A theory of

RAYMOND TANTER Department of Political Science, University of Michigan and

MANUS MIDLARSKY Department of Political Science, University of Colorado

Overview Finally, hypotheses are offered to account for some of the findings of this study. The focus of this study is an empirical examination of some causes of revolution.l A While prior studies have delineated dimen- Categorization of sions of conflict behavior within nations Prior to a discussion of theoretical signifi- (Rummel, 1963, 1966, 1967; Tanter, 1965, cance, it may be useful to inquire into the 1966) in order to discover the variables meaning of revolution itself. This exercise most representative of internal conflict, a in definition appears necessary primarily be- purpose of the present inquiry is to discover cause prior usage of the term has been their theoretical significance. A typology of somewhat ambiguous. To the Hegelian, the revolution is presented and two possible idea is equated with irresist- causes of revolution-changes in economic ible change-a manifestation of the world development and level of education-are spirit in an unceasing quest for its own ful- examined regarding the extent of their asso- fillment. Similarly, the Marxist, although ciation with certain characteristics of revo- opposed to Hegelian idealism, sees revolu- lution. Regional differences are found which tion as a product of irresistible historical imply a fundamental distinction between at forces, which culminate in a struggle be- least two of the categories in the typology. tween the and the proletariat. Hannah Arendt (1965, pp. 34-40), on the 1 This study is a part of a series on revolu- other hand, interprets the revolutionary ex- tions and their implications for international re- perience as a kind of restoration, whereby lations and some especially consequences for the insurgents attempt to restore liberties international business communities. Acknowl- and privileges which were lost as the result edgments are due to Harold Guetzkow of the International Relations Program at Northwestern of the government’s temporary lapse into University for his suggestions during formative . Indeed, aspects of the American stages of this inquiry. Thanks are due to the Revolution as well as some recent anti- Ford Foundation for a to the fellowship given colonial revolutions may be amenable to author and to the National Science Foun- junior Arendt’s The dation for its assistance to the senior author. interpretation. insurgents view the colonial elite as Helpful comments were made by Richard A. might strangers Brody, James C. Davies, and Robert C. North. who have usurped the freedoms which,

CONFLICT RESOLUTION VOLUME XI NUMBER 3 265

TABLE 1 CHARACTERISTICS OF FOUR TYPES OF REVOLUTION

mythically or otherwise, once belonged to pp. 22-24) constructs a similar classifica- the people now in revolt. tion, but instead of the palace revolution he Tocqueville (1955, p. 8), on the other discusses &dquo;caudillismo&dquo; (predatory milita- hand, has employed a more empirical ap- rism), which is a common form of the coup proach to the problem of revolution, and has d6tat in Latin America. These three forms defined it as an overthrow of the legally con- of revolution appear to reflect an increasing stituted elite, which initiated a period of in- degree of change initiated by the successful tense social, political, and economic change. insurgents, and may be placed on a rank- Crane Brinton (1952, pp. 3-4) has continued order of increasing political or social change. this empirical thrust by differentiating be- James Rosenau, in fact, constructs such a tween the coup d6tat, as a simple replace- classification of revolution. ment of one elite by another, and major Personnel wars are defined by Rosenau as revolutions such as the French or Russian, those which are fought over the occupancy which were accompanied by social, political, of existing roles in the structure of political and economic changes. Similarly, with re- authority (Rosenau, 1964, pp. 63-64). An spect to Latin America, George Blanksten example of this type is the palace revolution (1962, p. 72) suggests that we should dis- or Latin American caudillismo. A second tinguish between the coup d6tat and revo- category is what Rosenau calls the authority lutions such as the Mexican experience, wars, or those in which the insurgents com- which eventually had profound conse- pete not only for the occupancy of roles in quences for the structure of that society. the political structure, but for their arrange- The distinction between two forms of ment as well. Struggles to replace dictator- revolution may provide a basis for the de- ships with democracies would be classified velopment of further classifications. For as authority wars. The final classification is example, Harold Lasswell and Abraham that of the structural wars, in which the goal Kaplan (1950, p. 252) present a further re- of the insurgents is the introduction of social finement in the classification of revolution and economic changes in the society. Wars by the introduction of a three-category ty- involving Communist factions would fall pology in which they differentiate between under this heading. In addition, Rosenau palace revolutions, political revolutions, and notes that structural wars contain elements social revolutions. Edwin Lieuwen (1960, of both personnel and authority wars. 266

Changes in the occupancy of government major difference between the two forms is roles, as well as the arrangement of these in the degree of change initiated in the roles, would automatically take place if the structure of the political authority. The insurgents were successful in a structural &dquo;Young Turks&dquo; implemented a complete re- war. Similarly, the authority war is person- vision of the political authority which led to nel-oriented, because the arrangement of a truncation of the Ottoman Empire and the political roles would seldom, if ever, be establishment of a . The revolt altered without a change in the occupancy against Peron, on the other hand, was an at- of these roles. In Rosenau’s ranking of the tempt at reform, in that Peron’s mismanage- three types of internal war, personnel wars ment of the economy and the dissatisfaction would occupy the lowest rank with regard of major political forces, such as the Roman to the degree of societal change; the au- Catholic Church, led to a revolt against what thority wars would occupy an intermediate had become an oppressive political execu- rank, and the structural wars would receive tive. the highest rank. The existence of several types of revolu- Samuel Huntington (1962, pp. 23-24) tion suggests that we might be able to isolate has suggested a classification of revolution different characteristics of revolution. Karl in which four categories are enumerated: Deutsch (1964, pp. 102-104) proposes that the internal war, the revolutionary coup, the the degree of mass participation in a revo- reform coup, and the palace revolution. lution, as well as its duration, may be essen- Huntington’s use of the concept &dquo;internal tial to an adequate description of the revolu- war&dquo; differs from the meaning attributed to tionary experience. A third characteristic that concept in earlier systematic studies. may be the number killed as a result of the For that reason the term mass revolution2 revolution. Given a high degree of commit- will be substituted for internal war as used ment by the insurgents and the incumbents, by Huntington. The terms mass revolution the number of persons killed both during and palace revolution correspond respec- and after the revolution may be a measure tively to Rosenau’s structural and personnel of intensity. This measure will be discussed wars, while the revolutionary and reform more fully at a later point. Finally, the in- coups both may be placed under the head- tentions of the insurgents may be critical to ing of the authority wars. Kemal Ataturk’s the form of the revolution as well as to its revolution in Turkey, for example, illustrates eventual outcome. If the successful insur- what Huntington might call a revolutionary gents are ideologically committed to certain coup, whereas the 1955 coup in Argentina goals, then they may initiate changes in the might be classified as a reform coup. The societal structure to effect the realization of these goals. If, on the other hand, the in- 2 One may also distinguish between various surgents have no particular ideological orien- types of mass revolution. Lawrence Stone tation, then they might intend to replace the (1966, pp. 162-63), for example, presents a incumbents in the structure of political au- typology which differentiates among the Jac- to in the querie, the Anarchistic , the Jacobin thority without recourse changes and the Militaristic Mass societal structure. Insurrection. However, a purpose of this study Table 1 contains a rank-order of the cate- is to compare the categories of revolution larger gories of revolution based on their position (e.g., mass revolution vs. palace revolution) with to mass duration, rather than to engage in a detailed examination respect participation, of subcategories. domestic violence, and the intentions of the

CONFLICT RESOLUTION VOLUME XI NUMBER 3 267

TABLE 2 FOUR TYPES OF REVOLUTIONa

a Unsuccessful revolutions are indicated below the the dotted line. b The date in parentheses refers to the year in which the insurgents were either defeated or seized the roles in the structure of political authority. c The Algerian revolution of 1962 refers to the anti-colonial revolt of the Algerian independence movement, whereas the of 1958 applies to the fall of the French Fourth Republic. insurgents. We may now suggest a defini- within the political structure. This defini- tion of revolution which would be designed tion, then, sets a lower bound or minimum to include the categories in Table 1, as well criterion for the existence of revolution. as to allow for the possibility of a continuum. Table 2 provides examples of the four A revolution may be said to exist when a types of revolution. This table is not meant group of insurgents illegally and/or force- to be exhaustive, but serves to provide a list- fully challenges the governmental elite for ing of those revolutions judged by the the occupancy of roles in the structure of authors to be illustrative of the particular political authority. A successful revolution categories.3 The dotted line in Column 1 occurs when, as a result of a challenge to separates successful mass revolutions from the are even- governmental elite, insurgents those which were unsuccessful. In addition, tually able to occupy principal roles within instances of both successful and unsuccess- the structure of the This political authority. ful anti-colonial revolutions are included to is not to that once the have say insurgents indicate the possible similarity of this type occupied these roles, the structure of politi- to the general mass revolution. Having pro- cal authority will remain unchanged. As vided such illustrative material, a next step suggested above, changes in the personnel is to provide the operational referents for of the governmental elite often are the pre- the examination of causes of revolution. condition for meaningful changes in the and social structure. If the insur- political 3 Other writers such as Gross Hunt- intend and social (1958), gents major political ington (1962), and Silvert (1961) are in sub- changes, they must first occupy these roles stantial agreement with this classification. 268

Empirical Referents TABLE 3 DEATHS AND DURATION FOR SUCCESSFUL Two of the four measures-duration and REVOLUTIONS, 1955-60 domestic violence-may be interpreted as defining characteristics of revolution, and taken together may be seen as a measure of revolutionary intensity.4 Domestic violence, for example, is present in the majority of revolutions. If the population is indifferent to changes in the personnel of the govern- mental elite, then a palace revolution may be initiated by insurgents. They may en- counter opposition only from the incumbent elite. Because of the lack of large-scale, organized conflict behavior, the number of deaths would be at a minimum, and the duration of such a revolution would be short. If, however, the insurgents proclaim their intention to initiate changes in the social structure, then large segments of the popu- lation may be alienated from the insurgent a The data sources for Column 1 are the raw data for Rummel cause. The number of and (1963), Tanter (1965), and the New York deaths, possibly Times. Two countries which experienced successful the duration of the revolution, may increase revolutions, Sudan ( 1958 ) and Laos (1960), were excluded because of missing data on deaths. proportionally to the degree of societal b Per million population, 1950-62 (Russett et al., the Simi- 1964, pp. 97-100). change envisaged by insurgents. c Data relevant to the duration of revolutions were larly, the duration of a revolution has been drawn from the New York Times. d Both the 1957 and 1958 Thai revolts were initiated cited as a possible measure of revolutionary by the same individual, the commander-in-chief of the army, Sarit Thanarat. In addition, the 1958 coup took place with the express agreement of the &dquo;ousted&dquo; Kitti- kachom For these this latter in- 4 The degree of mass participation as well as government. reasons, stance is omitted from the subsequent analysis although, the intentions of the insurgents are less satisfac- for the sake of completeness, it is included in this listing. tory as empirical referents. Mass participation might range from popular dissatisfaction with- out direct involvement, on the one to hand, intensity (Rosenau, 1964, pp. 76-77). Many overt complicity and active support of the in- Latin American palace revolutions are of surgents, on the other. Thus, measurement of short and the number killed is mass participation presents a great deal of dif- duration, ficulty unless the researcher is present at the relatively small. Indeed, bloodless coups site of the revolution and is able to measure often occur in which no one is killed, but a at the time the revolution takes public opinion new set of elites assumes the major roles in place. Similarly, the intentions of the insurgents, the political authority structure. Revolu- although important as a descriptive variable, tions such as the French and the nevertheless present important problems of mea- Russian, surement. Intentions and ideological directions however, are of longer duration and a are subject to changes during the process of greater number of persons are killed. revolution. Fidel Castro, for example, pro- The two measures-duration and deaths claimed his for the goals of the Cuban support from domestic violence-are presented in middle class during the recent Cuban revolu- Table 3. The first column lists all success- tion, yet once in power he allied himself with the Communists. ful revolutions which occurred between

CONFLICT RESOLUTION VOLUME XI NUMBER 3 269

1955 and 1960. The criterion for the in- built up after a long period of repression. clusion of a revolution was that the revolt A revolution generally occurs after a period ended sometime within that period.5 Col- of instability, and it is suggested in this umn 2 lists the number killed as a result of paper that the form of a revolution is de- domestic group violence, per million popu- pendent on the degree of political instability lation. These figures are for the period which exists prior to its occurrence. 1950-62. An alternative would be to con- struct an index on the basis of the number Two Correlates of Revolution killed during the period of the revolution Having provided operational referents itself. the number killed However, during for the concept of revolution, we may now a a few revolutionary coup lasting days may turn to an examination of possible correlates. not be indicative of the characteristics of The problem of identifying the precondi- that revolution. For example, thirty persons tions of revolution has been of long concern were killed during the Iraqi revolution of to political theorists. For example, Plato 1958. Two of these were the king, Faisal, (1951, p. 54) proposed that differences in and his prime minister, Nuri. The intensity economic interests led to factionalism in of feeling against the government was mani- politics and contributed to the instability of fested when the populace dragged the king’s the city-state. , according to Plato, the streets of The body through Baghdad. produces revolution, meanness, and villainy, death of thirty people would seem to be an while riches produce luxury, idleness, and inadequate reflection of this degree of in- villainy. In substantial agreement with this if we examine the num- tensity. However, position, Aristotle (1962, p. 59) also pro- ber of deaths to the revolution and prior posed that poverty may be a cause of politi- we find that ex- a f ter its occurrence, Iraq cal revolution. Tocqueville (1955, pp. 22- perienced the second highest number of 23), however, dissented from this emphasis million of revolu- deaths per population any on poverty. He suggested that the French 1955-60. tion during the period Similarly, peasant prior to 1789 enjoyed a considerably Venezuela, which experienced the overthrow higher degree of economic independence of the Jiminez , also had one of than did the remainder of the European the highest numbers of deaths as a result of peasantry. Because of this independence domestic violence. The inclusion of the and security, those aspects of feudalism still number of deaths prior to and after a revo- remaining in French society, such as the lution might be indicative of the frustration corvée (a form of periodic forced labor), appeared all the more odious and con- 5 The duration of a revolution is bounded by temptible. two points in time: (1) when active hostilities Indeed, long-term economic depressions first broke out against the in power, and regime were not present prior to the outbreaks of (2) when the insurgents occupied the roles in either the American or Russian revolutions. the structure of political authority. A period of After an exhaustive of four unrest prior to the victory of the insurgents was study major included in the determination of the values in revolutions, Crane Brinton (1952, p. 264) Column 3. Riots and clashes between the police concluded that one uniformity in the oc- are these disturbances. and populace examples of currence of these revolutions was that the In addition, if two or more changes in the per- societies under investigation were all on the sonnel of the governmental elite took place before the outbreak within a three-month period of time, the period upgrade economically between changes was included in the duration. of revolution. Some statistics relevant to the 270

French economy prior to 1789 may be in- TABLE 4 structive. In the 6lectioit of Melun from RUSSIAN EXPORTS AND IMPORTS PER CAPITA: 1895-1913a 1783-85, the amount of uncultivated land was reduced from 14,500 to 10,000 arpents. Rouen in 1787 produced cotton cloth worth 50,000 livres, which was double the produc- tion of a generation before. In the dozen years since the death of Louis XIV, French trade had increased nearly 100,000,000 a Mazour, 1962, p. 323. livres (Brinton, 1952, p. 31). The English Revolution of the 17th cen- well as roads, factories, museums, and a tury and the were pre- higher standard of living, which ceded by similar economic experiences. Ac- generally accompany a high rate of achievement, also cording to Brinton, early Stuart England may tend to increase aspirations. It is posited was notably prosperous, as were the Ameri- that the rates of change of the achievements can colonies prior to 1775 (1952, p. 32). and aspirations are correlated. If achieve- Russia also was making significant economic ment is increasing at a given rate, then the progress prior to the outbreak of World War populace would most likely aspire to the I. Table 4 illustrates this growth in the acquisition of social commodities at the Russian economy. same rate as had been ac- The theories of Plato and Aristotle they previously appear quired. to contradict those of Tocqueville and If we were to plot the increase in aspira- Brinton. The first set asserts that poverty tions over time, the of leads to revolution while the second set slope (rate change) of this variable would approximate the rate claims that revolutions are preceded by a of change of achievement. The similarities significant increase in economic develop- in the slope of the two plots is illustrated in ment. James C. Davies (1962, pp. 6-7) sug- Figure 1. Now if, instead of increasing, the gests that a partial synthesis of these ap- rate of achievement were to decrease, we proaches may provide a more comprehensive may posit a third concept, expectations. The explanation than either of the two taken rate of change of expectations would ap- alone; major revolutions may be preceded proximate that of the decrease in the rate of by steady long-term increases in economic achievement. Rates of expectations may be development, followed by a sharp reversal affected more by immediate reality than are just before the outbreak of the revolution. rates of aspirations. Expectations represent In more general terms, economic develop- a change in outlook caused primarily by an ment might be viewed as a single aspect of immediate decrease in the of a society’s achievement. That is, political, production social commodities. are more in economic, and cultural development may to- Aspirations the nature of a and an gether comprise a single concept which, for hope optimism gen- erated our purposes, may be called achievement. by long-term past performance.6 The A second concept closely associated with distance between the two concepts (revolu- achievement is termed aspirations. The re- 6 The distinction between aspirations and ex- sults of political, economic, or cultural at- pectations may be analogous to the Feierabends’ tainment are visible to the ordinarily gen- (1966, pp. 256-57) use of the terms "want eral populace. Newly granted freedoms, as formation" and "want satisfaction."

CONFLICT RESOLUTION VOLUME XI NUMBER 3 271

FIG. 1. Relationship between rate of achievement, aspirations, and expectations. (Adapted and modified from Davies, 1962, p. 6.) tionary gap) may be seen as a measure of revolution broke out (Mazour, 1962, pp. the potentiality for the occurrence of a vio- 553-55). Apparently, these four major revo- lent revolution. The larger the revolutionary lutions provide evidence for Davies’ hy- gap, the longer and the more violent the pothesis. But whether this hypothesis holds revolution may be. This hypothesis is repre- true in all such instances is a question to be sented in Figure 1. investigated by means of a systematic analy- As evidence for the low rate of expecta- sis across a larger universe of revolutions. tions prior to major revolutions, Davies cites A second correlate of revolution might be the poor agricultural harvests of 1788-89 in the level of educational attainment prior to France which followed the economic ad- the outbreak of the disturbance. Seymour vances of the prior decade. The English M. Lipset (1959, pp. 73-80) has found that economy had suffered a reversal prior to the occurrence of revolutions is associated 1688, while there was a tightening up of negatively with the level of educational at- business conditions in America in 1774-75. tainment. Lipset categorized nations ac- Conditions in Tsarist Russia in 1917 provide cording to their stability and degree of a striking example of a breakdown in the democratic or totalitarian control and found economic machinery immediately prior to that Latin American , which the outbreak of a revolution. Industrial pro- are notoriously amenable to the occurrence duction had declined rapidly. Coal mining of palace revolutions, are also societies which had fallen to a level that threatened the rank lowest on his scale of educational nation’s transportation. The peasants had attainment. Evidence supporting Lipset’s grown sick of the war, and instead of trans- finding is provided in the Dirnensions off porting much-needed grain to Russia’s urban Nations (Rummel, Sawyer, Tanter, and centers, they were hoarding food in the ex- Guetzkow, forthcoming, 1967). These pectation of future disasters. The economy writers find that three educational variables was in a state of virtual collapse when the are correlated negatively with the occur- 272

rence of revolution. One of the three-the TABLE 5 ratio of number enrolled in primary schools DOMESTIC VIOLENCE, TIME RATE OF CHANGE OF GNP/CAP, AND PRIMARY ENROLLMENT RATIO to the total population aged 5-14-gave the highest negative correlation (r = -.84, N = 66). It appears from the data of the mid- 1950s that revolutions will occur more fre- quently in societies with the lowest levels of educational attainment. Having stated the principal concepts and their measures, it may be useful to summa- rize the definitions prior to a presentation of the results: ( 1 ) A revolution is operationally defined by domestic violence and duration. (2) Achievement and aspirations are de- fined by the rate of change of GNP/CAP over time (gross national product divided by population). (3) Expectations are defined by the drop or reversal in the rate of change of GNP/ a Same variable as in Table 3, Column 2. b The sources for this data column are the UN Sta- CAP. tistical Yearbooks, 1955-1962. The data are presented Educational level is defined the in time-series form with base year 1953 = 100. (4) by C Primary enrollment divided by total population primary school ratio-number enrolled in aged 5-14. These ratios were drawn from the World Survey of Education II, UNESCO, 1958, Table 15, pp. primary school divided by total population 58-60. The majority of these values are for the years 5-14. 1950-1954, with Brazil, Cuba, and Iraq for 1950- aged 1953, Haiti and Jordan for 1952-1954, and Turkey for 1950-1952. Given these definitions, the following hy- d The rate of change of GNP/CAP for Cuba can be seen in Figure 2. are potheses suggested: . The values for Iraq and El Salvador were not included in the standardized tabular form in the data (1) The higher the rate of increase of sources, and were therefore calculated separately by means of net national product and population data. the revolution and the f GNP/CAP preceding The values for Syria and Thailand are based on sharper the reversal immediately prior to four and six data points respectively, due to limited data availability. the revolution, the greater the duration and g The dashes in the columns indicate that the data violence of the revolution. were not available. (2) The lower the level of educational at- tainment prior to the revolution, the greater time in GNP/CAP is presented in Column the duration and violence of the revolution. 3. The figures in this column are regression coefficients which were calculated for the These hypotheses were tested across seven-year period preceding each revolu- seventeen cases of successful revolutions for tion.7 Here, the regression coefficient is a 1955-60.

7 The period of seven years prior to the out- Data and Results break of the revolution is an arbitrary choice, determined chiefly by the availability of com- Data for the years 1955-60 are presented parable data. It is possible that the investiga- in Table 5. Unsuccessful and colonial revo- tion of longer periods might yield different re- lutions are excluded. The variation over sults.

CONFLICT RESOLUTION VOLUME XI NUMBER 3 273 quantitative representation of a given vari- TABLE 6 able’s rate of change over time ( slope ) .8 CORRELATION COEFFICIENTS BETWEEN CHANGES IN LEVEL OF AND Gross national was WEALTH, EDUCATION, product per capita DOMESTIC VIOLENCEa plotted against time for each nation, and the slopes were approximately linear. The values of these coefficients are comparable, since the same base year (1953 = 100) was chosen for all cases.9 Column 4 lists the pri- mary school ratios. The product-moment correlation coeffi- a Domestic violence was logarithmically transformed cients are presented in Table 6. These co- for the first two values in the second row. No trans- formations were necessary for the calculation of the efficients were calculated first for the total remaining values. number of cases in Column 1, and then The dependent variable in the calculation of these coefficients was domestic violence. separately for the Asian and Latin American cases. The number of cases on which each coefficient is based is written in parentheses plotted and, where necessary, domestic vio- below the value of that coefficient. Before lence was logarithmically transformed to calculating the coefficients, curves were compensate for the effect of an outlier ( Cuba ) . 8 The data in Table 5 are presented with cer- Since these calculations are based on an tain reservations. The regression coefficients in entire population of revolutions in the given the second data column were calculated from period, it must be noted that the sample of UN statistical yearbooks, and in that source it is cases is not random. Although the conven- noted that gross national product data are in- dicative only of general trends. Indeed, different tional significance levels are thus not ap- yearbooks reported varying values of GNP/CAP plicable, they are utilized to provide some for the same country and the same year. In criterion for the tenability of hypotheses. to these order partially resolve inconsistencies, All correlation coefficients are considered the regression coefficients were calculated using significant at the .05 level of confidence; a single time series (from one yearbook) for each in 100 the coefficient. In addition, the criterion for choos- i.e., five times coefficient could ing a time series was that it was reported in the be due to chance. A second interpretation yearbook published one year subsequent to the of the correlation coefficient is the percent- occurrence of the revolution the 1959 (e.g., age of variance explained-the square of Yearbook for Pakistan For those cases [1958]). the coefficient. This mode of presentation is in which the time series was incomplete, such also in the discussion. as Syria or Thailand, a subsequent yearbook employed following two was chosen which contained the most complete The values in the first data columns time series. of Table 6 are statistically nonsignificant. 9 The term "base year" refers to the basis of The figures in Column 3, however, are some- the UN Statistical Yearbook’s calculation of the what more indicative of the presence of re- time-series In all of the series, 1953 is figures. As R. Alker set equal to 100 and the values for the other lationships. Hayward (Russett et has cul- years are expressed relative to that year. This al., 1964, pp. 322-23) suggested, term does not refer to the starting point for the tural differences between regions may ac- calculation of the regression coefficients. The count for an apparent lack of significant re- choice of a for the calculation of starting year lationships. When the total number of cases these coefficients is based solely on the fact that is considered, domestic violence in Asian the chosen year is seven years prior to the out- break of the revolution. countries in general, and Middle Eastern 274

countries in particular, may have an entirely relation coefficients was calculated for dura- different political meaning from the same tion, the rate of change of GNP/CAP, and degree of violence in Latin American coun- the level of the primary enrollment ratio; tries. This division of the countries in Table the results were not significant at p < .05. 5 into Latin American, Asian, and Middle Perhaps one reason for this lack of a rela- Eastern is a first approximation to a control tionship is the extreme difficulty in mea- for cultural differences. Only France, of the suring duration. As Brinton (1952, pp. 238- countries included in Table 5, did not fall 50) notes in his study of the Russian Revolu- into any one of these categories. tion, the notion of duration may be better When Asian revolutions are considered, suited to the making of qualitative judg- we find that the degree of association be- ments rather than quantitative distinctions. tween rate of change of GNP/CAP and For example, there is a large degree of un- domestic violence is significant at p < .011 certainty as to the duration of the Russian and accounts for approximately 88 percent Revolution. On the one hand, the period in of the variance. The proportion of students the early 1920s may be viewed as a &dquo;Ther- enrolled in primary schools is correlated midor,&dquo; hence a denouement of the revolu- negatively with domestic violence, and the tion. Later, on the other hand, the Soviet coefficient is nonsignificant at p < .10. Ap- Union under Stalin experienced a reign of proximately 58 percent of the variance is terror which rivalled the intensity of the accounted for by this relationship. For the French Revolution at its height. The Soviets Middle Eastern countries, the correlation may have undergone a &dquo;permanent revolu- between the rate of change of GNP/CAP tion&dquo; throughout the Stalin period, and only and domestic violence is significant at p < after Khrushchev’s condemnation of Stalin .05 and accounts for approximately 92 per- did the revolution subside into a period of cent of the variance. The rate of change of normalcy. Thus there are at least two possi- GNP/CAP prior to the revolution is almost ble measures of the duration of the Russian perfectly associated with the level of domes- Revolution. In his attempt to define the tic violence. The correlation between the duration of the English Revolution, Brinton primary enrollment ratio and domestic vio- encountered similar difficulties. Because of lence is nonsignificant at p < .10 and ac- the close-knit series of events prior to the counts for 85 percent of the variance. outbreak of the in 1642, the dura- The coefficients for the Asian and Middle tion of this revolution is essentially a matter Eastern revolutions appear to provide some of subjective determination (Brinton, 1952, of the explanation of variance for all p. 75). the cases. However, because of the small Brinton’s analysis and the results of this number of cases, these results should be ac- study suggest that the duration variable is cepted with caution. Before we can assume difficult to measure. Moreover, when com- that 92 percent of the variation of domestic paring durations we often commit the fallacy violence in the Middle East is accounted for of employing different units of analysis in by the rate of change of economic develop- the same investigation. By recording dura- ment, we clearly must have further verifica- tions we are, in effect, ignoring the events tion efforts based on a longer time-period. which occurred during these periods. A Deaths from domestic violence is one three-day revolutionary coup which is fought measure of revolution; another is the num- with bitter intensity should not be equated ber of days of the revolution. A set of cor- with a three-day palace revolution which is

CONFLICT RESOLUTION VOLUME XI NUMBER 3 275

FIG. 2. Plot of gross national product per capita (GNP/CAP) versus time for Cuba, 1947-53. (In the construction of this figure, the inception of the Cuban revolution was taken to be Castro’s first armed landing in 1953.) bloodless and leads only to a change in the The longer the revolution persists, the personnel of the governmental elite. A more greater the polarization of the society around adequate measure might be the number of the warring factions. Traditional associa- events of a similar type compressed into a tions are broken. Factions which at the given time-interval. Rather than treating start of the conflict might have been able to duration as a continuous variable, however, settle their differences by negotiation now we may more profitably view it as discrete find that, with the passage of time, enmities at a given threshold. Below a certain value have increased to the extent that only un- of this variable, the duration of a revolution conditional surrender by one of the factions may be irrelevant to the explanation of can terminate the revolution. Reference to either the antecedents or the consequences Table 3 reveals that the Cuban revolution, of its occurrence. Above that value, the with a duration of six years, is at the same duration of a revolution may be a meaning- time the only instance of mass revolution in ful characteristic. the chosen period. The duration of the 276

Cuban revolution is more than ten times one example of a mass revolution-but are greater than any of the other values in Col- left without an adequate test for the other umn 3, and the &dquo;threshold&dquo; may be some- instances of revolution. How do we mea- where between one and five years. Future sure expectations in cases of revolutionary attempts at the determination of this value coups, reform coups, and palace revolutions? may be fruitful for a better explanation of the revolutionary process. Differences in Expectations: The Gini Index A Mass Revolution: The Case Cuba of An operational definition of expectations We may recall that Davies’ hypothesis might be the degree of inequality in income suggests that the higher the long-term rate or land distribution.10 If the rate of achieve- of achievement and aspirations preceding ment is high and unequally distributed, large the revolution, and the sharper the reversal sectors of the population might be aroused (expectations) immediately prior to the revo- by this appearance of new wealth in a lution, the greater the intensity as measured limited segment of the population. Aspira- by duration and domestic violence. Rate of tions, which are based on long-term achieve- achievement is defined as rate of change of ments, may tend to be higher, whereas ex- GNP/CAP, whereas the reversal is equated pectations-more a function of immediate with expectations. The rate of change of reality-might be lower because of the pre- GNP/CAP for Cuba was not included in vailing economic inequalities. Where the Table 5. Instead, Figure 2 plots GNP/CAP &dquo;revolutionary gap&dquo; between aspirations and against time for the seven-year period pre- expectations is largely due to a high level of ceding Castro’s first armed incursion against inequality in land distribution, we would ex- Batista in 1953. pect an increase in the probability of revolu- If we compare Figure 1 (p. 271) with tion. Figure 2 we see a marked similarity. A The Gini index is a measure of inequality. high rate of achievement is followed by a It measures the area between the Lorenz reversal which, according to the theory, cor- curve of any distribution and the line of per- responds to the expectation of the society. fect equality. For high inequality, the area In addition, the Cuban revolution, with between the Lorenz curve and the line of 2,900 deaths per million and a duration of perfect equality would be correspondingly six years, falls into the mass revolution cate- large; hence, the Gini index would also have instance of such a revolution gory-the only a large value. For lower degrees of in- in the 1955-60. Cuba is a period Although equality, the reverse is true. it with single case, compares favorably Table 7 presents the Gini indices for all Brinton’s analysis, in that the four countries nations included in the data source (Russett in his study experienced economic condi- et al., 1964 p., 239ft). Table 7 (a) lists the tions similar to that of Cuba prior to the values for nations which suc- onset of revolution. experienced cessful revolutions in the 1955-60, The Cuban revolution is the only instance period whereas Table 7 ( b) presents the values for in the period under consideration where a severe reversal actually occurred immedi- 10 It should be noted that this treatment of ately prior to the revolution. We thus find "expectations" is cross-sectional in design, as a confirmation of the existence of a &dquo;revolu- opposed to the longitudinal nature of Davies’ tionary gap&dquo; in the Cuban example-our theory.

CONFLICT RESOLUTION VOLUME XI NUMBER 3 277 those nations which did not have successful TABLE 7 GINI INDICES OF LAND revolutions. The indices are for land, whose INEQUALITY’ distribution is probably correlated with in- come distribution. A one-tailed t-test of the significance of difference was performed and a t equal to 2.641 was calculated. With 48 degrees of freedom, this value is signifi- cant at p < .05. Thus degree of land in- equality might be interpreted as an opera- tional form of expectations; at any rate, revolutions occurred in those societies with a higher degree of land inequality. Russett (1964, p. 272) also reports a relationship between the Gini index of land inequality and domestic violence of r = .44 across 44 cases.

Conclusions The analysis of inequalities of land distri- bution and the Cuban case study provide empirical support for the hypothesized rela- tionship between the revolutionary gap and the domestic violence characteristic of revo- lutions. However, while the correlational analysis provides support for the hypothesis in the case of Asian revolutions, it was not a From Russett et al., 1964, pp. 239-40. substantiated for Latin American revolutions in this period. occurring in Asia are either reform coups The results suggest two alternatives: (1) or revolutionary coups; thus we may suggest If we assume that the level of domestic vio- that the palace revolution is little affected lence is indeed an indicator of the type of by domestic group violence, whereas the revolution in Latin America, then we are other forms of revolution may indeed be so forced to conclude that in Latin America affected. The rate of change of GNP/CAP there is virtually no relationship between and level of education may then be theoreti- the rate of change of GNP/CAP and the cally significant for the study of revolutions form of revolution. This alternative leads to other than the palace variety. Now we may a reliance on an explanation based on re- ask why the palace revolution should be gional differences. (2) If we assume that relatively unaffected by domestic violence. domestic violence bears little relationship to George Blanksten (1958, p. 143) has sug- the palace revolution in Latin America (most gested that one reason for the repeated of the Latin American revolutions in Table &dquo;palace revolutions&dquo; in Latin America is that 5 are of the palace variety), then the rela- the class-oriented social structure provides tionship between the rate of change of the upper class with a virtual monopoly GNP/CAP and the other forms of revolu- of participation in the political authority. tion is not invalidated. Those revolutions Indians are almost entirely excluded from 278

participation in politics, and the mestizos centuries witnessed a hardening of class (descendants of Spanish-Indian marriages) lines between the nobility, the small but are also inactive in politics but emerge at growing bourgeoisie, and the peasantry. Be- times as followers of a particular political fore 1780, class lines had become so rigid faction. Class distinctions in Latin America that Tocqueville referred to the nobility as then act as a relatively impermeable barrier a caste within the general class structure of between the elite (the socioeconomic class French society. We may also note that from which holders of high political office France during this period was free of revo- are drawn) and the mass (all those who are lutionary disturbances. At the same time, excluded from this highest socioeconomic the English were thought of as politically category). Because of this barrier, distur- unstable, whereas the French were consid- bances in the society may seldom affect the ered to be fortunate in having a government &dquo;normal&dquo; operations of government in most which exemplified political stability. Latin American nations. Thus palace revo- In the decade prior to 1789, however, lutions may frequently take place uncon- there was a significant decrease in the rigid- nected with violent changes within the so- ity of barriers between classes. Local parle- ciety. We find that the rate of change of ments were once again assigned a political GNP/CAP bears little or no relationship to function which they had virtually relin- domestic violence in Latin America, but- quished centuries before. The meetings of at the same time-domestic violence bears the parlements provided a social setting for little relationship to the palace revolution, this increasing fluidity across class lines. which until now has been the most frequent The case of the French Revolution may form of revolution in this region. provide a comparison within a single coun- The Asian countries included in this study try : during one period, barriers between the may not experience the same degree of class elite and the mass prevented social distur- rigidity as do the Latin American nations. bances from disrupting the &dquo;normal&dquo; opera- Barriers between the mass and the elite tions of government; later, according to might not be so rigid or impermeable. Tocqueville, this barrier became increasingly Therefore, economic changes which affect permeable and mass revolution occurred. domestic violence also have an effect on the Recent theories of mass society have also form of the revolution. A similar argument emphasized the importance of the relation- can be developed for the effect of education ship between the elite and the mass. LeBon on the form of revolution. As a general (1947, pp. 214 f f ) and Ortega y Gassett proposition we suggest that the more im- (1940, pp. 97 f f ) stress that for the govern- permeable the barriers between the mass mental elite to perform its proper function and the elite, the less the effect of domestic of judicious decision-making, it should be violence on the form of revolution. The con- protected from the effects of mass persua- verse may be equally valid. The more per- sion, which frequently manifests itself in meable the barriers between the mass and violent forms. On the other hand, Hannah the elite, the greater the effect of mass activ- Arendt’s view (1958, pp. 315-24) is that ity on the form of revolution. the mass, if unprotected from the elite, may The social structure of France prior to undergo a process of &dquo;atomization&dquo; in which 1789 provides an example of these relation- individuals are dissociated from each other ships. According to Tocqueville (1955, pp. and become easy prey for the totalitarian 97-108) the seventeenth and eighteenth designs of the elite.

CONFLICT RESOLUTION VOLUME XI NUMBER 3 279

William Kornhauser (1959, pp. 39 f f ) of revolution is essentially one of commit- suggests that if organized group activity in a ment to change. Revolutions are not pieces pluralistic society can act as an intermediary of historical machinery, as the Hegelians or between the elite and the mass, then neither Marxists may contend, but are deeply asso- will be accessible to undue penetration by ciated with the ideological orientations of the other. The intermediary group associa- the participants. Indeed, this study suggests tions can then be viewed as a second form that domestic violence may vary with the of the social barrier between the elite and degree of change intended by the insurgents. the mass. Therefore, the continuum we speak of is Both notions, that of class distinctions essentially a continuum of human desires for and that of the pluralistic group process, change, which-translated into a specific may be amenable to future research as to form of political action-make a revolution. their effect on the incidence and form of The insurgents in a palace revolution, how- revolution. With the possible exception of ever, intend only to occupy roles in the France, none of the countries in Table 5 existing authority structure. Changes in the have strongly developed pluralistic societies. social system or in the structure of political However, the state of their pluralistic de- authority are not included in their program. velopment may be related to the form of Thus, if we are measuring a continuum of revolution. political and social change, the palace revo- Finally, what do these results and our in- lution may not contribute to these measure- terpretation say for the scheme of classifica- ments. tion of revolutions presented initially? That REFERENCES is, we suggested that revolutions may fall The Totalitarian- on a rank-order: mass revolution, revolution- ARENDT, HANNAH. Origins of ism. Cleveland: Meridian, 1958. ary coup, reform coup, and palace revolu- —. On Revolution. New York: Viking, tion. In Asian such an societies, ordering 1963. was suggested by the relationship between ARISTOTLE. Politics. (Ernest Barker, ed. and changes in wealth, in educational level, and translator.) New York: Oxford University in domestic violence. Latin American so- Press, 1962. BLANKSTEN, GEORGE. "Revolutions." In H. E. cieties, with a preponderance of palace revo- DAVIS (ed.), Government and Politics in did not exhibit these lutions, relationships. Latin America. New York: Ronald Press, The palace revolution appears, then, to be 1958, 119-46. a breed apart from the others. —. "Latin American Revolutions." In Our definition was predicated on the The 1962 Carolina Symposium: Today’s Revo- lutions. Hill: of North palace revolution as a lower bound above Chapel University Carolina, 1962, 71-79. which lies a rank-order or continuum of BRINTON, CRANE. The Anatomy of Revolu- more intense changes in both government tion. New York: Vintage, 1952. and ex- society. Since the data appear to DAVIES, JAMES C. "Toward a Theory of Revo- clude the palace revolution from the con- lution," The American Sociological Review, cept of a continuum, the definition might 27 (Feb. 1962), 5-13. have to be revised so that the &dquo;lower bound&dquo; DEUTSCH, KARL W. "External Involvement in Internal Wars." In H. ECKSTEIN Inter- is raised and the palace revolution is ex- (ed.), nal War. Glencoe: Free 100-10. cluded from the definition. Press, 1964, FEIERABAND, IVO K., and ROSALIND L. FEIERA- This would reasonable if we exam- appear BAND. "Aggressive Behavior Within Polities, ine the insurgents’ orientations. A continuum 1948-62: A Cross-National Study," Journal 280

of Conflict Resolution, 10, 3 (Sept. 1966), RUMMEL, RUDOLPH J. "Dimensions of Conflict 249-69. Behavior Within and Between Nations," Gen- GROSS, FELIX. The Seizure of Political Power. eral Systems Yearbook, 8 (1963), 1-50. New York: Philosophical Library, 1958. —. "Dimensions of Conflict Behavior HUNTINGTON, SAMUEL. "Patterns of Violence Within Nations, 1946-59," Journal of Con- in World Politics." In S. HUNTINGTON (ed.), flict Resolution, 10, 1 (March 1966), 65-73. Changing Patterns of Military Politics. New —, JACK SAWYER, RAYMOND TANTER, and York: Free Press, 1962, 17-50. HAROLD GUETZKOW. Dimensions of Nations, KORNHAUSER, WILLIAM. The Politics of Mass forthcoming, 1967. Society. Glencoe: Free Press, 1959. RUSSETT, BRUCE M., HAYWARD R. ALKER, JR., LASSWELL, HAROLD, and ABRAHAM KAPLAN. KARL DEUTSCH, and HAROLD LASSWELL. Power and Society. New Haven: Yale Uni- World Handbook of Political and Social Indi- versity Press, 1950. cators. New Haven: Yale University Press, LEBON, GUSTAVE. The Crowd: A Study of the 1964. Popular Mind. London: T. Fisher Unwin, SILVERT, KALMAN H. Reaction and Revolution 1896. in Latin America. New Orleans: Hauser LIEUWEN, EDWIN. Arms and Politics in Latin Press, 1961. America. New York: Praeger, 1960. STONE, LAWRENCE. "Theories of Revolution," LIPSET, SEYMOUR M. "Some Social Requisites World Politics, 18 (Jan. 1966), 159-76. of American Political Science Democracy," TANTER, RAYMOND. "Dimensions of Conflict 53 69-105. Review, (March 1959 ), Behavior Within and Between Nations, 1958- MAZOUR, ANATOLE G. Russia: Tsarist and 60," Journal Resolution, 10, 1 Communist. New York: Van Nostrand, 1962. of Conflict (March 1966 ), 41-64. ORTEGA Y GASSET, JOSE. The Revolt of the Masses. New York: W. W. Norton, 1940. —. "Dimensions of Conflict Behavior PLATO. The Republic. ( A. D. Lindsay, transla- Within Nations, 1955-60: Turmoil and In- tor.) New York: Dutton, 1951. ternal War," Papers, Peace Research Society, ROSENAU, JAMES N. "Internal War as an Inter- 3 (1965), 159-83. national Event." In J. N. ROSENAU (ed.), TOCQUEVILLE, ALEXIS DE. The Old Regime International Aspects of Civil Strife. Prince- and the French Revolution. New York: ton : Princeton University Press, 1964, 45-91. Doubleday, 1955.

CONFLICT RESOLUTION VOLUME XI NUMBER 3