¹ ‘The Politics of Latin America,” in Almond, Gabriel A. and Coleman, James S., eds., The Politics of the Developing Areas (Princeton 1960), 502Google Scholar.

² Johnson, John J., The Military and Society in Latin America (Stanford 1964Google Scholar); Lieuwen, Edwin, Arms and Politics in Latin America (New York 1961Google Scholar) and Generals vs. Presidents (New York 1964Google Scholar).

³ Certain of these variables are, to be sure, not entirely constant throughout the area, but they are so nearly so as to warrant ignoring their effects.

⁴ The following works were consulted in preparing this inventory of theoretical propositions: Alexander, Robert J., “The Army in Politics,” in Davis, H. E., ed., Government and Politics in Latin America (New York 1958Google Scholar); Andrzejewski, Stanislaw, Military Organization and Society (London 1954Google Scholar); Finer, Samuel E., The Man on Horseback (New York 1962Google Scholar); Gutteridge, William F., Military Institutions and Power in the New States (New York 1965Google Scholar); Huntington, Samuel P., The Soldier and the State (Cambridge, Mass., 1957Google Scholar); janowitz, Morris, The Military in the Political Development of New Nations (Chicago 1964Google Scholar); Johnson, John J., ed., The Role of the Military in Underdeveloped Countries (Princeton 1962CrossRefGoogle Scholar); and the works cited in footnote 2.

⁵ Pp. 87-88. The term “social mobilization” (which Finer himself does not use) was introduced in this sense by Deutsch, Karl W. in “Social Mobilization and Political Development,” American Political Science Review, LV (September 1961), 493–514CrossRefGoogle Scholar.

⁶ Huntington's counterhypothesis linking social mobilization with increased military intervention is discussed later in this section.

⁷ Finer, 113–15; Alexander, 158.

⁸ Janowitz's attack on this proposition (pp. 18–20) is weakened by his failure to distinguish the military-civilian dimension from the democratic-authoritarian dimension and by his failure to recognize that a correlation can be important without being perfect.

⁹ Germani, Gino and Silvert, Kalman, “Politics, Social Structure and Military Intervention in Latin America,” Archives Européennes de Sociologie, 11 (Spring 1961), 62–81CrossRefGoogle Scholar.

¹⁰ P. 21.

¹¹ Finer, 87; Johnson, John J., “The Latin-American Military as a Politically Competing Group in Transitional Society,” in Johnson, , ed., Role of the Military, 127Google Scholar.

¹² Finer, 21, 87–88, 115; Alexander, 157.

¹³ Huntington, Samuel P., “Political Development and Political Decay,” World Politics, XVII (April 1965), 386–430CrossRefGoogle Scholar.

¹⁴ Lieuwen, , “Militarism and Politics in Latin America,” in Johnson, , ed., Role of the Military, 132–33; Martin C. Needler, Latin American Politics in Perspective (New York 1963), 76Google Scholar.

¹⁵ Huntington, The Soldier and the State, 84; Finer, 24ff.

¹⁶ For example, Finer, 32.

¹⁷ For example, Alexander, 153.

¹⁸ Ibid., 154–55.

¹⁹ Janowitz, 42.

²⁰ Edelmann, Alexander T., Latin American Government and Politics (Homewood 1965), 189Google Scholar.

²¹ “The Latin American Military,” in Johnson, Role of the Military, 129.

²² Lieuwen, “Militarism and Politics,” ibid., 134.

²³ See Blalock, Hubert M. Jr., Social Statistics (New York 1960), 273ffGoogle Scholar.

²⁴ Obviously, this study can consider only the range of variation in the independent variables which occurs in Latin America. For example, levels of social mobilization above or below the level achieved in Latin America might have effects on militarism which could not be detected in this study.

²⁵ Gilmore, Robert L., Caudillism and Militarism in Venezuela (Athens, Ohio, 1964), 4–5Google Scholar.

²⁶ In this connection it may be helpful to note Gilmore's distinction between “militarism” (as defined above) and “caudillism”: “Caudillism is a political process in which violence is an essential element. . . . [It] may be defined as the union of personalism and violence for the conquest of power” (pp. 5, 47). Caudillist regimes, such as Haiti's, are rated two.

²⁷ The fact that the differences in MI scores among Needler's third, fourth, and fifth categories are much less than the differences involving his first and second categories comports with one's intuitive notion of the “distance” between the levels of intervention indicated by the descriptions of his categories. Thus, for example, the difference between a military that habitually “intervenes” and one that has “veto power” is less than the difference between one that “intervenes” and one whose political role is “limited.” Needler's ratings of individual countries are given in Appendix I.

²⁸ Banks, Arthur S. and Textor, Robert B., A Cross-Polity Survey (Cambridge, Mass., 1963Google Scholar).

²⁸ See Ferguson, George A., Statistical Analysis in Psychology and Education (New York 1959), 289Google Scholar. Given the probable error margins for the data used here, reliability coefficients in the range .8-.9 would be expected. This would mean, for example, that an obtained coefficient of .20 understates the actual correlation by about .02-.05 and that a coefficient of .60 understates the actual correlation by about .06-. 15. For error estimates, see Russett, Bruce M. and others, World Handbook of Political and Social Indicators (New Haven 1964Google Scholar). For a detailed analysis of GNP error and a calculation of an approximate reliability coefficient, see Alker, Hayward R. Jr., “The Comparison of Aggregate Political and Social Data . . . ,” Social Sciences Information, v (September 1966), 1–18Google Scholar.

³⁰ For a few of the variables, data were not available for all twenty Latin American countries. In some cases, I have estimated the missing data and calculated coefficients including this “best guess” data. Unless these “guesses” are wildly off (and I do not believe that they are), the error introduced by including them is probably less than the error that would be introduced by ignoring the countries they represent. Precise data on Uruguay, for example, are often missing; yet it would be quite misleading to ignore the fact that this country fits many of our hypotheses remarkably well. Coefficients based on “best guess” data are indicated as such, and in all cases coefficients have also been presented without this “best guess” data.

³¹ There is considerable debate about whether tests of statistical significance are appropriate in cases, like the present study, in which we have not a random sample from a larger universe but a complete universe, viz., all contemporary Latin American countries. Strictly speaking, significance testing is merely a way of checking inferences from a random sample to the universe from which that sample is drawn. On the other hand, Blalock and Gold have argued that significance tests may help us sift important from unimportant findings, even when there is no question of inferring to a larger universe. See Blalock, 270, and David Gold, “Some Problems in Generalizing Aggregate Associations,” American Behavioral Scientist, viii (December 1964), 16–18. Gold, however, adds the qualification that when one is dealing with small APs (as we are here), “judgments of importance that can be made reasonably from the size of associations should take precedence over tests of significance.” Keeping in mind the problems associated with significance-testing in this situation, one may find the following figures helpful: assuming a one-tailed test and an N of 20, an r ≥ .38 is significant at the .05 level; an r ≥ .31 is significant at the to level; and an r ≥ .23 is significant at the .33 level. These significance levels are derived from the World Handbook, 262.

³² See Blalock, 295–99.

³³ In calculating this mean correlation and in compiling the E D index that follows, I have reversed the scoring for Variables 7–9, so that a large positive number always refers to a high level of development.

³⁴ The discussion that follows is not intended to be a complete presentation of the logic and methodology of causal path analysis. For introductions to this recently developed technique, see Hubert M. Blalock, Jr., Causal Inferences in Nonexperimental Research (Chapel Hill 1964); Boudon, Raymond, “A Method of Linear Causal Analysis . . . ,” American Sociological Review, xxx (June 1965), 365CrossRefGoogle Scholar–74; and Duncan, Otis Dudley, “Path Analysis: Sociological Examples,” American Journal of Sociology, LXXII (July 1966), 1–16CrossRefGoogle Scholar. For a readable and comprehensive introduction for political scientists, see Alker, Hayward R. Jr., “Causal Inference in Political Analysis,” in Bernd, Joseph, ed., Mathematical Applications in Political Science, 2nd Series (Dallas 1966Google Scholar).

³⁵ The equations for calculating the path coefficients (or p's) in this case are quite simple:^r21=p21

This is a simple algebraic system of three equations and three unknowns. For the general equation of path analysis, see Duncan, 5.

³⁶ The equation for calculating the residual variance is quite simple:

See Duncan, 6.

³⁷ R. Vekemans and J. L. Segundo present numerically identical data under the heading “Percent of population at intermediate and senior grades of employment,” in “Essay on a Socio-economic Typology of the Latin American Countries,” in de Vries, E. and Echavarria, J. M., eds., Social Aspects of Economic Development in Latin America, Vol. I (Paris 1963Google Scholar).

³⁸ Pp. 192–93.

³⁹ It is true that the accuracy of electoral turnout as an indicator of participation is limited by variations in the social and institutional context of the act of voting. Voting does not have the same meaning in the U.S. and the USSR, in the Netherlands and Uganda. Restriction of our attention to the Latin American countries minimizes this problem. For a fuller discussion of the problem, see Germani and Silvert.

⁴⁰ Edelmann argues that one particular sort of interest group is especially likely to inhibit military intervention: “The most serious threat of all to the power of the military is that posed by the labor unions . . .” (p. 194). Actually, the correlation between military intervention and the proportion of union members is —.07. Huntington argues that political decay varies directly with the extent of party fragmentation. The comparable Cross-Polity Survey variable is called “Party System: Quantitative”; this correlates .23 with the MI index; controlling for social mobilization reduces this to .16.

⁴¹ I have borrowed this weighting technique from Eldon Kenworthy, “Predicting Instability in Latin America” (unpublished).

⁴² This negative finding is independent of the particular period during which the various missions were in residence. The mean MI score for only those countries that had German missions in the nineteenth century—generally the period referred to in this connection—is lower still: 13.6.

⁴³ See Coleman, James S., Introduction to Mathematical Sociology (Glencoe 1964), 288–311Google Scholar.

⁴⁴ Attempted but unsuccessful coups were ignored, partly for the conceptual reason that the definition of an “attempted coup” is problematic, partly for the practical reason that adequate information on attempted coups is lacking. The period of a half-year is chosen as representing about the optimum length of time during which contagion might be expected to operate. An analysis using one-year intervals produced results exactly comparable to those reported. I tallied countries rather than coups so as to exclude the effects of “contagion” within a single country.

⁴⁵ The lower level of intervention in the earlier period probably reflects (1) the fact that after the turn of the century civilian government enjoyed a period of considerable success in Latin America and (2) my decision (see footnote 26) to distinguish “caudillism” from “military intervention.”

⁴⁶ This elimination process follows the technique suggested by Simon, Herbert and Blalock, Hubert M. Jr. See Blalock, Causal Inferences, 61–94Google Scholar and passim.

⁴⁷ Perhaps an analogous explanation would apply to the relationship between economic development and military intervention. W e lack the data necessary to carry this analysis further.

⁴⁸ Huntington argues that one aspect of political development is an increasing autonomy of politics from other social spheres. See “Political Development and Political Decay,” 401–30.

⁴⁹ Huntington's theory of political development (ibid.) implies that social mobilization leads to military rule and that civilian rule depends on strong political institutions; neither of these propositions is confirmed by the present study. It would be worth further investigation to determine whether the propositions d o apply to other underdeveloped areas.

⁵⁰ Among the other independent variables considered and rejected in the present study were (1) total population (r=.02), (2) racial composition (r=.16), and (3) rates of social change (for urbanization, r = —.18).