Hostname: page-component-cd9895bd7-7cvxr Total loading time: 0 Render date: 2024-12-24T17:37:05.646Z Has data issue: false hasContentIssue false
  • English
  • Français

Theory, Methodology and Results in Political Science – The Case of Output Studies

Published online by Cambridge University Press:  27 January 2009

George A. Boyne
Get access Rights & Permissions [Opens in a new window]

Extract

This review evaluates the success of output studies in explaining intranational variation in the policies of local governments. Output studies address the central question of political science: why do different governments adopt different policies? It is therefore important to examine the contribution of output studies to our understanding of the reasons for policy variation and the role of politics within the relevant causal processes.

Type
Review Article
Information
British Journal of Political Science , Volume 15 , Issue 4 , October 1985 , pp. 473 - 515
Copyright
Copyright © Cambridge University Press 1985

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 For details of these results see Table 3.

2 Dye, T. R., Policy Analysis (Alabama: University of Alabama Press, 1978).Google Scholar

3 See Appendix B. Of this total of nineteen studies, nine stem from discipline bases other than political science – five from economics, two from geography and two from social administration. Of the remaining ten studies, four are by American political scientists.

4 The major American critiques which show this concern are Coulter, P. B., ‘Comparative Community Politics and Public Policy’, Polity, III (1968), 22–43Google Scholar; Fenton, J. H. and Chamberlayne, D., ‘The Literature Dealing with the Relationships Between Political Processes, Socio-Economic Conditions and Public Policies in the American States: A Bibliographic Essay’, Polity, IV (1969), 1388–404Google Scholar; Jacob, J. and Lipsky, K., ‘Outputs, Structure and Power: An Assessment of Changes in the Study of State and Local Polities’, Journal of Politics, XXX (1968), 61–82Google Scholar; Rakoff, S. and Schaeffer, G., ‘Politics, Policy and Political Science: Theoretical Alternatives’, Politics and Society, I (1970), 51–77CrossRefGoogle Scholar; Munns, J. M., ‘The Environment, Politics and Policy Literature: A Critique and Reformulation’, Western Political Quarterly, XXVIII (1975), 646–67.Google Scholar British critiques which show the same concern are King, D. N. ‘Why Do Local Authority Rate Poundages Differ’, Public Administration, LI (1973), 165–73CrossRefGoogle Scholar; Alt, J. E., ‘Politics and Expenditure Models’, Policy and Politics, V (1977), 83–92CrossRefGoogle Scholar; Newton, K. and Sharpe, L. J., ‘Local Outputs Research: Some Reflections and Proposals’, Policy and Politics, V (1977), 61–82.CrossRefGoogle Scholar The most recent British work states that an ‘overriding objective has been to restore in output research the status of the political party’ (Sharpe and Newton, see Appendix B).

5 See Geise, J. P., ‘Theory Construction and Political Inquiry’, Canadian Journal of Political Science, IX (1976), 626–53.Google Scholar

6 Newton, and Sharpe, , ‘Local Outputs Research’, p. 79.Google Scholar

7 Easton, D., The Political System (New York: A. Knopf, 1953)Google Scholar; ‘An Approach to the Analysis of Political Systems’, World Politics, IX (1957), 383–400Google Scholar; A Framework for Political Analysis (Chicago: University of Chicago Press, 1979)Google Scholar; A Systems Analysis of Political Life (Chicago: University of Chicago Press, 1979).Google Scholar

8 Koestler, A., The Ghost in the Machine (London: Pan, 1975).Google Scholar

9 Dawson and Robinson, see Appendix A.

10 J. H. Fenton claims that he is the unheralded pathbreaker in this field. See Fenton, and Chamberlayne, , ‘The Literature Dealing with the Relationships Between Political Processes’, p. 389.Google Scholar Others could claim to have been performing output studies without knowing it – see, for example, Hawley, Appendix A.

11 On output studies in Europe see Aiken, M. and Depre, R., ‘The Urban System, Politics and Policies in Belgian Cities’, in Newton, K., ed., Urban Political Economy (London: Frances Pinter, 1981).Google Scholar On Canada see Falcone, D. and Mishler, W., ‘Legislative Determinants of Provincial Health Policy in Canada’, Journal of Politics, IXL (1977), 345–67CrossRefGoogle Scholar; Poel, D. H., ‘The Diffusion of Legislation among the Canadian Provinces: A Statistical Analysis’, Canadian Journal of Political Science, IX (1976), 604–26.Google Scholar On Israel see Torgovnik, E., ‘Local Policy Determinants in a Centrist System’, Publius, VII (1977), 61–84Google Scholar, and ‘Central Aid and Local Policy’, Public Finance Quarterly, VI (1978), 211–39.Google Scholar

12 See Foley, J. W., A Comparative Study of the Determinants of Public Policies (Cornell University, Program in Urban and Regional Studies, Occasional Paper 9.)Google Scholar For a discussion of studies of output variation in Britain which do not include political variables see Newton, K., ‘Community Performance in Britain’, Current Sociology, XXIV (1976), 49–86.Google Scholar

13 The term is from Hilton, G., Intermediate Politometrics (New York: Columbia University Press, 1976).Google Scholar

14 Hofferbert, R. I., ‘State and Community Policy Studies’, in Robinson, J. A., ed., Political Science Annual Volume III (New York: Bobbs Merrill, 1972).Google Scholar

15 The hypothesis rests on a number of assumptions about the composition and values of the electorate and the behaviour of politicians. These assumptions have been given detailed consideration in the rational choice literature. See Barry, Brian, Sociologists, Economists and Democracy (Chicago: University of Chicago Press, 1970).Google Scholar

16 IPC is specified by averaging three measures of majority party success over a 21 -year period: percentage of votes for the governor, percentage of state senate seats and percentage of state house seats.

17 Dawson, and Robinson, , p. 280Google Scholar, Appendix A. Problems of employing such background variables in statistical analysis are discussed in Blalock, H. M., Causal Inferences in Non-Experimental Research (Chapel Hill: University of North Carolina Press, 1964).Google Scholar

18 C. F. Cnudde and D. J. McCrone, Appendix A.

19 Dawson, and Robinson, , p. 289, Appendix A.Google Scholar

20 For example Cowart, Fry and Winters, Baer and Jaros, Appendix A.

21 See American critiques in footnote 4.

22 See discussion of political variables in Section III below.

23 Feyerabend, P., Against Method (London: Verso, 1975).Google Scholar

24 A useful summary of the concepts and their inter-relationship is provided in Hall, P., Land, H., Parker, R. and Webb, A., Change, Choice and Conflict in Social Policy (London: Heinemann, 1976).Google Scholar For more critical discussion see Astin, J. D., ‘Easton I and Easton II’, Western Political Quarterly, IV (1972), 726–38Google Scholar; Leslie, P., ‘General Theory in Political Science: A Critique of Easton's Systems Theory’, British Journal of Political Science, II (1972), 155–72CrossRefGoogle Scholar; Sozarno, J. S., ‘David Easton and The Invisible Hand’, American Political Science Review, LXIX (1975), 91–106.Google Scholar

25 A Systems Analysis of Political Life, Preface xiv.

26 A Framework for Political Analysis, p. 89 (emphasis added).Google Scholar

27 A Systems Analysis of Political Life, Preface xiv.

28 A Systems Analysis of Political Life, Preface xiv.

29 See, for example, the passage quoted from A Systems Analysis of Political Life, p. 133Google Scholar, in the discussion of ‘The Autonomy of the Authorities’ below.

30 The Political System, p. 158.Google Scholar

31 A Systems Analysis of Political Life, pp. 25–7.Google Scholar

32 A Systems Analysis of Political Life, p. 53.Google Scholar

33 A Systems Analysis of Political Life, p. 18.Google Scholar

34 A Framework for Political Analysis, p. 99.Google Scholar

35 A Framework for Political Analysis, p. 133.Google Scholar

36 A Systems Analysis of Political Life, p. 72.Google Scholar

37 A Systems Analysis of Political Life, p. 389.Google Scholar It is important to note that the ‘relative autonomy’ variant of Marxist theory refers to a conception of autonomy which is different from that outlined here. Relative autonomy concerns freedom of state action only from the short-term political constraints imposed by ‘fractions of capital’ who fail, unlike state policy makers, to see the long-term interests of capital as a whole. (To the extent that state policy makers are not consciously omniscient then some ‘true unconsciousness’ must presumably be at work.) However, the state possesses no such relative autonomy from the long-term ‘structural requirements’ of capital. For a discussion of state autonomy in general and the Marxist concept of relative autonomy in particular see Nordlinger, A. E., On the Autonomy of the Democratic State (Cambridge, Mass.: Harvard University Press, 1981).Google Scholar For a crude Marxist application of the relative autonomy concept at the local level see Cockburn, C., The Local State (London: Pluto Press, 1977).Google Scholar

38 On the role of axioms in the development of theories see Blalock, H. M., Theory Construction (Englewood Cliffs, N.J.: Prentice Hall, 1969).Google Scholar

39 Similar arguments are contained in other areas of political theory. For example, Schattschneider argues that the definition of conditions as ‘issues’ of legitimate relevance to political action is ‘the supreme instrument of power’. See Schattschneider, E. E., The Semi-Sovereign People (New York: Holt, Rinehart & Winston, 1964).Google Scholar This argument has also been prominent in the recent boom in ‘values’ literature. See, for example, Young, K., ‘Values in the Policy Process’, Policy and Politics, V (1977), 1–22.CrossRefGoogle Scholar An early application of the values perspective in output studies is Elau and Eyestone, Appendix A. The perspective has been developed most explicitly in the context of output studies by Hansen, T., ‘Transforming Needs into Expenditure Decisions’Google Scholar in Newton, , ed., Urban Political Economy.Google Scholar

However, while Hansen departs from this point he arrives at very different conclusions from this discussion. Hansen specifies a model analogous to that considered in the discussion of the estimation of interaction effects through sub-groups in Section IV below.

40 The phrase is adapted from Berger, T. and Luckman, K., The Social Construction of Reality (New York: Doubleday, 1966).Google Scholar For a more specific application of their argument to the relationship between organizations and their environment see Silverman, D., The Theory of Organisations (London: Heinemann, 1970).Google Scholar

41 Dawson, and Robinson, , p. 266Google Scholar, Appendix A (emphasis added).

42 See American critiques in footnote 4.

43 Dye (b), Appendix A; Grumm, Appendix A; Hofferbert, R. I., ‘Ecological Development and Policy Change in the American States’, Mid-Western Journal of Political Science, X (1966), 464–83CrossRefGoogle Scholar; and ‘Elite Influence in State Policy Formation’, Polity, II (1970), 316–44.Google Scholar

44 Dye (b), p. 5. Appendix A.

45 Dye (c). Appendix A. This approach is also used by Davies, O. A., ‘Empirical Evidence of Political Influences upon the Expenditure Policies of Public Schools’, in Margolis, J., ed., The Public Economy of Urban Communities (Washington, D.C.: Resources of the Future Inc., 1965)Google Scholar and Davies, O. A. and Haines, C. H., ‘A Political Approach to a Theory of Public Expenditure: The Case of Municipalities’, National Tax Journal, XIX (1966), 259–75.Google Scholar On tne dangers of inferring attitudinal and behavioural characteristics from socio-economic data see Obler, J., ‘The Dubious Link Between Democratic Politics and Redistributive Fiscal Policies’ in Dye, T. R. and Gray, V., eds, The Determinants of Public Policy (Lexington, Mass.: Lexington, 1980).Google Scholar

46 See Godwin, R. K. and Shepard, W. B., ‘Political Processes and Public Expenditures: A Re-examination Based on Theories of Representative Government’. American Political Science Review, LXX (1976), 1127–35CrossRefGoogle Scholar; Stonecash, J., ‘Assessing the Roles of Politics and Wealth for Public Policy’. Political Methodology, VI (1979). 463–83.Google Scholar

47 Stonecash and Hayes, Appendix A.

48 Rackoff, and Schaeffer, , ‘Politics, Policy and Political Science: Theoretical Alternatives’Google Scholar; Munns, , ‘The Environment, Politics and Policy Literature: A Critique and Reformulation’.Google Scholar

49 C. F. Cnudde and D. J. McCrone, Appendix A; Swant, F., ‘Linking Theory and Method in Urban Policy Analysis: Problems of Test Interpretation’. Political Methodology, IV (1977), 333–46.Google Scholar

50 Lockard, D., p. 212.Google Scholar Appendix A.

51 Coulter, , ‘Comparative Community Politics and Public Policy’Google Scholar: Jacob, and Lipsky, , ‘Outputs, Structure and Power: An Assessment of Changes in the Study of State and Local Polities’.Google Scholar

52 The case for testing for a ‘party of the right’ effect has been argued by F. G. Castles and R. McKinlay, D., ‘Public Welfare Provision. Scandinavia and the Sheer Futility of the Sociological Approach to Politics’. British Journal of Political Science, IX (1979). 157–71.Google Scholar For tests of the effect of Conservative/Independent variables in Britain see Sharpe and Newton, Appendix B.

53 The best critiques of the use of IPC in output studies are Riley, D., ‘Party Competition and State Policy Making: The Need for a Re-examination’, Western Political Quarterly, XXIV (1971), 510–13Google Scholar, and E. T. Jennings (a), Appendix A. Debates concerning the conceptualization and measurement of other political variables are few. Readers interested in particular variables should refer to the studies indicated under the variable headings in Table I.

54 Y. H. Cho and G. Frederickson (a), Appendix A.

55 L. J. Sharpe and K. Newton, Appendix B.

56 There has been some debate about whether region should be used as an environmental or as a political variable. See Sharkansky, I., ‘Regionalism, Economic Status and the Public Policies of American States’, Southwestern Social Science Quarterly (1968), 9–26.Google Scholar Two recent examples of the explicit use of region as a political variable are Hanson, R. ‘The Content of Welfare Policy: The States and Aid to Familes with Dependent Children’, Journal of Politics, XLV (1983), 771–85CrossRefGoogle Scholar; Sigelman, L., Lowery, D. and Smith, R., ‘The Tax Revolt: A Comparative State Analysis’, Western Political Quarterly, XXXVI (1983), 30–51.Google Scholar

57 Religion has also been used as a measure of political culture. See Hutcheson and Taylor, Fairbanks (a) and (b), Appendix A.

58 Dye (b), Appendix A.

59 See Danziger, Sharpe and Newton, Appendix B.

60 Lester, Regens, Perry, Appendix A.

61 See American critiques, footnote 4.

62 Ostrom, E., ‘The Need for Multiple Indicators in Measuring the Output of Public Agencies’, Policy Studies Journal, II (1973), 85–92CrossRefGoogle Scholar; Munns, , ‘The Environment, Politics and Policy Literature: A Critique and Reformulation’Google Scholar. For a general review of problems in the measurement of ‘policy’, see Greenberg, D. G., Miller, J. A., Mohr, L. B. and Vladeck, B. C., ‘Developing Public Policy Theory: Perspectives from Empirical Research’, American Political Science Review, LXXI (1977), 1532–43CrossRefGoogle Scholar

63 See Ashford, Danziger, Sharpe and Newton, Appendix B; Edwards, G. C., ‘Disaggregation in Public Policy Research’Google Scholar, in Dye, and Gray, , The Determinants of Public Policy.Google Scholar

64 Fry and Winters, Appendix A.

65 J. W. Clarke, Appendix A. For a broader discussion of the importance of symbolic policies see Edelman, M., The Symbolic Uses of Politics (Illinois: University of Illinois, 1964).Google Scholar

66 See Lester, Perry, Dye and Renick, Eisinger, Downs, Sigelman et al., Appendix A.

67 These sub-function results are analysed in Boyne, G. A., ‘Output Disaggregation and The Quest for the Impact of Local Politics’, Political Studies, XXXII (1984), 451–58CrossRefGoogle Scholar. On staffing levels see Storey, Appendix B; on housing see Boaden, Alt, Appendix B; on Education see Boaden and Alford, Boaden, Appendix B.

68 For example, Dye (b) and (f); Thompson, Appendix A.

69 See Sharkansky, I., ‘Governmental Expenditures and Public Services in the American States’, American Political Science Review, LXI (1967), 1066–77CrossRefGoogle Scholar; Christensen, J. and Taylor, G., ‘Determinants, Expenditures and Performance of Common Public Services’, Rural Sociology, XLVII (1982), 147–63.Google Scholar

70 Such as they are. For a review of the literature see Barret, S. and Fudge, C., Policy and Action (London: Methuen, 1981).Google Scholar

71 On path diagrams, path analysis and multiple equation models see Duncan, O. D., An Introduction to Structural Equation Models (New York: Academic Press, 1975)Google Scholar; Asher, H. B., Causal Modelling (Beverley Hills, Calif.: Sage, 1976).Google Scholar

72 Standardization is achieved by expressing the coefficient of each variable in units of standard deviation or ‘beta weights’. Where models are presented in path diagrams the beta weights are termed ‘path coefficients’. See Asher, , Causal Modelling.Google Scholar

73 Two ‘rules of thumb’ which have been suggested is to consider zero-order coefficients above 0·8 or above the coefficient of determination for the whole model as constituting ‘high’ collinearity. See Farrar, D. E. and Glauber, R. R., ‘Multicollinearity in Regression Analysis: The Problem Revisited’, Review of Economics and Statistics, IL (1967), 92–107.CrossRefGoogle Scholar

74 For a concise discussion of possible responses to the multicollinearity problem see Gujurati, D., Basic Econometrics (Auckland: McGraw Hill, 1979), Chap. 9Google Scholar. The specification of Model I slightly alleviates the collinearity problem to the extent that environmental variables do not appear in the same equation as policy-maker variables anywhere in the model.

75 See Atkins and Glick. Booms and Halldorson. Clark, Dye and Renick, Elsinger (a) and (b). Fisher. Jones. Meier and England. Pulsipher and Weatherby, Rose. Shaffer and Weber, Sigelman and Smith, Sullivan, Appendix A; see Ashford et al., Davies et al. (a) and (b) Danziger. Jackman and Sellars, Nicholson and Topham (a) and (b). Pinch, Sharpe and Newton. Storey, Appendix B.

76 The zero order correlations are not in themselves an adequate test for multicollinearity. This requires that all the other independent variables be regressed on each independent variable in turn. See Johnston, J., Econometric Methods (New York: McGraw Hill, 1972), pp. 163–4.Google Scholar

77 See Baker and Colby, S. E. Clarke, Fairbanks (a) Gibson et al., Grumm, Ulsaner and Weber. Appendix A; see Alt, Boaden, Appendix B.

78 In addition to the shared problem of imposing an inappropriate model each of these techniques present individual technical problems which introduce further unreliability to the estimates obtained. On partial correlation see Blalock, , Causal Inferences in Non-Experimental Research, pp. 85–8Google Scholar; on multiple partial correlation see Cnudde and McCrone, Appendix A; on stepwise analysis see Lewis-Beck, M., ‘Stepwise Regression: A Caution’, Political Methodology, IV (1978), 213–40.Google Scholar

79 For example, a £1 increase in rateable value per capita is assumed to have the same effect whether Labour holds 5 per cent or 55 per cent of a council's seats.

80 See M. Lewis-Beck, Appendix A; Lewis-Beck, M. and Mohr, L. B., ‘Evaluating Effects of Independent Variables’, Political Methodology, III (1976), 27–47Google Scholar; Swam, , ‘Linking Theory and Method in Urban Policy Analysis: Problems of Test Interpretation’.Google Scholar

81 It is important to stress that the source of the unreliability discussed here is specification error. The extent to which the estimates are inefficient because of multicollinearity is yet a further source of unreliability. The same point applies to the discussion of Model III below.

82 The exception is Klaas, Appendix A.

83 The overestimate in any particular study depends on the extent to which direct effects are included in the estimate of total effects for all environmental variables.

84 Most show partial correlations only for the political variables; some show ‘selected’ partial correlations on no apparent basis; Bingham. Appendix A, attempts to test Model III through a cumbersome sequence of partial correlations which defies interpretation.

85 See Wright, G., ‘Linear Models for Evaluating Conditional Relationships, American Journal of Political Science, XX (1976), 349–73.CrossRefGoogle Scholar

86 Lineberry and Fowler, Appendix A.

87 There is also a subtle theoretical difference between specifying a single equation model with a multiplicative interaction term and specifying a multiple equation model. In the single equation specification, policy makers are assumed never to act independently of environmental constraints. However, in the multiple equation model, this assumption is tested by estimating the value of B 1X 1 in Equation 3 of Model 1. If the coefficient is not significant then this is consistent with autonomy from direct environmental constraints, although not from the indirect constraints which operate through inputs and formal structure

88 See Coulter, , ‘Comparative Community Politics and Public Policy’Google Scholar. For a broader discussion of the meaning of ‘cause’ see Dahl, R. A., ‘Cause and Effect in the Study of Politics’, in Lerner, D., ed., Cause and Effect (London: Collier-Macmillan, 1965).Google Scholar

89 Strictly, a fully dynamic model requires a non-recursive specification which would involve linking theories of policy formulation to theories of policy implementation and identifying the unique impact of policy on the environment. Given the present state of the art of policy analysis this seems rather like running before we can crawl. However, two interesting attempts to move more quickly in this direction are Fowler, P. and Lineberry, R., ‘Comparative Policy Analysis and the Problem of Reciprocal Causation’, in Liske, C., eds., Comparative Public Policy (New York: Wiley, 1975)Google Scholar, and Shepard, W. B. and Godwin, R. K., ‘Policy and Process: A Study of Interaction’, Journal of Politics, XXXVII (1975), 576–82.CrossRefGoogle Scholar

90 See Van Meter, D. S. and Asher, H. B., ‘Causal Analysis: Its' Promise for Policy Studies’, Policy Studies Journal, II (1973), 103–9CrossRefGoogle Scholar; Dye, , Policy Analysis;Google Scholar Lewis-Beck, Tompkins, Appendix A.

91 For a general discussion of the problems of inferring causal processes from static crosssectional models, see Brunner, R. D. and Liepelt, K., ‘Data Analysis, Process Analysis and System Change’, Mid Western Journal of Political Science, XVI (1972), 538–69.CrossRefGoogle Scholar Their argument is developed in the output studies context in Gray, Appendix A. A useful alternative approach to this issue which arrives at substantially the same conclusions as Brunner and Liepelt is Hofferbert, R. I. and Schaeffer, G., ‘The Application of General Systems Methodology to the Comparative Study of Public Policy’, International Journal of General Systems, VIII (1982), 93–108.CrossRefGoogle Scholar

92 Dyson, J. W. and Angelo, D. St., ‘A Methodological Problem in the Socio-Economic Interpretation of State Spending’, Policy Studies Journal, II (1975), 131–6.Google Scholar

93 A notable exception is Jones, Appendix A. It is possible to justify the use of static measures where they represent variables of theoretical importance which were constant over the period in question. Of the studies which analyse output change this is argued only by Lyons and Morgan, Appendix A.

94 Exceptions are Morss, Jones, Roeder, Shaffer and Weber, Uslaner and Weber, Appendix A; Nicholson and Topham, (a) and (b), Karran, Appendix B.

95 Coulter, , ‘Comparative Community Politics and Public Policy’Google Scholar, Rakoff, and Schaeffer, , ‘Politics, Policy and Political Science: Theoretical Alternatives’Google Scholar; Munns, , ‘The Environment, Politics and Policy Literature: A Critique and Reformulation’.Google Scholar

96 Tucker, H., ‘It's About Time: The Use of Time in Cross-Sectional State Policy Research’, American Journal of Political Science, XXVI (1982), 176–96.CrossRefGoogle Scholar

97 Studies listed in Table 1 which report results neither for the relative importance of environmental and political variables nor for their combined explanatory power are omitted from Table 3.

98 A further problem in interpreting results for individual variables is that some studies fail to report the results of significance tests even where the cases represent a sample. There is considerable divergence of opinion within output studies on the utility of significance tests where the cases constitute a population. There is a good discussion of this issue in Sharpe and Newton, Appendix B. See also Henkel, R. E., Tests of Significance (Beverley Hills, Calif.: Sage, 1976).CrossRefGoogle Scholar

99 See Johnson, , Econometric Methods, pp. 35–8 and 129–30.Google Scholar

100 Booms, S. E. Clarke. Eisinger (b), Mazmanian and Sabatier, Stonecash and Hayes. Winters, Appendix A; Storey, Appendix B. The adjusted R 2 can be calculated provided information on the number of cases, number of variables and unadjusted R 2 is presented. See Johnson, , Econometric Methods, pp. 130 for the formula.Google Scholar

101 Fisher, Giertz, Appendix A; Danziger, Foster el al., Appendix B.

102 This greater ability to explain output levels than to explain output change is also true of incremental models. See Danziger, Hoggart, Appendix B.

103 The discussion of unstandardized regression coefficients on p. 492 noted that they are an inappropriate means of measuring the relative importance of the environment and politics because of the different measurement scales of the variables. However, the difference between the significance of the estimates of the unstandardized regression coefficients of environmental and political variables is so clear-cut as to allow a conclusion concerning their relative importance in the following studies: Cnudde and McCrone, Pulsipher and Weatherby, Jennings (b), Appendix A; Oliver and Stanyer. Nicholson and Topham (a), Danziger, Scholfield, Foster et al., Appendix B.

104 This is especially so in the case of results for total expenditures. Weicher and Emerine have shown that the statistical significance of an independent variable in the total expenditure equation depends on the signs, magnitudes and variances of the coefficients for that variable in the equations for functional expenditures of which total expenditure is composed. They conclude that the estimates in the total expenditure equation are likely to be so misleading that they serve no useful purpose. See Weicher, J. C. and Emerine, R. J., ‘Econometric Analysis of State and Local Government Expenditure Functions’, Public Finance, XXVIII (1975), 69–83.Google Scholar

105 See Aiken and Alford, Asher and Van Meter, Gary, LeMay, Appendix A; Downs, G. W. and Rocke, D., ‘Bureaucracy and Juvenile Corrections’Google Scholar, in Dye, and Gray, , The Determinants of Public Policy.Google Scholar

106 See Asher and Van Meter, Cho and Frederickson (a), Lyons and Morgan, Shaffer and Weber, Appendix A; a conflicting result is Eisinger (a), Appendix A.

107 Sullivan, Fry and Winters, Appendix A.

108 A similar analysis is not possible for the British studies because none use more political than environmental variables.

109 This pattern is also present within the broad output categories. For financial output levels 91 per cent of results from models containing more environmental than political variables show environmental effects dominant and 55 per cent of results from models containing more political than environmental variables show political effects dominant. The parallel figures for nonfinancial output levels are 94 per cent and 83 per cent and for output change are 50 per cent and 82 per cent. It is difficult to interpret the low figure for models containing more environmental than political variables in the output change category because it is based on only six results.

110 Danziger, Sharpe and Newton, Appendix B.