Formal Alliances, 1816-1965: an Extension of the Basic Data*

FORMAL ALLIANCES, 1816-1965: AN EXTENSION OF THE BASIC DATA*

By MELVIN SMALL Wayne State University and J. DAVID SINGER University of Michigan

In an earlier number of this Journal, we able to others whose work embraces the published our findings on the distribu- two decades following Hiroshima and tion of formal alliances among the mem- Nagasaki.3 bers of the interstate system during the While much of the theoretical and me- period between the Napoleonic Wars and thodological discussion found in the World War II. Our purpose there was to earlier paper need not be repeated here, present a systematic and quantitative some of the latter problems are suffi- description of all formal alliances, their ciently different to merit brief attention; membership, duration, and type, as well this is particularly true of our data sour- as the procedures we used in generating ces and their reliability, to which we will those data (Singer and Small, 1966b). address ourselves at the outset. Following While our major motivation was to that, we will identify and justify the com- provide the empirical basis for a number position of the post-1945 interstate sys- of inquiries into the correlates of war tem and its major power sub-system, during that 125 years, it also seemed describe the three classes of alliance with likely that these data might be of use to which we are concerned, outline the others in the scholarly community. 1 coding and measuring procedures, and Since completing the original paper, then present our results in a variety of however, we have been under some forms. Throughout, we will compare our compulsion to extend our data beyond procedures and results with those of the World War II and up to the quite recent original study and note any deviations past, so that the period on which we therefrom; those who utilize these data concentrate is now the 150 years from are urged to note such deviations, espe- 1 January 1816 to 31 December 1965. cially as summarized in the Appendix. Some of the pressure has been self- induced and some has come via encour- The sources of information agement from the increasing number of For a great deal of diplomatic informa- scholars who are now engaged in data- tion, one may readily turn to the diplo- based, quantitative research in interna- matic archives of many national govern- tional politics.2 At the outset, we had ments and to the published volumes planned to restrict the entire project to which subsequently embrace and codify the 1816-1945 period, but for a variety of a large portion of those archives. But scientific as well as policy reasons we later this only holds true for materials which decided to extend it up through the mid- are at least two (and often, more) decades 1960’s. Given this set of considerations, in the past; few, if any, governments it now seems appropriate to up-date the make such documents available until earlier study and make our findings avail- twenty or more years after the fact. For 258 the earlier study, then, we had the docu- foreign consumption - as strictly de- mentary evidence to make us quite con- fensive moves, undertaken reluctantly in fident that all relevant alliances had in- the face of potential aggression. Fourth, deed been identified. But for the more in light of the consequences of America’s recent years, it looked as if we might be failure to make explicit its commitments in somewhat the same situation as was to South Korea before June of 1950, the Wilhelmstrasse in 1910; it was known there has since been a strong desire to for example, that some sort of undertak- reduce the ambiguities and uncertainties; ing existed between France and Britain, secrecy would not be useful in such a but the German Foreign Office could not context. Finally, as the material which be at all certain what the specific com- follows will make abundantly clear, it is mitments were. Similarly today, Western nearly impossible to think of any allian- scholars know that the U.S.S.R. and ces that have not already been consum- North Vietnam enjoy some sort of frater- mated - and publicized. In every part of nal relationship, for instance, but cannot the world, just about any alliance that ascertain whether a formal alliance was one could reasonably expect to be made contracted, no less ascertain the nature since 1945 has been made. of the obligations involved. And even if Thus, despite the unavailability of the we know that a formal alliance does standard archival sources, we are per- exist, and have identified it, we still may suaded that the present compilation wonder whether there are secret provi- includes virtually every single alliance sions which significantly alter the pub- which satisfies the criteria which are licly stated arrangements, and which may described below. In addition to the not become known until the archives are United Nations Treaty Series and the eventually opened. League of Nations Treaty Series, we have The picture is not, however, quite as turned to the governmental and second- bleak as it might appear. First of all, in ary sources cited in Table 3 and in the the period since World War I, and even References for the texts of the sixty-two more since World War II, the League and qualifying alliances extant during the the United Nations have maintained a post-World War II period.5 registry wherein all treaties, conventions, and agreements may be recorded by the Membership in the system signatory governments. While registra- It may be recalled that in the earlier tion is not compulsory, the consensus paper we differentiated between the to- is that a very large percentage of all post- tal interstate system and its more re- 1945 agreements have been deposited stricted sub-systems: that comprising with the Secretariat.4 This gives us, at the most of the European and a few of the least, a single and comprehensive source most important non-European states with which to begin. Second, with the (which we called the central system), and many changes in the culture of diploma- that comprising the major powers only. cy, its increasing visibility, and the height- Those states which did not qualify for ened role of ideological appeals and pro- inclusion in the central system were paganda moves, governments are less assigned to the peripheral system. The and less prone to undertake secret com- central peripheral distinction might have mitments. Third, and closely related, the been quite justified during the period initiators of most of the alliances of the 1816-1919, but by the end of World past two decades have been eager to War I, most of the independent nations present them - both for domestic and of the world were sufficiently interde- 259 pendent, and the primacy of Europe was is the judgment and consensus of the sufficiently ambiguous to permit the ter- historians who specialize in the diplomacy mination of that distinction as of 1920.6 of the period, and who, in turn, largely In this paper, therefore, the only two reflect the consensus of the practitioners. types of nations are those which qualify The other might be more objective cri- for the interstate system, and those five teria, such as military power, industrial which comprise the major power sub- capability, or diplomatic status. system after 1945. Fortunately enough, both sets of cri- The justification and a detailed de- teria produce essentially the same set of scription of our coding procedures will nations. Thus, for most of the nineteenth be found in Singer and Small (1966a) and and that part of the twentieth century Russett, Singer, and Small (1968), but embraced in the Correlates of War pro- they may be summarized here. Essentially ject, we find that those states which score any putatively sovereign state with a at or very near the top in military-indus- population of at least 500,000 was includ- trial capability and diplomatic status are ed, provided that it enjoyed the de facto the same ones assigned to the major diplomatic recognition of the two ’legit- power category by those whose research imizers’, France and Britain. This latter focuses on the several epochs and regions requirement was only used up through involved.8 Out of this consensus comes 1919, and since then the basic criterion the following. Going back to the pre- has been either: (a) membership in the World War I decades, we included: Eng- League or the United Nations, or (b) a land, France, Germany, Austria-Hun- population of 500,000 or more and rec- gary, Italy, Russia, Japan, and the Uni- ognition by any two major powers.7 Be- ted States. When the debris and chaos of cause the 1816-1945 period was marked that war were cleared away (by the mid- by the consolidation and redistribution 1920’s) the Hapsburgs were gone, but of empires and by many major wars, the the other seven remained in (or had re- composition of the interstate system un- turned to) the ranks of the major powers. derwent frequent shifts. The post-1945 In the wake of World War II, the ranks system, on the other hand, shows greater were further reduced, leaving in 1946 stability. While we do see an appreciable only the U.S.S.R. and the United States upsurge in system size due to the ’liquid- plus England and France; and with the ation of colonialism’, the only other consolidation of the Communist revo- change is the disappearance of two mem- lution and their creditable showing in the bers. One case is that of Syria, which Korean War, China entered this oligar- ’federated’ with Egypt to become a part chy for the first time. By 1950, then, the of the United Arab Republic from 1958 major powers were exactly those nations to 1961, and the other is Zanzibar, which which had been assigned special status achieved independence in 1963 but which (via the veto power) in the United joined with Tanganyika to form Tanzania Nations Security Council, and which in 1964. would soon also become the five nuclear As to the major powers - whom we powers. must identify in order to treat their Having summarized our criteria and alliance patterns separately later on in line of reasoning, we can now turn to the the paper - the problem is more compli- system membership compilations which cated in the recent past than it was earlier. emerged. In Table 1, then, we list those There would seem to be two sets of cri- states which comprised the total inter- teria here, regardless of time period; one state system during all or part of the pe- 260 riod 1946-65. They are listed by regional ceded 1946, no date is shown, and if location with their standardized code they did not remain in the system for the numbers to the left and their date of entry entire twenty years (Syria and Zanzibar) into the system shown to the right; if the dates of departure and/or return are their qualification for membership pre- shown as well.9

Table 1. Membership in the Interstate System, 1946-65 261

Table 1. Membership in the Interstate System, 1946-65 (cont.)

Coding the alliances The treaty obligations were ascertained With the spatial-temporal domain iden- by a literal reading of the texts, supple- tified, we can now turn to the alliances mented (if there were any verbal ambi- entered into by the members of the de- guities) by the interpretations of the fined system during the twenty years diplomatic historians. In other words, under study, or carried over from prior the classification is not sensitive to the years. It may be recalled that we defined political relations of the governments three different classes of alliance in the involved, nor to interpretations made by original study: defense pacts, neutrality other governments. Second, no indirect and non-aggression pacts, and ententes, alliance obligations were inferred via with the following distinguishing char- overlapping memberships. That is, even acteristics. In the defense pact (Class I), if nation B was allied with both A and C the signatories obligated themselves to via separate treaties, A and C were not intervene militarily on behalf of one treated as allies unless they were both another if either were attacked. In the also signatories to the same treaty of neutrality pact (Class II), the commit- alliance. Third, a variety of more general ment was to remain militarily neutral if commitments were not classified as al- the partner were attacked. And in the liances. Among those excluded were : (a) entente (Class III), the only obligation charters of global or quasi-global in- was to consult with, or cooperate, in such ternational organizations, such as the a military contingency. Treaties of friend- League, the United Nations, or their ship, etc. (which we do not include) specialized agencies; (b) treaties of guar- merely involve a more general promise of antee to which all relevant parties regis- mutual cordiality.lo tered their assent, such as the 1960 Greek- 262

Turkish guarantee of Cyprus; (c) con- other purposes should take careful note - ventions or agreements setting out gen- not included any alliances which were eral rules of state behavior, such as the consummated by nations while participa- Geneva Conventions; (d) ’mutual secu- ting in war or within three months prior rity’ arrangements which involve bases, to such participation, unless those al- financial aid, and training programs ex- liances emerged from the war intact. clusively, such as the Spanish-American Likewise, no alliances consummated Treaty; (e) unilateral and asymmetric during either of the World Wars were guarantees, such as the 1951 Japanese- included unless they, too/continued in American security treaty, in which only force during the post-war period.ll One one signatory is committed to defend the effect of this particular coding rule is to other. On the latter rule, we must reiter- make it unnecessary for us to cover the ate that we are concerned exclusively 1914-18 and 1939-45 periods, and this is with the commitments and resultant what accounts for the gap between the cross-pressures which bind two or more dates in the original paper and those used states to concert their policies in time of here. crisis. An alliance, in other words, must Turning, then, to the effective dates of contain at least two member states. any alliance, the beginning date was a Without this distinction, any pronounce- relatively simple matter. Even though ment which declared that one nation some months may pass between the nec- would protect the territory of another essary signatures and ratification, the nation would have to be considered as an former date is always used; if, however, alliance. For example, in 1951, Egypt the treaty failed of ratification (such as denounced her 1936 alliance with Eng- EDC) it is of course not included at all. land, but England refused to accept this As to termination dates, the problem - unfriendly gesture and maintained that especially in more recent years - is more the alliance was still in force. Obviously, complex. That is, with the decreasing after 1951, the 1936 alliance became a incidence of formal (or even informal) unilateral (and unwanted) guarantee of abrogations or denunciations of alliances, the territory of Egypt by England, the termination of a treaty whose text signifying something quite different does not specify an expiration date can from reciprocal obligation and cooper- become difficult to pin-point. Thus, we ation. have in several cases had to make a po- Let us shift now from the nature of the litical judgment as to the year in which alliance commitments which concern us the obligations were no longer effectively here to the problem of identifying the binding on one, several, or all of the sig- span of time during which they are in natories. A good example might be force. In this connection, a preliminary Yugoslavia’s leaving the Cominform in point is in order, clarifying the connection 1948. Even though not all of the states between the data presented here and which had joined in the Soviet bloc’s alli- those shown in the original paper. Our ance system (via the several bilateral major purpose in gathering alliance data treaties of 1945 and 1946) formally abro- is to ascertain the extent to which the gated their commitments, it seems evident resulting clusters and configurations cor- that neither the Yugoslav nor the other relate with the onset of war in the years governments considered themselves following each set of observations of such bound after the Tito regime’s expulsion. alliance distributions. We have, there- So that the user will know the basic fore - and those who use our data for reason for the termination dates assigned 263

Table 2. Alliances which terminated during 1946-65

to those twelve of the sixty-nine qualify- entente would not be included in our ing alliances which did become, in our compilations, even though it followed the judgment, ineffective during the 1946-65 defense pact in the chronological sense. period, we indicate them briefly in Table A final question arises from cases in 2. which a number of bilateral treaties of a Another problem of a chronological given class and national membership is nature is that of new alliance agreements followed (or preceded) by a multilateral which are undertaken between and one of the same class and membership. among governments which were already In such cases, the multilateral alliance allied. The question here is one of deter- takes priority and the bilateral ones are mining which commitment takes pre- dropped from our computations. Typi- cedence, since we are concerned not cal of these would be the post-World merely with whether or not certain states War II bilateral arrangements which were are allied, but ascertaining the nature or superseded by the Warsaw and NATO class of that commitment. We begin with pacts. It should be stressed here that the assumption that - in terms of the these coding rules are not meant to imply obligations undertaken - a defense pact that the superseded alliance is considered imposes greater commitments than a neu- to be terminated and no longer in effect. trality or non-aggression pact, and that They serve only to make our indicators of each of these imposes a greater commit- alliance aggregation and alliance comment than an entente. Therefore, when- mitment, as discussed below, more con- ever any two or more signatories to a sonant with the empirical realities which treaty with lower level obligations subse- they are meant to measure, and hence quently join in one with greater obliga- more valid. tions - and this says nothing about the The results of these coding procedures probability of such obligations being ful- and classification criteria are shown in filled - the latter takes precedence and Table 3. In addition to the names of the former is no longer included for the signatory states, we show the dates computational purposes. Conversely, if of inception and termination, the alliance an entente were consummated between or class (defense, neutrality, or entente) and among states which were already mem- the place in which its text may most bers of a defense pact, for example, the conveniently be found. 264

Table 3. Interstate Alliances in Force, 1946-65

NOTES 1 Classes of Alliance are : 1-Defense Pact; 2-Neutrality or Non-Aggression Pact; 3-Entente. 2 Asterisk (*) following termination date indicates that the alliance was superseded by another arrangement. 3 Parentheses around a year indicate that it applies only to the state alongside which it appears. 265 266 267 268 269 270

Table 3. (cont.)

Before turning to the conversion pro- teenth century alliances, that number cedures by which these raw data are rose sharply to 37 (or 48 per cent) during made more useful for research purposes, the 1900-39 period. But the number it might be helpful to present some simple dropped sharply for the post-World War summaries. The most general summary, II period, with the seven neutrality pacts found in Table 4, shows the frequency accounting for only 9 per cent of all the distribution of those alliances which were alliances in force at any time during in force, according to our criteria, among those twenty years. The non-aggression the states which constituted the system pact, which is one variation of the tradi- during all or part of the twenty years tional neutrality pact, was clearly an which concern us here. It should be noted invention of the 1920’s and 1930’s, and that the total number of alliances shown if the amount of war which followed is here (80) comes to more than the 69 any indication, they were not particularly alliances actually in effect, since several effective. Given that experience, it is little of the multilateral ones link not only wonder that only four such alliances were major powers with non-majors (minors), consummated after World War II and all but majors with majors, and minors with of these involved China; the other three minors, thus falling into more than one were signed during the heyday of the of the rows. ’non-aggression’ era between the two world wars. Table 4. Distribution Alliances of It also be noted that the by C’lass and Signatories, 1946-65 might percentage of ententes - a modest consul- tative obligation - remained constant after the ceremonies in Tokyo Bay. Ententes accounted for 23 per cent of the nineteenth century alliances and 22 per cent and 23 per cent respectively, for both If we may be permitted one interpretive twentieth century periods; the latter fig- comment here, it is worth noting how few ures somewhat overstate their import- neutrality and/or non-aggression pacts ance, since all but a few of those since are found during these two recent 1945 were consummated among the mi- decades. While there were only four such nor non-Western states. Be that as it may, arrangements during the 1816-99 period, given the very low frequency of the class accounting for 11 per cent of the nine- II (non-aggression) alliances, we have, 271 for the aggregate computational purposes In addition to the computation pro- outlined in the next section, combined cedures, there are two specific coding them with those of class III. rules worth reiterating. First, any pair of An alternate way of summarizing the states may have more than one alliance data is to shift from the number of alli- commitment in force at any given time. ances (and alliance bonds) to the number But, second, we only count the strongest of national alliance commitments, count- or most dominant bond which any state ing each individual nation-to-nation com- has vis-A-vis any other. Defense pacts mitment. Applying the formula n(n-1) to take precedence over neutrality pacts, the 21-nation Rio Pact, for example, we and these take precedence over ententes, get 420 national commitments. The fre- and if A is in both a defense pact and an quencies resulting from this set of compu- entente with B, the latter bond is not in- tations are found in Table 5, as follows: cluded in the computation. To illustrate, then, if the Alliance Commitment Indi- Table 5. Distribution of Nation-to-Nation cator (ACI) for a given year is 7.12 (as in Alliance Commitments by Class and Signatories, 1946 for all classes of commitments 1946-65 among all states in the system) there was an average of 7.12 alliance commitments per state; a figure of 1.00 indicates an average of one such commitment per state, but tells us of course nothing about the concentration or dispersion of such commitments,.13 One virtue of both sets Annual alliance indicators of indicators is that they are normalized Returning once more to the presentation for system size, thus permitting com- of our data, there is the problem to which parisons across time. we alluded earlier: how can the raw Table 6, then, is divided into two parts, alliance figures be converted into a form with the Alliance Aggregation scores on which is useful for correlational analysis the left and the Alliance Commitment over time? That is, if our concern is to scores on the right. After showing, for ascertain the extent to which alliance each of the twenty years, the number of patterns predict to, and correlate with, states in the system and the number in fluctuations in the incidence of war, or the major power sub-system, we present any other types of event, the raw data four separate indicators of Alliance Ag- must be converted into a variety of annual gregation : the percentage of the system’s indicators. members who are in one or more allian- As in the original paper, we suggest ces of any class; the percentage in defense two different measures of this particular pacts only; the percentage of major pow- structural attribute of the interstate ers in any alliance; and the percentage system. One, called Alliance Aggregation, of majors in defense pacts only. On the reflects the percentage of states of a given right hand side, under Alliance Commit- type which belong to one or more allian- ment, we show the following ratios be- ces of any given class in each successive tween the number of national commit- year. The other, called Alliance Commit- ments and the system or sub-system size: ment, is a bit more complex, and reflects the number of commitments of any class the number of nation-to-nation com- per member of the total system; number mitments per system member for each of defense pact commitments of any class year. 12 by major powers (regardless of partner’s 272 273

status) per major; and number of major fore doing so, it might be useful to sum- power defense pact commitments per marize (verbally and statistically) the al- major. liance patterns of this more recent period Having computed these ten indicators by itself. The post-World War II scene for each of the twenty years, we and oth- divides rather naturally into three sub- ers may next wonder as to their utility periods. The first of these, extending up for analytical purposes. As independent, through 1951, saw not only the liquida- intervening, or dependent variables, we tion of the most severe war in human may find one or more of these several history and the establishment of a collec- measures useful, depending on the sys- tive security system which might prevent temicfocusandtheoreticalinquiryathand. another such holocaust, but the creation But for more general purposes, it might of an unprecedented number of alliances. be useful to have either a single combined While such ’collective defense’ treaties index, or to select one of the indices as were explicitly permitted by the United generally representative. The first could Nations Charter, the speed with which be generated by a variety of techniques, they were formed could only cast doubt among which the ’construct mapping’ on the expressions of confidence which version of factor analysis (Jones, 1966) accompanied the birth of the world or- would seem particularly appropriate. We ganization. By 1947, 71 per cent of the do not offer such combined measures nations in the interstate system were in here, but do suggest a basis for the second one or more alliances of one class or strategy. That is, if we find that the scores another, and 54 per cent of them were of all or most of the separate measures in the more concrete defense pacts; show a high correlation vis-h-vis one moreover, by the next year, 100 per cent another, one may then be justified in of the major powers were already com- using any one of them for certain purposes. mitted to defense pacts, and when this Whereas the picture for much of the flurry of alliance-making came to an end 1816-1939 period was a rather confusing in 1951, three-quarters of all the system’s and erratic one, the post-World War II members were allied, as were all of the pattern is remarkably clear. Even a major powers. cursory visual inspection of Table 6 The second period, extending from reveals that any rank-order correlation 1951 through 1959, was exceptionally would be extremely high; whether one stable in terms of alliance aggregation compares across alliance classes, nation and alliance commitment scores, with types, or alternative indices, the years no appreciable movement into, out of, would fall into essentially the same rank- or between, alliance blocs, even though ing. Likewise, if we treated our data in the size of the system rose from 75 to 89. interval scale fashion, any of the appro- This is not to say, however, that no new priate correlation coefficients would turn treaties of alliance were consummated. out to be remarkably high. Thus we do That decade saw the establishment of not include here any of the correlation CENTO, SEATO, and the Warsaw Pact, matrices which were quite necessary in and these certainly helped to further the earlier paper. institutionalize the cold war cleavage. On the other hand, most of the alliance Summary and speculation bonds represented in these three defense In this final section we want to compare pacts had already been established, al- the period under review here with that beit sometimes at the entente level, via embraced in the original inquiry, but be- prior bilateral treaties. Hence - and this 274 is precisely why our measures do not pacts, none of the system members which suffice for all theoretical purposes - began the cold war as announced neu- there was no appreciable increase in the trals defected from that position to join several indices during the 1950’s. the major power blocs. At one point A third period, from 1960 to the close (the Bandung conference of 1955) there of our study in 1965, was characterized was some discussion of an alliance of the by a momentary decline in both alliance nonaligned nations to formalize that aggregation and alliance commitment state of affairs, but it was not considered scores (largely as a result of the influx of necessary enough to justify the costs and new states) followed by a rapid rise in obligations which might be involved.15 these indicators to new heights. By 1965, In Latin America, Africa, and the the year we close our study, 81 per cent Middle East, likewise, the picture showed of the states in the system were allied, and little change over these two decades. The the Alliance Commitment Indicator former region’s members revived their showed an average of 16.8 alliances for ev- pre-war regional bloc affiliated with the ery member in the system. During the seven U.S. in the form of the Organization of years from 1959 to 1965 the system in- American States, and only Cuba failed creased in size from 89 to 124, leading to remain in it during the entire period. to a much larger denominator in our In Africa, despite the pro-Soviet inclin- ratio, but the several new alliances ation of the ’Casablanca group’, the (largely African) led to a comparable Organization of African Unity insti- increase in the numerator. tutionalized the neutrality (in the cold Turning from the sheer magnitude of war context at least) of almost every our ten different indices of aggregation state in the region. In the north, the and commitment, another striking ele- Arab League embraced all of the Moslem ment is the relatively ’natural’ as well as nations of North Africa and the Middle stable pattern which developed. That is, East. Thus, from the Dardanelles to the once the ’cold war’ confrontation became Cape, every system member was in one apparent, all but five of the European alliance or another, except for Israel and members of the system,14 and a good South Africa.16 many in the other regions, had cast their The above patterns, while they held lots with either the American or the for most of the period under review, do Soviet bloc. In Asia, two relatively estab- not tell the entire story. As we urged in lished states - Thailand and the Philip- the earlier paper and elsewhere in this pines - joined with two newly indepen- one, formal written alliances offer only dent ones - Pakistan and Malaysia - to one index of the system’s basic config- link up with SEATO. On top of this, urations. A more complete picture of the Japan, Taiwan, and South Korea had system at any point in time must certain- bilateral commitments with the U.S., ly take account of political alignments which was the major architect of SEATO. and predispositions which stem from On the opposing side, for much of these strategic, geographic, economic, and ide- two decades, the Asian states of China, ological factors. And while all of these North Korea, and Mongolia were linked factors do exercise some impact on the formally to the Soviet Union, while decisions which lead to formal alliance, North Vietnam was an informal member they do not all necessarily produce the of this anti-Western configuration. Equal- same alliance configurations. ly interesting in this regard is the fact As a matter of fact, one of the work- that, aside from two non-aggression ing - but not yet tested - assumptions 275 of this project is that the peacefulness of and out of alliances became increasingly the system depends very much on the inhibited, and what had formerly been existence of strong cross-pressures among thought of as rational diplomacy and states, varying as to which sectors of Realpolitik became a matter of perfidy activity and concern are involved. In and condemnation. The question, then, other words, we posit that high alliance is whether - assuming that this model is aggregation and commitment scores need an accurate reflection of reality - the not necessarily make the system more slight movement toward a loosening of war-prone by and of themselves. But if the cold war alliance bonds in the 1960’s such conditions are accompanied by will help make ours a more stable system. configurations in which many of the After all, the only other times in which states in the system divide up into two the alliance indices stood at levels even opposing blocs whose composition is approximately as high were 1912-14 and constant across a wide range of issues, 1937-39, and the consequences then were then we would expect the salutary effects disastrous. of the ’invisible hand’ to be seriously inhibited. With the pluralistic, cross- A. cutting bonds thus weakened, we hypo- Appendix Substantive Modifications of’ thesize that war becomes much more Original Data probable.17 In any enterprise of this kind, the re- This consideration leads, then, to our searcher is bound to discover new or concluding comments. We found that conflicting facts as the project unfolds high alliance aggregation scores in the and as comments come in from others in nineteenth century did not precede, or the field. This has certainly been our ex- predict to, increases in the incidence of perience in the Correlates of War pro- war. On the contrary, the most peaceful ject, and as a consequence, the following periods in the 1816-99 period were substantive modifications in our data largely those which were preceded by the have become desirable. First, these dates highest levels of alliance aggregation and of qualification for system membership commitment. In the twentieth century, have been changed: Cuba, from 1934 to however, quite the reverse obtained, with 1902; Hungary, from 1920 to 1919; high alliance levels predicting all too Czechoslovakia, from 1919 to 1918; Esto- regularly to sharp increases in the fre- nia, Latvia, and Lithuania, from 1920 to quency, magnitude, and severity of war 1918, and Yemen, from 1934 to 1926. (Singer & Small, 1968). Second, we are now persuaded that the One plausible explanation might be that Anglo-Portuguese defense pact of 1899 nineteenth century alliances were largely should be coded as surviving World War ’affairs of convenience’ rather than ’mar- I, and terminating only with the NATO riages of passion’, to reverse the conven- treaty (which supersedes it) rather than in tional idiom. That is, only as the tra- 1914, as originally coded. Third, new evi- dition of quiet diplomacy among cul- dence suggests that the 1933 treaty be- turally similar elites gave way to the tween Finland and eight Latin American welfare state, rising public and partisan states did not satisfy our criteria and involvement in foreign policy, and ex- should not be classified as an alliance in tensive use of psychological mobilization the sense used here; it is therefore ex- techniques, did alliance bonds become cluded from our revised compilation. increasingly inflexible and dysfunctional. And last, we had originally ignored a trea- Under such conditions, movement into ty partner, and also overrated the classifi- 276 cation of the Saudi-Arabian-Yemeni alli- As a further aid to readers who are ance of 1937. It should now be coded as interested in alliance patterns since 1816, an Iraqi-Saudi entente of 4/36, to which we offer again our basic listing (with the Yemen adhered in 4/37. Those who utilize above modifications) from the original our data decks will find these modifica- article. Sources and the distinctions tions already made, but those who are between central and total system mem- working from the article itself should note bers have been eliminated for the sake the changes. of simplicity.

Table 7. Inter-Nation Alliances, 1$16-1 ~45 with Commitment Class and Dates 277 278 279

Table 7. (cont.)

NOTES * We are indebted to a number of scholars who are using, and have commented upon, the earlier study of which this is a continuation. In addition to those identified in the References we would like to particularly thank Bruce Russett, who has helped considerably in our data acquisition, and has gone over this manuscript in detail in the course of our collaboration on the role of alliances in the international system. Since his theoretical concerns are somewhat different from ours, there will be appreciable disparities between our data and those which he will be reporting; see Russett (1968). 280

1 In addition to several master and doctoral dissertations which have utilized those materials, and some employment for teaching and simulation purposes, there are several other papers based on these data; see for example, Haas (1968) and Zinnes (1967). We have ourselves published two studies based on them; see Singer & Small (1967 and 1968). 2 Much of this work will be found in such journals as the present one, The Journal of Conflict Resolution, and Peace Research Society Papers, and a representative sampling is available in Quantitative International Politics (Singer, 1968). 3 The entire data deck may be had at nominal cost from the authors, or from Raymond Tanter, Director of International Relations Archive, ICPR, University of Michigan. 4 On the basis of his UN Treaty Series project, however, Rohn (1968, p. 177) concludes ’no government had ever checked whether all their treaties actually appeared in the UNTS,’ and that no hard evidence as to its completeness yet exists. In one such inquiry he found a 23 per cent gap between ’Canada’s own published treaty records and Canada’s treaties in the UNTS’ (1966, p. 116). 5 We have not had a chance to consult Treaties and Alliances of the World (1969), a new volume which may prove to be useful. 6 A small terminological change was also made. In order to differentiate between independent national entities which had all the earmarks of sovereignty and thus qualified for inclusion, and those which lacked one or more critical attributes of statehood, we now include both sets of nations in the international system, but only include the former in the more restricted interstate system. 7 There are a few minor exceptions: India, despite League membership was not included until 1947, and Byelorussia and the Ukraine have never been included despite UN membership. The alternative rule is necessary because several important states are not UN members: Switzerland and the two Germanies, Koreas, and Vietnams. 8 For diplomatic status data and rankings, see Singer & Small (1966a) and Singer, Handley, & Small (1969); for military-industrial data and rankings, see Singeret al. (1969). 9 The nation code numbers, which have been adopted by a number of projects other than those at Yale and Michigan, and by the Inter-University Consortium for Political Research, are presented in Russett, Singer, & Small (1968). 10 The designations Class I, II, and III suggest a hierarchy based upon levels of political commitment, with the defense pact a more serious commitment than the neutrality pact, and the neutrality pact a more serious commitment than the entente. While a Class I alliance obviously is more serious than a Class II or a Class III, a Class III may be more serious than a Class II. In the nineteenth century, a neutrality pact was generally a more serious commitment than an entente. In the twentieth century, however, the entente seems to be a more serious commitment than the non-aggression pact. 11 While we are not immediately concerned with such alliances, we are planning to gather data on them in the near future. Aside from alliances contracted during the two wars, this will most likely involve fewer than 10 alliances. 12 In the original study, we used the concept of Alliance Involvement, measuring dyadic bonds via the equation n(-1)/2. We believe that a more valid measure is the total number of nation-to- nation alliance commitments existing at any given time; thus, all such commitments, not only the dominant ones, are counted, and for any given alliance the number of commitments is n(-1). 13 Such concentration might be computed by use of the Gini index, for example, which reflects what fraction of the system’s members account for what fraction of the commitments in force. It should also be noted that whereas the decimal point was omitted from Tables 5 and 6 of the original paper, we decided that the Alliance Commitment score would be more meaningful if we did include it here. The other difference, as indicated earlier, is that we have not computed either the Alliance Aggregation or Alliance Commitment Indicators for neutrality and entente agreements separately, since there were so few of either in this post-1945 period. 14 These were the three traditionally non-aligned states (Sweden, Switzerland, and Ireland), Spain, whose exclusion from NATO was largely in deference to anti-facist views in some of the Western nations, and Yugoslavia, which left the Soviet bloc in 1948. 15 A provocative hypothesis regarding the costs and gains of alliance membership, and the coalition building strategies which might be expected to result, is in Riker (1962). That hypothesis, based largely on domestic political systems, is now being tested for the international system; see Singer and Bueno de Mesquita (1969). A more general model of the factors that go into alliance formation is in Russett (1968). 281

16 Another might be Rhodesia, but its failure to achieve any substantial diplomatic recognition after the unilateral declaration of independence in 1965, leaves it outside of our interstate system. 17 The reasoning behind this classical balance of power argument is summarized and partially operationalized in Deutsch & Singer (1964) and partially supported in Singer & Small (1968). For a critical re-analysis of our data, see Zinnes (1967); other discussions of the issue are Gulick (1955), Liska (1962), Rothstein (1968) and Waltz (1964). A suggestive alternative model is in Galtung (1964).

REFERENCES [1.] Benes, Eduard, Memoirs, London: G. Allen and Unwin, 1954. [2.] British Documents on the Origin of the War, I, London, 1927. [BD] [3.] British Foreign and State Papers. [BFS] [4.] Deutsch, Karl W., and J. David Singer, ’Multipolar Power Systems and International Sta- bility,’World Politics, 16 April 1964, 390-406. [5.] Documents on International Affairs. [DIA] [6.] Galtung, Johan, ’A Structural Theory of Aggression,’ Journal of Peace Research, 2, 1964, 95-119. [7.] Gantenbein, James (ed.), The Evolution of Our Latin American Policy, New York: Columbia University Press, 1950. [8.] Gulick, Edward V., Europe’s Classical Balance of Power, Ithaca: Cornell University Press, 1955. [9.] Haas, Michael, ’International Subsystems: Stability and Polarity,’ paper presented at APSA Meetings, Washington, D.C., 1968. [10.] Jones, Ronald, Construct Mapping, Kansas City, Mo.: University of Missouri, mimeographed, June 1966. [11.] Keesing’s Contemporary Archives, 1963-64. [K] [12.] Lawson, Ruth, International Regional Organizations, New York: Praeger, 1962. [13.] League of Nations, Treaty Series. [L] [14.] Legum, Colin, Pan Africanism, New York: Praeger, 1965. [15.] Liska, George, Nations in Alliance, Baltimore, Md.: Johns Hopkins Press, 1962. [16.] Rice, Berkeley, Enter Gambia, Boston: Houghton-Mifflin, 1967. [17.] Riker, William, Theory of Political Coalitions, New Haven, Conn.: Yale University Press, 1962. [18.] Rohn, Peter, ’Canada in the United Nations Treaty Series,’ Canadian Yearbook of Inter- national Law, 1966, 102-30. [19.] - ’The United Nations Treaty Series Project,’ International Studies Quarterly, 12(2), June 1968, 174-95. [20.] Rothstein, Robert L., Alliances and Small Powers, New York: Columbia University Press, 1968. [21.] Russett, Bruce, ’Components of an Operational Theory of International Alliance For- mation,’ Journal of Conflict Resolution, XII, No. 3, Sept. 1968, 285-301. [22.] Russett, Bruce, J. David Singer, and Melvin Small, ’National Political Units in the Twen- tieth Century,’ American Political Science Review, September 1968, LXII, No. 3, 932-51. [23.] Singer, J. David (ed.), Quantitative International Politics: Insights and Evidence, New York: Free Press, 1968. [24.] Singer, J. David, ’Modern International War: From Conjecture to Explanation,’ to appear in Albert Lepawsky, Edward Buehrig and Harold Lasswell (eds.), Essays in Honor of Quincy Wright, New York: Appleton-Century-Crofts, 1970 [forthcoming]. [25.] Singer, J. David and Melvin Small, ’The Composition and Status Ordering of the Interna- tional System, 1815-1940,’ World Politics, 18/2, January 1966a, pp. 236-82. [26.] Singer, J. David and Melvin Small, ’Formal Alliances, 1815-1939: A Quantitative Descrip- tion,’ Journal of Peace Research, 3/1, January 1966b, 1-32. [27.] Singer, J. David and Melvin Small, ’Alliance Aggregation and the Onset of War, 1815-1945,’ in Singer (ed.), Quantitative International Politics: Insights and Evidence, New York: The Free Press, 1968, 247-286. [28.] Singer, J. David and Melvin Small, ’National Alliance Commitments and War Involve- ment, 1815-1945,’ Peace Research Society, Papers, VI, 1967, 110-40. 282

[29.] Singer, J. Davidet al. ’The Military-Industrial Capability of Nations, 1816-1965: A Quanti- tative Assessment,’ Ann Arbor, Michigan, MHRI Prepint, (1969 forthcoming). [30.] Singer, J. David, David Handley, and Melvin Small, ’The Diplomatic Importance of States, 1816-1965: An Extension of the Basic Data,’ Ann Arbor, Michigan; MHRI Preprint (1969a forthcoming). [31.] Singer, J. David and Bruce Bueno de Mesquita, ’The Minimum Winning Coalition in In- ternational Politics: A Test of the Hypothesis, 1816-1965,’ Ann Arbor, Michigan: MHRI Preprint (1969b forthcoming). [32.] Treaties and Alliances of the World, New York: Scribners, 1969. [33.] United Nations, Treaty Series. [U] [34.] United States State Department. Bulletin [DSB] [35]. United States State Department. Documents and State Papers. [SDD] [36.] Waltz, Kenneth, ’The Stability of a Bipolar World,’ Daedalus, XCIII, 1964, 881-909. [37.] Zinnes, Dina A. ’An Analytical Study of the Balance of Power Theories,’ Journal of Peace Research, 3(1967), 270-88.