippled, or other health impaired children who by reason thereof require special education and related services."2 With this legislation the Congress made eligible clOse t<> 40,000 disadvantaged and handicapped preschool children _‘ (1 1Economic Opportunity Amendments of 1972, sec. 3(b), 36 Stat. 683 972). 2Education of the Handicapped Act, sec. 602, 84 Stat. 175 (1970). 100 Of ages three through five for participation in Head Start programs throughout the nation. Prior to this legislation Head Start guidelines did not specifically'exclude handicapped children from participation in the program. And, indeed, some local Head Start programs did enroll several disadvantaged handicapped preschoolers and served them along with the non- handicapped. However, with this Congressional mandate the Office Of Child Development was being told that they must make at least ten per cent Of the Head Start enrollment Opportunities available to handicapped children. LaVor traces the legislative history Of P.L. 92-424 to 1969 and the House Education and Labor Committee. In 1969 and in 1970 the Committee considered H. R. 13520 and later, H. R. 19362, a comprehensive child development act. The latter bill contained a provision directing the states to provide programs for the handicapped and to assure that at least seven per cent of the annual allotment Of each state 1 This bill was reported out Of are used for such programs. the Select Education Subcommittee before the adjournment Of the 9lst Congress, but no final action was taken on the bill during that session. In 1971, while the House Subcommittee was still working on a comprehensive child development legislation, the Senate was considering another child development proposal sponsored by Senator Mondale. Originally this bill, S. 1512, 1Martin L. LaVor, "Economic Opportunity Amendments of 1972, Public Law 92-424," Exceptional Children, Nov. 1972, pp. 249-253. 101 contained no language on behalf of the handicapped. An amend- ment to S. 1512 Offered by Senator Prouty contained the fol- lowing language: " . . . not less than 10 per cent of apprOpriate funds under Section 513 (c) be reserved by the Secretary in order to guarantee that handicapped children are included in child care programs." S. 1512 was incorporated into the Economic Opportunity Act legislation (S. 2007), but in December 1971 the President vetoed the entire bill. It was not until June 29, 1972 that the Senate passed S. 3010, which included an amendment con- taining the language present in P.L. 92-424. According to LaVor the rationale for the Senate action was summed up in the Committee report Of S. 2007: "The history of Headstart clearly shows that severely handi- capped children have been systematically excluded from pro- grams and, in fact, children with only moderate handicaps have generally been refused access to such services. These refusals have normally been based on the feeling that the national program is not primarily oriented toward treating handicapping conditions, and expertise is not available at the local level for developing effective programs. Thus, those children with the greatest need for help, and those parents with the heaviest burden to bear, are denied the help to which they should be entitled. . . . We find it dis- couraging that this program, created to help the children in need, has excluded or failed to help a significant portion of that population." There are two major implications Of this new legisla- tion. First, the Congressional mandate makes about 38,000 handicapped children eligible to receive federal assistance through the Head Start program. Prior tO this action, ——___ 11bid., p. 250. 21bid., p. 250. 102 according to an estimate by LaVor, less than 25,000 of the total population of about one million handicapped children under the age Of Six were receiving any type Of service at all as a result Of Federal funds.1 Thus the new law more than doubles the number Of handicapped preschool children eligible to receive federal assistance. Secondly, with the passage Of the new law a new concept was introduced in the Federal approach to the handicapped. For the first time a federal program for the handicapped had ng_funds earmarked for Special services to the handicapped. The ten per cent enrollment mandate was to be met within the budgetary limitations Of the regular Head Start grant. Head Start programs were expected to make the necessary arrange- ments with other community agencies to meet the Special needs Of handicapped children enrolled in their centers. The legislation clearly Spells out what outcome was tO be accomplished by HEW. It does not, however, specify 22!, HEW was going to go about enrolling and serving the handicap- ped without any additional funds. The mechanics of "how to do it" were left up to the Secretary Of HEW. Furthermore, 1Ibid., p. 253. LaVor's estimates are based on total dollar expenditures for preschool handicapped children in all Federal programs. According to his account, only four federal programs had any expendi- tures at all for handicapped preschool children: Title I of the Elemen- tary and Secondary Education Act ($465,000 per year for preschool handi- capped), Title VI-B of the Elementary and Secondary Education Act ($2,800,000 per year for preschool handicapped), P.L. 89-313, Children in State Supported Schools ($3,264,000 per year for preschool handicap— ped), and P.L. 90-538, Handicapped Children's Early Education Assistance 4Act ($7,500,000 per year for preschool handicapped). See: Ibid., II. 251. 103 the Secretary was mandated to report back to the Congress by March 19, 1973, within six months after the date Of enactment Of the Act, and at least annually thereafter on the status Of handicapped children in Head Start. Operationalization by OCD Operationalization Of the new policy by OCD can be best understood by reviewing the Transmittal Notice it sent tO local programs in Head Start services to handicapped chil- dren.1 This Transmittal Notice outlines the revisions tO be made in the Head Start Manual in lieu of the new law and, as such, it summarizes the manner in which the policy has -been Operationalized by HEW. The Notice starts with the claim that Head Start has never discriminated against handicapped children: "Head Start has always had a national policy of enrollment for all eligible children, including handicapped children. . . . As noted in the Head Start Manual of September 1967, 'Head Start encourages the inclusion of mentally or physically handicapped preschool children in an integrated setting with other Head Start children.‘ . . . "Mere recently, the Office Of Child Development has mounted initiatives to give priority to services to handicapped children. These relate, in part, to the overall thrust of the Head Start Improvement and Innovation effort which emphasizes the importance of individualizing services to meet the unique needs and potential of every child partic- ipating in the program." This is followed by a statement that explains OCD'S 10.8. Department of Health, Education and Welfare, OCD Notice N—30-331—l on Head Start Services to Handicapped Children (Washington, D.C.: U.S. Dept. of HEW, Office of Child Development, 1973). 2Ibid., p. i. 104 understanding of the reasons for the scarcity Of handicapped children in Head Start: " . . . there is reason to believe that there are substantial numbers of eligible children who are not registered in Head Start because families are not aware that they are accepted or because local programs do not encourage their participa- tion. Some grantees, contrary to Head Start policy, have even discouraged the participation of handicapped children."1 The Notice, according tO OCD, ”sets forth the policy to imple- ment this legislative mandate and requirements affecting each "2 grantee. Sanctions against not conforming with the con- tents Of the Notice are spelled out as follows: "This issu- ance constitutes Head Start policy and compliance is required as a condition Of further funding."3 Salient aspects Of the policy set by OCD for implement- ing the mandate, as indicated in this Notice, can be summed up in the following manner: 0 Eggional Offices.--The ten per cent mandate is passed down to the ten Regional Offices of OCD. Thus, each Regional Office is required to ensure an average of at least ten per cent enrollment in their geographic area. Furthermore, local programs are encouraged to "negotiate" with their respective Regional Office on their handicapped enrollment targets based on the make up of their target handicapped population, resources and capabilities of the program to serve the handicapped, and desires of the parents to enroll their children in Head Start. o Comprehensiveness of service delivery.--Local programs are asked to ensure that the full range of services normally available to non-handicapped children are made available to handicapped children as well. 1Ibid., p. 11. 2Ibid., p. 11. 3Ibid., p. 1. 105 o Integration.-Local programs are asked to serve the handi- capped in an integrated fashion with the non-handicapped. Exceptions to this policy could be made if it is judged that an integrated setting would not benefit the child. o Outreach and recruitment.--Head Start grantees are asked to develop, in cooperation with other community groups and agencies, plans and procedures for identifying, attract- ing and enrolling handicapped children from their respective service areas. o Screenipg and diagnosis.--Local programs are required to develop procedures for diagnosing a child's handicap and for assessing his special service needs. This is also to be accomplished in cooperation with other community agencies. These procedures are to cover all handicapping conditions. o Scheduling.-OCD set a three-phased schedule for implement- ing the new mandate. In Phase I local programs were asked to give immediate priority to handicapped children in fil- ling their normal enrollment vacancies. In Phase II, Summer 1973 Head Start Programs were asked to make comprehensive plans for serving handicapped children. In Phase III, Full Year Head Start Programs were required to implement the mandate to the fullest extent. In addition to these, local programs were reminded not to drop non-handicapped children from the program in order to make room for handicapped children. Other provisions of the OCD policy included a warning against mislabeling of chil- dren, a clause about the conditions under which a handicapped child could be terminated, and a reminder that some Of the program Options might be more suitable in serving the handi- capped than the standard Head Start model. Conspicuously missing from the OCD policy are specific definitions Of the handicapping conditions included in the mandate and "how to do it" guides for planning and implement- ing the congressional requirement. The National Office of OCD delegated the implementation responsibility to the regions and the regions, in turn, delegated it to each .— .1 {ad 106 grantee. Given the non-specific nature Of the policy guide- lines, this meant that the local programs were given full responsibility and flexibility in interpreting and implement- ing the mandate. In March 1973, as per the reporting requirements of the legislation, OCD presented its first Annual Report on the status Of handicapped children in Head Start to the Congress.1 The Report, which was principally based on responses Of 712 out Of 1,000 Head Start grantees to a mail-out survey con- ducted in August-September 1972, concluded that "roughly four to five per cent Of total Head Start enrollment con- sistS Of children with Special needs. . . . The principal handicapping conditions reported in the survey were Speech impaired (31%), seriously emotionally disturbed (14%), and mentally retarded (8%)."2 The Report did not include any other significant findings than these twO. Detailed analy- ses Of the survey data were not submitted to the Congress, nor were they made available to the general public. OCD'S First Annual Report to the Congress, having been prepared on the basis Of whatever data were available at the time of the Signing Of the legislation (September 1972) or could be gathered within Six months from this date, did not 1U.S. Department of Health, Education and Welfare, Head Start Services to Handicapped Children - First Annual Report of the U.S. Depart- ment of Health,_Education and Welfare to the Congress of the United States on Services Provided to Handicapped Children in Project Head Start (washington, D.C.: U.S. Dept. of HEW, Office of Child Development, 1973). 2Ib1d., pp. 7-8. 107 adequately assess the true status Of the handicapped in Head Start. Validity Of the reported figures was not assessed. Furthermore, in light Of OCD'S three-phased schedule for implementing the mandate, full compliance with the Congressional requirement was not expected tO occur before the start Of the 1973-74 program year. Thus, OCD started looking at the Second Annual Report to the Congress, which was scheduled for submission in or before March 1974, as the first comprehensive report on Head Start's efforts on behalf Of the handicapped. Thus, in order to have the data required for preparing the Second Annual Report col- lected OCD decided to procure the services Of an indepen- dent contractor and issued a Request for Proposals (RFP) for a major study on Head Start. The Project RFP 53-73-HEW-OS, entitled "Evaluation Of the Handi- capped Effort In the Head Start Program", was actually made up Of Six separate RFPS combined in one package. Studies on Six major aspects Of the handicapped effort, which were originally planned as separate projects with separate RFPS, were combined at the last minute into one RFP. Each of the Six tOpics was made a "major task" of this new project and a budget Of close to one-half million dollars was allocated for its completion. OCD estimated that completion Of the project would require about nine person-years of effort over a Sixteen month period starting 108 in July 1973. Following an extensive review process, OCD selected the Systems Research Incorporated-Syracuse University pro- 1 posal from among the responses to the RFP. HEW Signed a contract with Systems Research Incorporated (SRI), the prime contractor. SRI, in turn, subcontracted portions of the work to Syracuse University.2 1See: "Technical Proposal - Response to RFP 53-73-HEW-OS - Eval— uation of the Handicapped Effort in the Head Start Program" (proposal submitted to the Department of Health, Education and Welfare by Systems Research Incorporated, Lansing, Michigan, May 24, 1973). 2Systems Research Incorporated-a Lansing, Michigan based consult- ing firm with previous experience in management of large scale governs mental projects, evaluation of public programs, survey research, and computer applications--was to carry the bulk of the study. The Project Director, Mr. Alan F. Bogatay, was a vice-president of SRI. In addi- tion to Mr. Bogatay, the SRI study team included Mr. Selcuk Ozgediz, a Co-Manager of the study team, Mk. William B. Baucom, chief of the Special Surveys Group, Mr. Allan D. Dale, chief Of the Information Systems Development Group, and several other professional staff members. The Syracuse University study team consisted almost exclusively of the staff of the Division of Special Education and Rehabilitation. Professor Gail Ensher served as a Co-Manager Of the study team. Professors Robert Bogdan, Daniel Sage, and Burton Blatt headed up the Assessment and Evaluation Group, the Cost Analysis Group, and the Policy Analysis Group, respectively. Professor Blatt, Director of the Division of Special Education and Rehabilitation and of the Center for Human Policy at Syracuse University, also served as the chairman of the Senior Consultant Group assembled for the project. Serving with Professor Blatt in this group were some of the leading authorities on child development and special education. Among the persons in this advisory group were Dr. Edward Newman (former Commissioner of the Rehabilitation Services Administration at HEW), Professor Seymour Sarason (Yale University), Professor Merle Karnes (University of Illinois), Professor Howard Spicker (Indiana University), Professor Frank Garfunkel (Boston University), Professor Wolf Wolfensburger (Syracuse University), Dr. Julius Richmond, and Dr. John Johnson. In addition, graduate students from Syracuse University handled the bulk Of the fieldwork. 109 Major tasks Of the project The project consisted Of six major tasks, as indicated earlier. Two Of these tasks, Tasks II and III, involved extensive data gathering and analysis. The other four tasks relied heavily on the data generated in Tasks II and III. The data used in this dissertation were gathered, almost exclusively, within Task III. The following is a description Of the pusposes and the scope Of the three major tasks Of the project which are relevant to our work. Task II The primary purpose Of this task was tO make sure that the information required by Congress regarding the status of handicapped children in Head Start is collected and ana- lyzed in time tO develop the Annual Report due in March 1974. In addition, data collected in Task II provided a basis Of selection Of Head Start programs visited for in-depth assess- ment during Task III. Task II also helped establish several key parameters Of program costs analyzed in Task IV. Task II was completed in essentially seven stages. In the first stage, immediately after the contract award, a mail- Out survey was conducted to assess the status Of the handi- capped effort in the Summer. 1973 programs. During the second stage, definitions for the nine handicapping conditions were developed and a "master" list Of Head Start grantees and delegate agencies was developed. In the third stage, 110 the survey instrument and accompanying instructions were developed, pretested, and cleared with the Office Of Manage- ment and Budget. During the fourth stage the questionnaires were mailed (October 1974). The fifth stage involved cod- ing, keypunching, verifying, programming, processing, and analysis Of the returns. In the sixth stage two special telephone surveys were conducted. First, a random sample Of about 100 of the responding grantees and delegate agen- cies were telephone interviewed tO assess the test-retest reliability Of the survey instrument. Second, a telephone interview was conducted with each non-reSponding grantee to find out if the non-respondents differed from the respond- ing programs in terms of their handicapped effort. Finally, in the seventh stage, a draft and a final report was pre- pared and submitted to OCD. The draft report was used by OCD in preparing their Annual Report tO Congress.1 Task III The primary purpose Of Task III was tO collect in- depth, on-Site information on the implementation Of the Con- gressional mandate SO that an assessment could be made of the ways Of improving the implementation. Task III involved site visits to three types Of programs: a set of regular Head Start programs, a set of special "experimental 1See: "The Status of Handicapped Children in Head Start - Final Report Based on Surveys of Summer 1973 and Full Year 1973-74 Head Start Programs" (report submitted to the Office of Child Development by Systems Research Incorporated, October 1974). 111 projects," and a set Of "exemplary projects." The experi- mental projects, they were fourteen in all, were regular Head Start programs which had received special funding from either OCD or the Bureau Of the Education for the Handicapped for experimenting with and demonstrating new approaches to providing comprehensive services to pre-school handicapped children in a program setting with non-handicapped children. The exemplary projects consisted, primarily, Of ten non- Head Start preschool enrichment programs which could Offer new models for serving the handicapped within Head Start. Task III site visits were conducted in two rounds. In the first round, which took place during November-Decem- ber 1974, visits were made to sixteen regular Head Start programs and eleven experimental projects. During the second round, which was conducted in April-May 1974, thirty- Six regular Head Start programs, three experimental projects, and ten exemplary projects were visited. Approaches to the first and the second round Of visits to regular Head Start programs differed substantially. In the first round the visits were exploratory and Open-ended. Participant Observation was the principal technique used and data collection centered around eleven major "areas Of inqui- ry." An Open-ended interview guide was used to guide the field Observations. In contrast with the first round visits, the second round visits were more structured and they concentrated on better defined issues. A total Of thirty-six regular Head 112 Start programs were visited in this round. Since the data used in this dissertation comes primarily from these thirty— six programs, the procedures used in selecting them are dis- cussed in detail in Appendix A. Task V This task involved an extensive review Of the need for and the supply Of services for preschool handicapped chil- dren in the country. The overall aim Of the task waS to pro- vide OCD with data that can be used for future planning and policymaking. A literature review was conducted on serving the preschool handicapped and estimates were made Of the numbers Of preschool handicapped children by handicapping condition. Data collected from the exemplary programs were studied to identify some Of the best current approaches to serving the preschool handicapped. Also, an extensive telephone and mail survey Of agencies serving the preschool handicapped was conducted in each Of the fifty states. Finally, the data generated in the other tasks Of the study were examined by the Senior Consultant Group and their impressions and recommendations were forwarded to OCD.1 This completes our review Of the major tasks of the project that are relevant to our work. We now turn our attention to some Of the major findings from the study. 1See: "Statement on Policy Recommendations to the Office of Child Development" (statement prepared and signed by the members Of the Senior Consultant Group, September 1974). 113 Major findings We will not go over the detailed findings from all the tasks Of the project. Instead we will summarize the most important, and the most contraversial, findings from the work done by SRI and Syracuse University. Some Of these findings have led tO debates within as well as outside the project team. Let us first review the national survey conducted by SRI. The survey of Full Year 1973-74 programs yielded a response rate Of about 79%, with at least one response from 83% Of all grantees.1 Since a separate telephone survey was conducted with the non-respondents, if one combines the responses to key items Obtained through this survey with the responses tO the mail survey, the overall response rate for key items Of data reaches approximately 90%. A third survey, conducted in the form Of telephone interviews with 100 Of the responding programs for ascertaining the reliability of the responses to the mail survey, has yielded a correlation coefficient Of .97 between the total number Of diagnosed handicapped children reported by the sampled programs to the Full Year survey and the telephone survey. The following are the principal findings from the national survey: 1The unit of analysis for the SRI survey was a delegate agency. The 792 figure represents the response rate from all delegate agencies. The 832 figure refers to the grantees. This means that not all of the delegate agencies under a grantee may have responded to the survey. 114 c There were about 29,000 handicapped children enrolled in Full Year programs as of November-December 1973. "These 29,000 were children whose handicapping conditions were reported to have been diagnosed by qualified professionals. In addition, the handicapping conditions of 5,000 children had only been partially diagnosed as of November-December 1973; and 4,000 other children were believed by Head Start staff to be handicapped, even though their handicaps had not yet been confirmed."1 Thus, percentagewise the handi- capped enrollment in Head Start was at least 10.1%; this figure could be as high as 13.2% if all the children partially diagnosed or suspected to be handicapped are eventually confirmed as "handicapped." 0 About one of every five handicapped children were reported to have multiple handicaps. The primary handicapping con- ditions 8f children reported as handicapped were as follows: Speech impaired 35.0% Health or developmentally impaired 20.0% Seriously emotionally disturbed 12.22 Physically handicapped 9.52 Hearing impaired 7.92 Mentally retarded 7.42 Visually impaired 6.62 Deaf 1.02 Blind 0.52 e 46.62 of the handicapped children were reported to require "a fair amount" or "practically constant" special assistance. Since the amount of special assistance required by a child was used as a measure of the severity of the child's dis- ability, roughly one-half of the children reported as handi- capped were only "mildly" handicapped. The other half were judged to be "moderately" or "severely" handicapped. 0 Handicapped children, in general, were served in physically integrated settings. "In 57% of all Head Start programs at least one handicapped child was present in 80 to 1002 of the classrooms of these programs." 1See: "The Status of Handicapped Children in Head Start — Final Report Based on Surveys of Summer 1973 and Full Year 1973-74 Head Start Programs", p. 2. 21bid., p. 5. 3Ibid., p. 6. 4Ibid., p. 7. 115 o Roughly 60% of the programs serving the handicapped provided or arranged special pre-service training, the same percentage of the programs provided in-service training to improve the abilities of their staff to work with handicapped children. Special equipment and materials were provided for handicapped children by one out of six programs serving the handicapped. About 352 of the handicapped children in Head Start were referred to Head Start as handicapped (by their parents or by other agencies); and special diagnostic services were pro- vided by Head Start for an estimated 75% of the handicapped children enrolled.1 0 Finally, "the average handicapped child in Head Start was receiving about two of the seven special services listed below."2 - provision of special counsel to parents of handicapped children related to their child's handicap - provision of planned special experiences to handicapped children to increase "adjustment skills" related to their handicap - special, individualized counseling — speech or physical therapy - medication and/or drug therapy - prosthetic devices - other special services Let us now turn to a review Of the principal findings from Task III Of the study. Findings from the first round Of visits to sixteen reg- ular Head Start programs were based primarily on the impres- sions Of the numbers Of the research teams who have conducted the visits. ”Participant Observation" being the main tech- nique used in these visits, results from each visit was reported through a narrative "site visit reports," which 11b1d., p. 8. 2Ibid., p. 9. Special services provided are listed in order of decreasing frequency, i.e., the first listed service is reported to have been provided to more number of children than any Of the Others. 116 sometimes reached seventy-five typewritten pages in length. Although these insightful reports were extremely useful in getting a feel for the manner in which the mandate was being implemented, they were tOO heterogeneous and non-uniform to permit adoption Of any systematic analysis scheme. Thus, the main conclusions from the first round visits were arrived at through a modified delphi exercise. Participating in these lengthy analysis sessions were close to all Of the field work staff. The prime Objective of these sessions was to reach a consensus on Observations about the implementa- tion Of the mandate. The following were some of the key conclusions reached in the analysis sessions. 0 Head Start staff were confused by and unclear of the term "handicapped." As a result, the mandate had the deleterious effect of mislabeling some children as "handicapped," o All of the programs visited reported that they had always served the handicapped and that they had not excluded any handicapped child from participation in the program, with the exception of a few severely disabled children. e The handicapped effort at the local level proceeded more as an "evolving" than a pre—planned process. 0 Overall, the number of handicapped children enrolled in the Full Year 1973-74 program.was only slightly higher than the number of such children enrolled the previous program year. 0 The majority of the children identified and reported as handicapped were only mildly handicapped; severely handi- capped children constituted only a small percentage of all children identified as handicapped. 0 Programs, in general, were making more special efforts to serve the handicapped this year than heretofore. Signifi— cantly greater efforts were being made this year to diagnose handicapping conditions professionally than in earlier years. In addition, the mandate resulted in more detailed individualized assessments of the developmental needs of handicapped as well as non-handicapped children. 117 0 Parents of both handicapped and typical children expressed very positive attitudes towards the efforts of Head Start on behalf of the handicapped. The handicapped effort also resulted in increased relationships with community agencies. 0 Head Start staff expressed a strong need and desire for additional training which would help them better serve the handicapped. Receptivity of the staff to severely handi- capped children seemed to increase as their contacts with such children increased. 0 Staff of most of the programs visited indicated that they disagreed not so much with the intent of the mandate but with the manner in which the National and the Regional Offices of OCD "passed it down" to the local evel, with very little guidance, support, and direction. We need tO clarify that all of the programs visited in the first round had previous experiences in serving the handicapped. If at all, as a group they are more represen- tative Of programs with relatively larger numbers Of handi- capped children. Quoting from the Task III Interim Report, ” . . . if the findings from the first round Of visits are to be generalized, they can be generalized not to total Head Start, but to top 30 per cent Of the programs as far as the handicapped effort is concerned."2 Findings from the second round visits, which were based on a larger sample Of programs and a more rigorous methodology, did not differ significantly from the find- ings from the first round visits. The most prevalent issue was mislabeling Of children. Quoting from the Final Report, " . . . we found that almost without exception the programs 1Sources for the conclusions listed here are the Interim Report and the Final Report on Task III. 2See: "Interim Report . . . ", p. 22. 118 we visited were applying the term 'handicapped' tO some children with very minor difficulties, who required no Spe- cial classroom assistance or services, who have always been enrolled in Head Start without such classifications. . . . In our view, this has been the most serious and deleterious effect of the new legislation. . . . ”1 TO the question "How successful has Head Start been in providing new services for handicapped children this year?" the Task III Final Report provides the following answer: "Head Start services for children with special needs have basically remained the same this year and in order to really fulfill the intent Of serving more serious disabled children, the legislation needs to be further clarified and new approaches with greater resources developed."2 Conclusions of the Syracuse University team from both the first and the second round visits are summarized as follows: "Severely retarded children comprised a very small percentage of the total enrollment of Head Start centers visited, said population significantly less than the 10 per cent Congres- sional demand; there was great variation.among programs vis-a—vis attitudes towards the handicapped--especially the severely handicapped--and program opportunities for the handicapped. By and large, most mildly and moderately handi- capped children were physically and psychologically inte- grated in Head Start programs, with such integration usually assured upon admission; exclusion or exemption was the more serious problem than the integration of those admitted. The mandate appeared to have positive effects in increasing a coordinated involvement with families and other community agencies. Lastly, Head Start staff continued to feel very 18cc: "Final Report . . . ", p. 202 2Ibid., p. 230 (emphasis deleted). 119 strong general needs for both in-service training and improved and increased technical assistance and consulta- tion." The conflict between the findings reported in the Task II and the Task III final reports, prepared by the SRI and the Syracuse University teams, respectively, is quite appar- ent. While the Task II report claims that, based on the figures reported by the local programs in November-December 1973, the ten per cent Congressional mandate has been sur- passed, the Task III report makes an Opposite claim, based on data collected during site visits to a total Of fifty-two programs, that the ten per cent mandate has not been met in ‘terms Of the severely handicapped children enrolled and that imposition of the mandate has led to unnecessary and improper labeling of some children with very minor difficulties as "handicapped." The issue, clearly, is a definitional matter. In fact, there would even be no conflict between the two reports if one assumes that the intent of the mandate covers only the moderately and the severely handicapped. If one were to measure "moderately" and ”severely" in terms Of numbers of children requiring ”a fair amount of" or "almost constant" special assistance, the Task II survey yields a handicapped enrollment figure between 4.7 and 6.1 per cent. And this 11bido, pp. 252-252. 120 is clearly well below the mandated 10 per cent figure.1 There is considerable evidence that the intent Of the mandate lies in the direction Of the severely handicapped. Congressman Albert Quie (R-Minnesota) remarked, in a floor Speech on the day the House voted on the 8.3010 Conference Report, as follows: " . . . It is my hope that as a result Of this amendment that Headstart will now provide services for children who are totally blind, totally deaf and severely physically or mentally handicapped. . . . It has been said for years that a child who is poor is handicapped. I would add to that that a child who is poor and also severely handicapped is doubly handicapped . . . "2 The OCD policy on the handicapped effort also excludes mildly disabled children from the SCOpe Of the mandate: "While children with milder handicapping conditions (e.g. children with visual problems correctable with eye- glasses) will continue to be identified and receive appro- priate Head Start services, they fall outside the scope of this issuance. The intent is rather to insure that Head Start serves more fully children who have severe vision and hearing impairment, who are severely physically and mentally handicapped and who otherwise meet the legis- lative definition of handicapped children in terms of 1The 4.72 figure takes into account only children diagnosed as handicapped as of November-December 1973 222 reported as requiring "a fair amount of" or "almost constant" special assistance. The 6.1% figure includes these children plus children whose diagnosis had not been completed at the time of the survey and children who were judged to be handicapped by Head Start Staff prior to any diagnosis. In com- puting the 6.12 figure we assumed that the same percentage (i.e., 46.62) of these additional children would require "a fair amount of" or "almost constant" special assistance once their diagnoses were completed. 2See: LaVor, "Economic Opportunity Amendments of 1972, Public Law 92-424", p. 253(emphasis added). 121 their need for special services."1 Finally, the individual definitions Of the nine handi- capping conditions used in the surveys Of Summer and Full Year programs, which were develOped jointly by the SRI and the Syracuse University teams, clearly refer only to chil- dren with moderate or severe handicaps.2 In light of these interpretations of the policy, there- fore, one cannot readily accept the claim that the ten per cent mandate has been met. By the same token, it cannot be conclusively asserted that Head Start programs, nationally have failed in implementing the mandate. It can be stated with no or few reservations, however, that, in general, local programs with severely handicapped children were more success- ful in implementing the Congressional mandate than programs with the same number Of mildly or moderately disabled children. This measure Of policy success, in our Opinion, is more in line with the intent of the policy than one which does not distinguish between levels of severity. And the primary success measure we have used in the dissertation assigns a greater weight to each severely handicapped child enrolled. 13cc: OCD Notice N. 30-331-1 on Head Start Services to Handicapped Children, p. 3. ZSee: "Appendix A: Survey Questionnaire and Definitions - Full Year 1973-74 Programs" in "The Status of Handicapped Children in Head Start - Final Report Based on Surveys of Summer 1973 and Full Year 1973-74 Head Start Programs". 122 The'Data In this section we concentrate on methodological and data-related issues. Our main concern is the development Of empirical indicators for the six primary variables Of the conceptual model. This measurement process involves the con- struction of indices for each primary variable. The elements or secondary variables entering each index are generated from the data collected during the second round visits to the thirty-six Head Start programs. Our choice Of index construc- tion as the principal method, as we will further discuss below, was dictated by the nature Of the data that were col- lected in the project. We have divided this section into three parts. We first present an overview Of the methodology and the proce- dures used in the second round visits. This overview con- sists mainly Of a summary Of the detailed description Of the research design and the methodology presented in Appendix A. In the second part Of the section we describe and discuss the Specific procedure used in constructing the indices. The final part Of the section is devoted mainly to presenta- tion Of the empirical results from the index construction process. An Overview of Methodology and Procedures Research design The research design used for generating the data is primarily a cross-Sectional survey design. The data were 123 collected at approximately one point in time (April-May 1974) from a sample Of thirty—Six regular Head Start programs. Longitudinal data were generated by asking the respondents tO report information about earlier points in time and by having the respondents describe certain processes over time. Due to the lack Of a control group and pre and post measure— ments, the design suffers from the phenomena that typically limit the internal validity of all non-experimental designs. Questionnaire design The questionnaire was designed in order to find answers to the following three questions: (1) How well are Head Start programs presently serving the handicapped, in partic- ular the severely handicapped? (2) Why are some Head Start programs presently not serving the severely impaired? (3) What are the factors which account for the differences between programs in enrollment Of severely handicapped chil- dren? The instrument was develOped over a two-month period after a careful analysis Of the exploratory data collected during the first round visits to sixteen programs. Data from only the ”Program-Level Information" part of the ques- tionnaire are used in the dissertation; child—specific and classroom Observation data did not serve our purposes. The finalized version Of the instrument and its accompanying documents were fielded only after Obtaining formal approval from HEW and the Office Of Management and Budget. 124 Selection of programs A total of thirty-seven regular Head Start programs were originally selected for the second round visits. One Of these programs withdrew from participation in the study and we drOpped one program due tO problems caused by missing data. Empirical analyses found in the dissertation are, therefore, based on data from a sample Of thirty-five pro- grams. The sampling frame consisted Of the list of 1,353 Full-Year Head Start programs which responded to the mail survey conducted during November-December 1973. The original sample size Of thirty-seven was determined on the basis Of resource constraints and contractual Obligations. The final sample was selected through a complicated process which involved stratification Of programs based on their Size and severely handicapped enrollment, reduction Of the Size Of the sample frame through random sampling, further reduction of the reduced sample frame through application Of a set Of screening criteria, random selection Of programs with no reported handicapped children, and purposive selection Of the remaining programs. Although randomization has entered the selection process at several points, the final sample cannot be considered truly random. However, there is no reason to believe that the sample is not representative Of the overall population Of Head Start prOgrams. 125 Pre-fieldwork and fieldwork activities Almost all Of the interviewers/Observers who conducted the fieldwork during the second round had previous fieldwork experiences, mainly as a result of their participation in the first round visits. In addition, they were subjected to a two-day formal training session and they participated in a three-day classroom Observation program conducted in care- fully selected preschool and Special education settings. All of the ten persons who conducted the fieldwork were fac- ulty members and graduate students from Syracuse University. Visits to the selected programs lasted one to three days and, except for five large programs, were conducted on a one interviewer per program basis. The principal respon- dent tO the "Program-Level Data" section of the questionnaire was the Head Start director at the sampled program. NO major problems were encountered during the fieldwork. Post-fieldwork activities These activities included preparation Of the codebook, which included codes for 1,035 variables, editing Of the collected data, reinterviews with the respondents, coding Of the quantitative and the anectodal data, data processing, and analysis. Analysis Of the data by the Syracuse Univer- sity team was mainly descriptive and did not involve any model building or multivariate analysis. Our conceptualization Of policy implementation and analysis Of the second round data were done independently of 126 the work Of the project team from Syracuse University. The only thing common to our work and that Of the Syracuse Uni- versity team is the data base generated during the second round visits. Index Construction Methodology The second round visits to the sampled programs were conducted in order tO assess the status Of the implementation Of the policy at the field level and to investigate the fac- tors which may account for the differences in the success levels achieved by programs in implementing the congressional mandate. The questionnaire was not designed and the data Were not collected with a specific model of implementation in mind. Instead, the data collection effort centered around key issues and concerns which surfaced as hypotheses at the conclusion Of the exploratory research conducted dur- ing the initial stages Of the project. For this reason, the correspondence between the data at hand and the vari- ables Of our conceptual model is not as close as we would have liked to have. It is possible to measure two of our variables, level Of implementation success and self-evaluation Of capability, directly, i.e., without extensive data manipulation. This is because the available data allow us to develOp Simple indicators for measuring these two variables. For the other four variables, however, it is possible to construct a vari- ety of indicators which relate to one or more aspects Of 127 Of each variable. Unfortunately, none Of the indicators that relate to a given variable is "superior" to the others in terms of validity or reliability. Thus, in developing measures for these four variables we are faced with a dif- ficult choice between (1) selecting only one Of the relevant indicators and using that as our measure for that variable, and (2) developing a new measure by combining all the indi- cators relevant to that variable. We have chosen the second alternative because it allows us to keep the "measuring power” Of all indicators relevant to a variable. This alter- native involves construction Of an index for each of the fol- lowing four variables: implementation potential, level Of effort, environmental forces for implementation, and policy support. All Of the secondary variables which were consid- ered in developing the indices for these variables are defined, described, and discussed in Appendix B. Index construction is a data reduction technique. It is similar to and at least as widely used as scaling, although there is much less methodological literature on index construction than scaling. Both techniques yield ordi- nal measures and there are slight differences between them. Babbie summarizes these differences as follows: "An index is constructed through the simple cumulation of scores assigned to specific responses to the individual items comprising the scale. A scale is constructed through the assignment of scores to response patterns among the several items comprising the scale. A scale differs from an index by taking advantage of any intensity structure that may exist among the individual items."1- 1Earl R. Babbie, Survey Research Methods (Belmont, California: wadsworth Publishing Company, Inc., 1973), p. 254. 128 Construction of an index typically involves completion Of the following steps: selection of the items considered for inclusion in the index, examination Of the relationships among the items selected, determination Of the scoring procedure, construction Of the initial composite index, internal validation of the through item analysis, and external validation Of the composite index index. These steps correspond roughly to the procedure we used in constructing the indices for EFRT, IPOT, ENVR, and PSUP. However, because Of the unavailability Of a criterion measure we could use for external validation Of each index, we decided to construct each index twice, using different indi- cators for each item considered in the index. Our rationale for doing this is that if the same items end up in the two alternative indices and if the two final indices correlate strongly with one another, our confidence in both indices would be greatly increased. If, however, the indices devel- Oped from two different sets of indicators for the same ini- tial items seem to differ from one another Significantly, we would have less confidence in the index and would have to explain the differences between the two indices before using either index in the statistical analyses. There is a second reason for our choice to develop two alternative measures for each item considered in the index. As can be Observed in Appendix B, most Of the indicators we 129 have for the items (secondary variables) initially included in the indices for EFRT, IPOT, ENVR, and PSUP are measured at the ordinal level.1 Theoretically, we cannot add two or more ordinally measured items and Obtain a composite score. However, if we can transform these ordinal measures into an interval measure, then we can conduct the arithmetic Operations necessary for the construction Of the composite index. And if the index based on interval-level measures is found to be very strongly correlated with the one based on ordinal measures, we would have reason to argue that the two indices can be used interchangeably. How can we transform an ordinal scale into an interval measure? The only alternative that seemed reasonable is tO assign to each ordinal code of an item the cumulative per- centage which corresponds to that code in the univariate fre- quency distribution of that item. Cumulative percentile scores can be interpreted as cumulative probabilities (which are measured at the ratio level). With this transformation, each cumulative percentage shows the probability that a pro- gram randomly selected from the sample at hand would achieve a given ordinal score or lower. Since the original ordinal codes are all previously arranged such that the lowest code corresponds to the least desirable outcome (from the view- point Of the policy's success), the cumulative percentage 1We consider the "yes-no" type dichotomies we have as ordinal measures because the two categories of all of our dichotomies are directional and ordered. 130 distribution preserves the original order among the programs. The range Of the cumulative percentage distribution is between zero and one. Although the minimum cumulative per- centage would never equal zero, it could be very close to it. However, the distribution always attains the maximum value Of one. The transformation described above yields, for each item considered for inclusion in the index, a new distribu- tion which has a measurement level higher than ordinal. These new sets of values can now be used in constructing what can be called an "alternative index" for each Of the four primary variables. This index is alternative to the "orig- inal index" which is based on a simple linear transformation of the original codes. The linear transformation is neces- sitated because the range Of the ordinal values assigned to each item differ from one another. In the case Of dichoto- mies, for example, the originally assigned codes are zero and one. For a trichotomy these codes are zero, one, and two; and in one instance the values Of an item assumes integer values between zero and eight. Simple addition Of these codes in constructing the index amounts to assigning a weight to an item on the basis Of the number of values it assumes. In order to correct for this imbalance we linearly transformed the original codes of each item into the [0,1] interval. This transformation assigned a value of zero to the minimum and a value of one to the maximum original code. Codes inbetween the minimum and the maximum were transformed 131 into the [0,1] interval on a linear basis. This transforma- tion maintains the original order among the programs and assures equality Of the intervals between succeSsive values Of the item. In summary, we constructed two indices, an original and an alternative index, for each Of our four primary variables which are measured through indices. The original index was based on a simple linear transformation Of the ordinal code for each item into the [0,1] interval. The alternative . index was based on a transformation which assigned each orig- inal ordinal code the cumulative percentage corresponding to that code in the univariate frequency distribution Of the item. The two indices were constructed independently Of one another and the same procedures were used in arriving at the sets Of items which are included in the two final indices. The following is an outline Of the specific procedures used in developing the two indices: 0 The items tO be included in the item pool for each variable were selected on the basis Of their face validity after a careful review Of the data. A conscious attempt was made to include as many valid items as possible in the item pOOl. The full list Of items (secondary variables) initially included in the item pools appears in Appendix B. O A correlation matrix was constructed in order to examine the bivariate Pearson correlations among the variables and their associated significance levels. 0 The signs of the bivariate correlation coefficients were .llli.’lit‘l.ii[lll'||‘ll‘llilll'1l'l|i l| iii I [it irill‘ll‘ll‘lll 132 studied. Since the codes for the items were arranged such that in all cases the lowest value Of the item corresponded to the least desirable outcome, items which had consistently negative correlations with the others were dropped from the the item pool. 0 An initial index was constructed by adding the scores Of items remaining in the item pOOl. This rep- resents assignment Of equal weight to each item. We had nO a priori reasons for assigning differen- tial weights to the items. 0 An item analysis was conducted by studying the bivariate correlations between each item and the initial index minus that item. If this correla- tion coefficient was not significant at the 0.10 level the bivariate coefficients between that item and the others which entered the initial index were examined. If the item did not correlate with any Of the others (again at the 0.10 level Of Sig— nificance) the item was drOpped from the item pool. 0 A final index was Obtained by revising the initial index based on the results of successive item ana- lyses. In all instances, items were dropped one at a time to Observe the effects of dropping a variable on the correlations between a remaining item and the new index minus that item. We compared the alternative index and the original index, which were independently constructed using the above procedure, in terms Of the final set of items remaining and the bivariate correlation coefficient between the two final indices. In all fourcases the two final indices were almost identical. They yielded the same final set Of items and the 133 Pearson correlations between the two indices were all signif- icant at the 0.001 level. We review these results and the measurement Of the other two variables (i.e., LIS and CAPB) in the next part of this section. Measurement of Variables We describe how we have measured the six primary vari- ables Of the model in the order these variables were intro— duced and defined in Chapter III. Level of implementation success The overall implementation Objective Of the policy is to enroll and serve in Head Start sufficient numbers Of hand- icapped children to account for at least ten per cent Of the total enrollment. This is the goal specified in the legisla- tion and the Objective passed down to local programs by OCD. Thus, the Simplest success measure we could use is a dichot- omy based on breakdown Of the programs into two groups on the basis Of their percentage handicapped enrollment. How- ever, this measure is unsatisfactory because the use Of the phrase ”at least ten per cent" in the legislation implies that a program with, say, twenty per cent handicapped would be considered more successful than one with ten per cent handicapped. Adding to this the fact that a great number Of programs did not in fact reach the ten per cent figure, it is reasonable to measure successin terms Of the actual per- centage Of the handicapped enrollment. Enrollment Of handicapped children reflects one aspect itiiili.‘([ 134 of policy success. Another success related issue is the question Of how well each program is serving the population of handicapped it has enrolled. This is a very difficult question tO address empirically. Eventually the question boils down to how well the program is serving each handicap- ped child enrolled. This could be perhaps best measured by estimating the extent to which Head Start is meeting the special needs Of each handicapped child. Although we have some child-specific data on this issue, the sample Size per program is too small to lead to any meaningful empirical analyses.1 Thus, the success measure used in our analyses can only be based on the enrollment Of handicapped children. How can one determine if a child is handicapped? This is a highly complex and technical issue. The legislation Specifies nine handicapping conditions2 and in the course Of the project definitions were develOped for each Of these impairments.3 In addition, a distinction was made between levels. Children who required little or some Special assis- tance within Head Start, i.e., assistance that would not be normally provided tO non-handicapped children, were referred 1Seventy-four child-specific case studies were developed in the second round. These children were enrolled in twenty-five of the thirty- five programs in our sample. This yields only a sample of three children per program. 2These are: blindness, visual impairment, deafness, hearing impair- ment, physical impairment, speech impairment, health or develOpmental impairment, serious emotional disturbance, and mental retardation. 3These definitions were sent to local programs as a part of the Task II questionnaire. See: "Appendix A: Survey Questionnaire and Def- initions-Full Year 1973-74 Programs" in "The Status of Handicapped Chil- dren in Head Start-Final Report Based on Surveys of Summer 1973 and Full Year 1973-74 Head Start Programs." 135 to as "minimally handicapped;” those that required a fair amount Of assistance were called "moderately handicapped?" and children who required practically constant attention and/or assistance were classified as "severely handicapped." These measures Of severity, however, were only suggestive, the Specific handicapping condition and its severity level was to be determined by a qualified professional. In sum, although some attempt was made tO provide guidance and assistance to local programs regarding the definitional question, the final decision was left to the judgment Of qualified medical personnel affiliated with the program. Measurement Of implementation success in terms Of per- centage Of handicapped, where the term "handicapped" covers all handicapping conditions and severity levels, has a major disadvantage. It places mildly or moderately handicapped childrenmon a par with the severely impaired. AS we pointed out in the first section Of this chapter, the intent Of the policy clearly lies in the direction of the severely handi- capped. Given two identical programs each with only one handicapped child with the Same impairment, the program serving a severely handicapped child would be considered as more successful than the one serving a mildly or a moderately handicapped. This is not only because Of the intent Of the mandate but also because of the fact that, given the insuf- ficiency of educational and other services for the preschool handicapped nationwide, the severely handicapped would in all likelihood benefit more from participation in Head Start 136 than the mildly or the moderately handicapped. Adding to these the consideration that in many Of the programs children classified as mildly handicapped were in fact either non- handicapped or were clearly outside the sc0pe Of the policy, we reach the conclusion that the success measure should be based more heavily on the percentage severely handicapped. One alternative is tO use as our success measure only the severely handicapped percentage. This has two disadvan- tages. First, it does injustice to programs serving "legiti- mate” moderately handicapped children, which, for one reason or another, have enrolled only few or no severely handicapped children. Secondly, as illustrated in Table l, the distribu- tion Of the severely handicapped percentage is extremely skewed. If we were to use this measure, our dependent vari- able would have very little variance. In particular, we would not be able to meaningfully study the twelve programs with no severely handicapped children. The Specific measure we are using here is based on dif- ferential weighing Of the percentage severely handicapped and the percentage mildly/moderately handicapped. We assigned a weight Of two to the percentage severely handicapped and a weight of one to the percentage mildly/moderately handicap- ped and added the weighed percentages to Obtain the success scores illustrated in Table 2. Differential weighing of the two percentages reflects bOth the assignment Of a higher level Of success to a program Witt: severely handicapped children and the discounting Of I '1 ll [-l.[t'fi ll lnIl’l [.{fiii’l‘ll 137 TABLE 1 PERCENTAGE OF SEVERELY AND MILDLY/MODERATELY HANDICAPPED CHILDREN Programs with the Programs with the Percenta e of Indicated Percentage of Indicated Percentage Of g Severely Handicapped Mildly/Moderately Severely or Handica ed Mildly/Moderately pp Handicapped Number Percentage Number Percentage 0 12 34.3 4 11.4 0.1 - 5.0 11 31.4 7 20.0 5.1 - 10.0 6 17.1 8 22.9 15.1 " 20.0 1 2.9 4 11.4 20.0 0 O 4 11.4 Total 35 100.0 35 100.0 the reported numbers Of mildly/moderately handicapped. The weight Of two for severely handicapped is not totally arbi- trary. It is based, in part, on the assumption that roughly half Of the children reported as mildly/moderately handicap- ped can be considered as "normal" children with only minor ‘disabilities. This is supported by the findings from the Task II survey as well as the Observations of the interview- ers who conducted the fieldwork. According tO this procedure, then, a program with three per cent severely handicapped and seven per cent 138 TABLE 2 IMPLEMENTATION SUCCESS SCORES Success Number of Percentage of Score Programs Programs 0 11.4 0.1 — 5.0 5.7 5.1 — 10.0 8.6 10.1 - 15.0 14.3 15.1 - 20.0 25.7 20.1 - 25.0 5.7 25.1 - 30.0 2.9 30.1 - 35.0 5.7 35.1 - 40.0 14.3 40.1 or over 5.7 Total 35 100.0 mildly/moderately handicapped would be assigned a success score Of thirteen. The potential range of the success score is between zero and 200. The actual range is between zero and 51.6. The mean score is 19.1 and the median is 15.4. The distribution is somewhat skewed to the left, however, the skewness coefficient for this distribution is considerably less than those for the distributions Of percentage severely handicapped and the percentage mildly/moderately handicapped (0.59 vs. 1.26 and 1.18, respectively). SO is the coefficient of variation. It is 72.5% for the distribution Of success 139 Scores, as compared tO 131.4% and 79.3% for the distribu- tions Of percentage severely handicapped and percentage mildly/moderately handicapped, reSpectively. Level Of effort We do not have any direct measures Of the level of "human effort" expended by each sampled program during the implementation Of the mandate. Instead we have eight indi- cators which roughly measure the amount Of task-specific effort that has gone into eight different aspects Of the implementation. These eight items, which are defined and dis- cussed in Appendix B, relate to the following aspects Of the implementation: ORGS : effort to organize for the implementation of the policy; PLAN : planning effort as indicated by the compre- hensiveness Of the plan prepared for imple- menting the policy; COOR : effort to coordinate the implementation of the policy with other agencies within OCD; MOBR : effort to mobilize the resources of the community; FUND : effort to secure additional funds for the implementation; SIZE : effort to increase the size Of the imple- mentation staff; TATA : policy-related training and technical assistance effort; MFAC : effort tO modify the physical facilities in order to make them more appropriate for serving handicapped children. These eight areas do not fully exhaust all possible 140 policy-Specific activities which were undertaken by the implementing agencies. Activities for which we do not have indicators Of level Of effort include, in particular, recruit- ment and enrollment Of handicapped children; screening, test- ing, and diagnosis Of the handicapping conditions; and activ- ities directed towards the parents Of handicapped children. Although we do not have indicators for these variables, it is quite likely that the items eventually included in the index would account for most Of the variation in each Of these. Let us now review the results Of the index construction process. 0 The two indices yielded identical results. Only one vari- able, COOR, was removed from the initial item pool in both instances. The bivariate correlation coefficient (Pearson) between the two final indices is 0.96. e The two correlation matrices were almost identical. For this reason we illustrate in Table 3 only the matrix based on the original codes. The largest correlation coefficient among the 28 shown in Table 3 is .52 and the smallest is .04 (absolute value). The variable with the poorest cor- relation with the others is COOR, which was eventually dropped from the index. COOR is also the only variable with two negative correlation coefficients. 0 Before removing COOR from the item pool we examined the cor— relation coefficients between each variable and the initial index (made up of eight items) minus that variable. The two item analyses yielded correlation coefficients which were significant at the .10 level for all variables with the exception of COOR. e The two item analyses were repeated after removing COOR from the item pool. The new set of correlation coefficients were all significant at the .10 level. These results are illus- trated in Table 4. The final index based on the original codes Of the variables is the one we chose tO use in further analyses. 141 .HO>OH 0H.o msu um uanfiwacwwm ma beefiowmmooo ozu spam pusposes“ seem HN. zmmzfimm mezmHonmmoo ZOHHzm 4 220 the agency resulting from the success level attained at the end of the implementation period. The dotted lines show lagged relationships between the variables of the model. Given a set of initial values for ENVR, PSUP, CAPB, and IPOT the complete model in equation form can be stated in terms of the following six functions: PSUP(t) = f[PSUP(t-l), REWD(t-1)] CAPB(t) = f[CAPB(t-l), LIS(t-l)] IPOT(t) = f[IPOT(t-1), EFRT(t-l)] EFRT(t) = f[ENVR(t), PSUP(t), CAPB(t)] LIS(t) = f[IPOT(t), EFRT(t)] REWD(t) = fILIS(t)] where (t) and (t-l) denote successive implementation periods. According to this model rewards are viewed solely as a function of the level of success achieved during the imple-. mentation. We realize that variables other than LIS, such as those depicting one or more characteristics of the reward allocation mechanism in Operation, could affect these rewards. Thus, depending upon the policy being examined, it may be appropriate to add these to the model as exogenous variables affecting REWD. The model includes three time-lagged relationships. First, it is hypothesized that the success level attained during a particular year influences the leadership group's perception of the agency's capability to implement the policy before the start of the following year. Second, it is claimed that actual rewards received at the end of a specific 221 implementation period affects the leaders' level of support of the policy during the following period. Finally, the model postulates that implementation efforts during a given period modify the potential of the agency to implement the policy during the next period. Needless to say, the model requires further conceptual- ization before it can be tested empirically. We believe that further research along the lines of this model would prove to be quite fruitful in understanding and improving policy implementation. APPENDICES APPENDIX A METHODOLOGY AND PROCEDURES In this appendix we describe in detail the Specific methodology and the procedures used in generating the data analyzed in the body of the dissertation. This description is organized along the following lines: research design used questionnaire design selection of programs for the second round visits pre-fieldwork activities fieldwork and data collection post-fieldwork activies. Research Design We have basically used a cross-sectional survey design for the second round visits. The data were collected at approximately one point in time (April-May, 1974) and the selected sample was intended to represent the larger popula- tion of all Head Start programs at that point in time. The period of data collection corresponded roughly to the final two months of full-year program Operations. The cross-sectional survey design, although it is mostly used to arrive at generalizations for the point in 222 223 time at which the data were collected, can also be used to generate longitudinal data. This is accomplished either by asking the respondent to report information about an earlier point in time or by having the respondent describe a process over time. In the case of the second round visits both of these methods were used to arrive at estimates of longitu- dinal data. In the first instance, the respondents to the second round interviews were asked to provide us with data about the previous year's enrollment of handicapped children, the specific month they were notified of the new policy, and their estimate of the implementing agency's capability to execute the mandate. Secondly, a significant portion of the interviews concentrated on collection of data about the spe- cific processes instituted and the activities undertaken by the implementing agency to get ready for or to facilitate the implementation of the policy. Thus, the specific design used for the second round visits was primarily cross-sec- tional, with provisions for approximating some longitudinal data within the overall framework of the cross-sectional design. The use of a cross-sectional design, with provisions for approximating longitudinal data, was necessitated by the specific requirements of the project. Initially, the overall design of the project called for two rounds of visits to approximately twenty-five programs in each round and did not rule out the possibility of visiting the same twenty-five programs twice. However, during the initial 224 segments of the project it became evident that the project team needed first-hand, on-site information about Head Start and the initial stages of implementation of the handicapped policy before formulating the specific hypotheses that could be tested empirically. Consequently, the approach to Task III was revised early in the project to accomodate this need. Thus, the first round visits covered sixteen programs with an unstructured, exploratory data collection technique. This allowed the project team to increase the sample size for the second round by about fifty per cent (from twenty-five to thirty-seven) and to field a more relevant and comprehen- sive questionnaire than what would have been possible had the original design been implemented. The cross-sectional survey design used in the study is clearly a non-experimental design and, therefore, suf- fers from many of the phenomena that typically limit the internal and the external validity of all non-experimental designs. Using Campbell and Stanley's terminology, the design of the second round visits can at best be identified "1 And this would as a "one-group pretest-posttest design. only be possible if we can assume that approximation of Ion- gitudinal data through the cross-sectional survey covers measurement of our primary dependent variable at two points in time: before the start and at the conclusion of the 1See: Donald T. Campbell and Julian C. Stanley, Experimental and Quasi-Experimental Designs for Research (Chicago: Rand McNally 225 program year during which the policy was implemented. Even this provision is not sufficient to raise the validity of the design to a level comparable to what one would achieve with the so-called "true" experimental designs. At this point it would be apprOpriate to discuss two of the reasons for the unapplicability of a true or quasi- experimental design to the problem we have at hand. First, it should be kept in mind that we are attempt- ing to explain the difference between successful and unsuc- cessful implementors of a given policy, wherquii implement- ing agencies are mandated to implement the policy by law. It is, therefore, impossible to manipulate some variables and observe their effects on the dependent variable. By the same token, the circumstances at hand precludes prior selec- tion of some programs as the "experimental" and others as the "control" group. Success can be observed only arpmst jean; and the programs can be divided into "successful" and "unsuccessful" groups after completion of the implementation of the policy. Second, the research problem at hand is highly explor- atory. There are no clear and well-formulated theories of implementation which suggest potentially explanatory vari- ables which could be easily manipulated or varied in an experimental or quasi-experimental situation. Instead, we have a conceptual model with a set of independent variables. Relationships between each independent variable and the dependent variable and the dependence of the independent 226 variables upon one another have been postulated only at an exploratory level. If these relationships are not supported by the data at hand, we need to search for alternative expla- nations which can be verified. Use of an experimental or quasi-experimental design, however complex it might be, precludes exploration and limits the search for alternative explanations. The basic superiority of the experimental designs over the others is that through experimental work the researcher can effectively control for other variables which can poten- tially influence the dependent variable. This often becomes an insurmountable task to achieve through non-experimental designs such as those based on survey research. In these instances the primary alternatives available to the researcher are: control of background factors by random- ization, by specification, or by some combination of these two.1 In the case of the present study, as we shall see below, the alternatives we have used are similar to these three. Questionnaire Design The instrument used in the second round visits to regular Head Start programs was develOped over a two-month period following analysis of the data generated during the 1See: Robert T. Holt and John E. Turner, "The Methodology of Comparative Research," in The Methodologyiof Comparative Research, ed. by Robert T. Holt and John E. Turner (New York: The Free Press, 1970), pp. 1-20. 227 first round. The most critical issues to be explored during this phase of the project were identified, discussed, and finalized by the principals of the project team and the mem- bers of the Senior Consultant Group. It was agreed that three types of data should be collected in this round: data on the implementing agencies, data on children with Specific handicaps, and data on classrooms within which handicapped children were being served. The final version of the ques- tionnaire reflects this breakdown. It has four distinct parts: Part I : Identifying Information Part II : Program-Level Information Part III : Child—Specific Information Part IV : Guide for Classroom Observations. Data from the last two parts of the questionnaire served very specific purposes and are not used in this dissertation. Design of the survey instrument was completed in five stages. The following is an overview of the activities undertaken in each stage. Firgr, the primary purposes of the second round visits were identified to be the following: (1) How well are Head Start programs presently serving the handicapped, in partic- cular the severely handicapped? (2) Why are some Head Start programs presently not serving the severely impaired? (3) What are the factors which account for the differences between programs in enrollment of severely handicapped children? 228 The answer to the first question requires collection of data on level of overall implementation success as well as on specific services provided to individual children with- in and outside the classroom. The second question involves in-depth study of programs with little or no severely handi- capped children and their comparison with programs with higher enrollment figures. The third question requires gen- eration of data on a series of potentially explanatory vari- ables. Second, the general structure of each part of the ques- tionnaire was formulated and discussed. Key issues and poten- tially explanatory factors were identified. Tentative find- ings from the first round visits played an important role in this phase of questionnaire design. irirg, a draft questionnaire was prepared. This draft was reviewed and revised twice by a core group from the proj- ect team. The revised draft was reviewed by all project prin- cipals and some graduate students. Coding, data processing, and analysis implications of each question were discussed. This led to the finalized version of the questionnaire. Fourth, the finalized version of the questionnaire was mailed to Washington, D.C., along with a statement describ- ing how it will be used, for approvals by the Department of Health, Education, and welfare and the Office of Management and Budget. No changes were suggested by the reviewers from the two federal agencies,1 indicating that the questionnaire lThe om: Approval Number was 0MB-85—S-74-005. 229 and the data collection plan satisfied their minimum criteria of acceptability. Eirrh, implications of the survey instrument for train- ing the interviewers, coding the responses and processing the data were identified and discussed. Selection of Programs A total of thirty—seven regular Head Start programs were originally selected for the second round of visits. One of these programs withdrew from the sample two days prior to the scheduled visit to that program and was not replaced due to logistical problems. Of the remaining thirty-six programs thirty-three were distributed among the ten OCD regions and three were Indian programs. The sample was selected through a complicated scheme. Here we will describe the essential features of the proce- dures used in identifying thethirty-seven programs. These features include: a sampling elements 0 survey population 0 sample size 0 sampling frame 0 stratification of programs 0 screening criteria 0 initial selection 0. telephone interviews 0 final sample 230 We will now discuss each feature in the order they appear above. There were three types of sampling elements considered. First, regular Head Start programs, with or without handi- capped children, comprised the primary unit of analysis. If a grantee agency had no delegates, the grantee was considered as a sampling element. If the grantee agency did not admin- ister any Head Start center, i.e., if all centers were run by one or more delegate agencies, each delegate agency was given a sampling element status. Finally, if a grantee agency ran some centers and delegated others, both the grantee and the delegate agencies were considered as sepa- rate elements. This definition coincides exactly with the definition of a "respondent” used in the mail survey con- ducted in Task II. Each handicapped child enrolled in Head Start was considered as a second type of sampling element. The third element consisted of each Head Start classroom with at least one handicapped child. Data from the children and the class- room samples are not studied here. The survey populations were defined as the total pOpula- tion of 1973-74 Full-Year Head Start programs, the total population of handicapped children enrolled in these pro- grams, at the time of the survey, and the population of Head Start classrooms with at least one handicapped child. The sample size was determined on the basis of resource (constraints and contractual obligations. Given that sixteen 231 programs were visited during the first round visits and that a total of fifty regular Head Start programs were to be vis- ited in Task III, the contract with HEW demanded completion of visits to at least thirty-four programs. This figure was increased to thirty-seven by adopting some cost-saving measures such as visiting the selected programs in a round-robin fash- ion instead of returning back to Syracuse after completion of each visit. The contract also demanded that at least ten of the visits were to be made to programs with no or very few handicapped children. The sampling frame consisted of the computerized list of 1,353 Full-Year Head Start programs which had responded to the mail survey conducted during November-December 1973. 79 per cent of all grantees and delegate agencies responded to this survey and at least one response was obtained from 83 per cent of all Head Start grantees. The closeness of the sampling frame to the survey population is quite important in judging the external valid- ity of the findings based on a sample drawn from this sampl- ing frame. For this reason we will briefly review here the results of two special surveys conducted within Task II for assessing the reliability of the responses to the mail survey and for estimating the effects of nonresponse. The test of the reliability of the Full-Year survey was based on telephone interviews with a randomly selected sample of respondents to the mail survey. This telephone survey was conducted during the last half of January 1974, 232 approximately one month after the responses to the Full-Year survey were received. A simple random sample of one hundred programs were selected for telephone interviewing. After three attempts, there were thirty-three programs which did not respond to the survey and, of the remaining sixty-seven, three were incomplete and were dropped from the analysis. The usable sample of sixty-four responses were analyzed to assess the differences between the two surveys in terms of the reported number of handicapped children, their handi- capping conditions, and the severity of their handicaps. The principal measure used to estimate the reliability of the Full-Year survey was the correlation coefficient between the total number of diagnosed handicapped children reported by the sampled programs to the Full-Year survey and the telephone survey. This coefficient was estimated at .972 which indicated that the two surveys did not differ statis- tically from one another in terms of the total number of diagnosed handicapped children.1 The second special survey in Task II was conducted in order to ascertain whether the nonrespondents to the mail survey differed significantly from the respondents in terms of enrollment of handicapped children. The survey was con- ducted in January 1974 and covered the 120 nonresponding 1For a detailed account of the two special surveys conducted within Task II see: "Appendix E: Verification of the Findings From the Survey of Full-Year 1973-74 Head Start Programs," in "The Status of Handicapped Children in Head Start - Final Report Based on Surveys of Summer 1973 and Full Year 1973-74 Head Start Programs." 233 grantee agencies. After three attempts responses were obtained from sixty-nine grantees. Responses from the gran- tees covered the handicapped enrollments in all centers administered by the grantee, including those Operated by delegate agencies. Two statistical tests were conducted to assess the significance of the differences between the reported handi- capped enrollments of the respondents and the nonrespondents to the Full-Year survey. First, the difference in average percentage handicapped in the two groups was not found to be significant at the .01 level. Second, a Kolmogorov-Smirnov one-sample test revealed that the percentage distributions of handicapping conditions did not differ from one another at the .01 level of significance. This led to the following conclusion: ". . . On the basis of the above discussion, we have no reason to believe that the findings from the Full Year Survey would have changed had we been able to Obtain a greater response rate. In other words, programs that did not respond to the Full Year Survey were not at all different, in terms of the variables used to measure the status of OCD'S handicapped effort, from those that did respond to the mail survey." Thus, based on the evidence presented above, we can safely assume that the sampling frame used for selecting the sample of programs visited in the second round is a good approximation of the true survey population studied here. The first major step in selecting the sample involved stratification of programs in terms of percentage of severely 1Ibid., p. 18 234 handicapped enrolled and size. Percentage severely handi- capped was chosen because it represented the major focus of the second round visits. Program size, measured in terms of the total enrollment, was selected as a stratifying variable because it accurately reflects other dimensions such as economies of scale and the urban-rural dichotomy. Let us review, first, the stratification of programs into clusters in terms of enrollments of severely handi- capped children. One of the clusters was identified prior to starting the stratification as a result of the contractual requirement that we study at least ten programs with no handi- capped children. This meant that it was necessary to divide the remainder of the sample into at least two clusters if we intended to make any meaningful comparisons between groups of programs. Recognizing that dividing the remainder of the sample into three clusters would result in a very low aver- age of nine programs per cluster, it was decided to split the remaining twenty-seven programs to only two clusters. This decision was executed by first Obtaining a frequency distribution of all programs in terms of the percentage of severely handicapped children enrolled. This distribution was examined to arrive at‘a breakpoint which divides the programs with handicapped children into two clusters. Of all poten- tial breakpoints considered-~which included the mean, the median, the upper and lower quartiles, and a few "natural" breakpoints inIthe frequency distribution--a natural breakpoint corresponding to 3.5 per cent of total enrollment 235 was chosen as the dividing line between the two clusters. Thus, according to this stratification criterion, the three clusters we defined in the following manner: Cluster I : Programs in which the severely handi- capped constitute 3.5 per cent or more of total enrollment Cluster II Programs with handicapped children in which the severely handicapped consti- tute less than 3.5 per cent of total enrollment Cluster III Programs reporting no handicapped children enrolled. Stratification of programs in terms of their total enrollment involved a similar process. In this instance, initially two basic alternatives were explored. First, the total population of Head Start programs was divided into three groups containing equal numbers of programs. This Option was found unsatisfactory since it resulted in allo- cation of a very large proportion of the sample to small programs. The second alternative involved division of programs into groups in a manner that would equate the total number of children enrolled in each group. This sec- ond Option was also found unsatisfactory because one of the groups included only very few extremely large programs. A workable solution was found after a study of the interval between the breakpoints which corresponded to these two extremes. The solution consisted of selecting two natural breakpoints within the two intervals. The resulting size groups were as follows: 236 Small programs . l - 120 enrolees Medium programs : 121 - 300 enrolees Large programs : 301 or more enrolees Distribution of the 1,353 programs into the three clusters and the three size groups is diSplayed in Table A.1. TABLE A.1 DISTRIBUTION OF NUMBER OF HEAD START PROGRAMS BY SIZE AND CLUSTER Size Cluster I Cluster II Cluster III Total Small 138 566 128 832 Medium 48 281 22 351 Large 15 144 11 170 Total 201 991 161 1,353 Source: Systems Research Incorporated, Head Start Project files, Lansing, Michigan, 1974. Selection Of the thirty-seven programs was preceded by the introduction of a constraint which dropped many of the programs from the sampling frame. The need to obtain a sample of programs with "adequate” number of children repre- senting all handicapping conditions and severity levels was the primary reason for the adoption of the two screening criteria. The first criterion applied only to programs in Cluster I and demanded representation of at least three of the nine handicapping conditions, with at least two severely 237 handicapped children in each of the three handicapping con- dition categories. The second criterion applied only to programs in Cluster II and required representation of at least three handicapping conditions, with at least two minimally or moderately handicapped children in each of the three handicapping condition categories. These two criteria ensured that the field workers would be able to collect child-specific data at programs in Clusters I and II on at least three children with different handicapping conditions and that they would also have the option to select from at least two children within a given handicapping condition and severity level. These criteria were not at all applied to programs in Cluster III, since no child—Specific or class- room observation data were to be collected from programs with no handicapped children. The sample frame reduced by the two screening criteria is displayed in Table Ao2. Initial selection of the thirty-seven programs involved introduction of two additional constraints. First, it was felt that the programs selected from Clusters I and II Should provide ”minimal" representation of each handicapping condi- tion. Second, it was agreed that during the selection proc- ess an attempt should be made to attain an "adequate" repre- sentation of programs across OCD regions. The ten programs selected from Cluster III were drawn through a two stage process. First, a Simple random sample of ten programs was selected. Second, the regional 238 TABLE A.2 DISTRIBUTION OF NUMBER OF HEAD START PROGRAMS WHICH MET THE SCREENING CRITERIA BY SIZE AND CLUSTER Size Cluster I Cluster II Cluster III Total Small 6 76 128 210 Medium 14 149 22 185 Large 6 * 100 11 117 Total 26 325 161 512 Source: Systems Research Incorporated, Head Start Project files, Lansing, Michigan, 1974. distribution of this sample was studied to observe if it dif- fered significantly from the regional distribution of the pop- ulation of 161 programs. In this instance the regional dis- tribution of the first random sample selected was judged to be "adequate" and, therefore, no replacements were made. Selection of programs from Cluster II involved a dif- ferent procedure. First, a quota of twelve programs was established for Cluster I, leaving a total sample size of fifteen for Cluster II. Second, a twenty-five per cent sim- ple random sample was drawn from each size group. Programs falling into this sample of eighty-one were studied individ- ually in terms of their region and distribution of mildly/ moderately handicapped. Third, an initial sample of fifteen programs was selected purposively from the reduced universe. 239 Selection of these programs was made collectively by three of the project principals after a discussion of the advantages and the disadvantages Of including each of the eighty-one programs into the sample of fifteen. The quota of twelve for programs in Cluster I was dic- tated by the need to obtain a sample of at least four severely handicapped children for each handicapping condition. Given that the minimum of four children with a particular handi- capping condition were to be drawn from at least four differ- ent programs, and given also that no more than four children were to be studied at each program selected from Cluster I, it was deduced that it would be necessary to select at least twelve programs from Cluster I. Selection of the twelve programs from Cluster I was made on a purposive basis. Since there were only twenty-six programs left after application of the two screening criteria, there was no room for further reduction of the parent popu- lation. The purposive selection process for Cluster I was the same as the one used for Cluster II. The next step in the selection process involved comple- tion of telephone interviews with the directors of the thirty- seven programs. These interviews had four specific purposes: (1) to verify the questionnaire data on the numbers of mildly/ moderately and severely handicapped children, (2) to verify the presence of children with specific handicaps selected to the case-study sample, (3) to Obtain the consent of the pro- gram director for studying his program in the second round, 240 and (4) to schedule the visit and make the initial logistical arrangements. All telephone interviews were conducted by the project principals at Syracuse University. As a result of these interviews four programs were dropped from the sample because of scheduling difficulties. No program was drOpped due to the absence from the program of previously reported handicapped children. Each of the four programs dropped from the sample was replaced with one from the same cluster-size group. In addi- tion, an attempt was made to select the substitute from the same OCD region and from among the programs with the same or similar combination of handicapping conditions. Interviews were also conducted with the directors of the four new pro- grams. The thirty-seven programs thus identified constituted the final sample of programs to be visited in the second round. One of the programs in the sample withdrew two days before the visit as a result of scheduling difficulties and unanticipated commitments. This program was not replaced with another one since most programs were drawing close to the end of their program year. Geographical distribution of the thirty-six programs in the final sample is displayed in Table A. 3 O Pre-fieldwork Activities Two types of pre-fieldwork activities deserve our atten- tion here. The first one relates to the training of the 241 Amy OSOH .cowaom O Amv.unoz.mmcwummm huwo mmmcmM .HH> sz.xu<.sueo amputee o museum .H> . c . no so O Advamvmomz.wm”wcow O AAV.:O«z.wchcmA O Amv.ccfizmhauo>m3 owmuwno .> AAv.ccoa.hauoucoz O . . AAV.mmaz.comxumn O . . AAV Wham nonoumz O sz.ccoa.mwoocmuum£o O AAN %M.OHHH>Oxam mucmau< .>H sz mu amouohmz O Amv.o.z.ocoom O Amv OH< mwoomaams Amv.pu.p=na:Hoo . .O .wc mo O Amv.pm.=oucpuom . _Amvepmflw.umSpommpew . sz.p>.s.aoupapafiafi= peaaapppaena .HHH AmVOyOZQAHomvSm . O O a O Amv.w.z.xcoum O AAV H z xumsmz Mao» 3oz HH szpaepz .papauuoa Amv.m.z.ocoox O szocamz .numm goumom .H Amv.ommz .OSOumOHumso HHH waspsao HH umumaao H “armada cosmos moo KMHmDAU 92¢ m.mNHm z QZDom 9260mm mmh Mom QNHUNANm mMHHm nah mo ZOHHDQHMHmHn m.< MAQpz.oaoe o Aav.maapo.peflppfi> . confiscate ppm .xH Amv.oaoo.sufio coupe . uo>apn .HHH> HHH ppumsflo HH “mousse H “pupaao cosmos coo Aposcfiucoov m.< mam