International Integration
JAMES A. CAPORASO
228
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 229
answer the important questions facing us. 3 1 have already indicated how com-
plementary Lindberg's and Nye's suggestions are, and I feel that implementa-
tion of their suggestions would aid in making research findings touch base
with one another and thus impart to evidence its proper multiplying effect.4
occurred since the EEC came into existence, they argue that these increases
are expected "from mere random probability." They utilize die RA index
developed by Deutsch and I. Richard Savage to take account of these abso-
lute increases and to derive scores which express levels of integration in terms
of departures from a null model.8
Research on mass and elite attitudes toward integration has been less con-
troversial. Jacques-Rene Rabier, director of die Press and Information Service
of die European Communities, has presented evidence from opinion polls
conducted during 1952 and 1962. The figures for 1962 show marked increases
in favorable attitudes toward Europe.9 A more recent study, suggested by
the Press and Information Service, but, significantly, carried out under odier
professional auspices, comes up widi extremely favorable reactions toward die
formation of a "United States of Europe." Results ranged from a low of 60
percent favorable in Belgium and Italy to a high of 75 percent in Luxem-
bourg. These figures become even more astonishing if one discounts the large
number^of people who did not respond.10 Similarly, Ronald Inglehart, pre-
senting survey data on the attitudes of die younger generation in France,
the Federal Republic of Germany (West Germany), die Netherlands, and die
United Kingdom, has come to die conclusion diat "European integration
may have moved into full gear only since 1958."11
That mass attitudes toward past activities of the EEC and future possibili-
ties for federation are favorable does not seem to be disputed. However, when
one attempts either to assess die influence of odier variables (such as die crea-
tion and operation of die EEC) on attitudes or, conversely, to assess the im-
pact of changing attitudinal patterns on transactions or political institutions,
many difficulties emerge. Donald Puchala has attempted to show diat, al-
diough diere is a generally high level of support for die EEC, this
support is not generalized into support for political union.12 Using survey
data collected for 1957 and 1962, Puchala has observed diat "increasing ap-
proval for the Common Market did not generally coincide widi increasing
support for political federation on die Continent between 1957 and 1962 "13
8
For the details of working out the RA index, in addition to some of its mathematical properties,
see I. Richard Savage and Karl W. Deutsch, "A Statistical Model o£ the Gross Analysis of Transaction
Flows," Econometrica, July i960 (Vol. 28, No. 3), pp. 551-572.
8
Jacques-Rene Rabier, L'Injormation des europeens et I'integration de I'Europe (No. 10) (Brussels:
Instirut d'etudes europeennes, Universite libre de Bruxelles, February 1965), p. 38.
1<!
Les Europeens: "Out" a I'Europe, Resultats commented d'un sondage d'opinion realise en janvier-
fevrier 1970 dans les six pays de la Communaute europeenne et en Grande-Bretagne (Brussels: Direction
g£n£rale de la presse et de Pinformation, Commission des Communautes europiennes, May 1970), pp.
4 and 11.
11
Ronald Inglehart, "An End to European Integration?" American Political Science Review, March
1967 (Vol. 61, No. 1), p. 91.
12
Donald J. Puchala, "The Common Market and Political Federation in Western European Opinion,"
International Studies Quarterly, March 1970 (Vol. 14, No. 1), p. 32. Puchala's data applies only to
France, West Germany, and Italy.
13
Ibid., p. 53.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 231
This has led him to conclude that, attitudinally at least, economic integration
and political federation were quite distinct processes. Similarly, Roger Cobb
and Charles Elder have noted the general noncongruence of indicators of
integration and call into question the relations between transactions and mass
perceptions."
While most observers concede progress in economic integration, the picture
of political integration is much less optimistic. The defeat of British applica-
tion for entry into the EEC in 1963, the agricultural crisis of 1965-1966 and
the subsequent Luxembourg agreements, Charles de Gaulle's second "no" to
the United Kingdom in 1967, the increased importance of the Committee of
Permanent Representatives in the decisionmaking process,15 and the generally
attenuated role of the Commission of the European Communities all point to
this conclusion.
The optimism which argued that an unobtrusive series of limited steps would
eventually lead to political union has been eclipsed by a new pessimism which
fears that past progress toward integration may have brought the EEC to a
new equilibrium point far removed from both the model of the jealous nation-
state and the authoritative supranational organization. In short, by success-
fully segregating economic rewards from cultural and political preferences
the new community system has now made it possible for Europeans to enjoy
the fruits of a large market and customs union while at the same time sacri-
ficing neither cultural identity nor political autonomy. Discussion of the com-
mission as an embryonic federal executive has all but ceased and scholars in-
creasingly view it in its administrative capacity. In a provocative analysis of
the administrative-political role of the EEC commission David Coombes has
come to the conclusion that the commission's role has evolved toward that
of a classic bureaucracy.16 To state the pessimism more strongly, the fear is
that the EEC will become little more than an intergovernmental secretariat
dedicated to implementing the details of the common agricultural policy.
This picture contrasts sharply with the conclusions reached by William
Fisher and Carl Friedrich. Both scholars, working independently and em-
ploying very different techniques, have come to similar conclusions. Fisher,
defining political integration as the growth of the decisionmaking capacity
14
Roger W. Cobb and Charles Elder, International Community: A Regional and Global Study (New
York: Holt, Rinehart & Winston, 1970), p. 138.
15
Although the Committee of Permanent Representatives is not specifically mentioned in the Treaty
Establishing the European Economic Community (Rome Treaty), it has come to occupy a central role
in the community decisionmaking machinery. It is a delegate body for the Council of Ministers, im-
parting day-to-day administrative continuity to the irregular council sessions and preparing the work for
its meetings. It also acts as a clearinghouse for commission proposals and is in this sense akin to an
"early warning system," alerting the nationally minded council to any possible supranational maneuvers
on the part of the commission.
16
David Coombes, Politics and Bureaucracy in the European Community: A Portrait of the Commis-
sion of the E.E.C. (Beverly Hills, Calif: Sage Publications [in cooperation with Political and Economic
Planning], 1970), p. 327.
232 INTERNATIONAL ORGANIZATION
in which
- i < RAij < oo .
Aij refers to the actual trade between two countries and Eu refers to their
expected trade. By allowing expected trade to enter into the computation of
the RA index both absolute trade levels and trends are controlled. (Trends
are viewed as increases or decreases in the mean level of world or regional
trade.) 21
The RA index is interesting to us because its use is tied to a particular sub-
stantive controversy with respect to European integration. On the basis of RA
scores Karl Deutsch and his associates have presented evidence to show that
European integration has come to a halt:
European integration has slowed since the mid-1950's, and it has stopped or
reached a plateau since 1957-58. In the 1957-58 period, Europe reached the
highest level of structural integration that it has ever had.22
The authors add that:
There are, to be sure, absolute increases after 1958 in trade, travel, postal cor-
respondence and the exchange of students, but all this can be accounted for by
the effects of prosperity and the general increase in the level of these activities
in the 1950's and early 1960's. There have been no increases in integration in
regard to all these transactions beyond what one would expect from mere ran-
dom probability and increase in prosperity in the countries concerned.23
These findings concerning the leveling-off of European integration since
1957-1958 have been criticized by Ronald Inglehart as well as others.24 Ingle-
hart has noted that "far from finding a stagnation of integrative processes
since 1958, I would argue that, in some respects, European integration may
have moved into full gear only since 1958."25 Lindberg has noted that he and
Ernst Haas have similarly argued that "it is since 1957 that integration has
made its greatest strides."26
This disagreement is not in itself disturbing. Conflict over theory may
lead to "crucial tests" which may produce change, cognitive reshuffling, and
theoretical refinement. For the most part, however, parties to the controversy
have failed to meet one another on common ground. Inglehart has made a
21
Paul Smoker, "A Time Series Analysis of Sino-Indian Relations" (Paper presented at the Second
International Peace Research Association Conference, Tallberg, Sweden, June 17—19, 1967), p . 3.
22
Deutsch et al., p. 218.
23
Ibid., p. 219.
24
Inglehart, American Political Science Review, Vol. 6 1 , No. 1, p . 9 1 . See also Fisher, International
Organization, Vol. 23, No. 2, p. 273.
26
Inglehart, American Political Science Review, Vol. 6 1 , No. 1, p . 9 1 .
26
Lindberg, Journal of Common Market Studies, Vol. 5, No. 4, p . 344.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 235
serious attempt to deal with the RA index but he has criticized so many of
Deutsch's findings in terms of his own political socialization data that it is
impossible to identify the locus of the disagreement. Do the differences arise
from the respective variations in data (socialization data versus transactional
data), from different scoring procedures (the RA index versus simple per-
centage data), or purely from differences in interpretation ?
This controversy highlights four important problems of general interest for
the social sciences. The first problem is the lack of criteria by which to assess
evidence. In the absence of such criteria it is impossible to evaluate opposing
arguments and, likewise, it becomes difficult to assess progress toward the
solution of the problem. Both Inglehart and Deutsch present persuasive cases:
Deutsch says we must control for size; Inglehart says we must not penalize
nations just because they account for a large proportion of world trade. While
these arguments appeal to our sense of plausibility and reasonableness, they
are not necessarily desirable in terms of theory construction. In addition, we
do not assess the arguments in terms of identical and explicit standards.
The second problem arises because the controversy is carried on outside of
a common mutually accepted theoretical framework. The underlying assump-
tion of Deutsch's model is one of scarcity. There is a certain amount of inte-
gration to be divided among various countries. The RA index is based on a
"share of the pie" model. The total size of the pie does not increase; only the
allocations to individual countries change. Inglehart's notion of integration
seems to be based on an expanding pie model.27 Not only may the respective
portions of the pie change but the total size may also increase or decrease.
Adopting Inglehart's assumptions it is possible to imagine four or five regions
of the world becoming more integrated whereas this is very improbable with
the RA index.28
27
Sec Ronald Inglehart and Robert Schoenbcrgcr, "Communications and Political Mobilization: De-
velopment o£ a European Orientation in Great Britain, France, and Italy" (Paper delivered at the an-
nual American Political Science Association Convention, Washington, 1968), especially pp. 2-6.
28
The assumption here is that the world is composed of only four or five regions. The RA index
technically is not based on perfect zero-sum assumptions in the sense that increases in integration be-
tween any two actors must be canceled out elsewhere in the system. However, it is based on assumptions
that may be described as "highly competitive." For example, imagine a trading system of three actors
(A, B, and C). Imagine that A's RA score with B increases. It is technically possible for A's RA score
with C to increase also. Roughly, the conditions under which this can occur require: 1) an approximate
equality of actor A's raw trade in imports and exports with actors B and C; 2) an approximately equal
increase of A's imports and exports with B and C; and 3) A's increases with B and C must be greater
than the increases between B and C.
A rough idea o£ the empirical competitiveness o£ the RA score is provided by our data. We have a
three-actor system, West Germany, the EEC, and the world. If the RA score has no scarcity biases, we
should expect West Germany's trade (measured in RA scores) with the EEC and the world to be un-
correlated. In fact, West Germany's exports to the EEC and to the world correlate at -.836 and imports
correlate at -.960. Thus, despite the fact that West Germany's total trade increased with both the EEC
and the world, the RA scores for trade with these two systems are negatively related.
I am grateful to Raymond Duvall of Northwestern University for working out the mathematical
properties of the RA index as well as providing simulated data on the behavior of this index in a variety
of international trade systems.
236 INTERNATIONAL ORGANIZATION
theoretically significant if they are precisely defined and measured. But while
operationalism was fairly successful in providing criteria for precise, inter-
subjective concepts, it was not nearly as successful in terms of leading to theo-
retical terms with broad scope. The present controversy concerning the level
of European integration can never be resolved with respect to research and
measurement considerations alone, just as an exclusive focus on theory or de-
scriptive argument is futile. Yet, the argument has been carried out in just
these terms as if some self-evident descriptive answer were waiting there to be
found by the straightforward application of some neutral measurement lan-
guage.
Every research method carries a theoretical bias in its assessment of reality.
If countries are considered more integrated when the value of their trade in-
creases, then integration is viewed in terms of growth characteristics. If the
RA index is used, then integration is viewed in terms of changes in preferen-
tial behavior. Most of the methods used to measure international integration—
from absolute trade figures and percentage scores to measures which control
for national income and the RA index—frequently argue that these control
terms are needed since they neutralize factors that otherwise would lead to
artifactual results. But what is artifactual is far from a settled matter. One will
probably want to control for increases in trade due to inflation, but whether
one wants to control for changes due to higher gross national product (GNP),
increased growth rates, or general increases in the level of world trade is by
no means a closed matter. There is simply no extratheoretical solution to the
problem.
We mentioned that the RA index could be roughly thought of as a zero-
sum measure whereby increases or decreases in integration between any two
actors would probably be offset elsewhere by changes with other actors.
Deutsch's communication perspective leads him to view integration in terms
of the relative density of communication flows, i.e., transactions. There are
thus two areas in which one must attempt to introduce controls, between
members of some integrating system and third parties and between the supra-
national system and its component units, in this case nation-states. Thus, it is
important to note not only that integration has reached a plateau with respect
to diird members but also that there has been a relative decline in interna-
tional life compared with the intensity of domestic activity and transactions.34
In contrast to this view of integration which stresses that the development
of domestic and supranational systems are distinctive, even competitive, con-
cerns other authors choose to view integration as a process of mingling and
blending activity and institutions at different levels. Lindberg and Stuart
Scheingold have discussed in some detail how the institutional system of the
34
Karl W. Dcutsch, "The Propensity to International Transactions," Political Studies, June i960
(Vol. 8, No. 2), pp. 147-155.
238 INTERNATIONAL ORGANIZATION
My argument has been that the crux of the problem of measuring integra-
tion is the lack of criteria by which to assess the state of integration. I offer
two criteria both of which are based on the belief that what is involved here
is really a question of validity. These two criteria are 1) the convergence be-
tween one particular scoring procedure and others in attempting to measure
the same construct and 2) the differential predictive and explanatory capacity
of a variety of different scoring procedures for the same raw variable. The
35
Lindberg and Scheingold, pp. 32, 37. See also the excellent articles on the preparation of commu-
nity decisions at the national level by Theodor Holtz, Pierre Gerbert, Marco Olivetti, Guy de Muyser,
and Robert De Bruin in La Decision dans Us Communautes europeennes (Brussels: Presses universitaires
de Bruxelles, 1969).
36
Parsons's conception of power is one that is seen as infinitely expandable depending upon the
number of functional contexts in which it operates. See "On the Concept of Political Power," in Talcott
Parsons, Sociological Theory and Modern Society (New York: Free Press, 1967), pp. 297-355.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 239
former criterion utilizes the notions of both reliability and validity especially
as formulated by Donald Campbell and Donald Fiske.3' The latter borrows
from a variety of sources but especially from the literature of construct
validity.38
Convergence
The criterion of convergence is satisfied when, in the attempt to measure
a construct, a variety of independent methods arrive at similar results. The
independence of methods is crucial here in distinguishing validity from re-
liability.39 William Scott has correctly argued that "it is prudent to intercorre-
late quite different instrument types, if their correspondence is to be attrib-
uted primarily to common content."40 To the extent that a particular method
produces results which are artifactual these results will tend not to agree with
those derived from highly different measurement techniques. Thus, given a
set of scoring techniques, we will place more confidence in those which show
highest agreement with all others, ceteris paribus. To the extent that none of
the various scoring procedures displays significant (in the nonstatistical sense)
overlap we must conclude either that each technique taps a different con-
struct, that the construct is not unidimensional, or that a substantial portion of
the variance results from variance in the methods and not from variance in
the substantive variables that we are attempting to measure.
The construct of interest here is "economic integration," particularly the
economic integration of West Germany into die European Economic Com-
munity. Two indicators have been selected to measure this construct: West
German imports from and exports to EEC countries. Four different scoring
techniques have been applied to these two indicators.41 Each of the scoring
procedures is treated as a separate method.
The four scoring procedures are as follows: i ) Raw data for imports and
exports is used. This technique does not control for internal economic growth,
total growth of trade in the world or region, or for West Germany's changing
share of EEC and world markets. 2) Preoccupation ratios are used.42 A pre-
occupation ratio (PR) is a measure of relative integration. It assesses the
37
Donald T. Campbell and Donald Fiske, "Convergent and Discriminant Validation by the Multitrait-
Multimethod Matrix," Psychological Bulletin, March 1959 (Vol. 56, No. 2), pp. 81-105.
38
Lee Cronbach and Paul Meehl, "Construct Validity in Psychological Tests," Psychological Bulletin,
May 1955 (Vol. 52, No. 3), pp. 281-302.
39
Campbell and Fiske, Psychological Bulletin, Vol. 56, No. 2, p. 83.
40
William A. Scott, "Attitude Measurement," in Gardner Lindzey and Elliot Aronson, eds., The
Handbook, of Social Psychology, Vol. 2: Research Methods (2nd cd.; Reading, Mass: Addison-Wesley
Publishing Co., 1968), p. 254.
41
This results in a "multimethod matrix," a variant on the Campbell-Fiske theme. Technically, there
are four methods and one trait since imports and exports are viewed as indicators of the same construct.
We thus expect high interindicator correlations and are not concerned here with discriminant validity.
42
James V. Toscano, "Transaction Flow Analysis in Metropolitan Areas: Some Preliminary Explora-
tions," in Philip E. Jacob and James V. Toscano, eds., The Integration of Political Communities (Phila-
delphia: J. B. Lippincott Co., 1964), pp. 102-103.
24O INTERNATIONAL ORGANIZATION
the "greatest convergence" and the least unique method variance. In this sense
we would also have confidence that the method is a most valid one.
Method 1
Imports 1.000
Exports (-955) 1.000
•
Method 2
Imports •936 .884 1.000
Exports •796 •913 (.798) 1.000
Method 3
Imports •557 421 .740 .308 1.000
Exports -•509 -•329 -•679 -.169 (-.980) 1.000
Method 4
Imports •833 •830 •834 •736 .486 -425 1 -ooo
Exports •777 .78. •765 .688 •4'7 --357 (-964) 'ooo
Method i
CO CO
•798
O) r-
NOTE: XPr refers to the mean of the purified correlation coefficient. It is equal to the average inter-
corrdation coefficient omitting the two RA scores. Pearson's r is used. N=35.
242 INTERNATIONAL ORGANIZATION
The average correlation for all methods except the RA method appears to
be quite high especially if we consider that the low RA scores are included
in the composite scores. The XPr ("average purified correlation") indicates
how much higher these correlations are when calculated without the RA co-
efficients. Table 3 ranks these methods in terms of the strength of the inter-
indicator correlation.
a
Exports and imports combined.
The highest scores are produced by method i, indicators i and 2 (.621 and
.636). Method 2, indicator 1 ranks second with .611, then method 4, indicator
1 with .608, and method 2, indicator 2 with .581. The RA method has the low-
est average intercorrelation with exports and the second lowest with imports.
It is striking to note that the highest composite score (considering exports and
imports combined) is yielded by the raw data on exports and imports, a value
of .629. Method 2, the preoccupation ratio, is close behind with a combined
score of .596. Russett's method is next (.592), and the RA index is the least
convergent with an index of -.107.40
Perhaps a bivariate correlation matrix cannot give us the information we
need. For example, two bivariate correlations (rAB, rBC) may be high in-
ternally and yet rAC will be low. Any two variables (A and C) may share
variance with a construct by virtue of common variation with a third variable
(B).47 Factor analysis may help us with this problem. Table 4 presents the
results of an orthogonal rotation.
40
I hasten to add that this does not invalidate the RA index although it does decrease our confi-
dence in it. The criterion o£ convergence could be looked upon as a kind o£ "halfway house" for validity.
In principle, though, any construct should be amenable to measurement through a variety of methods.
If, after exhaustive attempts to find other techniques sensitive to the same variance as the RA index,
the RA index still stands alone, we would be tempted to draw the conclusion that the variance is due
to special properties of the method. (In this line of reasoning, in particular with respect to the view of
convergence as a "halfway house," I am indebted to Thomas Milburn, professor of social psychology,
DePaul University.)
47
On this point see Benjamin Fruchter, Introduction to Factor Analysis (Princeton, N.J: D. Van
Nostrand Co., 1954), pp. 5-6.
THEORY AND METHOD INT INTERNATIONAL INTEGRATION 243
Method 1
Imports .881 .371
Exports .954 .181
Method 2
Imports .809 .558
Exports .919 .035
Method 3
Imports .256 .955
Exports --158 --986
Method 4
Imports .881 .285
Exports .855 .222
Percentage variance 61.161 3 • -939
NOTE: The loadings in this table are for the rotated matrix.
All variables have high loadings on factor i except imports and exports
measured by die RA mediod (.256 and -.158). This factor explains over 61
percent of die variance in die correlation matrix. Also, on factor 2 die two
RA scores behave very differendy from the odier variables. Imports load at
.955 and exports at -.986. These two variables define die second factor and
account for most of die variance. Aldiough it is true diat die two RA variables
show strong loadings in the opposite direction, diis is not necessarily incon-
sistent widi die interpretation diat diey are bodi tapping the same dimension.
If die RA index is sensitive to changes in preferential behavior, it would be
sensitive to bodi positive and negative changes.
This supports our interpretation of die correlation matrix diat die RA index
is least convergent widi odier mediods. It also suggests diat die RA mediod
may be sensitive to a dimension of integration different than diat tapped by
odier scoring procedures. Theoretically, we may diink of factor i as measur-
ing economic integration where integration includes growth characteristics
in die individual country, die region, and die world. Factor 2 may be sensi-
tive only to changes in preferential behavior, what Deutsch calls "structural
behavior." It probably is not sensitive to size-effect changes, volume changes,
and growdi variations.
Deutsch's view of integration comes close to Kennedi Boulding's idea of
structural growdi/ 8 Odier views of integration which stress die growdi of
48
Kenneth Boulding has developed a threefold typology of growth: simple growth, population growth,
and structural growth. Simple growth and structural growth are relevant for our purposes. Boulding
thinks of simple growth as "the growth or decline of a single variable or quantity by accretion or de-
244 INTERNATIONAL ORGANIZATION
structures and functions at the suprasystem level are more directly concerned
with the "mere effects of size and prosperity," i.e., simple growth. The growth
of political structures, institutions, and functions at one level (e.g., the EEC)
may leave patterns of political activities and capacities at the nation-state level
quite undisturbed. Indeed, it may even encourage further growth of these
individual nation-states.
While the evidence here suggests that raw data, preoccupation ratios, and
the Russett method may all be valid techniques for scoring transaction data,
there are two caveats we should mention. First, convergent validity is ex-
tremely impressive when one has great confidence in one of the measures
beforehand. In the absence of this confidence one must accept commonality
of multiple measurements as the only acceptable evidence. The risk of circu-
larity here is decreased to the extent that the methods are independent. Sec-
ond, as Johan Galtung has stated,
"the proof of the pudding is in the eating," i.e., what can be explained and
predicted from the data collected is the important thing, not how much "con-
sistency" tiiere is between forms of data collection with perhaps no theoretical
reason for consistency at all between them.49
Construct Validity
The question arises as to whether a set of methods can yield consistent re-
sults and still be invalid. One interpretation is that the three methods whose
results converge are contaminated (e.g., by method variance) in the same
direction. The only uncontaminated method in this case would be the RA
index and its unique noncontamination would explain its lonely nonconver-
gence. In order for this argument to be acceptable the nonconvergent method
(the RA index) should have greater predictive ability (according to a theory)
than the convergent methods. This brings us directly to our second criterion,
the predictive and explanatory capacity of various measures. This approach is
pletion" ("Toward a General Theory of Growth," in Joseph John Spengler and Otis Dudley Duncan,
eds., Population Theory and Policy: Selected Readings [Glencoe, 111: Free Press, 1956], p. 109). Struc-
tural growth, on the other hand, consists not of a scale or volume change but of a change in the com-
plexity of a system or in the relationships of the parts of the system to one another. We notice a strik-
ing parallel between this concept and Deutsch's conception of integration as reflected in the RA index:
This index thus measures by how many more or fewer percent these two countries deal with one
another than they could be expected to do according to random probability and the mere size of
their total foreign trade. The RA index separates, therefore, the actual results of preferential be-
havior and structural integration from the mere effects of the size and prosperity of countries.
[Deutsch et al, p. 220.] :
I do not, however, mean to imply that Deutsch's concept of integration and Boulding's idea of struc-
tural growth are identical. Boulding's notion of structural growth also includes changes in system com-
plexity in which the system may gain new components without these additions being offset elsewhere in
the system. This kind of growth is epigenetic in Etzioni's sense. See Amitai Etzioni, "The Epigenesis of
Political Communities at the International Level," American Journal of Sociology, January 1963 (Vol.
68, No. 4), pp. 407-421.
49
Johan Galtung, Theory and Methods of Social Research (New York: Columbia University Press,
IQ
67)> P- i 2 7- This quote should be seen in light of a position on which Galtung was commenting; it
is not necessarily his own position.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 245
straightforward: "To make clear the meaning of a concept is to set forth the
laws within which it occurs."50 Thus, we can simply embody the alternatively
scored data in the same substantive propositions and ask what the explanatory
import of each is. This approach to validation, unlike the convergence ap-
proach, immediately raises considerations about the structure of science and
the laws in terms of which it is composed. The construct validation approach
simultaneously involves the validation of a construct and a theory. The ra-
tionale of this approach is concisely summarized by Helen Peak:
The essence of the approach to validation through testing predictions from
theory may be stated briefly. The meaning of any measured process is given
not only by a description of operations used in isolating it from other processes
and in assigning some index of quantity but also by knowledge of its influence
on other processes and their influence on it. . . . This involves all the problems
of formulating theory, deducing consequences, and testing the deductions un-
der conditions of controlled observation.51
It will be convenient to apply this test in four steps: i) specifying the con-
cept, 2) embedding the concept in a theory, 3) deducing testable theorems
from it, and 4) presenting and evaluating the evidence. The concept of inter-
est here is economic integration and the theory with which we will be dealing
is that of "spillover." According to this theory the greatest amount of spill-
over is expected from the least technical sectors. It is expected that integration
in technical sectors will remain more or less isolated while integration in non-
technical areas will generate effects in other sectors.52 We deal with the third
step of our test, deducing testable theorems, by selecting two sectors within
the EEC, transport and agriculture, as examples of technical and nontechni-
cal integration. According to the theory more spillover will occur between our
measures of economic integration and political integration in agriculture, a
nontechnical sector of the EEC, than in transport, a technical sector.53
A word should be mentioned concerning the operationalization of the
variable political integration. Political integration is viewed as the emergence
and development of political structures and activities around supranational
units. Thus, the EEC can be said to become more integrated as it acquires
decisionmaking capabilities, the power to make binding decisions, the power
to implement these decisions, and the power to adjudicate disputes between
member states. We can thus utilize decisions, regulations, recommendations,
50
Cronbach and Mcehl, Psychological Bulletin, Vol. 52, No. 3, p. 290.
51
Helen Peak, "Problems of Objective Observation," in Leon Festinger and Daniel Katz, cds., Re-
search Methods in the Behavorial Sciences (New York: Holt, Rinehart & Winston, 1965), pp. 288-289.
52
This is almost the opposite relationship from the one suggested by David Mitrany in his A Word-
ing Peace System: An Argument for the Functional Development of International Organization (Chi-
cago: Quadrangle Books, 1966).
63
There is no space to fully defend the selection of these two sectors here. Briefly, I considered the
extent to which different groups were involved in the issue area, the extent to which autonomy is given
to experts, the degree of involvement of key political decisionmakers in the decisions of the sectors, and
the degree of problem-solving versus purely political (e.g., satisfying group interests) activity.
246 INTERNATIONAL ORGANIZATION
and other indicators of the political activity of the Commission and the Coun-
cil of Ministers of the European Communities to measure integration. Any
composite indices which are formed on the basis of these indicators can be
thought of as "institutional output scores" as suggested by William Fisher.54
The following procedure was utilized. There are basically four community
institutions empowered to make decisions that have some political and legal
force in the EEC: The Council of Ministers of the European Communities,
the Commission of the European Communities, the Court of Justice of the
European Communities, and the European Parliament. Both the Council of
Ministers and the Commission of the European Communities have the power
to issue regulations, decisions, directives, and recommendations; the European
Parliament can submit questions to the commission and the council although
the Rome Treaty provides no legal basis for submitting questions to the coun-
cil.55 The following list explains the differences between the four types of
action:
1) Court communications—these are letters and other information concern-
ing a case which the court receives and sends out.
2) Court judgments—judgments are true judicial acts resulting from con-
sideration of alleged infractions of community laws. Judgments of the Court
of Justice are directly enforceable in all member states.
3) Court rulings—rulings of the court are, for the most part, procedural in
nature and concern the manner in which a case is litigated.
4) Court decisions—decisions are preliminary in nature and usually concern
interpretation of the Rome Treaty or acts of community institutions.
5) Number of new cases considered by the court—this indicator is intended
to reflect the volume of new legal activity undertaken by the court.
6) Parliamentary questions—the number and kind of written questions
•submitted by the members of the European Parliament to members of the
•Council of Ministers or the Commission of the European Communities.
7) Council regulations—these do not differ in legal status from regulations
•of the Commission of the European Communities. However, in substantive
terms, their content is different; they usually apply to more general, political
•concerns than commission regulations which are usually reserved for "those
technical areas in which the principles and the general policy have been de-
fined in the Treaty."56
8) Council information—this indicator reflects the amount of internal (to
the EEC) communication which takes place between the Council of Ministers
of the European Communities and EEC institutions, especially the Economic
and Social Committee.
54
Fisher, International Organization, Vol. 23, No. 2, p . 273.
55
Murray Forsyth, " T h e Parliament of the European Communities," Political and Economic Planning,
March 1964 (Vol. 30, N o . 478), p . 52.
56
Leon N . Lindberg, The Political Dynamics of European Economic Integration (Stanford, Calif:
.Stanford University Press, 1963), p. 35.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 247
Indicator
Indicator 1 2 3 4 5 6
1. Parliamentary questions 1.000
2. Council regulations •679* 1.000
3. Council information •509* .466* 1.000
4. Commission regulations •332* .710* .221 1.000
5. Commission directives •528* •553* •723* •369* 1.000
6. Commission recommendations .408* .420* •548* •'73 .690* 1.000
NOTE: Values statistically significant at the .05 level are indicated by an asterisk. N=4O.
The first thing to note is that almost all the indicators are highly correlated,
i.e., there is a great deal of shared variation which is statistically significant.
Yet, there are hints that there may be subgroupings within the total correlated
set, e.g., council regulations have a very high correlation with both parlia-
mentary questions (-\-.6yg) and commission regulations (+.710) while com-
mission recommendations correlate at +.690 with commission directives.
There appear to be complex, multivariate relationships here with which a
bivariate correlation matrix is not equipped to deal. Factor analysis may help
us discover some of these more subtle relationships.62 The following factor
structure offers interpretations of the data that were barely hinted at in the
correlation matrix:
TABLE 6. FACTOR ANALYSIS OF INDICATORS OF ALL POLITICAL INTEGRATION I N THE
EEC
Indicator Factor i Factor 2 Factor 3
NOTE: The loadings in this table are for the rotated matrix.
62
Factor analysis is a multivariate technique particularly helpful in determining whether a given set
of data (the data may be correlation coefficients) contains one, two, or "n" sets of organized properties.
In short, it is very useful when one wants to dimensionalize an array of data.
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 249
The factor structure reinforces the suggestion that all indicators of politi-
cal integration do not belong to the same dimension. It seems clear that
factor 1 represents a rule-adjudication dimension. This dimension is defined
by communications of the court, judgments of the court, and new cases
brought before the court. Indicators 8 and 10 (council information and com-
mission directives and decisions), also part of this factor, probably reflect the
beginnings of implementive activity in the EEC. The fact that two "adminis-
trative" indicators are part of the same dimension as the activity of the court
may reflect the relatively low level of structural differentiation existing in the
community today.
The interpretation of the second factor is clearer. Indicators 6, 7, and 9
(parliamentary questions, council regulations, and commission regulations)
define this factor. We remember that council and commission regulations are
legally binding and have the force of real community law. In addition, they
are directly enforceable in all member states. We thus label this factor our
rulemaking dimension. The procedure from here was to standardize these
indicators and add them together to form a composite index of political
decisionmaking.
The above propositions will be tested through a correlation analysis. Before
turning to this evidence, however, it should be made clear just what it is the
theory "expects." First, we expect higher absolute correlations between method
1 (raw data) and our dependent variables. We expect lower correlations with
methods 2, 3, and 4. This is because the theory on which the empirical test is
based is not grounded in assumptions about scarcity or relative share of the
market. We postulate a relationship between economic and political integra-
tion which is not concerned with the "share" of a particular country's trade
with a particular region. If West Germany increases its trade with the EEC
by 15 percent and with the rest of the world by 30 percent, we still expect
the hypothesized spillover processes to occur. Second, the theory suggests that
we are interested in the difference in the respective coefficients of strategic
sets of variables in addition to their absolute values. In other words, we are
interested in which predictor variable is most discerning among our integra-
tion measures for agriculture and transport. Correlations may be high for
both agriculture and transport but this by itself would neither confirm the
theory nor validate any of the scoring procedures. Table 7 presents evidence
for our hypotheses.
The first thing to note is that method 1 does yield very high correlations
as predicted. Method 2 also produces high correlations but methods 3 and 4
both are lower. A useful summarizing measure is the mean intervariable cor-
relation (XVr) which is presented in the right-hand portion of the matrix.
The highest predictive capacity is given by method 1, variables 1 and 2. Again
25O INTERNATIONAL ORGANIZATION
Variable
Method 1
Imports .886 •743 •705 .824 .717 .487 .809 •374 •693
Exports .889 .620 •793 .871 .770 •577 .788 •322 •704
Method 2
Imports .834 •653 •707 .783 •695 •532 •775 •453 .680
Exports .821 •442 •873 .862 •777 •679 .708 .284 .681
Method 3
Imports .428 •397 •254 •33O •304 .216 •445 •379 •344
Exports -•358 -.424 -•139 -•239 -.218 -.110 -•390 -.362 -.280
Method 4
Imports .728 •556 .656 •739 .629 •536 .611 •365 •603
Exports .661 •503 •637 .666 .631 •598 .562 •333 •574
NOTE: Variables i through 8 are identified as follows: i ) political integration within the agricultural
sector of the EEC; 2) political integration within the transport sector of the EEC; 3) total political inte-
gration in the EEC. Variables 4-6 are based on the Almond-Coleman output functions of a political
system, adjudication, rulemaking, and external affairs, respectively. Each of these is a measure of the
degree to which the institutions of the EEC perform these functions; 7) functional (trade) integration
in agriculture; 8) functional integration in transport. The political integration measures are based on
political activities of four institutions: the Court of Justice of the European Communities; the Council
of Ministers and the Commission of the European Communities; and the European Parliament. In all
there were fourteen indicators of political integration on which our final measures here are based. Func-
tional integration in agriculture is simply a composite score of trade in food, beverages, and tobacco
between the six. Functional integration in transport was measured by five indicators, two of them based
on railroad activity between the six, two on inland waterway shipping, and one on tourism. Again,
Pearson's r is used and N = 3 5 .
closely behind is method 2 with .680 and .681 for imports and exports, re-
spectively. Russett's technique ranks third with .603 and .574, and the RA
method ranks last with the lowest predictive power, .344 and -.280, for im-
ports and exports. It is interesting to note that the outcomes of this predic-
tive test exactly parallel those of the convergence test. It will be recalled there
that method 1 ranked first, just barely above method 2. Here method 1 ranks
first but again the difference between methods 1 and 2 is negligible. Methods
3 and 4 are ranked in the same order as in the convergence test. This "match-
ing up" of predictive and convergence tests is in itself evidence of a kind of
"convergent" validity.
Political Political
Scoring Integration Integration Significance
Procedure (agriculture) (transport) Difference t-value at .05 level*
Method i
Imports .886 •743 +•'43 2.31 Yes
Exports .889 .620 + .269 3-8i Yes
Method 2
Imports .658 2.23 Yes
CO CO
+.176
CO «
ID CO
Imports .728 •556 + .172
CO CO
Exports .661 •5°3 +•158 No
V. CONCLUSION
The primary purpose of this article has been to suggest a means by which
we may settle the particular methodological dispute in question. It has not
been my intent to "prove" or "disprove" the worth of a particular scoring pro-
cedure such as the RA index. The two criteria suggested, convergence and
predictive capacity, seem to be two helpful criteria which are generally ac-
cepted as sound methodological tools by the scientific community. They
should help to provide a broad common ground on which to base a fruitful
exchange of ideas and perhaps to lessen the "dialogue of the deaf quality
which is so much a result of different (or implicit) criteria for evaluation.
Convergence is really a specific application of the general scientific mandate
that inquiry be intersubjective and yield findings invariant with respect to the
method employed. We recall that three of the methods (raw score, preoccu-
pation ratios, and Russett's method) are highly convergent. In interpreting the
process of integration the notion of construct validity is very helpful. We have
hypothesized that the process measured by the three convergent methods is
one of "simple growth" in which growth is thought of as a general (i.e., aver-
age) increase in the value of a variable. We think of growth models as largely
concerned with increases in the capacity or frequency of a variable. Non-
growth models may be concerned only with the distribution of an existing
capacity. Perhaps the RA index would be fruitful for such a model. It seems
that for integration, however, it is precisely the growth of the capacity of a
new system (the suprasystem) to build new structures and perform new func-
tions that is important. We are generally interested in monotonic increases
and not in structural changes. The essence of an integrative relationship is
system growth, the creation and development of new behavioral patterns, not
the reallocation of existing ones.
We have hypothesized that if we were measuring "simple growth" our mea-
sures should behave in certain specified ways. The theory is capable of pre-
dicting substantial portions of the variance as well as ranking the different
scoring procedures in terms of predictive capacity. The most convincing case
was made when our methods were set up competitively to test the hypothesis
about differential rates of spillover for agriculture and transport. The results
of this test have reinforced our confidence in the previous two tests and have
provided independent evidence of the validity of our first two scoring proce-
dures. This test has required precise measurement and clarity, demanding not
only high correlations but also the ability to ascertain differential spillover
rates in the two issue contexts. In this regard we should note that method 4
(Russett's technique) has not satisfied the requirements of this test although
it did meet the requirements of the other two tests.
I want to emphasize that no inference has been drawn with respect to the
THEORY AND METHOD IN INTERNATIONAL INTEGRATION 253
63
Similarly, I have pointed out how this problem poses a limitation on the conclusions drawn by
William Fisher in his test of Deutsch's theory of integration. See James Caporaso, "Fisher's Test of
Deutsch's Sociocausal Paradigm of Political Integration: A Research Note," International Organization,
Winter 1971 (Vol. 25, No. i ) , pp. 120-131.