¹ See Cowart, Andrew T., “Anti-Poverty Expenditures in the American States,” in State and Urban Politics, Readings in Comparative Public Policy, ed. Hofferbert, Richard I. and Sharkansky, Ira (Boston: Little, Brown, 1971), p. 415 Google Scholar.

² For a concise discussion of this problem, see Przeworski, Adam and Teune, Henry, The Logic of Comparative Social Inquiry (New York: Wiley, 1970), pp. 51–53 Google Scholar.

³ Walker, Jack L., “The Diffusion of Innovations Among the American States,” in Hofferbert, and Sharkansky, , pp. 377–412 Google Scholar; a somewhat revised version of this article, found in Politics in the American States, ed. Jacob, Herbert and Vines, Kenneth N., 2nd ed. (Boston: Little, Brown, 1971), pp. 354–387 Google Scholar, indicates (Tables 7, 8, 9) that about half the diffusion depends on four states, one in each of four innovation regions, which provide innovations to states. It would appear, though it is not entirely clear, that California and New York both provide policy models for virtually all other states, that the regionalism largely occurs in accepting innovation from what might be called second-order innovators—Wisconsin, Florida, Illinois—and in the sequence of diffusion (rather than the ultimate source of innovation).

⁴ See Sharkansky, and Hofferbert, , “Dimensions of State Policy,” in Jacob, and Vines, , p. 346 Google Scholar, and Sharkansky, , “State Administrators in the Political Process,” in Jacob, and Vines, , p. 266 Google Scholar.

⁵ Dye, Thomas R., Politics, Economics, and the Public (Chicago: Rand McNally, 1966), pp. 303–309 Google Scholar.

⁶ It is standard statistical practice to control for the influence of a third variable on both variables in a relationship. Controlling for effects on just one of the variables produces anomalies, especially in the correlation coefficients. Controls reduce the variance in the “dependent” and “independent” variables by the amount of covariance (weighed) with the “control” variable. The reduction in the variances properly adjusts the base—the estimated maximum covariance—against which the independent-dependent covariance is measured in a correlation coefficient. When only one of the variables is controlled, only one of the variances is properly reduced, leaving an established maximum covariance which is artificially high, so the correlation coefficient is estimated to be lower than it is.

⁷ Walker, , “Diffusion of Innovations,” p. 384 Google Scholar.

⁸ A number of authors use the coefficient of variation to establish that there are large differences among states. The statistic is not convincing. The coefficient of variation is meaningful only with a normal distribution and a true zero point, in which case the coefficient of variation for the American population of 200 million (where the lowest case is expected to be over 5.6 standard deviations from the mean) should be less than .18 at the maximum, when the lowest case is zero; if the lowest case is more than zero, then the coefficient of variation should decrease. The coefficient of variation for fifty states, however, is expected to have a maximum of .429 (the lowest case is expected to be about 2.33 standard deviations from the mean). We might expect, then, to find fairly high coefficients of variation in the states. The coefficient of variation is useful only as a scale device for comparing standard deviations on variables in which everything is similar except the magnitude of the scores. Any other type of difference (number of cases, modality of cases, absence of a ratio scale, etc.) will make coefficients of variation noncomparable.

⁹ Dye, , Politics, Economics and the Public, p. 98 Google Scholar.

¹⁰ Dishman, Robert B. and Craig, Robert F., “Regionalism, State Policy Outputs, and General Explanatory Models,” a paper delivered at the Annual Meeting of the American Political Science Association, New York, September 2–6, 1969 Google Scholar; tables between pages 27 and 28.

¹¹ Blalock, Hubert M. Jr., Causal Inferences in Nonexperimental Research (Chapel Hill: University of North Carolina Press, 1964), pp. 103–114 Google Scholar.

¹² The literature on level fallacies is considerable. For a basic discussion, see Alker, Hayward R. Jr., “A Typology of Ecological Fallacies,” in Quantitative Ecological Analysis in the Social Sciences, ed. Dogan, Mattei and Rokkan, Stein (Cambridge, Mass.: The MIT Press, 1969), pp. 69–86 Google Scholar.

¹³ For summary discussions, see Blalock, , Causal Inferences, pp. 180–181 Google ScholarPubMed and Hannan, Michael T., “Problems of Aggregation,” in Causal Models in the Social Sciences, ed. Blalock, Hubert M. Jr. (Chicago, Aldine, Atherton, 1971), pp. 491–496 Google Scholar.

¹⁴ Hannan, , “Problems of Aggregation,” pp. 487–489, 496–505 Google Scholar.

¹⁵ Ibid., pp. 499–500.

¹⁶ Causal Inferences, p. 180.

¹⁷ Hannan, p. 501 and note.

¹⁸ Causal Inferences, p. 108.

¹⁹ Ibid., p. 180.

²⁰ Ibid., p. 180.

²¹ Under such a scheme, the party share of offices would be decided by the party share (to the nearest even percentage) of the national vote, with the N Democratic governors going to the N most Democratic states.

²² Munger, Frank, “Opinions, Elections, Parties, and Policies,” a paper delivered at the Annual Meeting of the American Political Science Association, New York, September 2–6, 1969 Google Scholar.

²³ Munger, from Table 1, pp. 7–12. The marginals used here to generate expectations are those for an entire class of questions. Results are similar if marginal expectations are generated separately for each question.

²⁴ Ibid., Table 1.

²⁵ The uniform national law, while improving opinion-law democracy, removes the possibility of further innovation on that policy. Although a variety of mechanisms (such as time limitations on the uniform law) could remove this disability, the whole suggestion is an academic standard rather than a practical proposal.

²⁶ “Public Opinion and Taxes,” United States Advisory Commission on Inter-governmental Relations, (Washington, May, 1972)Google ScholarPubMed, Table A-3, p. 15, Table A-1, p. 13, Table A-4, p. 16.

²⁷ Expected deviation, calculated from the state variance components reported in Stokes, Donald E., “Parties and the Nationalization of Electoral Forces,” in The American Party Systems, ed. Nisbet, William and Burnham, Walter Dean (London: Oxford University Press, 1967), Table 1, p. 186 Google Scholar.

²⁸ For discussions of the changes over time in voters' orientations to parties, see Pierce, John C., “Party Identification and the Changing Role of Ideology in American Politics,” Midwest Journal of Political Science, 14 (February, 1970) 25–42 CrossRefGoogle Scholar; RePass, David E., “Issue Salience and Party Choice,” American Political Science Review, 65 (June, 1971), 389–400 CrossRefGoogle Scholar; Field, John O. and Anderson, Ronald E., “Ideology in the Public's Conceptualization of the 1964 Election,” The Public Opinion Quarterly, 33 (Fall, 1969), 380–398 CrossRefGoogle Scholar; and Pomper, Gerald M., “From Confusion to Clarity,” American Political Science Review, 66 (June, 1972), 415–428 CrossRefGoogle Scholar.

²⁹ “The Nationalization of State Politics,” in State and Urban Politics, pp. 469, 472 and “Dimensions of State Policy,” p. 322.

³⁰ “Socioeconomic Dimensions in the American States,” in State and Urban Politics, pp. 456–457.

³¹ Walker, , “Diffusion of Innovations,” pp. 360–361, and column 1, Table 6, p. 377 Google Scholar.

³² Competition among states might also be a source of convergence. The states are unequal competitors, however, so highly similar policies might not result from competition.

³³ For a similar suggestion, see Jacob, Herbert and Lipsky, Michael, “Outputs, Structure, and Power,” in State and Urban Politics, p. 21 Google Scholar.

³⁴ Hofferbert, , “Socioeconomic Dimensions in the American States,” Table 6, p. 460 Google Scholar.

³⁵ Sharkansky, and Hofferbert, , “Dimensions of State Policy,” pp. 331–333 Google Scholar.

³⁶ Ibid., pp. 338–339.

³⁷ Rose, Douglas, “State Election Systems.” (Ph.D. dissertation, University of Minnesota, 1970), chapters 4 and 5Google Scholar.

³⁸ Accomplished mainly through strong party organization and at the expense of individual demogogic and coalition-building skills. See Rose, “State Election Systems,” Chapter 5.

³⁹ Munger, , “Opinions, Elections, Parties, and Policies,” pp. 30–31 Google Scholar.

⁴⁰ Rose, , “State Elections Systems,” p. 215, pp. 158–159Google Scholar.

⁴¹ Wisconsin's innovation and progressive policies on welfare-education appear rooted in earlier state-national differences associated with the Progressive movement and continued through the state bureaucracy which is one of the major monuments of progressivism. The bureaucracy itself can continue as “too” innovative for the state because, as discussed in the text, it is highly functional for a state to be innovative. The origins of the Wisconsin progressive deviancy can speculatively be traced to the deviant “administrative” culture of the large German population, the availability of the progressive movement, the alienation from the Democratic party in World War I, and the feudal electoral politics associated with the La Follettes.

⁴² One possibility is that the legitimacy that the national government has for a state's citizens may correspond (partially) to the legitimacy of the state government, for the same conflictual process affects both of them.