Professional Documents
Culture Documents
Historically, half of the interstate wars identified by Singer and Small have been
preceded by arms racing; therefore, not all wars stem from weapons competition.
Similarly, there are arms races which end pacifically. Here, considering only arms race-
related conflicts, the author argues that war may be anticipated at the end of an arms race
on the basis of time-constrained mathematical stability characteristics of the involvement.
Simple linear models of arms racing are applied to a sample of historically founded
weapons competitions. Coefficients thus provided are used to determine the stability of
the arms races, using a time-restricted definition of interesting stability rather than
stability in the limit, to obtain a moderate rate of successful predictions. Stability analyses
with time constraints appear to provide a method for assessing part of the risk of a general
course of armament.
INTRODUCTION
AUTHOR’S NOTE: The author gratefully acknowledges the financial support of the
Political Science Department and the Graduate School of the University of Minnesota for
a prior set of analyses, and the Political Science Department and Research Council of
Rutgers University for these. The author is considerably indebted to Brian L. Job and
Hector Sussman for comments on this and earlier drafts.
253
(For example, the Korean war apparently began without an arms race,
if the author’s definition is used.)
On the basis of argument and accumulated research findings
(Caspary, 1967; Smoker, 1964, 1965, 1969; Milstein, 1972; Wallace,
1970, 1971, 1972, among others), we may then reject the most absolute
views-that racing always ends in war (there are also races which end
without war-cf., Table 1) and that arms rivalries and wars are totally
unrelated. This leaves us considerable middle ground. Military com-
1. Note that by definition catalysts may serve not only to initiate a reaction but also
to accelerate an ongoing process.
DEFINITIONS OF TERMS
case the tide should turn, but presumably if racing posed a genuine or
PAIRING OF RIVALS
assertion for the great powers before the world wars, but without
exception one link required for a complete multilateral race is missing.
For example, while Germany armed against both France and England
before World War I, France and England did not race against each
other. Usually where there is multinational involvement in racing, as in
the Arab-Israeli case, the ongoing competition is still essentially
bilateral, with a coalition of sorts forming one side.
For these reasons, in this inquiry I have grouped adversaries or sets of
adversaries by pairs. Perhaps the recent case most likely to argue for an
exception to this bilateral treatment is provided by the Chinese-Soviet
and Soviet-American but even here, the third part of the
involvements,
triangle, a U.S.-Chinese race, would be extremely difficult to sub-
stantiate. It is true that the U.S. ABM was once justified in this context
by then Secretary of Defense Robert McNamara, but this was largely in
response-apparently-to an objection that the ABM system as
envisioned would be an inadequate foil for Soviet attack. Since the
Chinese still have no missiles capable of reaching the United States (the
recent development of a longer-range ballistic missile in China indicate
this may not be so for long) this rationale appears flimsy. (Since the
ABM system is now dismantled, foreign policy generally seems less
antagonistic; and the United States is providing the PRC with a
surveillance satellite to observe Soviet Central Asia [cf., a number of
New York Times reports beginning with NYT, June 9, 1978, in which
US administration officials note an agreement to provide to the PRC
&dquo;airborne geological survey equipment using infrared scanning&dquo; which
would be denied to the USSR].) There seems to be at least as much
evidence for cooperation as for rivalry, especially since U.S. recognition
(January 1, 1979) and Chinese requests for increased U.S. military
presence in the western Pacific.
But this example does show that a country may race against more
than one opponent at a time; in this illustration consider that the USSR
apparently races with the United States and with the PRC simulta-
neously. When this occurs, the question of what proportion of the
military budget is directed toward each arms race naturally arises. Here
I have counted 100% of a multiple racer’s military spending against each
adversary in the multiple race. In doing so I probably exaggerate the
racing behavior of the multiple race participant by some unknown and
variable amount, but not to do so probably underestimates. The 100%
figure presumably does represent an accurate picture of worst case
THE SAMPLE
2. Note that there is no practical method for obtaining more accurate estimates, since
even if these are matters of official record, which appears not generally to be the case, there
is some overlap in weapons, installations, and armed forces which could be employed with
approximately equal facility against either (or any) adversary. If this overlap exists, even if
"true" percentages of funds can be obtained, they should sum to more than 100.
no arms races before this time, simply that nothing can be said about
them in the quantitative terms I would like to use.) Banks Cross-
National Time Series figures were supplemented with early military
budget estimates from the Almanach de Gotha and the Staleman’s
Yearbook. Post-World War II figures are taken from the Stockholm
International Peace Research Institute Arms Control and Disarmament
Yearbooks. A fairly complete and continuous record of military budgets
for most countries in the world at any given time in the 120 years from
1857 to 1977 was thus obtained. All increasing military appropriations
sequences of at least four years’ duration within this period were then
identified and investigated through historical sources. In the military
spending data there were well over 200 such sequences for individual
nations. When the foreign policies of the relevant countries were
researched, a list of thirty-two involvements which met the definitional
criteria was obtained. This list appears in Table 1. These involvements
clearly vary considerably in length, cost, commitment, and evident
intensity, but their differing among themselves in not, for example,
involving only rivalries inter pares does not alter their adherence to the
general definition proposed. (Readers may prefer an alternate or more
restrictive definition, or a reading of different historians. There is of
course reasonable scope for disagreement among area specialists,
TABLE 1
Arms Races, 1860-1977
a. Military competition may be centered in only one part of the military, with a
particular service and/or weapons type receiving the most money and/or showing
rapid change in generations of weapons, transportation and delivery vehicles, or
sheer numbers of items deployed. If this appears to be so, the emphasized service
is noted under the &dquo;principal mode&dquo; heading. Development of nuclear weapons is
also noted here.
b. Cf., Hurewitz, 1969: 79; NYT, March 15, 1976. In the Times, a CIA official
addressing members of the American Institute of Aeronautics and Astronautics on
March 11, 1976 estimated Israel had 50 to 20 nuclear weapons &dquo;available for use.&dquo;
Then Director of CIA George Bush later accepted responsibility for the disclosure
of what was purportedly privileged information. Additionally, former U.S. Ambas-
sador to Morocco Robert E. Neuman has suggested to the author that Soviet war-
heads might have been transferred to Egypt for a brief period in the fall of 1973,
preceding the alert of U.S. forces. This is a rather durable rumor. (See SIPRI Year-
book, 1977 : 6.)
the most revealing work in this area would be that of the crisis decision-
making researchers. On the other hand it is possible that there are
certain large-scale deterministic factors which appear generally to be I
associated with the onset of war in the arms racing context, irrespective
of individuals in power. If so, some concepts which are helpful in
examining questions in the physical sciences may be applicable to this
problem, particularly the ideas of stability and instability in mechanical
systems. One way to examine arms races with a view to predicting their
outcomes could then be to consider them as mechanical systems which
may tend either toward a point of no further change, or to greater and
greater activity. (That is, consider arms races not as results of conscious
decisions but as physical systems working within certain bounds as a
result of some initial disturbance, as though arms racing were roughly
analogous to billiards.) One of the simplest ways to do this, if it is
assumed that a linear treatment is an adequate one for these purposes
and for the spending patterns under study, is to use the basic Richardson
models of arms races,3 shown later. If these models are accepted as
reasonable approximations of the arms racing process, the stability
coefficients which can be computed from them to determine their
mechanical stability may provide a useful indicator of the anticipated
outcome.
processes, but because this introduces the possibility of multiple points of equilibrium (?) it
seems excessively complex at this stage of the inquiry.
To consider why this may be so, the reader should note that in
Richardson’s view (1960a), unstable races which tend toward ever-
increasing armaments exacerbate international tensions and eventually
incite violence. (&dquo;Eventually&dquo; is not defined.) This suggests that
Richardson’s notion of stability, regardless of its intrinsic importance
or the theoretical validity of the basic linear model, may yield a useful
indicator of dangerous races; unstable races may be more risky than
others. But this is only an assumption in Richardson’s work. To see
whether or not Richardson’s stability conditions criteria allow war
prediction in some instances, I will use his stability conditions with the
alterations made necessary by use of discrete rather than continuous
data to identify stable and unstable races in the mechanical sense
employed here. The groups of races thus labeled stable or unstable-for
stability is a characteristic of the race itself, not of individual nations’
participation in it-will then be separated by outcome to see whether
their stability characteristics have given any guide to outcomes in the
,
past, and whether the hints they make about continuing races seem
persuasive.
To explain why the stability characteristics of the models given below
might provide a useful indicator of oncoming war, the definition of
stability as Richardson adopts the term from mechanics must be
elaborated. Stability in this sense refers to the existence within some
mechanical system of a point of equilibrium at which change in the
process under study ceases. If a given physical system, or, in this case, a
set of equations important in predicting arms racing for two countries, is
in equilibrium, its rate of change is zero. In the arms racing context, this
point is not necessarily a point of disarmament, but represents the point
at which change in the arming process ceases, regardless of the absolute
levels it may have attained. If, on displacement, the system returns to
equilibrium, even though this takes a long time, it is stable.4 If the arms
race tends to drift (or &dquo;run&dquo;) away from equilibrium either positively or
axes, the acceptable region for x is comprised of all points on or above the line
describing x’s current funding. The acceptable region for y is made up of all points
on or to the right of the line describing y’s current funding. If a race begins relatively
slowly there may be a large intersection of these areas within which the race may
proceed.
simply as stable or unstable races; that is, races whose every solution
remains bounded or approaches zero, and races whose solutions
from which
are derived. (The x and y variables are in the same sequence for both
racers for notational convenience.)
Then the next step, writing the characteristic equation of the resulting
matrix5
gives
Two roots are now possible, depending on the sign on the second term in
the numerator above.6To approximate these lambdas
race, one which tends to die down. (It is theoretically possible to obtain
borderline cases in which either one or both of the roots is exactly equal
RESEARCH STRATEGY
To produce some results, the author took the list of arms races from
Table I and obtained estimates of military expenditures for as many
racers as possible, using the sources described in the section above titled
&dquo;the sample.&dquo; An attempt has been made in the SIPRI sources to include
disguised military expenditure from other budget headings, such as a
certain proportion of the All Union Science Budget to represent military
research and development spending in the USSR. Military aid is
included in the spending figures of the recipient nation. (Veterans’
benefits and payments on war debts are excluded.) These figures have
been converted to deflated U.S. dollars.
The research strategy then will involve fitting these data to the models
to estimate the stability coefficients, calculating the roots of the
characteristic equations from the matrices of coefficients to discover
whether stability or instability prevails for each race, and comparing
these expected results with the known outcomes of races which have
ended. Some speculations on outcomes of continuing races will
also be advanced. Readers who do not wish to be inconvenienced by a
discussion of statistical considerations may prefer to skip to the results
section.
ESTIMATION PROCEDURE
7. A relatively new test, Durbm’s h, has been developed precisely to fill the gap-to
estimate serial correlation when one or more lagged endogenous variables is present in a
regression. See Durbin ( 1970), Hopmann and Smith (1977), Ostrom ( 1978).
that they may be either positive or negative and practically any size.
These coefficients may, for example, be positive if arming is actually a
stimulus to the economy and hence to racing, and k and 1 may be
negative if the opposition appears overwhelming to the extent that
increments in weaponry are depressed. Of course, the coefficients may
also have the signs ordinarily given to them by Richardson.
RESULTS
8. The author notes that the small sample properties of this technique are evidently
unexplored. The assumptions underlying it include that the error structure approximates
AR (1). While this may not be the case, it is not a refutable assumption in these arms races
because the error structure cannot be determined empirically in small samples (in less than
30). Malinvaud (1966: 472) notes that because no methods exist for precise determination
of the form of serial correlation in the errors of brief time series, the choice of statistical
technique presents difficulties which are not encountered when only lagged endogenous
variables are involved (Malinvaud, 1966: 449). The Ostrom procedure seems to be one of
the few which takes some account of both of these issues. Since AR (2) and other more
complicated processes are relatively rare, a technique which makes the simplest possible
assumption in the absence of information should at a minimum provide a useful starting
point. There is apparently no known solution to the small sample bias problem.
TABLE 2
Fit for Models in Final Step of IV-OLS Procedure
Table 2 (Continued)
a.Though Japan was building a modern navy at a furious pace, hence the race,
construction was evidently funded partially through scrappmg and recychng of the
old fleet. There is no change in the official Japanese military budget during the race,
m this anomalous case.
may be less meaningful for policy than determining whether races are
stable or unstable within the time span of the competition (&dquo;interestingly
stable&dquo;). This kind of qualification may also be of interest theoretically
since Richardson probably would not have predicted peace for an arms
race which was not appreciably damped until the year became
arbitrarily large.
A breakdown of arms races by model fit is made in Table 3 on the
basis of the goodness of fit for the nation receiving the lower R2 of the
two competitors. For category I, the well-modeled races, the average R2
is .91. In category II, moderately modeled races, the mean R2 is .61. In
the category of poorly fit races, III, the average R is .76, because of the
inclusion of some races with extremely good fit for one rival. Since much
of the dynamics of racing is not captured by linear models for at least
one side in the involvements listed under category III, calculations of
TABLE 3
Division of Arms Races by Accuracy of Linear Treatment
for the Poorer-Fit Nation of Each Pair
Pakistan. This pair of results probably makes sense in view of the growing Chinese
military potential.)
treating the United States and the UK as a unitary actor, when there
were some interesting interactions between them during the period
under study.
......
.,..,
== Q)
3
>-
’õ)
cq3
....
b
0
,~t
~’1:1
a
aa r.
c::
~ ~
« I
E-<::::= Q)
3
0
c;I)
U
w
.¡:
N
(.)
0
H
C,3
4.1
u
>1
s
Pm
tI)
274
Z0
0
c
b
~5
o
0
’3
’S
d
~.
’&.
_S
c
0
b
>.
I
<u;:j
(1)
~!
C
z >.
¡::
0 ~5
nt
U
1-
,It
(1) S
:D
co
0
H
N
t~
w
~5
’M
.fi
N
0
e
~~! 8 «
*&dquo;
~
Xj
~w
S ’3
N e
wo :;
’S
N
s-~
c~3 (U
* ~
275
Two of the three races continuing as of 1977, the Korean one and the
Eastern European one, have approximated or passed points of inter-
esting stability. In the Albanian case this seems plausible. Even though
Albania and the USSR still have no diplomatic, economic, or party
relations and periodically exchange perfunctory denunciations, the
involvement appears unlikely to end in war. The race seems to have had
essentially domestic driving forces in Albania and not to have exercised
a deciding effect on foreign military policy. Nor does Albania presently
seem physically capable of the sort of threat which would intrigue the
USSR sufficiently to overcome the considerable logistical and other
difficulties of a Soviet-Albanian war.
In Asia, the Korean case does not seem as clear. The volatility of the
Korean situation should not be underestimated, since the Koreas are
focal points of interest to the PRC as well as the United States and the
USSR, and since the region is one of the most heavily armed (per capita
or per square mile) in the entire world. However, it does appear that the
interested major powers have reached an implicit (if temporary)
agreement not to exacerbate Korean relations. Although there seems to
have been no overt settlement of the outstanding political issues in
Korea which would correspond to approaching the bounds of the arms
race, both the North and the South appear increasingly preoccupied
with domestic concerns: the North with the political future and the
health of Kim 11 Sung; and the South with dissent and most recently with
the assassination of Park Chung Hee, which occasioned a military alert.
According to these analyses, the U.S.-USSR competition has also
passed a point of interesting stability, about 1959-1960, which rep-
resents a period of intense foreign political contention and a prelude to
renewed active confrontation. This coincidence of damped solutions
and confrontation only serves to indicate that there are risk factors in
racing which are not encompassed at the level of abstraction of these
deterministic models. Although SALT II and other negotiations may
indicate a greater potential in the long term for real limits on weapons
competition between the superpowers, the military policies of each
appear inconsistent and unpredictable to the other, an impression
heightened by hardware and software failures (cf., the accidental US
NORAD nuclear alert on Nov. 9, 1979) and by &dquo;adventurism&dquo; abroad.
As perceived or actual advantages in preemption multiply, these
uncertainties may contribute to pressures for escalation in crises,
10. An incomplete record of nuclear weapons accidents and delivery system accidents
appears in the SIPRI Yearbook 1977 (pp. 52-82). To date these have evidently been little
or remediable ones. We may, nonetheless, anticipate the patter of little fates.
is necessarily poorly described in the linear models used in this study and
had to be removed from the present analyses, preventing a direct
comparison with Rattinger. (The first Middle East race, 1949-1955, is
uninterestingly stable in that its size is not appreciably diminished for
about 57 years.) For his second group of races Rattinger finds relatively
little reactivity, which is not entirely inconsistent with a race dying down
in 12 years, as calculated here.
From Wallace’s inventory of disputes (1979) it can be inferred that his
study and this one have identified a number of the same arms races and
some of the same results. The 1903-1904 Russo-Japanese war was
CONCLUSIONS
APPENDIX:
FURTHER COMMENTS ON IDENTIFICATION OF CASES
11. See, for example, the nonlinear (and discontinuous) treatments provided by
Wallace and Wilson, 1978.
-The drive to arm China after 1860 can be documented through the government’s
REFERENCES
ALCOCK, N. (1972) The War Disease. Oakville, Ontario: Canadian Peace Research
Institute Press.
BAUMOL, W. J. (1970) Economic Dynamics. New York: Macmillan.
BROWN, T. (1971) "Models of strategic stability." Rand Corporation, Southern Cal-
ifornia Arms Control and Foreign Policy Seminar.
CARR, E. H. (1939) The Twenty Years’ Crisis 1919-1939. New York: Harper & Row.
CASPARY, W. R. (1967) "Richardson’s model of arms races: description, critique and an
alternative model." Int. Studies Q. 2, 1.
CHOUCRI, N. and R. C. NORTH (1975) Nations in Conflict. San Francisco: W. H.
Freeman.
DURBIN, J. (1970) "Testing for serial correlation in least squares regression when some
of the regressors are lagged dependent variables." Biometrika 38, 3.
DHYRMES, P. J. (1971) Distributed Lags: Problems of Estimation and Formulation.
San Francisco: Holden-Day.
GRAY, C. S. (1976) The Soviet-American Arms Race. Lexington: Lexington Books.
(1975) "Arms races and their influence upon international stability with special
———
reference to the Middle East," in G. Sheffer (ed.) Dynamics of a Conflict. New York:
Rand McNally.
(1974) "The urge to compete: rationales for arms racing." World Politics 36,
————
2: 207-233.
---(1973) "Social science and the arms race." Bull. of the Atomic Scientists 29, 6: 2.
(1971) "The arms race phenomenon." World Politics 24: 39-79.
————
Systems 16.
MALINVAUD, E. (1966) Statistical Methods of Econometrics. Chicago. Rand McNally.
McCLELLAND, C. A. (1968) "Access to Berlin: the quantity and variety of events
1948-1963," in J. D. Singer (ed.) Quantitative International Politics. New York:
Free Press.
McGUIRE, M. (1965) Secrecy and the Arms Race: A Theory of the Accumulation of
Strategic Weapons and How Secrecy Affects It. Cambridge, MA: Harvard Univ.
Press.
MILSTEIN, J. S. ( 1972) "American and Soviet influence, balance of power, and Arab-
Israeli violence," in B. M. Russet (ed.) Peace, War and Numbers. Beverly Hills: Sage
Publications.
MOLL, K. (1968) The Influence of History Upon Seapower. Stanford: Stanford Research
Institute.
MORRISON, J. L. (1970) "Small sample properties of selected distributed lag estimates."
Int. Econ. Rev. 2.
OSTROM, C. W., Jr. (1978) Time Series Analysis: Regression Techniques. Beverly
Hills: Sage Publications.
RATTINGER, H. (1976) "From war to war: arms races in the Middle East." Int. Studies
Q. 20, 4.
RICHARDSON, L. F. (1960a) Arms and Insecurity: A Mathematical Study of the
Causes and Origins of War. Pittsburgh: Boxwood.
(1960b) Statistics of Deadly Quarrels. Pittsburgh: Boxwood.
———
onset of war, 1820-1964," in B. M. Russett (ed.) Peace, War and Numbers. Beverly
Hills: Sage Publications.
(1971) "Power, status and international war." J. of Peace Research 8, 1.
———
25, 2.
WOHLSTETTER, A. (1974) "Is there a strategic arms race?" Foreign Policy 15.
WRIGHT, Q. (1942) A Study of War. Chicago: Univ. of Chicago Press.
ZINNES, D. A. (1976) Contemporary Research in International Relations: A Perspective
and a Critical Appraisal. New York: Free Press.