¹ Easton, David and Dennis, Jack, “The Child's Acquisition of Regime Norms: Political EfEcacy,” this REVIEW, (March, 1967), p. 27 Google Scholar.

² Easton and Dennis, op. cit., p. 38.

³ Almond, Gabriel A. and Verba, Sidney, The Civic Culture (Princeton: Princeton University Press, 1963), p. 257 CrossRefGoogle Scholar.

⁴ Barnes, Samuel H., “Leadership Style and Political Competence,” in Edinger, Lewis J. (ed.), Political Leadership in Industrialized Societies (New York: John Wiley & Sons, 1967), p. 60 Google Scholar.

⁵ Barnes, op. cit., pp. 78–83; see also Ch. 13 of Barnes', Party Democracy: Politics in an Italian Socialist Federation (New Haven: Yale University Press, 1967)Google Scholar.

⁶ Easton and Dennis, op. cit., p. 25.

⁷ Campbell, Angus, Gurin, Gerald, and Miller, Warren E., The Voter Decides (Evanston: Row, Peterson, 1954), p. 187 Google Scholar.

⁸ Their measures of political competence—also referred to as citizen competence and subjective civic competence—and its components of local and national competence, as well as local subjective competence and national subjective competence, all include the item pertaining to whether the individual thinks he can personally influence governmental decisions. See Chs. 7–9 in the Civic Culture.

⁹ Initially, seventeen items were included in the analysis. Variable #127 from the ICPR Five-Nation study issue, measuring respondents' agreement-disagreement with the sense of efficacy proposition, “the way people vote is the main thing that decides how things are run in this country,” did not load as hypothesized and was dropped from the analysis reported in this paper. Apparently, the belief that voting decides policy is not a component of the political efficacy dimension which is equivalent cross-nationally. The sixteen variable solutions shown in Table 1 were derived by specifying rotation to three factors. The eigenvalue for the third factor in these solutions is quite close to 1.0.

¹⁰ The items hypothesized as defining a Political Involvement dimension, as they appear in the ICPR Five-Nation study codebook, are: #23, Follow Government Affairs in Newspapers: 4-point scale, “Nearly Everyday”—low, “Never”—high; #34, Follow Government Affairs on Television or Radio: 4-point scale, “Nearly Everyday”—low, “Never”—high; #25, Follow Government Affairs in Magazines: 3-point scale, “Once a Week or More”—low, “Never”—high; #81, Pay Attention to Campaign: 3-point scale, “Much”—low, “None”—high; #99, Name Party Leaders: 7-point scale, Seven or Six Correct—low, “None Correct”—high; #135, Name Cabinet Positions: 6-point scale, “Five Correct”—low, “None Correct”—high; #26 (Do You) Talk About Government Affairs With People: 4-point scale, “Nearly Everyday”—low, “Never”—high, #31, Understand Local Issues: 5-point scale, “Very well”—low, “Not at All”—high; #30, Understand Issues Facing the Country: 5-point scale, “Very Well”—low, “Not at all”—high.

¹¹ The hierarchy of political involvement is presented in Figure 3 at page 18 of Milbrath, Lester W., Political Participation (Chicago: Rand McNally and Company, 1965)Google Scholar.

¹² The items hypothesized as defining an Ability to Influence Government dimension, as they appear in the ICPR Five-Nation study codebook, are: #39, Could Change Bad Regulation: 5-point scale; #40, Would Act on Regulation: 5-point scale; #43, Could You Change Law: 5-point scale; #44, Would You Act on Law: 5-point scale. Variables 39 and 43 were recoded as follows: “Very Likely” = 1; “Moderately Likely” =2; “Likely Only If Others Joined In” =3; “Some-what Unlikely” =4; “Not at All Likely—Impossible” =5. Variables 40 and 44 were recoded as follows: “Very Likely” = 1; “Moderately Likely” =2; “Depends on the Issue” =3; “Some-what Unlikely” =4; “Not at All Likely—Impossible” = 5.

¹³ A situation was defined as follows: “Suppose a regulation were being considered by (specify most local government unit—town, village, etc.) which you considered very unjust or harmful, what do you think you could do.” After respondents had reported whether they could do something they were then asked: (1) “If you made an effort to change this regulation how likely is it that you would succeed”; (2) “If such a case arose, how likely is it that you would actually do something about it”; (3) “Have you ever done anything to try to influence a local decision.” The initial item in the series (both local and national), whether respondents believed they could do something, was not included in the factor analyses because it formed a 2-point scale; the last item, have you ever done anything, was not included because it is a measure of a more intensive level of participation than Political Involvement, and this level of participation (what Milbrath terms “transitional activities,”) is not the subject of this analysis.

¹⁴ The items, as they appear in the ICPR Five-Nation study codebook, are #29, Average Man Understand Government: 3-point scale; #130 Candidates Act After Election: 3-point scale; #132, People Have No Say in Government: 3-point scale. In order to avoid performing factor analysis on unstable phi coefficients, DK responses to all these items were coded as the midpoint on these scales. These variables were recoded so that all scales would run in the same direction; in these instances “Disagree” is low, “Agree” is high.

¹⁵ See Harmon, Harry H., Modern Factor Analysis (Chicago: University of Chicago Press, 1967, 2nd Ed.), pp. 268–272 Google Scholar. The coefficients are shown in Appendix A.

¹⁶ See the review of the Civic Culture by Rokkan, Stein in this REVIEW (September, 1964)Google Scholar.

¹⁷ The Mexican sample only included towns with populations of 10,000 or more.

¹⁸ The methodology of scale construction is discussed in Appendix B. The Political Efficacy scales are less equivalent across the samples than the other scales because the contribution of the Average Man Understand Government component is negligible in West Germany, Italy, and Mexico. Thus, the equivalence of the Political Efficacy scales for all the countries is mainly in terms of beliefs about the responsiveness of government to the membership in general.

¹⁹ The multiple t test method has the least to recommend it because there is no way to determine how many of the comparisons which achieve statistical significance are due to chance alone, or to the circumstance that, since the t tests cannot be regarded as independent, some results dictate, others. For a useful discussion see Hays, William L., Statistics (New York: Holt, Rinehart and Winston, 1963), pp. 375–376 and 471–472 Google Scholar.

²⁰ The Scheffé method is relatively insensitive to departures from normality and homoscedasticity, and is applicable to groups of unequal sizes. Its drawback is that it is probably too conservative with regard to Type I error; as compensation, Scheffé recommends that a significance level of .10 be selected. A brief but lucid presentation is in Ferguson, George A., Statistical Methods in Psychology and Education (New York: McGraw-Hill Book Company, 1966, 2nd Ed.), pp. 295–297 Google Scholar.

²¹ For all a posteriori comparisons in Tables 2–4, the obtained F must be equal to or greater than 4.60 if the null hypothesis is to be rejected according to the Scheffé procedure (α = .10). In reporting the F values for the comparisons, those which are statistically significant have been italicized. The break in Political Efficacy by region among Italians is obvious. The relevant comparisons on Political Efficacy by size of place among Germans are: <20000, 20000–100000—F = 2.89; 20000–100000, > 100000—F = 6.25.

²² <20000, 20000–100000—F = 15.21; 20000–100000, > 100000—F = 1.69.

²³ East, West—F = 0.38; South, West—F = 12.96.

²⁴ England, Scotland—F = 0.74; Scotland, Wales—F=12.25.

²⁵ North, Central—F = 15.21; Central, South—F = 0.18.

²⁶ East, West— F = 5.29; Sout, West— F = 1.96.

²⁷ Where groups are of unequal sizes and a single test is desired of the hypothesis that σ₁ ²=σ₂ ²= … =σ _k ²=σ², the appropriate method is the Bartlett test for homogeneity of variance, described at pages 193–194 of Walker, Helen M. and Lev, Joseph, Statistical Inference (New York: Holt, Rinehart and Winston, 1953)CrossRefGoogle Scholar. The test yields a statistic B which has a chi-square distribution with k — 1 degrees of freedom. The B value for Ability to Influence Government by region is 27.23. This is substantially greater than the chisquare value required for rejection of the null hypothesis at the .001 level, which is 13.8. Therefore, considerable lack of homogeneity is indicated.

²⁸ North, South—F = 9.00; Central, South—F = 8.41.

²⁹ See Norman Kogan's review of Tarrow, Sidney G., Peasant Communism in Southern Italy (New Haven: Yale University Press, 1967) in this REVIEW (December, 1968)Google Scholar. Kogan's criticism is confined to Tarrow's presentation of the dualistic culture thesis, but LaPalombara's view is apposite with that of Tarrow; see LaPalombara, Joseph, “Italy: Fragmentation, Isolation, and Alienation,” in Pye, Lucian W. and Verba, Sidney (eds.), Political Culture and Political Development (Princeton: Princeton University Press, 1965)Google Scholar.

³⁰ See Neubauer, Deane E., “Some Conditions of Democracy,” this REVIEW (December, 1967)Google Scholar.

³¹ Alford, Robert R., Party and Society (Chicago Rand McNally & Company, 1963), pp. 143 and 142 note 26, respectivelyGoogle Scholar.

³² Easton and Dennis, op. cit., esp. pp. 29–33. As the authors note at page 31: “The most important conclusion emerging from our principal component analysis is that by grade 3 children have already begun to form an attitude, as revealed in the five high-loading items, which we could call a sense of political efficacy. This basic orientation is likely to become crystallized early in the life of the individual and to be maintained at least through these grades. This does not say, of course, that the third-grade child has developed a high sense of political efficacy, nor does it assert that any particular proportion of them experience this sentiment at any level of intensity. It only says that an attitude structure has begun to take shape among the children in their early years.” Data on the direction (positive-negative)of efficacy beliefs among children in the United States, the United Kingdom, Italy, and West Germany is reported in Dennis, Jack, Lindberg, Leon, McCrone, Donald, and Stiefbold, Rodney, “Political Socialization to Democratic Orientations in four Western Systems,” Comparative Political Studies (April, 1968)Google Scholar These authors are concerned with whether there is an upward or downward trend in the aggregate level of positive efficacy feelings among children from youngest to oldest in the four countries. Their study does not focus on the age at which children acquire a coherent structure of beliefs about the efficacy norm.

³³ See Duncan, Otis Dudley, “Path Analysis,” American Journal of Sociology (July, 1966)Google Scholar.

³⁴ Ibid., p. 6.

³⁵ Presumably, they would all load on one principal component.

³⁶ Indirect effects are computed by multiplying the path coefficients along a given chain of causation. See Duncan, op. cit.; also see Blau, Peter M. and Duncan, Otis Dudley, The American Occupational Structure (New York: John Wiley & Song Inc., 1967), Ch. 5Google Scholar.

³⁷ An excellent discussion of hierarchical models, as well as causal modelling in general, is in Forbes, Hugh Donald and Tufte, Edward R., “A Note of Caution in Causal Modelling,” this Review (December, 1968)Google Scholar.

³⁸ See the discussion in ibid., at p. 1261, note 11 and the literature cited therein.

³⁹ Tarrow, op. cit., p. 74.

⁴⁰ The comparability of beta coefficients across different samples, because the beta coefficients share the same dimensionality as correlation coefficients, can be distorted by unequal variances. Some of the estimates of the population variance for the competence scales do show statistically significant differences among the samples, but the magnitude of these differences is not large. Examination of the partial regression coefficients, which are not affected by unequal variances, suggests that the patterns of impact inferred from the beta coefficients (standardized partial regression coefficients) are not substantively affected by any departures from homogeneity of variance among the samples.

⁴¹ See the discussion in Almond and Verba, op. cit., pp. 493–494.

⁴² See Tables 18–1A and 18–1B at pages 623–25 of Feierabend, Ivo K., Feierabend, Rosalind L., and Nesvold, Betty A., “Social Change and Political Violence: Cross-National Patterns,” in Graham, Hugh Davis and Gurr, Ted Robert (eds.), Violence in America (New York: Signet Books, 1969)Google Scholar.

⁴³ This scale appears as Table 7 at page 181 of Nesvold, Betty A., “Scalogram Analysis of Political Violence,” Comparative Political Studies (July, 1969)Google Scholar.

⁴⁴ A study utilizing the same data base but with different theoretical concerns is the research reported in two parts by Nie, Norman H., Powell, G. Bingham Jr., and Prewitt, Kenneth, “Social Structure and Political Participation: Developmental Relationships, I,” this Review (June, 1969)Google Scholar and “Social Structure and Political Participation: Developmental Relationships, II,” ibid., (September, 1969). The Ability to Influence Government dimension identified here is similar to the construct which they label Political Efficacy; the exception is that the first two items of their scale were found to load on the factor labelled Political Involvement in this analysis. The Political Involvement dimension also includes items which Nie et al. used to define measures of Political Information and Political Attentiveness, as well as one item of their Political Participation construct. The fact that their Political Information and Political Attentiveness constructs are unidimensional within each sample is not surprising, and this finding is certainly useful from the point of view of parsimony in causal analysis.

⁴⁵ In “Social Structure and Political Participation: Developmental Relationships, II,” Nie et al. report at page 819 that size of place shows a slight but consistently negative relationship with their attitude items: i.e., the greater the community size the lower the citizen's involvement and sense of personal political potency (see Appendix III). However, most of the path coefficients are less than .100 and are probably not statistically significant.

⁴⁶ See the two Nie et al. articles for investigation of the social structure antecedants of political competence.

⁴⁷ Harmon, op. cit., p. 271.

⁴⁸ See ibid., pp. 350–60.

⁴⁹ Ibid., p. 350.

⁵⁰ Fruchter, Benjamin, Introduction to Factor Analysis (Princeton: D. Van Nostrand Company, 1954), p. 195 Google Scholar.

⁵¹ On this point see Harmon's, discussion at page 290 of Modern Factor Analysis Google Scholar.