1. Perspective And Summary
16A. On Causes of International Conflict
Democratic Peace page
Six phasing propositions are presented in Table 15A.1. Each will be treated separately below in terms of supporting theory, predictions, evidence, and conclusion. This approach, as well as the manner in which the evidence is handled, is described in Chapter 14 and in Appendix 16C.
Major evidence for each proposition will be discussed in order to communicate the quality and nature of the analyses drawn upon and the basis for the ratings. Space does not allow that this be done for all propositions in this book, however, and in subsequent appendices in which propositions are presented, if the evidence is not discussed in the text they still will be presented collectively in a table, such as Table 15A.2.
Preparations for conflict form a separate and distinct pattern of Conflict Behavior.
Theory: As discussed in Section 11.2 and Chapter 13, preparations manifest the uncertainty--initiation--phase of the balancing of powers, the girding of the will. No further Conflict Behavior may be manifest, for the underlying conflict may be resolved through accommodations or abnegation, or preparation may be followed by an actual confrontation of some kind. What Conflict Behavior will follow preparations is unclear at the initiation stage (thus, the uncertainty). It may only be status quo testing; or possibly negative actions, such as boycotts, embargoes, severance of diplomatic relations; or even coercive violence.
Therefore, (1) preparations should comprise specific behaviors, such as strengthening forces, alerts, cancelling leaves, troop movements, partial or full mobilizations; (2) preparations may or may not eventuate in confrontation; and (3) what kind of confrontation (balancing) will occur, if any, is unknown. It follows that preparations should form empirical patterns (latent functions) of behavior separate and distinct from other forms of Conflict Behavior.
Prediction: A factor or component analysis of diverse, static or dynamic Conflict Behavior data should delineate empirical patterns of preparations (on factor analysis, see "Understanding Factor Analysis"). Moreover, preparations and other kinds of Conflict Behavior should be relatively uncorrelated (bivariate or multivariate--on these terms see Understanding Correlation).
Moreover, if such data are included, alliances and increase in armaments (such as in military expenditures) should form a separate pattern from short-run preparations (such as alerts or mobilizations).
Evidence: See Table 15A.2.
In crises periods, a "warning and defensive acts" component is statistically independent of other types of behavior, and involves a variety of preparatory-type actions (Phillips and Hainline, 1972; Phillips and Lorimore, n.d.). In precrises phases the same component also exists (Phillips and Lorimore, n.d.), as it must if there is such a preparatory phase.
Considering preparations to involve a variety of military growth or decline factors (e.g., growth in defense expenditures), then U.S. and USSR preparations by year, 1948-1968, have no high correlation with the cold war. Indeed, most correlations are negative (Väyrynen, 1973: 132).
For states, a statistically independent component almost uniquely delineating prepatory actions and called "warning and defensive acts" existed in 1955 (Rummel, 1967; Keim and Rummel, 1967); and in 1963 (Phillips, 1973; Oliva and Rummel, 1969). Moreover, such a pattern exists in both the conflict behavior a state directs towards others (previous citations) and that received from other states (Phillips, 1973).
In other factor analyses studies on conflict behavior, an insufficient number of preparatory variables were included for such a pattern to emerge, but in both Rummel (1963) for 1955-1957 data and Tanter (1966) for 1958-1960 data, troop movements and mobilizations have low correlations with other conflict behavior, no high loading on conflict components, and low communalities (h2).
Dynamically, Keim and Rummel (1967) found that a preparatory pattern of conflict showed the most stability between 1955 and 1963. A super-P component analysis of state conflict behavior for 13 months, 1976-1977, delineated a statistically independent, preparations pattern (Project 15). A dynamic (super-P) analysis of six Middle Eastern states, separately and together, for 217 months (1949-1967) delineated a preparations pattern (Wilkenfeld, et al. 1972).
For dyads, a preparatory pattern also existed in the conflict behavior of states towards specific others in 1950 (Project 4), 1955 (Project 27; Rummel, 1965; Hall and Rummel, 1970), in three-month periods of 1963 (Phillips, 1969), and in 1967 (Project 5).
However, in the analysis of 24 types of conflict behavior of 305 dyads for 1965 no preparatory pattern emerges: it combines with a pattern of violence (Project 4). Moreover, when common factor analysis is done on 16 variables for 1955 (for apparently 341 dyads), only two distinct patterns emerge: military violence and negative communications. Finally, a warning and defensive actions variable was included in the analysis of conflict for 1950, 1955, 1960, 1963, and 1965. Its communality tended to be relatively low, although in 1955 and 1965 it correlated highly with a violence pattern.
Dynamically, at the dyadic level a preparations pattern (Hall and Rummel, 1970) is stable between 1955 and 1963 (product moment is .89). When dyadic conflict behavior is analyzed by month for 1963 (using super-P factor analysis), a distinct preparations pattern is found (Phillips, 1969). Such a statistically independent pattern also exists in the dyadic behavior between the United States, USSR, and China for 72 months, 1962-1968 (Phillips and Hainline, n.d.). Finally, super-P analysis of the years 1950-1965 (Project 3) showed a warning and defensive action variable to have relatively low correlation with other conflict variables (h2 = .59; highest loading = .68); in a common super-P, factor analysis of the same data, its variance in common with the other conflict behavior was shown (Omen, 1975) to be quite low (h2 = .27).
In sum, and including those studies not mentioned above, there are 50 analyses bearing on the Uncertainty Proposition. These are classified by category in Table 15A.5. (for discussion of this method of classification, see Appendix 16C), located near the end of the appendix. Twenty-one of these studies are direct and strongly positive; six of them are from important studies; 12 are my own analyses. An additional 22 analyses support the proposition, 19 of which are direct, 10 important, and three my own. Only five analyses are at all negative-two strongly so.
Conclusion: In those studies including a number of preparations-type conflict behavior variables an empirical, statistically independent pattern of short-run preparations was found to exist in all but one study, regardless of level and whether static or dynamic. Moreover, 86% of the evidence supports the proposition (almost half of this strongly so). Therefore, it appears safe to conclude that: the evidence strongly supports the proposition.
Conflict Behavior manifests different coercive versus noncoercive paths in the balancing of powers.
Theory: The balancing of powers includes coercion (which is the path to force), bargaining power, and intellectual power, as manifested in their means: threats or deprivations, promises, and persuasion. Fundamentally, therefore, we can divide the process of conflict along two tracks: coercion (and force) and noncoercion.
This does not mean that Conflict Behavior moves exclusively along one path or the other. Rather, these are two time-components of such behavior. For example, a person can both work at a career (such as journalism or the military) while getting a degree in law. These would be two separate paths underlying and explaining the complex of the person's day-by-day behavior. But they are divergent paths, not leading necessarily to the same goal. As an example of convergent paths, I both teach and do research. Each is a separable path of behavior, but each for me is a convergent process towards a better understanding of my subject. Similarly, in a violent conflict between states, as in the Vietnam or Korean Wars, along with the threats negative actions, and violence, also may occur offers of aid, attempts at persuasion, and tacit bargaining: convergent paths in the balancing of powers. See Section 12.4 for a discussion of the noncoercive path.
Prediction: Two separate, empirical patterns of static and dynamic Conflict Behavior should exist, regardless of level. One should be predominantly of coercive (and forceful) conflict behavior; the other of noncoercive, cooperative behavior.
This cooperative, noncoercive, behavior is part of the balancing of powers, and therefore manifested through particular cooperative events, such as a promise, offer, compliment, explanation, assurance; or aid, support, help, assistance. Such behavior is different from that constituting accommodations and negotiations, such as signing agreements, which are the end--the accommodations--subphase of the noncoercive path. Moreover, such cooperative events are different from familistic and contractual flows and structures, such as the flow of mail, trade, tourists, or comembership in intergovernmental international organizations.
Therefore, the noncoercive path in the balancing of power should be manifested by a pattern of primarily exchange and intellectual type of behavior, distinct from accommodative-negotiative behavior and familistic-contractual flows and structures (see Appendix II, Section II.1 ).
Moreover, we should find little correlation between coercive behavior and force on the one hand, and cooperative balancing on the other.
Evidence: There is a terminological problem here that may create confusion in surveying the evidence, and which I anticipated in Section 4.6. Most of the literature (including my own studies) label coercion and force as conflict behavior, and noncoercion within a conflict as cooperation. We thus have the set of conflict behaviors constituting the balancing of power, with the subsets of conflict behavior and cooperation. Surely, this is confusing.
Therefore, in presenting the evidence I will continue to use the following translation rules. The term "Conflict Behavior" will refer to the balancing of powers, encompassing both what the sources call conflict and cooperation (events); "conflict behavior" will refer to coercion and force, as it usually does in the literature. Now, for the relevant sources of evidence for Proposition 15.2, see Table 15A.2.
For 16 international conflicts involving violence or its threat, Fitzsimmons (1969) found one-third of the events to be cooperative, a number of which are conciliatory in the abatement stages of a conflict. In the beginning, duration (of violence), and abatement stages of conflict, two of three correlations between violence variables and conciliatory acts are nonsignificant.
Burrowes and Garriga-Pico (1974) did a variety of factor analyses of 4,500 events for 63 two-week periods in the Middle East, 1965-1967. They found an Arab conflict behavior pattern, three separate Israel-Arab conflict behavior patterns, and three distinct, Arab cooperative patterns. They also (using O-factor analysis) determined distinct periods of Middle East conflict, and then did an analysis within each, finding again separate cooperation patterns.
Still at the case level, Väyrynen's (1973) super-P component analysis of East-West conflict, 1948-1968, found a cold war pattern, a global violence pattern, and a separate US-USSR cooperative behavior pattern (involving also flows and structures).
Moving now to the state-level, Buff owes and Spector (1973) analyzed a variety of Syrian behavior over 74 four-week periods (1961-1967) and found that Syria's foreign conflict behaviors have higher intercorrelations than with cooperative behavior. Moreover, a component analysis of Syria's foreign behavior showed conflict and cooperative (event) behavior to be statistically independent.
The WEIS coding approach to event data involves both cooperative and conflict behavior and has been much used (see Appendix II, Section II.2). McClelland and Hoggard (1969) found one-third of foreign events (WEIS) data to be cooperative for 83 states in 1966. Component analysis revealed a separate, statistically independent cooperation pattern among 47 types of Conflict Behavior. Similar log transformed (WEIS) data for 73 nations, 1966-1969, showed a distinct cooperative pattern to exist, regardless of how the data was aggregated (Salmore and Munton, 1974). Moreover, when differently collected or aggregated data sets are analyzed, cooperation still emerges (Salmore, et al, 1974). Moore (1970) also found for 40 months, 1966-1969, for 86 nations, a distinct cooperation pattern. However, it also had some conflict behavior (deny, warn, and reduce relationship) correlated with it. Finally Kegley, et al (1974) factor analyzed WEIS coded CREON data (see Appendix II, Section II.2) and found a statistically independent negotiations pattern, with cooperation behavior also highly correlated with it.
For 14,500 acts of 32 black African states, 1964-1966, cooperative events comprise 35% of their behavior (McGowan, 1973). Moreover, cooperation forms two distinct patterns (accept; increase relationship) and is independent of nonmilitary and diplomatic conflict patterns. However, there is a mixture of reward behavior with force and subversion, and participatory with conflict behavior. In no case do conflict and cooperative behavior appear negatively correlated.
I am aware of only one state-level, dynamic study (Wilkenfeld, 1977). A super-P component analysis over the years 1966-1970 shows a clear cooperation pattern (with a few correlated conflict behaviors) in the behavior of 56 states. However, the Conflict Behavior received by states involves a large, mixed conflict-cooperation pattern (accounting for 58% of the variance) and a separate yield and reward pattern. These results for received Conflict Behavior are ambiguous, and may be due to an insufficient number of factors being extracted from the data.
At the dyadic level, separate positive communications and bargaining, cooperative patterns exist for 182 state-dyads in 1967 (Project 5). Moreover, for these dyads conflict behavior has little ability to predict cooperative behavior (canonical analysis, trace correlation squared is .10; highest squared canonical correlation is .18).
As tabulated in Table 15A.5, there are in total nine strongly positive, four positive, and four ambiguous analyses, all of which are direct. There are no unambiguous negative results. Of the important studies, two are strongly positively, two positive and two ambiguous.
Conclusion: No unqualified negative results appear. In McGowan (1973) a combined cooperative-conflict pattern appears, but so do purely cooperative patterns; in Moore (1970) some conflict behavior and cooperation are intercorrelated, but this is within a dominant cooperation pattern, and in Kegley, et al. (1974) negotiation and cooperation go together, but independently of conflict behavior.
These ambiguous results should be compared to the nine different sources of unambiguous, positive evidence for the existence of a clear and distinct cooperative pattern. The conclusion follows: the evidence strongly supports the proposition.
Conflict patterns manifest separate status quo testing, nonviolent coercion, violent coercion, force, and accommodations subphases in the process of balancing powers.
Theory: The balancing of powers involves two paths and a number of phases initiated by status quo testing, as described in Chapter 12 and Chapter 13. The noncoercive path leads to accommodation, while the coercive one has a number of phases:
No one necessarily follows another and, indeed, the process may return to previous phases (e.g. force may deescalate to coercion). Moreover, accommodations may follow from any nonforce phase. Accordingly, the separate phases should be manifest in separate patterns of conflict behavior.
Predictions: Separate, empirical patterns of conflict behavior exist, discernable as status quo testing, nonviolent coercion, violent coercion, force, and accommodations.
Evidence: Assessing the evidence presents problems. No study has used data coded specifically to test the proposition. We can, however, locate relevant studies with results bearing on the proposition, but their results require interpretation. Moreover, the studies differ in whether they include data which would allow a specific pattern to emerge. It is no negative result to find a pattern missing when the appropriate variables are omitted.
Another problem is that at the level conflict events are usually coded, the same event may be a manifestation of different phases, as for example, a threat may reflect status quo testing and nonviolent or violent coercion; military action may mirror status quo testing, coercion, or force. Escalation to more intense conflict behavior does not mean a termination of actions at a lower level. Nonviolent negative actions may continue even after escalation to coercive violence.
What makes the proposition empirically tractable, however, is that each phase is reflected in a deescalation of some kind of coded-conflict events and escalation of some other kinds, as shown in Table 15A.3. Probing and testing behavior should be most intense during the status quo testing period, and although there will be testing in other phases, it will be dominated by other forms of behavior.
Negative communications may reflect some status quo testing (as in probing with a verbal threat to test a new leader), but should be intense in the nonviolent coercion and even more in the violent coercion phase, when such communications are essential to conveying the threat and desired response. But when force is used, the aim is to overcome and there is less concern with reaching the mind of the other. Thus, negative communication should decrease.
For negative actions and sanctions, such as severance of diplomatic relations, boycotts and embargoes, they may be used in testing, but should be the primary behavior during the nonviolent coercive stage, and continue as the emphasis shifts to violence.
Low-level military action, such as clashes and discrete military action (e.g., shooting down another's aircraft) will usually reflect military coercion, although it may also sometimes simply be a testing operation. Coercion may also be manifested in extreme violence (as in the Vietnam war), but force will always be reflected by intense violence.
Considering then that these different kinds of behavior will serve different functions and correlate differently in different phases of conflict, we then should expect--if indeed these are actually phases--that five empirical patterns of conflict behavior should be found: testing, negative communications, negative actions and sanctions, low-level violence, and intense violence.
Moreover, there should also occur a separate accommodations pattern involving negotiations, concessions, yielding, agreements, and the like.
To best organize the evidence, Table 15A.4 presents most sources of static or dynamic results by level. A number of elements are involved in the assessments shown: number and types of variables, factor-patterns, communalities,
Different social scientists probably would evaluate the evidence differently in a number of cases. Overall, I have attempted to be conservative, and therefore trust that others would on balance arrive at the same conclusions, although perhaps even with more conviction.
Overall, looking at the totals in Table 15A.5, there are forty relevant analyses, of which ten are strongly positive (eight are direct and four from important studies). There are five negative analyses, two strongly so.
Conclusion: The most rigorous test of the proposition is whether all possible patterns emerged in the different studies, which it did in 38% of those listed in Table 15A.4 The majority of negative results were due to violence forming a single pattern in most of the studies in which it could have divided into the stipulated high and low patterns. Otherwise, the results would usually support the proposition.
Considering, then, the above and the overall analyses tabulated in Table 15A.2, 67% were at least positive and 13% negative, the conclusion is that the evidence supports the proposition.
One final note to provide statistical perspective. The evidence results from diverse factor analyses of usually a dozen or so types of conflict behaviors. There is nothing in the method to force out of the data artificial patterns even dimly approximating those postulated. In one study of two dozen types of conflict variables the probability of getting by chance even one of the expected patterns for 50 cases, states, or dyads is surely over a trillion to one (the one-tailed probability of just one correlation over .44 between two variables for 50 cases in over 2,000 to one.) Then to find two expected patterns in most of the studies and all in about one-third, surely can only be explained by some underlying, substantive factors--such as the theoretical subphases in the balancing of power.
Manifest hostility reflects the situation of conflict, uncertainty, and the balancing of powers.
Theory: International conflict usually engenders animosity, enmity, antagonism. While leaders themselves can be cool and calculating, racial and religious hatreds, cultural animosities, invidious propaganda, and ideological-nationalistic fervor, may create a naked, mass hostility against the "enemy."
Such hostility is not necessarily correlated with any phase of conflict, and may even occur in the situation of conflict, before leaders act overtly.
This hostility is a psychological variable, a strong emotional mix of hatred, anger, and self-righteousness. It may be manifested against another state in many ways: mass demonstrations, private boycotts (e.g., refusal to load the other's ships or buy their products), harassing their nationals, destruction of their property, attacks on their embassy, assassinations, bombings, and the like.
Prediction: Because hostility may be manifest at any phase or subphase in the conflict process, hostile acts--"antiforeign behavior"--will form an empirical pattern uncorrelated with other empirical conflict patterns.
Evidence: Table 15A.2 presents the sources of evidence.
At the case level, Phillips and Hainline's (1972 component analysis of 21 crises showed antiforeign behavior involved in separate patterns with protests and with severance of diplomatic relations. This is a clear negative result.
When the stages of 47 crises are component analyzed, Phillips and Lorimore (n.d.) found antiforeign behavior mixed with expelling and recalling ambassadors in the precrisis period, mixed with severance of diplomatic relations during the crisis period, and a purely separate pattern in the postcrisis. These are ambiguous results.
At the state level, I found (1967) a clear antiforeign behavior pattern for 26 conflict variables for 82 states in 1955. But for the same year and 13 variables, Keim and Rummel (1967) delineated antiforeign behaviors mixed with the incidence of violence, although for 1963 a separate and clear antiforeign pattern was found. For 1963 also, 24 conflicting variables, and all states, Oliva and Rummel (1969) found that same uncorrelated pattern; Phillips (1973) also uncovered the same pattern in behavior sent by states, but for received behavior antiforeign behavior was mixed with negative communications.
Some dynamic, state-level evidence is also available. Phillips' (1970) super-P component analysis of 12 months in 1963 for 65 states showed a clear antiforeign pattern in both sent and received conflict behavior. A similar analysis of 13 months, 1976-1977, also identified the same separate patterns (Project 15).
Moving now to the dyadic level, a separate antiforeign behavior pattern appears for all dyads manifesting conflict for four of the five different years analyzed (Project 4) and for all years in a more refined data analysis (Project 3). This pattern also was delineated by different studies for 340 dyads in 1955 (Rummel, 1967), 289 dyads in 1960 (Project 2), 275 dyads in 1963 (Hall and Rummel, 1970), and 182 dyads for 1967 (Project 5). For common factor analyses (Analysis 1.5) of 16 conflict variables, 1955 data, antiforeign behavior was unique--uncorrelated with common patterns of conflict behavior.
Finally, some dynamic evidence is available. A super-P component analysis of the years 1950-1965 delineated separate antiforeign patterns (Project 3), which became unique in a similar common factor analysis (Omen, 1975). Phillips' (1969) super-P analysis of 12 months in 1963 for 267 dyads delineated an uncorrelated antiforeign pattern. Such a pattern was also found by Park and Ross (1971) for Asian dyadic monthly conflict behavior, 1962-1968; and for separate analysis of foreign behavior was shown to be uncorrelated with other patterns. And a similar analysis (Phillips and Hainline, n.d.) of 72 months of the dyadic behavior between China, USSR and the United States also brought out a separate antiforeign pattern.
Overall, as shown in Table 15A.5, of the 24 analyses bearing on the proposition, most (16) are strongly positive and direct (four are from important studies). There are three negative analyses, two strongly so.
Conclusion: The evidence strongly supports the proposition. Even the negative findings lend support: there is no pattern across studies to the correlation between antiforeign behavior and the other conflict patterns. The few correlations that did occur involved warning and defensive actions (Project 27; Rummel, 1965), negative communications (Phillips, 1973), violence (Keim and Rummel, 1967), and negative actions (Phillips and Hainline, 1972; Phillips and Lorimore, n.d.).
Negative communications, negative actions and sanctions, and violence, are reciprocated.
Theory: Structures of expectations enable mutually reliable predictions between two actors. When this structure is disrupted it involves both actors; both are then parties to the balancing which establishes new expectations. Much of their confrontation will then involve an action-reaction process: coercion will be met with coercion, violence with violence. Moreover, by definition accommodative (negotiating) behavior will involve the opposing parties.
However, preparations may not be met with opposing preparations (the other side may try appeasement or simply be unaware of such activities), status quo testing may not be met by opposing behavior--actually, no response may result. Moreover, while violence may be met by violence, one side may be applying force, the other coercive violence.
Nonetheless, certain types of conflict behavior which reflect different phases of conflict (as shown in Table 15A.3) should be reciprocal. These are negative communications, negative actions/sanctions, low-level violence, and high-level violence.
The behavior of the parties to a conflict may be momentarily out of alignment as one side moves to escalate or deescalate the confrontation, but this should soon be met by an adjustment by the other side, and an action-reaction, lock-in at a new or prior phase will occur.
Prediction: Empirical patterns of negative communications, negative actions/sanctions, low-level violence, and high-level violence will be reciprocal. That is, each party tends to manifest behaviors similarly during the same period.
Evidence: Table 15A.2 presents the sources of evidence. The nature of the evidence can be illustrated by selecting a few of the analyses. At the case level, McClelland (1965) did a super-P factor analysis of the Taiwan Strait crises during 1950-1964. He found a high reciprocity in behavior, although the study does not make clear how this breaks out along the patterns. In the 12 days of the Sino-Indian Conflict in 1962, Field (1972) found a high correlation between the actions on each side.
For states, Phillips (1970) compared the canonical correlations between their sent and received conflict behavior to the (1) canonical correlations of present and past received, and (2) present and past sent Not only was the canonical correlation of sent and received behavior high (r = .80), it was much higher than for the prediction of present from past behavior.
In a similar study, Phillips (1973) did a canonical analysis of sent and received behavioral patterns and found that military violence sent best linked to military violence received (r = .98); negative communications and actions/sanctions to negative communications (r = .94); negative communications minus negative actions/sanctions to negative sanctions (r = .74) and warning and defensive actions to warning and defensive (r = .35).
For dyads, Wilkenfeld, et al. (1972) found that over 217 months, the best predictor of the conflict behavior of a Middle East state was the conflict received. Also dealing with the Middle East, and the United States and USSR in addition, Azar (1975) found a state's past behavior and that received from the other jointly predict its behavior (89% fit between observed and predicted). In a more comprehensive and detailed study of 12 dyads, Azar, et al (1974), showed that past behavior accounts for only about 25% of the variance and that behavior received is the best predictor (but the most effective model includes both an actor's past behavior and that received from the other).
Finally, in Phillips and Hainline's (n.d.) canonical time-series analyses of conflict behavior between the major powers, they found that an object's behavior explained more of the actor's behavior to it than did the actor's past behavior; and that the mutual behavior of actor and object in the same month tended to be similar.
In sum, considering the twenty-nine analyses rated in Table 15A.2 in regard to the proposition and totaled in Table 15A.5, most are strongly positive and direct (12). Five of seven important studies are strongly positive. There are four negative analyses.
Conclusion: Considering that 48% of the evidence is strongly positive and 86% at least positive, then I argue: the evidence strongly supports the proposition.
The threat of imminent escalation in a subphase is manifested through an intensification and diversification of Conflict Behavior.
Theory: A coercive phase in the balancing of powers is a behavioral lock-in by the parties. As the parties perceive that escalation is likely, they will intensify current behavior in the hopes of avoiding escalation and to signal its imminent possibility. Moreover, behavior will become more diverse. Warning and preparatory actions (troop movements, alerts, and the like) may be involved, as well as attempts at bargaining in order to avoid escalation. The imminence of the escalation also means that time is short, decisions must be made soon, and this will also tend to increase the tempo of activities.
Crises may occur in any phase; it involves a threatened escalation to a next and higher phase. Thus, a crisis may occur in a situation of conflict, threatening an outbreak of overt conflict; in the situation of uncertainty, threatening violence; or in the coercive violence subphase, threatening force. In other words, crisis is a behavior pattern separate and distinct from the phases and subphases of conflict.
Predictions: A crisis period during a conflict phase will manifest intensified (volume) and more diverse behavior. In studies employing crises variables, crises should be delimited as a distinct behavioral pattern.
Evidence: Table 15A.2 presents the evidence. Generally, studies of crises have not subdivided conflict into the theoretical phases posed here and isolated the crises within each. Such studies have been 6f imminent threats of violence. Nonetheless, they usually provide direct evidence on the proposition.
Again, some of the analyses will be selected to illustrate the evidence.
To begin with the case level, in the Bendix analysis ("United. . ." ; 1968) of conflict, 1945-1966, no crises pattern was found for 52 conflict variables and 351 conflicts. But many of these conflicts were riots, bloodless coups, international incidents, and the like. When only conflicts involving more than 99 killed were analyzed, a separate crises pattern of behavior emerged. Phillips and Hainline's (1972) component analysis of 21 crises found seven components, more than usually found for conflicts and thus indicating greater diversity. Moreover, for all except the first, the variance accounted for is fairly level across the components, again suggesting diverse behavior.
At the dynamic-case level, McClelland (1965) showed an increase in positive acts during the Taiwan Straits crises. McClelland and Hoggard (1969) also found an increase in conflict behavior in the Berlin and Taiwan crises; and McClelland's (1972) study shows that the change from noncrisis to crisis is in a greater variety of behavior. Azar's (1972) analysis of the 1956 Suez conflict showed that the crisis involved an increase in reciprocity and an increase in behavior ("events") above a normal range. Finally, Phillips and Lorimore's (n.d.) study of precrisis, crisis, and postcrisis periods found that crises involve more conflict patterns and more diverse behavior (they could use 22 conflict behavior/variables to analyze a crises, but only 15 for precrises and 16 for postcrises).
In total, there are 13 relevant crises analyses, as totaled in Table 15A.5, of which six are direct and strongly supportive, and three of these are from important studies. There are no ambiguous or negative analyses.
Conclusion: The evidence strongly and unanimously supports the proposition.
As can be seen, regardless of category, the totals come out clearly the same across the six phasing propositions. This indicates that the conclusions are independent of the directness of an analysis, importance of a study, or whether I did the analyses or not. Moreover, even when the one-level decrement is applied, the results are still favorable by more than 2 to 1.
Of course, the totals in Table 15A.5 are only a rough guide. Qualitative considerations also bear on the assessment, and these were taken into account in assessing each proposition. Table 15A.6 presents the quantitative-qualitative judgment on each proposition. As can be seen, when judged individually, the propositions are with one exception strongly supported.
One aspect of the results should be noted. The balancing of powers is a dynamic process involving several subphases. It is therefore interesting and important to note in Table 15A.2 that out of 47 relevant dynamic conflict analyses, only six were negative.
On balance, then, the evidence supports all the propositions. And by virtue of the propositions being theoretically imbedded in the field perspective the evidence supports this view of international conflict as well.
* Scanned from Appendix 15A in R.J. Rummel, War, Power, Peace, 1979. For full reference to the book and the list of its contents in hypertext, click book. Typographical errors have been corrected, clarifications added, and style updated.. Even if a statistically independent pattern of, say, negative actions does not emerge, low communalities for the variable(s) indicate that negative actions/ sanctions is separate from the other patterns.
1. "Conflict Behavior" refers to that behavior involved in the balancing of power; "conflict behavior" refers to that Conflict Behavior which is antagonistic. See Section 4.6.
2. Herein lies a source of terminological confusion. Cooperation and conflict are commonly seen as either opposed, or at least separate. It therefore seems nonsensical to talk of cooperative conflictful behavior (i.e., the noncoercive path). In the section on evidence I will try to straighten out this terminology. See also Section 4.6.
3. Unlike the coercive-force path, cooperation leads to accommodation. Force, however, may simply lead to a termination of the conflict without accommodations, such as the total victory of the Allies in World War II.
4. These are subphases in the balancing of power phase of the conflict helix. However, because the focus is on the dynamics of the balancing, I will simply refer to them as phases, leaving the "sub" understood.
5. A review of Chapter 12, especially of Figure 12.1 should be helpful at this point.
6. This refers to h2
7. Note especially that preparations, such as increasing armaments, is excluded from this action-reaction process.