¹ “Innovations in the States: A Diffusion Study,” American Political Science Review LXVII (December 1973), 1174–85

² “The Diffusion of Innovations Among the American States,” American Political Science Review LXIII (September 1969), 880–99

³ Most of this sociological research is discussed in Rogers, Everett M., Diffusion of Innovations (New York, 1962)Google Scholar; Rogers, Everett M. and Shoemaker, F. Floyd, Communication of Innovation: A Cross-Cultural Approach (New York, 1971)Google Scholar; and Hamblin, Robert, Jacobsen, R. Brooke, and Miller, Jerry L.L., A Mathematical Theory of Social Change (New York, 1973).Google Scholar

⁴ Walker, “Diffusion of Innovations,” 881

⁵ By “ ‘the determinants of state policy’ area” I mean the comparative state research of which the following are examples: Dawson, Richard E. and Robinson, James A., “Inter-party Competition, Economic Variables and Welfare Policies in the American States,” Journal of Politics XXV (May 1963), 265–89CrossRefGoogle Scholar; Dye, Thomas R., Politics, Economics and the Public: Policy Outcomes in the American States (Chicago, 1966)Google Scholar; Sharkansky, Ira, Spending in the American States (Chicago, 1968)Google Scholar; Cnudde, Charles F. and McCrone, Donald J., “Party Competition and Welfare Policies in the American States,” American Political Science Review LXIII (September 1969), 858–66CrossRefGoogle Scholar; and Booms, Bernard H. and Halldorson, James R., “The Politics of Redistribution: A Reformulation,” American Political Science Review LXVII (September 1973), 924–33.CrossRefGoogle Scholar

⁶ Walker, Jack L., “Comment: Problems in Research on the Diffusion of Policy Innovations,” American Political Science Review LXVII (December 1973), 1187Google Scholar

⁷ Ibid.

⁸ “Diffusion of Innovation,” 884

⁹ Ibid., 887

¹⁰ Ibid., 898

¹¹ “Innovations in the States,” 1175

¹² The end-product of Gray's work is a model “whose parameters can be estimated by the following equation:

A_t = (1 + bL)A_t−1 − bA_t−1² + c + e.

The concept of adopting units interacting is indicated by the portion (1 + bL)A_t−1, where bL is some proportion of the maximum proportion of possible adopters of a particular law and A_t−1 is the proportion of adopters at time t − 1. The squared term allows the fit to be curvilinear (applied to the rate of change curve). The quadratic r ² will always be equal to or greater than the linear r ² for the same data set. The interesting question in Gray's research is whether or not the quadratic r ² is significantly greater than the linear.

I would like to acknowledge the technical (and substantive) assistance of Leo Jonker of the Department of Mathematics, Queen's University. He cannot, however, be blamed for what I have done with his assistance.

¹³ “Innovations in the States,” 1179

¹⁴ An earlier version of this paper pursued both directions. In that paper, and in subsequent polynomial regression analysis, the distinction between the logistic and decaying exponential diffusion patterns was not strong. The results of the polynomial regression suggest that only three of the twenty-five policy items approximate a logistic curve. See Poel, Dale H., “The Diffusion of Legislation Among the Canadian Provinces: A Mathematical and Statistical Analysis,” paper presented at the Annual Meeting of the Canadian Political Science Association, Edmonton 1975Google Scholar

¹⁵ “Determinants of Innovation in Organizations,” American Political Science Review LXIII (March 1969), 111–26

¹⁶ Ibid., 114

¹⁷ I would like to thank the Queen's University students in Richard Simeon's policy seminar for reactions to earlier analysis of this data. Their comments were important for this work.

¹⁸ By this I mean all Canadian political scientists west of Moncton, New Brunswick.

¹⁹ In a similar vein, an assumption was made regarding the equivalence of the respective provincial legislation for each item. It would be devastating if the words following similar legislative titles were random in their meaning. For the most part the legislation does not differ radically across the provinces. In many instances the respective statutes are nearly identical. Sometimes judgment was required to decide upon an adoption date. For example, Quebec has prohibited a person from throwing “into any river, rivulet, or watercourse in the Province any slab, bark, waste stuff, or other refuse of any sawmill (except sawdust) or any stumps, roots or wasted timber, [if the person] allows the same to remain in and to obstruct…” since at least 1888 (RSQ 1888, S. 5555). Not until 1964, however, did the concern shift from obstruction which would interfere with other people's exploitation of the waterway to the broader notion of “pollution” (SQ 1964, C. 51). This latter point was chosen as Quebec's adoption date for the Environmental protection – water policy item. Professor G.I. Gow, University of Montreal, has suggested that the Quebec water protection legislation “has been judged to be very weak” (personal correspondence). Again, there is no assumption made regarding enforcement of legislation once it is passed into law. Enforcement is an important question but is beyond the scope or intention of this particular research project.

²⁰ An N of 10 is usually considered extremely small for factor analysis. However, the size of the zero-order correlation coefficients in the inter-item matrix is substantial. Only eight of fifty-one coefficients contributing to the factors are below .49 and most range from .70 to .92. The actual factor analysis procedure is “blind” to the number of cases involved or any question of statistical significance. The strength and clustering of the initial coefficients are more important than the number of cases per se.

²¹ Thanks, but not responsibility for use, are due to Professor Hugh Thorburn of Queen's University for suggesting Lowi's concept of interest-group liberalism for the second cluster of items. See Lowi, Theodore, “The Public Philosophy: Interest-Group Liberalism,” American Political Science Review LXI (March 1967), 5–24.CrossRefGoogle Scholar The “Socialism” label might imply more than I want it to. Other alternatives were considered but rejected.

The reader might wonder why I bothered with the factor analysis at all, given the liberties taken to arrive at the socialism and interest-group liberalism groupings. Factor analysis was introduced as an appropriate means of testing a conceptual hypothesis concerning distinct dimensions of policy adoptions. The hypothesis was simply that the twenty-five items were not unidimensional. The initial results of Table III could have been shelved and, while still benefiting from the analysis, the policy items could have been reconstructed into “clean” dependent variables with an after-the-fact rationale for their existence. Table III and its discussion simply represent what happened as I followed through on an initial measurement decision.

²² By “fad” I mean something like the legislation with respect to the mailing of unsolicited credit cards which either prohibits such action or removes from the recipient any legal obligation if he does not personally use the card. This legislation diffused through most of the provinces in a fairly short period of time. It did not cost the provinces any money, it made some people happy, and did not make anyone really mad.

²³ “The Public Philosophy,” 12

²⁴ In this factor analysis the policy items are the units of analysis and the provinces are treated as variables. The beginning inter-item matrix is a 10 × 10 matrix and the factor analysis looks for correlated provinces; that is, provinces which have similar adoption patterns for the twenty-six cases.

²⁵ See the research cited in footnote 5 above.

²⁶ The socioeconomic variables are all measured at the year 1961. An argument can go two ways. Using 1961 will allow some lag effect which could be appropriate for these long-term secular trends since over 81 per cent of the adoption dates are 1961 and after. On the other hand, the relative position of the provinces on these secular trends is fairly stable for the period from 1945 to 1961 so that an indicator based on a 1961 measurement point can stand for the variables as they would be measured in previous years. The actual measure used is the province's proportion of the highest province's value. This yields variables which range from 1.00 for the highest province to some proportion greater than .00 for the lowest province.

The political variables are averages calculated over the time period of 1945 to 1971. The first party competition measure (PC 1−2) is simply the difference between the first and second party's share of votes subtracted from one. The second (PC 10 per cent) is the percentage of ridings for a given election which were won by a 10 per cent margin or less. The malapportionment average is derived from data found in Pasis, Harvey E., “The Inequality of Distribution in the Canadian Provincial Assemblies,” Canadian Journal of Political Science V (September 1972), 433–6.CrossRefGoogle Scholar The malapportionment measure is not reported for all elections within our time span. The average is based on the information available.

²⁷ “Party Competition and Welfare Policies in the American States”

²⁸ Sharkansky, Ira and Hofferbert, Richard I., “Dimensions of State Politics, Economics, and Public Policy,” American Political Science Review LXIII (September 1969), 867–79CrossRefGoogle Scholar

²⁹ “Diffusion of Innovations Among the American States”

³⁰ Spending in the American States. Sharkansky's conclusions are a result of his decision to use state expenditures per capita only and not state and local expenditures combined. Most economists (and Statistics Canada) take the opposite view and opt for state/provincial and local expenditures combined “for the sake of comparability.” The comparability sought, however, must be seen as level of functional policy delivered within a geographic area and not comparability between political systems. There are more satisfactory ways of handling this problem than simply adding state/provincial and local expenditures together. For a survey of the “determinants” literature from an economist's point of view see Bird, Richard, The Growth of Government Spending in Canada (Toronto, 1970)Google Scholar, appendix B.

³¹ “Determinants of Innovation in Organizations”

³² I realize less complimentary interpretations can be given to high politician salaries. The use of cabinet ministers' salaries is not of any special significance since the correlation between salaries (MLA, Leader of the Opposition, Speaker, etc.) is so high as to make them nearly identical indicators.

³³ The concept and the literature are briefly discussed in Mohr, “Determinants of Innovation in Organizations,” 122–3

³⁴ Cited in ibid., 122

³⁵ Ibid.

³⁶ The best known is probably Lipset's, SeymourAgrarian Socialism: The Cooperative Commonwealth Federation in Saskatchewan (expanded ed., New York, 1968).Google Scholar

³⁷ See Poel, Dale H., “Canadian Provincial and American State Policy: A Qualitative Explication or an Empirical Difference,” paper presented at the Annual Meeting of the Canadian Political Science Association, Montreal 1972Google Scholar

³⁸ The step-wise regression program in the SPSS package was used to generate these equations. For Regression I the final solution with wealth and malapportionment entered is significant at the .05 level (F = 6.37, df = 2/7). For Regression II the same holds through industrialization and malapportionment (F = 5.11, df = 2/7) but falls below the F required for the .05 level with the third variable entered (F = 3.85, df = 3/6, s < .05). For Regression III the F obtained with wealth, malapportionment, and industrialization entered is close enough to satisfy me (F = 4.46, df = 3/6, s just > .05 where 4.76 is required). Again, this drops considerably below the .05 level when civil service size is entered and df = 4/5. One rather quickly runs out of degrees of freedom with an N of 10. The assistance of Professor Ralph Joyce of Queen's University is acknowledged in connection with these regression solutions.