Article contents
A Causal Model of Civil Strife: A Comparative Analysis Using New Indices1
Published online by Cambridge University Press: 01 August 2014
Extract
This article describes some results of a successful attempt to assess and refine a causal model of the general conditions of several forms of civil strife, using cross-sectional analyses of data collected for 114 polities. The theoretical argument, which is discussed in detail elsewhere, stipulates a set of variables said to determine the likelihood and magnitude of civil strife. Considerable effort was given here to devising indices that represent the theoretical variables more closely than the readily-available aggregate indices often used in quantitative cross-national research. One consequence is an unusually high degree of statistical explanation: measures of five independent variables jointly account for two-thirds of the variance among nations in magnitude of civil strife (R = .80, R2 = .64).
It should be noted at the outset that this study does not attempt to isolate the set of conditions that leads specifically to “revolution,” nor to assess the social or political impact of any given act of strife except as that impact is reflected in measures of “magnitude” of strife. The relevance of this kind of research to the classic concern of political scholarship with revolution is its attempt at identification and systematic analysis of conditions that dispose men to strife generally, revolution included.
- Type
- Research Article
- Information
- Copyright
- Copyright © American Political Science Association 1968
Footnotes
This is a revised version of a paper read at the 1967 Annual Meeting of the American Political Science Association, Chicago, September 5–9 The research was supported in part by the Center for Research in Social Systems (formerly SORO), The American University, and by the Advanced Research Projects Agency of the Department of Defense. This support implies neither sponsor approval of this article and its conclusions nor the author's approval of policies of the U.S. government toward civil strife. The assistance of Charles Ruttenberg throughout the process of research design, data collection, and analysis is gratefully acknowledged. Substantial portions of the data were collected by Joel Prager and Lois Wasser-spring. The author owes special thanks to Harry Eckstein for his advice and encouragement. Bruce M. Russett and Raymond Tanter provided useful criticisms of the paper in draft form. Research was carried out at the Center of International Studies, Princeton University.
References
2 Gurr, Ted, “Psychological Factors in Civil Violence,” World Politics, 20 (January 1968), 245–278CrossRefGoogle Scholar.
3 Coercive potential is labelled “retribution” in ibid. The theoretical model also stipulates a set of variables that determines the intensity of deprivation. In the research reported in the present article, deprivation was operationalized directly rather than by reference to its component variables. The causal mechanism of the theory is the frustration-aggression relationship, which the author has attempted to modify and apply to collective strife in the light of recent empirical and theoretical work, e.g., Berkowitz, Leonard, Aggression: A Social Psychological Analysis (New York: McGraw-Hill, 1962)Google Scholar, and Yates, Aubrey J., Frustration and Conflict (New York: Wiley, 1962)Google Scholar.
4 Gurr, Ted with Ruttenberg, Charles, The Conditions of Civil Violence: First Tests of a Causal Model (Princeton: Center of International Studies, Princeton University, Research Monograph No. 28, 04 1967)Google Scholar.
5 See Bwy, Douglas, “Governmental Instability in Latin America: The Preliminary Test of a Causal Model of the Impulse to ‘Extra-Legal’ Change,” paper read at the American Psychological Association Annual Convention, New York, September 2–6, 1966Google Scholar; Walton, Jennifer, “Correlates of Coerciveness and Permissiveness of National Political Systems: A Cross-National Study,” (M.A. thesis, San Diego State College, 1965)Google Scholar; Gurr, and Ruttenberg, , The Conditions of Civil Violence …, 81–84Google Scholar.
6 See, for example, Johnson, Chalmers, Revolution and the Social System (Stanford: The Hoover Institution on War, Revolution and Peace, 1964), pp. 14–22Google Scholar.
7 Huntington, Samuel P., “Political Development and Political Decay,” World Politics, 17 (April 1965), 386–430CrossRefGoogle Scholar; Kornhauser, William, The Politics of Mass Society (New York: The Free Press, 1959)Google Scholar; and Ross, Arthur M. and Hartmann, George W., Changing Patterns of Industrial Conflict (New York: Wiley, 1960)Google Scholar, among others.
8 Gurr, “Psychological Factors…”
9 Ibid.
10 Gurr, and Ruttenberg, , The Conditions of Civi Violence …, 100–106Google Scholar.
11 Merelman, Richard M., “Learning and Legitimacy,” this Review, 60 (September 1966)Google Scholar; see also the work of Pastore and of Kregarman and Worehel, reviewed in Berkowitz, op. cit., passim.
12 Five polities meeting these criteria were excluded: Laos on grounds that at no time in the 1960's did it have even the forms of a unified regime, and Albania, Mongolia, North Korea, and North Vietnam for lack of sufficient reliable data. The universe nonetheless includes polities with more than 98 percent of the world's population.
13 Blalock, Hubert M. Jr., Causal Inferences in Nonexperimental Research (Chapel Hill: University of North Carolina Press, 1964), pp. 166–167Google Scholar, italicized in original.
14 In each of a number of analyses by Rummel and others a set of “domestic conflict” measures was factor analyzed. Turmoil, indexed by riots and demonstrations, is found to be a distinct dimension in all the analyses; two other factors, labelled by Rummel “revolution” and “subversion,” are in some cases separate and in others combined. Principal components of the “revolution” dimension are coups, palace revolutions, plots, and purges; the category is labelled here conspiracy. Guerrilla war and terrorism are major components of the “subversion” dimension, here labelled internal war. See Rummel, Rudolph J., “A Field Theory of Social Action With Application to Conflict Within Nations,” Yearbook of the Society for General Systems Research, X (1965), 189–195Google Scholar; and Tanter, Raymond, “Dimensions of Conflict Behavior Within Nations, 1955–1960: Turmoil and Internal War,” Peace Research Society Papers, III (1965), 159–183Google Scholar. The subcategories used here are adapted, with their operational definitions, from Rummel, , “Dimensions of Conflict Behavior Within and Between Nations,” Yearbook of the Society for General Systems Research, VIII (1963), 25–26Google Scholar.
15 See, for example, Rummel, op. cit.; Tanter, op. cit.; Russett, Bruce M., “Inequality and Instability: The Relation of Land Tenure to Politics,” World Politics, 16 (April 1964), 442–454CrossRefGoogle Scholar; Tilly, Charles and Rule, James, Measuring Political Upheaval (Princeton: Center of International Studies, Princeton University, 1965)Google Scholar; and Ivo, K. and Feierabend, Rosalind L., “Aggressive Behaviors Within Polities, 1948–1962: A Cross-National Study,” Journal of Conflict Resolution, 10 (September 1966), 249–271Google Scholar.
16 Information coded, in addition to that required for the three measures specified, included the socio-economic class (es) of the initiators, the social context in which they acted, the category of events, the targets and apparent motives of action, the number and role of coercive forces, and the extent and types of external support for initiators and regime, if any. Although no formal reliability tests were undertaken, the four coders did extensive practice coding on the same set of materials prior to coding and reviewed points of disagreement, and the author reviewed all coding sheets for internal consistency and, where necessary, recoding or search for additional information. It should be noted that the 1100 “events” include many cumulated reports, e.g., civil rights demonstrations in the U.S. were treated as a single set of events, all European-OAS terrorism in Algeria as a single event, etc.
17 It has been suggested that strife in countries with press restrictions is under-reported. As a check on this type of systematic error a nine-point measure of press freedom was incorporated in initial analyses; the measure is from Nixon, Raymond B., “Freedom in the World's Press: A Fresh Appraisal With New Data,” Journalism Quarterly (Winter 1965), 3–14CrossRefGoogle Scholar. The correlations of this measure, in which high scores reflect low press freedom, with some measures of strife are: Duration, + 19; Intensity, +17; Pervasiveness, −16; Total magnitude of strife, + 11. The first two are significant at the .05 level, the third at .10. In effect, more strife tends to be reported from polities with low press freedom, not less, as might be expected. The results almost certainly reflect the association of high levels of economic development and press freedom in the Western nations, which tend to have less strife than the developing nations.
18 The missing-data procedures gave implausibly-high estimates for initiators and casualties for a number of events. In subsequent and comparable analysis it seems advisable to rely on estimates of deaths alone, rather than casualties, and to insert means derived from comparable events in comparable countries rather than such events in all countries.
19 Tables are available on request from the author listing the 114 countries, their strife scores, the summary measures of deprivation and mediating conditions discussed below, and the data sources.
20 The 48 deprivation measures, with only one statistically significant exception, were positively associated with strife, most of them at a relatively low level. The thirteen were selected with regard to their representativeness, relatively high correlations with the dependent variables, and low intercorrelations.
21 Coding judgments for both discrimination indices and for separatism were made on the basis of country studies. The proportionality measures are versions of indices reported in Gurr, Ted, New Error-Compensated Measures for Comparing Nations (Princeton: Center of International Studies, Princeton University, 1966), 67–90Google Scholar.
22 A crude measure of the proportion of each polity's population engaged in the monetary economy, to the nearest .10, was constructed for the purpose of weighting this and some other measures. The measure was based primarily on labor census data.
23 The two measures will be used in subsequent analyses to examine time-lag relationships between short-term economic deprivation and strife. The trade data, obtained primarily from United Nations sources, was converted to U.S. currency when necessary to maintain comparability over time.
24 The Hispanic-American Report is much more comprehensive a source, hence the mean deprivation scores for Latin America were much higher than those for other polities. As a crude adjustment, the Latin American polity scores were divided by a constant so that their mean approximated that of other polities. The same procedure was followed for indices 6 and 7, below. Analyses of regional clusters of polities, not reported here, provide a check on the adequacy of the procedure.
25 Types of restrictive actions, and their scale values, are as follows:
1 Amalgamation of splinter party with larger party
1 Restriction or harrassment of splinter party
2 Banning of splinter party
2 Amalgamation of minority party with larger party
2 Restriction or harrassment of minority party
3 Banning of minority party
3 Amalgamation of a major party with another major party
3 Restriction or harrassment of major party
4 Banning of major party
4 Improper dismissal of regional representative body
4 Improper dismissal of elected regional executive
5 Ban on party activities, parties allowed to continue their organizational existence
5 Improper dismissal of national legislature, with provision for calling new one within a year
5 Improper dismissal of elected chief executive, with provision for replacement within a year
6 Dissolution of all parties, ban on all political activity
6 Improper dismissal of national legislature, no short-term provision for reestablishment
6 Improper dismissal of elected chief executive, no short term provision for reelection
26 The annual scores for (5), (6), and (7) are being used in a series of time-lagged and crosspanel correlation analyses, not reported here, in further tests of causal relationships.
27 These are product-moment correlation coefficients, the strife measures including measures of duration, pervasiveness, intensity, and total magnitude of strife for 1961–65. The last two strife measures are defined differently from those employed in the present analysis, but are derived from the same 1100-event data bank.
28 If one or the other ratio was missing, it was assumed equal to the known ratio. Internal security force ratios for 94 polities are reported in Gurr, , New Error-Compensated Measures for Comparing Nations, 111–126Google Scholar.
29 The first two indices are reported in ibid., 33–66, 91–110. Correlations among all three and strife measures are reported in Gurr and Ruttenberg, The Conditions of Civil Violence, passim. The party characteristics are recoded from Banks, Arthur S. and Textor, Robert B., A Cross-Polity Survey (Cambridge: MIT Press, 1963)Google Scholar, raw characteristics 41 and 43.
30 Eckstein, Harry, “Internal War: The Problem of Anticipation,” in Pool, Ithiel de Solaet al., Social Science Research and National Security (Washington, D.C.: Smithsonian Institution, March 5, 1963)Google Scholar.
31 Inaccessibility appears to be an almost-but not-quite necessary condition for protracted internal wars. With one exception all such internal wars in the post-1945 period occurred in polities with high or very high scores on this index; the exception, a notable one, is Cuba.
32 The following rescaling was used, the sum of the “durability” and “character” scores being given on the upper line, the final legitimacy score on the lower:
![](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20151030062257483-0281:S000305540022488X_eqnU1.gif?pub-status=live)
33 The S-shape of this relationship is considerably more pronounced when coercive-force size is related to total magnitude of turmoil; see Gurr, Ted, “Why Urban Disorder? Perspectives from the Comparative Study of Civil Strife,” American Behavioral Scientist, 10 (March-April 1968)Google Scholar.
34 Significant computational errors in internal war and TMCS scores of several countries were identified and corrected after completion of the analyses reported here. Robert van den Helm of Princeton University has analyzed the corrected data, using the combined short-term deprivation measure in lieu of the two separate measures, with these multiple regression results: for TMCS, R2 = .638; conspiracy, R2 = .391; internal war, R2 = .472; and turmoil, R2 = .284. The significant increase in the degree of explanation for internal war is the result of increased correlations between magnitude of internal war and short-term deprivation (from .28 in Table 1 to .34); facilitation (from .57 to .61); and legitimacy (from −.23 to − .26). Their between magnitudes of turmoil and internal war increases from .17 to .23, their between TMCS and internal war from .79 to .86. No other results of the analyses reported here are significantly affected by the reanalysis. The actual TMCS scores shown in Table 3 are corrected ones.
35 These and other fundamental arguments about causal inference are well summarized in Blalock, Causal Inferences …, Chapters 2 and 3. A partial correlation coefficient can be most easily regarded as the correlation between X and Z after the portions of X and Z that are accounted for by Y are removed, or held constant. The results discussed below are based on the use of only one of a variety of related causal inference techniques and are open to further, more refined analysis and interpretation. For other applicable approaches see, for example, Alker, Hayward R. Jr., Mathematics and Politics (New York: Macmillan, 1965)Google Scholar, Chapters 5 and 6.
36 To simplify evaluation of the effects of the control variables, the summary short-term deprivation variable was employed rather than its economic and political components separately.
37 Analysis of the correlation coefficients does not indicate definitively that legitimacy contributes to coercive potential rather than vice versa; nor would it be impossible to argue, on the basis of the partial r's alone, that short-term deprivation is a weak intervening variable between coercive potential and facilitation, on the one hand, and strife on the other. It is the plausibility of the theoretical arguments, in each case, that gives deciding force to the interpretation proposed. For a comparable argument see Hugh Donald Forbes and Edward R. Tufte, “A Note of Caution in Causal Modelling,” elsewhere in this issue of this Review.
38 Tanter has examined time-lag effects between a number of measures of foreign economic and military assistance for the regime and magnitude of civil violence in 1961–63 for Latin American nations and finds generally weak relationships. The only consequential positive relationship, an indirect one, is between levels of U.S. military assistance and subsequent strife. Tanter, Raymond, “Toward A Theory of Conflict Behavior in Latin America.” (Paper read to the International Political Science Association, Brussels, 09, 1967Google Scholar).
39 In a reanalysis using corrected data (see footnote 34), four variables—the combined short-term deprivation measure, persisting deprivation, legitimacy, and facilitation—given an R2 of .629.
40 The partial r's for these five variables are: economic deprivation, .27; political deprivation, .13; persisting deprivation, .39; legitimacy, .36; facilitation, .61.
41 For example Boulding, Kenneth E., Conflict and Defense: A General Theory (New York: Harper and Row, 1962)Google Scholar; Coser, Lewis, The Functions of Social Conflict (New York: The Free Press, 1956)Google Scholar; and Schelling, Thomas C., The Strategy of Conflict (Cambridge, Mass.: Harvard University Press, 1960)Google Scholar.
42 See Gurr, “Why Urban Disorder?” for a causal inference analysis of the sources of turmoil. The turmoil model differs principally in that “past strife levels” has the primary mediating role that facilitation has in the TMCS model.
43 The test is less than precise because the measures are not comparable; the past strife measure is based on an arbitrary weighting of counts of number of events, whereas the magnitude of strife measures reflect levels of participation, duration, and intensity.
- 333
- Cited by
Comments
No Comments have been published for this article.