Download View PDF - CiteSeerX

Jari Eloranta
European University Institute, Department of History and Civilization (Florence, Italy)
and
University of Jyväskylä, Department of History (Jyväskylä, Finland)
E-mail: [email protected] or [email protected]
Reassessing the pre-First World Military Spending: The Role of the Great
Powers in the International System, 1870-1913
ABSTRACT:
This working paper explores the interaction of military spending and economic development among the
eight recognized Great Powers (Austria-Hungary, France, Germany, Italy, Japan, Russia, UK, USA) of
the period from 1870 to 1913. The purpose is to find: 1) whether economic growth Granger-causes
military spending or vice versa, at various lag lengths; 2) what different explanatory variables affected
the demand for military spending among the individual Great Powers; 3) whether the systemic demand
for military spending can be explained in terms of leadership (or the lack of it) and other systemic characteristics in the short run. For example, the hegemonic perspective advocated by many suggests that
the economic leader of a system has to dedicate growing resources on its security, eventually becoming
harmful to economic growth. The first question outlined above was here analyzed by employing
Granger non-causality tests between the military spending and economic growth variables for these
nations. The results suggest that it was the economy that influenced military spending levels rather than
vice versa. The second question pursued here suggests that the democracies and autocracies, in broad
terms, behaved differently, and hegemonic competition was indeed the norm. The declining leadership
of the UK and the absence of military challenge by the new economic leader, the USA, were conducive
for such challenges. The third hypothesis was tested with a system comprising 16 nations. Systemic
impacts were indeed ascertained, and the results suggest that these nations responded to the declining
leadership by the UK and the lack of leadership by the U.S. with an overall military challenge. Also, the
inclusion of alliance dummies improved the fits of the pooled equations. In the individual regressions
alliance impacts seemed less important.
INTRODUCTION
This working paper attempts to explore the complicated phenomenon of interaction between military
spending, economic development, and systemic characteristics among the eight recognized, based on
1
the assessment of major power status via the Correlates of War database utilized in this paper, Great
Powers of the period 1870—1913: Austria-Hungary, France, Germany, Italy, Japan (perceived as
achieving major power status from 1895 onwards), Russia, the United Kingdom, and the United States
(perceived as achieving major power status from 1898 onwards). Thus, this paper focuses on analyzing
the demand characteristics of the military spending, especially in interaction with economic development, as monads and as system participants. Moreover, in the analysis of a 16-country system, I will
also utilize the following countries: Belgium, Denmark, the Netherlands, Norway, Portugal, Spain,
Sweden, and Switzerland. Thus, the system will feature both major powers and weaker states, as well as
autocratic (defined as either maintaining the score of four in autocracy index in the Polity IIID 2000
1
See Singer-Small 1993. This measurement is mostly based on diplomatic representation and missions in various
countries; see Ray-Singer 1973.
2
dataset or becoming even more repressive) and democratic (defined as achieving at least the score of six
2
or more in the democracy index in the Polity IIID 2000 dataset). These questions are approached by
analyzing especially the impact of the decline of British leadership before the First World War and the
lack of American military leadership during the first half of the 20th century, emerging through the respective military spending behavior of these nations.
This paper is divided into five parts. The second section will feature discussion on the theoretical premises arising out of the interdisciplinary literature on the origins of wars and military spending, as well as
contemplation on the challenges posed by the data. The third section will evaluate the military spending
patterns for individual states as well as its systemic developments, including a re-evaluation of the
hegemonic assumption of close interrelationship between military exertions and economic stature via
Granger non-causality tests on the various indicators. The fourth section will feature analyses of the
determinants of military spending for the eight Great Powers individually and at the level of the 16country system. I shall conclude the paper with a set of conclusions and contemplation on further research challenges. The questions that I pursue here are: 1) whether economic growth Granger-causes
military spending or vice versa, at various lag lengths; 2) what different explanatory variables affected
the demand for military spending among the individual Great Powers; 3) whether the systemic demand
for military spending can be explained in terms of leadership (or the lack of it) and other systemic characteristics in the short run.
RESEARCH CHALLENGES AND DATA SOLUTIONS
The research that has focused on the impacts of war and military spending on economic development,
and/or vice versa, has been vast and highly interdisciplinary by nature. Nonetheless, certain characteristics can perhaps be distinguished from the efforts to study this complex topic among different sciences
(mainly history, economics, and political sciences). Historians, especially diplomatic and military historians have been keen on studying the origins of the two World Wars. Nonetheless, many of the historical studies on war and societies have taken the form of explaining developments at an elusive macrolevel, often without a great deal of elaboration on the quantitative evidence behind the assumptions on
2
See Polity IIID 2000 - The democracy score, as an aggregate, country-based index, indicates the general
openness of political institutions. Thus, the ten-point index has been constructed additively on the basis of the
sub-categories listed therein.
3
the effects of military spending. For example, Paul Kennedy’s “The Rise and Fall of the Great Powers.
Economic Change and Military Conflict from 1500 to 2000” (1989), as criticized by many economists,
is not comprehensive enough quantitatively to support his notion of interaction between military spending and economic growth.3 Quite frequently, as emerging from the classis studies by A.J.P. Taylor and
many of the current works, historians tend to be more interested in the impact of foreign policy decision-making and alliances, in addition to resolving the issue of “blame”, on the road towards major
4
conflicts , rather than how reliable quantitative evidence can be mustered to support or disprove the key
arguments. Economic historians, in turn, have not been particularly interested in the long-term economic impacts of military spending. Usually the interest of economic historians has centered on the
economics of global conflicts — of which a good example of recent work combining the theoretical
aspects of economics with historical case studies is “The economics of World War II”, a compilation
edited by Mark Harrison — and the immediate short-term economic impacts of wartime mobilization.5
The study of defense economics and military spending patterns as such is related to the immense expansion of military budgets and military establishments in the Cold War era. It involves the application
of the methods and tools of economics to the study of issues arising from such a huge expansion. One
could perhaps distinguish at least three aspects in defense economics that sets it apart from other fields
of economics: 1) the actors (both private and public spheres of influence, for example in contracting);
2) theoretical challenges introduced by the interaction of different institutional and organizational arrangements, both in the budgeting and the allocation procedures; 3) the nature of military spending as a
tool for destruction as well as providing security.6 One of the shortcomings in the study of defense economics has been, at least so far, the lack of interest focused on other periods than the period since the
Second World War.7 Within peace sciences, a broader yet overlapping school of thought compared to
3
See Kennedy 1989. Kennedy calls this type of approach, following David Landes, “large history”; see Kennedy
1994, 7, 26. On criticism of Kennedy’s “theory”, see especially Sandler-Hartley 1995 and the studies listed in it.
Other examples of long-run explanations can be found in, e.g., Pearton 1982 and McNeill 1982.
4
The two classic studies on the origins of World Wars I and II are Taylor 1954 and Taylor 1961. See also e.g.
Watt 1977 for just one review of Taylor’s works. On “typical” historical literature on the origins (as historians
are rightly wary of using the word cause) of these conflicts, see especially Kaiser 1983 (on Germany’s role),
Gordon 1974 (on German and British cases), and on the issue of balance of power see e.g. Kraehe 1992.
5
See The economics of World War II 1998. Classic studies of this type are Alan Milward’s works on the
European war economies; see e.g. Milward 1965, Milward 1970, Milward 1977.
6
Sandler-Hartley 1995, xi; Eloranta 1998.
7
One of the few exceptions is e.g. Conybeare-Sandler 1990, which analyzes the period before the First World
War. Rasler-Thompson 1989, especially with its theoretical and quantitative emphasis, could perhaps also be
4
defense economics, one of the most significant of the interdisciplinary efforts has been the Correlates
of War (COW) project, which started in the spring of 1963. This project became the largest research
program to trace ”the intellectual history of the ‘peace research movement’” as well as to try to discover
some explanatory models relating to the birth of conflicts. This project and researchers loosely associated with it, not to mention its importance in producing comparative statistics, have had a strong impact
on the study of conflicts.8 As Daniel S. Geller and J. David Singer have noted, the number of territorial
states in the global system has ranged from fewer than 30 after the Napoleonic Wars to nearly 200 at
the end of the twentieth century, and it is essential to test the various indicators collected by peace sci9
entists against historical record until theoretical premises can be confirmed or rejected. Yet, a typical
feature in most studies of this type is that they are focused on finding those sets of variables that might
predict major wars and other conflicts, in a way similar to the historians’ origins of wars -approach,
whereas studies investigating the military spending behavior of monads (single states), dyads (pairs of
states), or systems in particular are quite rare.
The first facet of this investigation relate to the question of whether military spending “drives” economic development and/or vice versa. For example hegemonic theorists, such as Robert Keohane, Joseph S. Nye, Paul Kennedy, Charles Kindleberger, and Robert Gilpin, are among those who claim a
strong relationship exists between the pursuit of leadership and economic development. According to
Keohane and Nye, a state is likely to provide hegemonial leadership in the international regime if there
are benefits to be gained from such action, with the hegemonial power being able to change the rules of
the game rather than having to adapt to changes imposed by others.10 The hegemon may use coercion
(=stick) or positive incentives (=carrot) to achieve the goals that it seeks. This hegemon’s economic/political leadership can erode due to crises or shifts in the overall balance of power between the
states in the international regime. At such time, the so-called secondary powers, the followers, respectively react by altering their goals to challenge the leader’s position.11 Needless to say, this very abstract
one. On importance of time-specific frameworks, see Singer 1979, xvi—xix; McCloskey 1987, 14—15, 24;
Komlos 1992.
8
Singer 1979, xi—xviii; Singer 1990.
9
Geller-Singer 1998, e.g. 1—7.
10
Keohane-Nye 1977, 44—45.
11
Keohane-Nye 1977, 45—46; North 1990, 147—149; Eichengreen 1990.
5
theoretical framework has attracted both criticism12 as well as further theorizing in regards to more precise applications. All in all, most historical studies utilizing these arguments have focused specifically
on monetary markets and trade regimes (especially competing trade blocs).13 As far as historical instances of hegemonic leadership are concerned, there seems to be unanimous agreement that the post1945 period has been one of American hegemony, and with considerable agreement on the 19th century
being that of British hegemony.14
One of the less explored aspects in most studies of hegemonic patterns is the military expenditure component in the competition between the states for military and economic leadership in a system. According to Paul Kennedy, uneven economic growth levels cause nations to compete for economic and military prowess. The leader nation(s) thus has to dedicate increasing resources to armaments in order to
maintain its position, while the other states, the so-called followers, can benefit from greater investments in other areas of economic activity. Thus, the follower states act as free riders in the international
system stabilized by the hegemon. A built-in assumption in this hypothesis is that military spending
15
eventually becomes harmful for economic development; a notion that has often been challenged.
The
challenge of the leader’s economic/military position would begin when the hegemon has overreached
itself. Overall, the assertion arising from this framework is that economic development and military
spending are closely interdependent, with military spending being the driving force behind economic
cycles.
Moreover, based on this development pattern, it has been suggested that a country’s poor economic
performance is linked to the ”wasted” economic resources represented by military expenditures. However, as recent studies have shown, economic development is often more significant in explaining military spending rather than vice versa.
16
One may refer to the latter effect as the so-called war chest hy-
pothesis. As some of the hegemonic theorists reviewed above suggest, economic prosperity might be a
12
Rapkin 1990, 3—5; North 1990, 142—144; Kindleberger 1988; Eichengreen 1990, 273—275.
Kindleberger 1973.
14
Rapkin 1990, 8—9. See also Kennedy 1989 — interpretations based on military might or trade dominance
alone are more contested, such as the case of the Netherlands in the 17th century. Modelski-Thompson 1996 on
leadership cycles in general; Modelski-Thompson 1988 on seapower. On criticism of the imperial overreach
argument in the British case, see especially Hobson 1993.
15
Kennedy 1989, xiii; Hodne 1992, 80—82. See also Maddison 1991; Kennedy 1991; Kindleberger 1973.
16
On criticism of this mechanism, see the references in Eloranta 2001a.
13
6
necessary prerequisite for war and expansion. Thus, as Brian M. Pollins and Randall L. Schweller have
indicated, economic growth would induce rising government expenditures, which in turn would enable
higher military spending — therefore military expenditures would be “caused” by economic growth at a
certain time lag.
17
In order for military spending to hinder economic performance, it would have to
surpass all other areas of an economy, such as is often the case during wartime. In this paper, firstly, the
hypothesis that economic growth is “caused” by military expenditures (=the Kennedy argument, yet
only in the short run) and/or vice versa is explored by utilizing Granger non-causality tests on the economic development and ME (=Military Expenditure) variables for the selected countries in 1870—
1913.
What about systemic variables and changes — did they have an impact on the military spending patterns of the Great Powers in this period? Following the framework outlined by Buzan et al. (1998), the
levels of military spending analysis include: 1) International system, meaning the largest conglomerates
of interacting or interdependent units that have no system level above them; 2) International subsystems, such as alliances, meaning groups of units within the international system that can be distinguished from the entire system by the particular nature or intensity of their interactions with or interdependence on each other; 3) Single units, here referring to states, meaning actors composed of various
subgroups within a unit, sufficiently cohesive and independent to be differentiated from other such
units; 4) Subunits, meaning organized groups of individuals within the units that are able or wish to
affect the behavior of the unit, such as bureaucracies, lobbies. Buzan et al. also include a fifth level of
analysis, that of an individual, which is not pursued here.18 Here in this paper I will explore only the
levels 1—3.
As Daniel S. Geller and J. David Singer have put it: “…the global/international system is an evolving
one, with some of its properties changing slowly over time, others rapidly fluctuating, and still others
remarkably constant over the decades and centuries”.19 Hypotheses arising from relevant research on
systemic factors should therefore indeed be investigated to determine their role in the military spending
demand fluctuations. It is impossible to understand, for example, the rearmament of the 1930s without
17
Pollins-Schweller 1999, e.g. 445—446. E.g. Mintz-Huang 1990 suggest an indirect (negative) growth effect
via investment at a lag of at least five years.
18
Buzan et al. 1998, 5—6. See also Geller-Singer 1998, 20.
19
Geller-Singer 1998, 9.
7
the analysis of the preceding failure of centralized cooperation on disarmament among the members of
the League of Nations.
20
Yet, what kinds of hypotheses have been put forward concerning systemic
forces, whether representing increased or decreased stability?
It is often maintained in conflict studies that the warproneness of a system is contingent on the distribution of capabilities within the system; in essence, this is an extension of the realist argument of selfinterested behavior by the states in the system, with all states reacting similarly given the same resources and strategic opportunities.21 According to Geller-Singer (1998), factors increasing the probability of war at the system level include polarity (such as weak unipolarity and/or declining leadership), unstable hierarchies, the number of borders, and the frequency of civil/revolutionary wars. Factors increasing the severity of war at this level include especially high polarization between alliances.22
Coincidently, such effects and ensuing hypotheses have rarely been investigated as possible independent variables in explaining military spending behavior.
The effect of polarity in a system seems to be one of the key issues to analyze. As a concept polarity is,
however, quite contentious and ambiguous. Whereas some argue that a system dominated by a single
state is the most stable, it is not entirely clear how multipolar systems compare with periods of bipolar
hegemonic competition. It may be important whether the hierarchies in a system are well defined or not,
since challenges may be directed against the leading state or lesser states within an increasingly unstable international order. Thus, the best solution may be to analyze the concentration of resources, especially total resources available to a state as well as its military resources. It is possible that as the international system moves from a high concentration of resources in the leading state or among the leading
states towards multipolarity, as has been discovered empirically in terms of war occurrence, military
spending rivalry among the states is more likely to occur.23 Questions relating to these concepts are
pursued in connection with the quantitative analyses in the following sections. In short: a) what different explanatory variables affected the demand for military spending among the individual Great Powers; b) whether the systemic demand for military spending can be explained in terms of leadership (or
the lack of it) and other systemic characteristics in the short run?
20
See Eloranta 2002a.
Geller-Singer 1998, 26. For an overview, see also Levy 1985.
22
Geller-Singer 1998, 27—28.
23
Rapkin 1990, 3—8; Geller-Singer 1998, 115—118.
21
8
It is not, nonetheless, unproblematic to gather or construct reliable indicators for such endeavors. The
definition of military expenditures utilized in this paper abides by Frederick L. Pryor’s (1968) definition. Military expenditures by his definition include all expenditures for the recruiting, training, and
maintenance of an army, navy, air and rocket forces, and national security troops. He also excludes on
the basis of his selection of nations such items as expenditures on civil defense, veterans, military research and development, interest payments on war debts, reparations, military assistance abroad, and
military construction. Here we have also included civil defense measures and military construction if
they have been reported separately.24 In certain isolated cases it is possible to employ an economically
more precise definition, arising out of national accounting procedures. Using the expenditure approach
in classifying the GNP, military consumption includes all elements of military expenditure: wages and
government pensions (not war pensions), the purchases of non-durable goods and services (such as
heat, lighting, food, clothes), and the purchases of durable military goods (such as armaments). However, studies providing a breakdown of government consumption are not available for most countries
concerning, for example, the interwar period.25 For the period before the First World War, I will employ
the figures presented by John M. Hobson for most of the Great Powers due to their reliability and consistency for comparative purposes, as well as figures arising out of various historical national account26
ing projects and statistical compilations. All in all, it is possible to gather quite reliable military spending figures for countries previously often overlooked, such as Spain and Portugal. Only in cases where
specific statistics were not available did we employ such a statistical database as Singer-Small (1993),
which has been found relatively reliable on the basis of interwar period comparisons, or Banks
(1976).
27
In order to posit the military spending efforts of the selected countries in relative terms, we need to
utilize other indicators to construct: 1) military burden (=nominal ME of nominal GDP, at market prices
or at factor cost); and 2) defense share (=nominal ME of nominal central government expenditures,
24
Pryor 1968, 85—86.
On classifications of government consumption, see Clement 2000, 22—35. Studies on government
consumption concerning the states selected exist only for Sweden and Belgium. On further time series data
concerns, see Eloranta 2002a.
26
Hobson 1993. Further details can be found in Appendix 1.
27
In these databases the sources are usually not listed in a specific enough manner to “satisfy” a historian, yet at
least Singer-Small 1993 provides a certain level of detail on them and fare better in comparisons with the ME
figures collected by this author and the ones offered by the League of Nations than e.g. Banks 1976. Details can
be found in Eloranta 2002a.
25
9
CGE). These indicators can be obtained from similar sources as described above (see Appendix 1 for
details), although I have had to resort to indirect estimation of GDP for some of these countries (for
example: the pre-1885 period). As advised by many previously, one needs to exercise caution in order
28
to keep the numerators and denominators the same for all sample countries.
Here this has been possi-
ble for most of the countries. On the whole, the data has been deemed less reliable for Japan, Norway,
and Russia for the period 1870—1913.
Additionally, the nominal MEs and nominal GDPs have been converted to real terms to derive country
shares in the 16-country system. For the real GDP, I have relied mainly on Angus Maddison’s (1995)
dataset, with a few exceptions. However, the method which Maddison has used to come up with the
real GDP series in 1990 Geary-Khamis dollars can be criticized for distorting the results in comparison
with PPPs arising from the actual period in question, which are, of course, very difficult to come across
for a large set of countries. Here I have, instead, “corrected” the Maddison figures with the indirect
PPPs calculated by Leandro Prados de la Escosura (2000) by using his 1913 benchmark results on the
real GDP (per capita) for the period 1870—1913.29 The deflation and conversion of military expenditures into a common currency is an equally challenging and controversial task. One way is to adopt the
method used by SIPRI (Stockholm International Peace Research Institute) instead. The steps include at
first deflating the currencies respective of certain year, and then converting them all to a common currency for that year.30 In this study the deflator used in the conversions to real terms for the military expenditures has been the combination of the wholesale price index and the consumer price index, despite
their obvious weaknesses. Yet, in order to keep the error inherent in the approach for every country the
same, this approach seems to be the least problematic solution. In addition, the interwar comparisons
undertaken in Eloranta (2002a) indicate that this solution may in fact produce very similar results as
other, theoretically more sound deflators. The error in using the consumer price or the wholesale price
indices becomes more pronounced in long-run comparisons.31 Furthermore, two problems have
emerged concerning the use of the SIPRI method on conversion to a common currency: 1) The lack of
comprehensive set of PPPs for the interwar years or older periods; 2) Military expenditures do not nec-
28
See e.g. Singer 1990; Kennedy 1983; Hawke 1980, e.g. 27—36; Cullis-Jones 1987, 64—73.
See Prados de la Escosura 2000 and Appendix 1 here. On a recent analysis of PPPs since the late 19th century,
see especially Taylor 2000.
30
See Cars 1987, 78—82; Blackaby-Ohlson 1987, 3—24.
31
Hirvilahti 1993; Eloranta 2002a.
29
10
essarily comply to the price trends of other goods; thus, one would need to construct PPPs specific to
military expenditures.32 Here I have, nonetheless, employed a PPP-correction on the military expenditures, again utilizing the indirect PPPs calculated by Leandro Prados de la Escosura.
EXPLORING THE “WORLD SYSTEM”: Military Spending and Economic Development,
1870—1913
As mentioned briefly before, peace scientists in particular have studied countries forming a system,
often a “world system”, in regards to how prone these systems are to aggressive behavior. Here we will
construct an analogous system, based on the selected 16 countries, and recalculate some of the common
variables used in conflict studies as well as construct new ones. One could say that these countries in
fact did form a “world system”, since they formed 84,8 per cent of the “world” ME in 1913 and 87,5
33
per cent of the “world” ME in 1929.
More precisely, Europeans or the former colonies of Europe in
the Americas controlled 84 per cent of the earth’s land surface in 1914. As described by Samuel Huntington, “by 1910 the world was more one politically and economically than at any other time in human
34
history”. They were naturally equally dominant economically as well.
The 19th century introduced fiscal reforms for Western states such as reliance on balanced budgets,
innovations in public debt management, and direct taxation. These reforms were supported by the industrial revolutions and rising productivity levels, as well as accompanied by an industrialization of war
and armaments production from the mid-century on. The economic challenges posed by these changes
differed. In the French case, the mean defense share remained roughly the same as before in 1870—
1913, at circa 35 per cent, whereas its military burden rose modestly to 4,2 per cent. In the British case,
the mean defense share in 1870—1913 declined slightly to 36,7 per cent compared to the early 19th
century. However, the strength of the British economy actually enabled a slight military burden decline
to 2,6 per cent, a similar figure incurred by Germany in the same period. For most countries the period
leading to the First World War meant comparatively higher military burdens. The United States, the
new economic leader, incurred a meager 0,7 per cent average military burden, similar to the interwar
period. In the First World War (1914—1918), this military potential was unleashed in Europe, with a
32
See e.g. Herrera 1994, 18; West 1993.
Calculated using the most comprehensive military spending database available: Singer-Small 1993.
34
Huntington 1997, e.g. 50—53; Maddison 1995. See also McNeill 1982 in particular.
33
11
war of attrition causing immense suffering and property damage amounting to perhaps 36 billion
USD.
35
Table 1. Estimates of Military Exertions for Five Great Powers in the First World War
A.
FRA
GER
RUSSIA
UK
USA
B.
1914—1918
1914—1918
1914—1916
1914—1918
1917—1919
C.
79
83
76
63
54
D.
43
..
..
22
7
E.
77
91
..
49
47
F.
11
7
4
7
2
Sources: Column C from Hardach 1977; column D calculated from Fontvieille 1976 (ME for France), Morgan 1952 (ME for
UK), Historical Statistics 1975 (ME for USA), Mitchell 1998b (GDP for France and UK), Historical Statistics 1975 (GNP
for USA); column E calculated from the same ME sources as column D except National Capabilities database (Singer-Small
1993) (ME for Germany), as well as Mitchell 1998b (CGE for France, Germany, and UK), Historical Statistics 1975 (federal
government outlays for USA); column F calculated from the National Capabilities database (Singer-Small 1993). A=country;
B=years; C=percentage, total war expenditures of total public expenditures; D=mean military burden; E=mean defense share;
F=mean percentage, armed forces of population.
The military spending implications of this “global war”36 have not been explored yet. The militaryeconomic dimension of the war has not been studied very thoroughly so far, compared to the Second
World War for example. As seen in Table 1, it is possible to compile some preliminary indicators on
the extent of the economic mobilization, central government spending, and the demographic strain
placed on the population of the Great Powers by the mobilization of large standing armies. It seems that
columns C and E, by and large, match one another. The highest public sector or central government
military spending share was borne by Germany, with France and Russia following close behind it. The
American defense share, for example, was low in comparison, due to the brevity of its involvement in
the war. The same observation could be made, more or less, for the mean military burdens of the war
years. Notably, however, the British share was quite low, below fifty per cent. Finally, in the demographic sense, the French mobilization respective of the population was the most extensive, eleven per
cent of the population on the average, whereas the United States had to mobilize only circa two per cent
of its population to its armed forces. Still, Germany and Russia suffered similar numbers of casualties
35
E.g. Eloranta 2001b; Webber-Wildawsky 1986; McNeill 1982. On earlier analyses of public and military
spending, see especially The Rise of the Fiscal State in Europe c. 1200—1815 (1999). The First World War
damage estimate from de Groot 2001.
36
On the concept of global war and its occurrence in history, see especially Rasler-Thompson 1989, 127—151.
12
in the war (1,8 and 1,7 million respectively), whereas the British lost significantly less men in this conflict (0,9 million).37
The First World War imposed its own constraints on the budgetary decision-making of Western states.
Many states, like Germany and France, were unwilling to fund the war with high taxes, contrary to for
example the United Kingdom, but preferred the sale of debt as a solution instead. Even the United
Kingdom and the United States discarded central controls and high taxes after the war. The war also
enhanced preexisting tendencies for higher public spending, as social programs, for example in France
and Britain, were expanded and new programs added. Aggregate spending levels thus tended to remain
higher after the war compared to the prewar period. Another factor contributing to this development
was the growth of state bureaucracies. According to the displacement hypothesis first expounded by
Peacock and Wiseman, people get used to paying higher taxes during wars, which in turn leads to
higher spending and taxation level afterwards. This indeed seems an adequate description of the fiscal
impact of the First World War. Another feature of the period was the persistence of balanced budgeting
and, especially in the U.S., efforts to introduce public spending cuts based on conservative policies. The
factors outlined above also tended to make public spending quite rigid, due to the counterbalancing
tendencies.38
What about the role of the so-called hegemons in this period? It has often been argued that the British
leadership in the 19th century became too expensive to maintain, a reference to the so-called imperial
overstretch argument. At the heart of this argument lies the proposition that Great Britain simply had to
devote too extensive military resources into defending the Empire. Nonetheless, it is equally often argued that the British army was simply just too small compared to her continental rivals. In an illuminating analysis John M. Hobson has proven conclusively that the military expenditures incurred by Great
Britain were small in relative terms compared to the other Great Powers of the period, with the exception of the United States, and that it is difficult to maintain that the British military commitments caused
39
the decline of her hegemonial status.
George Modelski and William R. Thompson perceive the roots
of the British decline not only in the economic catch-up of its rivals but also in erosion of its lead in
37
On the casualties, see Correlates of War, Inter-State War Data 2000 — France lost circa 1,4 million and the
United States “only” 0,1 million men in the First World War.
38
See for example Webber-Wildawsky 1986, 431—451; Rasler-Thompson 1989; Eloranta 1998.
39
See especially Hobson 1993 and the key studies scrutinized in it.
13
naval technologies, leading to a naval armaments race in the late 19th century and early 20th century.
Thus, despite being able to maintain a strategic dominance in the naval standoff during the war, “Britain emerged from the First World War no longer the world power and too poor to maintain its longstanding naval leadership”.
40
The role of the United States was that of a reluctant hegemon, especially
in terms of providing foreign policy leadership in this period. Prior to the last decade of the 19th century, the United States adhered to an isolationist foreign policy stance, but subsequently U.S. companies
embarked on a search for foreign raw materials. Thus, between 1896 and 1941, the United States pursued an expansionist policy, yet directed its activities toward economic and strategic goals in the Caribbean and the Pacific.
41
Table 2. Real Military Expenditure (=A) and Real GDP (=B) Percentage Shares of France, Germany, Russia/Soviet Union, the United Kingdom, and the United States in a System of 16 Countries, 1870—1938
FRA
A.
B.
YEAR
12,17
15,35
1870
12,05
15,63
1875
13,95
13,75
1880
15,33
13,18
1885
11,44
12,58
1890
11,04
11,70
1895
10,09
11,06
1900
4,77
9,78
1905
9,41
8,51
1910
10,10
8,88
1913
Sources: see Appendix 1.
GER
A.
B.
22,15
13,58
13,79
13,20
18,07
14,11
12,92
7,34
15,00
15,33
8,32
8,63
7,79
8,13
8,28
8,32
8,33
8,15
7,94
7,90
A.
RUS
27,40
28,07
31,16
27,01
26,95
27,86
25,27
44,96
27,56
30,78
UK
B.
A.
15,74
15,53
14,44
13,74
14,10
16,33
18,22
18,94
21,99
23,21
10,95
10,46
12,44
12,99
12,38
11,40
17,29
7,84
11,87
10,21
B.
16,17
15,75
15,16
14,66
15,07
13,90
13,28
12,17
10,93
10,47
USA
A.
B.
6,67
5,42
4,84
5,06
5,02
5,16
9,31
5,22
9,87
8,20
18,88
19,52
24,10
25,17
25,61
26,07
26,70
29,11
28,91
28,68
Table 2 displays the real GDP and real ME shares of the most important Great Powers in this period.
Similar to the modified CINC-scores discussed below, the British real GDP share demonstrated a
steady decline in the period and the U.S. share exhibited rapid growth respectively, bringing it to a clear
lead according to all economic indicators over Britain in the latter part of the 19th century. Russia’s real
military spending share in turn was clearly the highest in the period 1870—1913, usually equaling and
surpassing the ME shares of any two other Great Powers combined. The American real ME remained
low comparatively throughout the pre-First World War period. Quite surprisingly, though, the Russian
real GDP share was relatively high in the period preceding the First World War, although not as high as
40
41
Modelski-Thompson 1988, 210—211.
North 1990, 117—118.
14
the CINC-scores would suggest (see Figure 1). The total resource share, the so-called CINC
(=Composite Index of National Capabilities), is usually calculated as an average of six series: the share
of military personnel, the ME share, the energy consumption share, the iron and steel production share,
42
the total population share, and the urban population share.
Here I have decided to replace the energy
consumption share, which may be a poor proxy for economic stature in a system, with the real GDP
share explained earlier.
43
Equally, the ME share is the real ME share of a country in the system (of 16
states). Thus, the military resource share of a country (abbreviated: MILCINC) is calculated as an average of only the military components in the CINC (the military personnel share and the real ME share).
The dilemma of Russia’s strong showing in the CINC-scores and the military resource shares has been
astutely addressed by William C. Wohlforth. An important aspect in these measures is whether they
mirrors perceptions of, in this case Russia’s, power among the Great Powers. As he argues persuasively, none of the others thought of Russia as a superior power to Germany or Great Britain before the
First World War. It seems, nonetheless, that Russia’s allies actually possibly overestimated Russian
power, whereas Germany and Austria underrated it. Wohlforth places the most explanatory value on the
variables comprising the military resource share, although even for the military components several
other aspects affected the credibility of Russia’s (military) power: 1) Russia’s political and military
inefficiencies; 2) slow mobilization capabilities; 3) lack of internal societal cohesion; 4) Russia’s difficulties of withstanding a long war (which was not thought to occur anyhow); 5) Russia’s inability, due
44
to many of the factors already mentioned, to wage an offensive war.
On the other hand, it may be
difficult to separate the perception of defensive and offensive capabilities in macro-level estimations.
And, to be certain, Russia possessed immense defensive capabilities that were displayed in the two
World Wars and did engage in significant military reforms after its shocking defeat against Japan in
1905. This critique of the use of the CINC-scores notwithstanding (especially regarding the futility of
trying to assess the probability of war with them), it may however be plausible that these countries reacted to such perceptions in their military spending decision-making. In addition, following Wohlforth,
42
See Appendix 1 for details.
In addition, energy consumption appears to be highly correlated with economic growth (see Smil 1994, e.g.
206), yet it is hard to argue it would represent national economic resources better and more accurately than the
concept of (real) GDP.
44
Wohlforth 1987. This article is an illustrative critique of the use of these aggregate indices of power
distribution.
43
15
it is here expected that the military resource shares will more likely be statistically significant than the
aggregate CINC-scores in the system estimations.
Figure 1. Modified CINC-scores for Germany, Russia, the United Kingdom, and the United
States, 1870—1913
50
%
45
40
35
30
25
20
15
10
5
0
1870
1876
1882
1888
1894
Year
1900
GER, MILCINC
USA, MILCINC
UK, MILCINC
RUS, MILCINC
1906
1912
Sources: see Appendix 1.
In terms of MILCINCs, the Russian share was again the largest, eclipsing the others in terms of quantity at least. Comparatively, for example, Germany and the UK maintained rather steady military resource shares in the system. Of the other Great Powers investigated here, the Austrian MILCINC declined steadily during the period, the French share fluctuated yet showed a declining trend (like the
Italian one), and Japan in turn emerged as a rapidly growing Great Power from the 1890s onwards. As
seen in Figure 2, the aggregate real military spending of the period showed a clear upward trend, accelerating from the early 1890s onwards. Moreover, the war between Japan and Russia produced a peak in
the years 1904—1905, possibly destabilizing the system and feeding the quickening arms race in the
last ten years before the First World War. This also emerged through the military spending shares of
16
democracies versus autocracies in this period. The ME share of autocracies on the aggregate (in the
system of 16 states) decreased until the Russo-Japanese war, and then began to increase again.
Figure 2. Aggregate Real Military Spending (in GBP) for the System of 16 States, 1870—1913
SYSTEM ME (IN
1913 GBP),
BILLIONS
1,4
1,2
1,0
0,8
0,6
0,4
0,2
0,0
1870
1876
1882
1888
1894
1900
1906
1912
Year
SYSTEM ME (IN 1913 GBP)
Sources: see Appendix 1.
It has also been suggested, for example, that the effect of system-level capability concentration, with
capabilities concentrating in the hands of major powers (or just one hegemon), might have an enhancing decision-making certainty effect, although there is no consensus on this. A standard way in the conflict research literature to measure capability concentration is:
Nt
CONC t =
∑(S
i =1
it )
2
− 1/ Nt
1 − 1/ Nt
(1)
where Sit equals the proportion of the aggregate capabilities possessed by a major power in year t;
Nt=the number of major powers in the system in year t. This index takes a value from 0 to 1. Although
many studies have indicated that system-level capability concentration is unrelated to the occurrence of
17
a major power war, this system indicator has not previously been tested as a determinant of military
spending.
45
Figure 3. Concentration of Total Resources, Military Resources, and System ME, 1870—1913
INDICES (0-1)
CV
0,50
2,0
0,45
1,8
0,40
1,6
0,35
1,4
0,30
1,2
0,25
1,0
0,20
0,8
0,15
0,6
0,10
0,4
0,05
0,2
0,00
1870
0,0
1876
1882
1888
1894
1900
Year
INDEX OF CONCENTRATION, CINC
INDEX OF CONCENTRATION, MILCINC
SYSTEM ME, CV
1906
1912
Sources: see Appendix 1.
What in fact took place in the international system during this period? Did economic or military resources become more concentrated among the Great Powers? If we look at the index of concentration
for the CINCs, it seems to confirm the notion of hegemonic leadership emerging in the economic
sphere in the latter part of the period (see Figure 3). However, military spending (measured by a coefficient of variation) and MILCINCs displayed a fairly flat trend, with the exception of the war years in
1904—1905. Whereas economic leadership by the United States became evident towards the end of the
period, military leadership was not forthcoming and hegemonic competition was the outcome.
45
Geller-Singer 1998, 122.
18
The hegemonic framework, in the form advocated by Paul Kennedy, implies that military spending and
economic growth are interdependent due to the wasted economic resources embodied by military expenditures, often presuming a causal influence of military expenditures on economic development.
How much is too much? Although the analysis above provides us with some clues to those who might
be overspending, it is certainly only indicative. Nonetheless, for example, the American interwar military burden was, except for 1920—1922, between 0,6 and 1,3 per cent, whereas during the 1950s the
46
American military burden was often over ten per cent.
Thus, it must be that the meager burden im-
posed by the military spending of the interwar years could not have been very significant in the development of the American economy. The conclusion could be the exact opposite: military spending was,
in fact, in line with the war chest argument, dependent on the development of the economy and economic rivalry in general. Firstly, we can attempt to verify the “causal” links between economic devel47
opment and ME by applying Granger non-causality tests.
Here these relationships were tested for the following three pairs of variables for the 16 countries included in the systemic estimations: defense share and real GDP per capita; military burden and real
GDP per capita; individual country real ME share and individual country real GDP share. Of these
variables, the defense share can be expected reveal a “budgetary response”, the military burden a “direct economic impact”, and the real ME share a “systemic response” to economic changes. In order to
provide a consistent framework, I have applied the following rules to the analysis: 1) two out of the
three pairs of Granger-causality relationships would have to indicate the same direction of causation for
the results to be deemed credible; 2) Granger-causality relationships found at more than one lag structure, especially since such relationships at t-1 may be poorly representative, were deemed more reliable
than others.
46
48
See e.g. Stiglitz 1988, 41—42.
Granger non-causality tests have been applied to military spending analysis e.g. in Eloranta 2001a; Chowdhury
1991. For long run applications, see especially Rasler-Thompson 1991.
48
The rating scheme is as follows: 1) ECONOMY→ ME: 2 pairs of variables at more than one lag in the same
direction = weak evidence; 2 pairs of variables at more than one lag in the same direction plus at least one of pvalues below 0,01 or 3 pairs at more than one lag in the same direction = strong evidence; 2) ME →
ECONOMY: 2 pairs of variables at more than one lag in the same direction = weak evidence; 2 pairs of variables
at more than one lag in the same direction plus at least one of p-values below 0,01 or 3 pairs at more than one lag
in the same direction = strong evidence; 3) INTERDEPENDENCE: 2 (in one direction) + 1 (in the other
direction) pairs of variables at more than one lag = weak evidence; 2 (in one direction) + 2 (in the other
direction) pairs of variables at more than one lag = strong evidence; 4) INDEPENDENCE: 1 (in one direction) +
1 (in the other direction) pairs of variables at more than one lag = weak evidence; all other cases not meeting
47
19
Due to potential problems of autocorrelation and nonstationarity, the logarithmic forms of these variables were preferred. The assumption of stationarity, based on ADF-unit root tests, holds for most of
the variables in this period, with the exception some of the time series (details on differencing can be
found in the tables).
49
All Granger non-causality tests were applied to a maximum amount of time lags,
given the period (14 lags in 1870—1913). The results are summarized in Table 3, whereas the detailed
findings can be observed in Appendix 2, Table 2.
Table 3. Results of the Granger Non-causality Tests, 1870—1913: Summary of the Findings
ECONOMY→ME
ME→ECONOMY
INTERDEPENDENCE
1. 1870—1913:
AUT**, NED**,
ITA*
FRA**, GER**, NOR*,
POR**, RUS**
SPA*, SWE**, SWI*
Sources: See Appendix 2, Tables 2—3 for details.
Note: ** = strong evidence; * = weak evidence.
INDEPENDENCE
BEL**, DEN**, JAP*, UK*,
USA*
For the period 1870—1913, the results suggest a cautious validation of the war chest hypothesis inasmuch that the economy seems to influence military spending or that there is two-way causation between
the variables. The notion that military spending might have an influence on the economies of these
countries is clearly rejected for both periods, at least in the short run. Moreover, the war chest hypothesis is markedly less supported by the interwar comparisons. These results cast considerable doubt on the
idea that “high” military spending may be the driving force in economic development, at least in the
short term. Moreover, it may be possible that economic development was an important variable in explaining the demand for military spending, both at the individual country level and at the level of the
constructed system.
THE DEMAND FOR MILITARY SPENDING: Individual Great Powers and the System,
1870—1913
There is a plethora of ways of estimating the war-making capabilities of states, as we have reviewed
earlier. Different indicators measure different qualities of the coercion potential (excluding the exercise
these minimum requirements = strong evidence. Thus, the ratings in Table 3 should be reviewed in connection
with Appendix 2, Table 2. For more on these tests, see Eloranta 2002a.
49
ADF = Augmented Dickey Fuller. The results of the tests on the stationarity of the series can be obtained from
the author by request. On other applications, see especially Harris 1995. Using the so-called Johansen test, no
cointegration vectors were found.
20
of coercion within states) that states possess, aimed at securing their external interests. Military expenditures capture only one dimension of a state’s military activities and preparedness. The determination
50
of the level of ME is always a domestic choice, based on domestic concerns and external perceptions.
As Karen A. Rasler and William R. Thompson have suggested, the estimation of the war-making capabilities of states should include variables such as the cost and size of government, power concentration,
cultural heterogeneity, military organization, physical range, resource endowment, and the economy’s
51
level of commercialization. Defense economists in turn often employ demand models arising from the
theory of public goods in estimating the demand for military spending. Based on the utility maximization of an individual between a private good and a public good52, the subsequent basic (single-equation)
linear function is estimated:
MEit = β i 0 + β i 1 INCOMEit + β i 2 PRICEit + β i 3 SPILLINi , t −1 + β i 4THREATi , t −1 + ε it
(2)
in which ME stands for military expenditures for agent i in year t; INCOME for example GDP per capita; PRICE for the price development of military goods; SPILLIN (lagged) for spillovers from both actual defense alliances and free-riding based on perceived increased security; THREAT (lagged) is the
perceived defense expenditure of a potential enemy or enemies.53 The demand for ME is then estimated
either for a single state or a system of states. These types of estimation efforts have not usually included
systemic variables or attempted to investigate the possibility of system-wide common characteristics.
There are certain characteristics that piece scientists have discovered regarding systemic behavior and
the probability of war, which may or may not have an effect on the demand for military spending at the
level of a system. The aim here is to attempt to combine some of those characteristics with the basic
demand equation outlined above, and to test whether these countries exhibited certain common characteristics in their military spending behavior, respective of the various perceptions. Firstly, I will test the
demand for military spending individually, respective of the two dependent variables: defense share and
50
See e.g. Eloranta 2002b; Eloranta 2001c; Morrow 1993. On a survey of the vast literature on defense
economics and public choice implications, see Sandler-Hartley 1995. The impact of domestic forces, e.g. the
political markets is not pursued here due to the systemic perspective adopted. On domestic market interaction,
see Eloranta 2001c.
51
Rasler-Thompson 1989, 8.
52
For the appropriate derivation of this demand function, see Sandler-Hartley 1995, 53—60, and Cornes-Sandler
1996, 484—487.
53
For further details, see e.g. Sandler-Hartley 1995, 60—62.
21
military burden. Then I will evaluate the demand for military spending, utilizing the same dependent
variables in pooled format, in a system of 16 states. Thus, the following equation was chosen as the
54
basis for the estimations in this period :
+
−
+
?
?
?
ME = β 0 + β 1 INCOME+ β 2 CINC+ β 3 MILCINC+ β 4UKCINC+ β 5USACINC+ β 6UKMILCINC
+
?
?
?
... + β 7USAMILCINC+ β 8 SYSTOTME+ β 9 SYSTOTMECV+ β 10CONCCINC
?
?
?
?
?
(3)
... + β 11CONCMILCINC + β 12 DEMCINC+ β 13 DEMTOTME+ β 14 ATCCINC+ β 15 ATCTOTME
−
... + β 16− 31 ALLIANCEDUM + ε
The income variable was represented either by the individual or pooled real GDP per capita, similar to
Equation (2), and should have exhibited a positive coefficient if the ME variable was a normal good.
Prices were not brought in as a separate variable, since the variables were already in real terms. The
individual country CINCs and MILCINCs were included because these countries may have responded
to changes in their economic and/or military status in the system. The CINC is hypothesized as having a
positive coefficient, thus implying a military spending adjustment to declining system stature (along the
lines of the hegemonic theorists). Correlation is in turn expected between the ME and the MILCINC
variables. It is equally important to analyze the determinants of ME in terms of leadership, or the lack
of it. Thus, I have included the CINCs and military resource shares of the two systemic economic giants, the United Kingdom and the United States. The key variables to watch are the UK variables for
1870—1913 and the U.S. variables for 1920—1938. If the coefficients of these variables were to be
negative, the other states acted as challengers (with larger states reacting more strongly and smaller
states following suit). Systemic threat should emerge especially via the total system ME variable, with a
positive threat adjustment to be expected.
In addition, it may be that as the military spending of the 16 countries becomes more dispersed
(SYSTOTMECV), the countries in question will experience a mounting perception of threat (=positive
coefficient). Nonetheless, the data-based evidence on the impact of polarity is definitely mixed, which
is perhaps better embodied by the concentration of resources (as represented by Equation (1)). Hegemonic theorists often argue that a system dominated by a single state in a unipolar system is the most
stable as it reduces the necessity of states to spend more on the military establishment (resulting in a
54
Details on the variable and other abbreviations can be found in Appendix 2, Table 1.
22
negative coefficient for the concentration variables). It may be, however, that this assumption depends
55
on whether the hierarchies among the states are clear. Thus hegemonic competition in the system may
be implied by a positive coefficient for the concentration of CINCs (CONCCINC) and MILCINCs
(CONCMILCINC).
Here, in Equation (3), I will additionally analyze possible sensitivity to the aggregate average CINCs
(weighted by the individual country military resource shares) and total military spending of democracies
and autocracies, as either possible sources of spillovers or threats in the system level. The corresponding coefficients will be reviewed in connection with the results, as it is not obvious which signs they
would incur. And, in order to bring the impact of alliances at least superficially into the analysis, indi56
vidual country alliance dummies are utilized.
It seems that evidence regarding system-level alliances
and warfare is clear in conflict studies: the onset of war (occurrence/initiation) of war is unrelated to
either alliance formation (aggregation) or configuration (polarization), yet the magnitude, the duration,
57
and the severity of war are consistently correlated with alliance configuration.
Nonetheless, the alli58
ance impacts of military spending have hardly been investigated before for these two periods.
It is
here argued that the individual country alliance dummies will enable us to estimate the common systemic responses better, especially in terms of reducing the impact of specification errors.
The individual country regression results warrant some observation on the aggregate (see Table 4).
Firstly, it seems that the two dependent variables produced largely different results, which would suggest that the structural variable (MILBUR) and the budgetary variable (DFSHARE) did not have the
same demand structure in these political economies. Secondly, the income variable did not consistently
produce a positive coefficient, which may at first seem surprising. However, as the interwar comparisons undertaken in Eloranta (2002a) prove, military expenditures were not an income-normal good in
most cases. Thirdly, two of the variables, MILCINC and SYSTOTME, incurred the expected coefficient in all of the statistically significant cases, thereby confirming the hypothesized effect of these variables. Fourthly, all in all, Austria, Germany, Italy, and Russia in many instances reacted to the same
55
See e.g. Geller-Singer 1998, 115—117.
See Appendix 1 for details. Alliance dummies applied to specific countries’ data only, not the cross-section.
57
Geller-Singer 1998, e.g. 119—120.
58
The only thing that comes close is Conybeare-Sandler 1990 for the pre-First World War period, yet the
dependent variable in it is military personnel instead of ME. On alliance dynamics, see also Morrow 1993.
56
23
variables (with also UK and Japan joining the group for some of the variables). On the other hand,
France and the USA (with UK aligning with them at times) seemed to form the “opposing team”.
Nonetheless, these impressions were not consistently confirmed by the individual cases, since the coefficient signs differed. Fifthly, the impact of alliances was not recorded in the individual cases.
Table 4. Signs of the Independent Variables in the Individual Country Military Spending Demand Regressions, Eight Great Powers, 1870—1913
COEFFICIENTS SIGNS
(DEPENDENT VARIABLE
MILBUR)
FRA (+), GER (-), ITA (+), RUS (-),
UK (-), USA (+)
AUT (+), RUS (-)
INDEPENDENT VARIABLE
(EXPECTED SIGN IN
BRACKETS)
INCOME (+)
COEFFICIENT SIGNS
(DEPENDENT VARIABLE
DFSHARE)
FRA (+), UK (-)
CINC (-)
AUT (+), GER (-), JAP (-), RUS (-),
UK (-), USA (+)
FRA (+), GER (+), ITA (+), JAP (+), AUT (+), FRA (+), GER (+),
RUS (+), UK (+), USA (+)
ITA (+), JAP (+), RUS (+), UK (+),
USA (+)
AUT (+), GER (-), RUS (-), USA (+) AUT (+), GER (-), RUS (-), USA (+)
AUT (-), GER (-), JAP (-)
AUT (-), GER (-), JAP (-)
FRA (-), USA (-)
GER (+), JAP (-), RUS (-), USA (-)
AUT (+), FRA (+)
FRA (+), UK (+), RUS (+)
AUT (+), FRA (+), GER (+),
FRA (+), GER (+), ITA (+),
ITA (+), JAP (+)
RUS (+), UK (+), USA (+)
JAP (+), RUS (+), UK (+)
AUT (-), GER (-), ITA (-), JAP (+),
RUS (+)
JAP (-), USA (+)
AUT (+), GER (+), JAP (+),
RUS (-), UK (+)
ITA (-), JAP (-), RUS (-), UK (+),
AUT (+), GER (+), ITA (+), UK (+),
USA (-)
USA (-)
AUT (+), RUS (-), UK (-), USA (-)
AUT (-), ITA (-), JAP (-), UK (+)
AUT (-), GER (+), RUS (-), UK (-)
AUT (-), ITA (-), JAP (-), UK (+)
AUT (+), FRA (-), ITA (-), JAP (-), FRA (-), GER (-), ITA (+), UK (-)
RUS (+)
JAP (-), RUS (+)
AUT (+), GER (+), ITA (-), UK (+)
AUT (-), FRA (+)
FRA (-)
MILCINC (+)
UKCINC (?)
USACINC (?)
UKMILCINC (?)
USAMILCINC (?)
SYSTOTME (+)
SYSTOTMECV (?)
CONCCINC (?)
CONCMILCINC (?)
DEMCINC (?)
DEMTOTME (?)
ATCCINC (?)
ATCTOTME (?)
ALLIANCEDUM (-)
Sources: see Appendix 1.59
Note: on variable specifications and definitions, see Appendix 2, Table 1. Only statistically significant variables’
signs listed here.
Of the individual countries, Austria displayed challenger behavior towards the democracies and followed the British lead. France in turn was one of the few to respond to military expenditures as an income-normal good; it also seemed to react as a challenger towards the British MILCINC. Germany also
59
Individual country results in their entirety available from the author by request.
24
behaved as a challenger of the declining British hegemony (hence the negative sign for the UKCINC
variable, with a large coefficient). Italy’s position in the system was quite ambiguous, as already suggested in its alliance behavior by Conybeare-Sandler (1990), and it seemed to react opposite to the democracies in its military burden in particular. Japan adjusted its ME upwards when its stature in the
system would slip, as expected. Both the American and the British systemic changes brought forth a
reaction in the Japanese military spending, which is consistent with its fast rise to the group of Great
Powers and its awaking hegemonic ambitions in the Far East. Russia, with its extremely high military
spending levels, quite logically did not display statistical significance to changes in income. It was also
one of the states to act in a predatory fashion when SYSTOTMECV increased, along with Japan and
UK. Britain, in its choices, took the role of both a follower and a challenger towards democracies in
general. Additionally, it was clearly one of the states involved in the hegemonic power struggle of the
period. This seemed to hold for most of these Great Powers.
In order to estimate Equation (3) jointly, I decided to apply both the seemingly unrelated regressions
(SUR) method, which estimates the parameters of the 16-country system by correcting for heteroskedasticity and contemporaneous correlation in the errors across equations, as well as three-stage least
squares (3SLS), which allows for a correction with instrumental variables if one of the right-hand side
variables is endogenous in the estimated SUR system. As Todd Sandler and Keith Hartley have noted,
the SUR technique may be appropriate when a nation is a member of an alliance and demand equations
60
are estimated for multiple allies.
In order to improve the coefficient estimates, especially since there
61
was ample reason to believe that the country-specific MILCINC was endogenous , I also estimated the
3SLS version of the equation
62
(using the exogenous variables as well as the real GDP share and the
military burden or the defense share as instruments).
Furthermore, in order to verify whether the inclusion of the countries with more dubious data (1870—
1913: JAP, NOR, RUS influenced the underlying SUR system, I additionally estimated the SUR for the
13 other countries. If the variables did not appear statistically significant on the level, up to two lags
were introduced. The equations were corrected for autocorrelation, with AR(1) arising as the most
60
Sandler-Hartley 1995, 62. On an application of this method, see e.g. Murdoch-Sandler 1986.
Suggested initially by the Hausman test, performed on the military burden variant of Equation (3).
62
One example of using the 3SLS, as well as other methods, in military spending demand estimations is Hewitt
1996.
61
25
common additional variable. Although the estimated systems were expected display certain joint responses, one would have to be careful not to place too much emphasis on these estimates alone, due to
us forcing common response coefficients for most variables.63
Table 5. 3SLS (Three-Stage Least Squares) Estimation of the Determinants of Pooled Defense
Share and Military Burden in a 16-State System, 1870—1913
INDEPENDENT VARIABLE
COEFFICIENTS
(DEPENDENT VARIABLE
DFSHARE)
1.130***
0.234***
-0.619***
0.808***
0.587***
-0.183**
-0.310***
-0.096*** (t-1)
0.222***
-0.081***
0.161***
0.071***
-0.273***
0.092***
-0.379**
-0.038*
0.073***
-0.100***
†
COEFFICIENTS
(DEPENDENT VARIABLE
MILBUR)
†
-0.128***
-0.227***
1.077***
0.447***
-0.106*** (t-1)
-0.132***
0.559***
-0.215***
0.441***
0.071*** (t-1)
-0.047***
0.153***
-0.143***
0.057***
0.103***
-0.044**
0.822***
CONSTANT
POOLED GDPCAP
POOLED CINC
POOLED CINC
UKCINC
USACINC
UKMILCINC
USAMILCINC
SYSTOTME
SYSTOTMECV
CONCCINC
CONCMILRES
DEMCINC
DEMTOTME
ATCCINC
ATCTOTME
AUT ALLIANCE DUM
BEL ALLIANCE DUM
DEN ALLIANCE DUM
FRA ALLIANCE DUM
GER ALLIANCE DUM
ITA ALLIANCE DUM
JAP ALLIANCE DUM
NED ALLIANCE DUM
NOR ALLIANCE DUM
POR ALLIANCE DUM
RUS ALLIANCE DUM
SPA ALLIANCE DUM
SWE ALLIANCE DUM
SWI ALLIANCE DUM
UK ALLIANCE DUM
USA ALLIANCE DUM
AR(1)
Sources: see Appendix 1.
Note: on variable specifications and definitions, see Appendix 2, Table 1. Only statistically significant coefficients are listed. † = cross-section specific, coefficients not listed here. * = null hypothesis of no correlation rejected at 10 per cent level; ** = null rejected at 5 per cent level; *** = null rejected at 1 per cent level.
63
In spite of EViews 3.1 allowing us to introduce some cross-correlational elements, the SUR system is most
often used as a way of pooling only some of the countries within the sample, according to coefficient equality
tests. Here, however, we were more interested in evaluating joint systemic determinants of ME.
26
Nonetheless, the 1870—1913 equations did not seem to be suffering from extensive specification errors. The results of the 3SLS estimation for the period 1870—1913 can be found in Table 5.64 In order
for the results to be more credible, in both of the estimations (on DFSHARE and MILBUR) the coefficients had to contain the same sign for us to interpret the results as convincing. Thus, for example, in
the estimations for both periods the results for the real GDP per capita were deemed ambiguous. Firstly,
I evaluated the likeness of the 16-country SUR systems against the results of the 13-country SUR systems, then the similarities between the 16-country SURs and the 3SLS. By comparing the sizes and
signs of the coefficients, the comparisons between the systems equations in the period 1870—1913
indicate that they remain quite stable from one system to another. Also, the 16-country system’s adjusted R2 seems much lower than the one incurred by the 13-country system, indicating perhaps a degree of error introduced by the additional data or by the exclusion of variables relating to the more specific immediate threats and spillovers perceived by the individual countries. Another important set of
65
variables missing from these estimations relate to the political markets of these countries.
For the period 1870—1913, the individual country CINCs displayed a negative coefficient, indicating a
military spending adjustment in response to a loss of economic stature in the system. The MILCINC
incurred a positive coefficient, which is perhaps logical due to this variable being endogenously determined with the ME variable. The CINC of the United Kingdom, as we have seen, declined during this
period, inducing a large positive ME response, indicating challenger behavior. The situation was reversed in the case of USACINC, with a smaller coefficient, perhaps due to more limited impact of the
perception of American leadership. The UKMILCINC also points to the same conclusion as in the case
of UKCINC above: responsiveness to declining British leadership in the system. Increased total ME in
the system was perceived as a threat, as expected. The concentration of resources in the system, measured by the various variables, did not end up statistically significant for this period. Yet, when we take
the individual country results and those of the pooled system, hegemonic power struggle was definitely
taking place in the international system in 1870—1913. Nonetheless, the extent of this struggle varied
from one Great Power to another. Additionally, similar to the individual country results, the alliance
dummies pointed to mostly conflicting conclusions as far as the signs are concerned.
64
The SUR system equation results corresponding to the 3SLS results, both for the 16 and the 13 country
systems, can be found in Appendix 2.
65
See Eloranta 2002a for further discussion.
27
CONCLUSIONS
It is highly problematic to attempt analysis of military spending from the perspective of leadership or
military-economic interaction alone. Military expenditure analysis should try to combine both external
factors and internal factors in the explanations. This also corresponds well with the analysis of any public
66
good. The definition of a public good , arising from the free-rider dilemma in its production, excludes
simple explanations. Military spending decision-making, often a controversial issue within the political
spectrum, is also subject to the same historical and institutional continuities and discontinuities as other
fields of public policymaking. Furthermore, unlike national defense, military spending should be exam67
ined as an impure public good.
It should also be noted that in most countries military establishments also obtained funding from various sources, which were not always included in the aggregate spending figures. In the United States, for
example, the Navy’s rearmament drive was, in addition to the official federal expenditures dedicated to
national defense, funded from the NIRA (=National Industrial Recovery Act) funds in the mid-1930s.
The Army received numerous indirect benefits, due to no doubt pressure activity as well, from the New
Deal building and employment programs.68 The influence and importance of domestic political markets,
budgetary patterns, and perceptions of threat can hardly be overlooked in the study of these two periods. Thus, the results achieved have to be considered somewhat tentative, yet they offer interesting
insights on the interdependence of military spending and economic development.
Consequently, in this article I investigated: 1) whether economic growth Granger-causes military spending or vice versa, at various lag lengths; 2) what different explanatory variables affected the demand for
military spending among the individual Great Powers; 3) whether the systemic demand for military
spending can be explained in terms of leadership (or the lack of it) and other systemic characteristics in
the short run. For example, the hegemonic perspective advocated by many, for example Paul Kennedy,
suggests that the economic leader of a system has to dedicate growing resources on security, eventually
harming its economy. The first question outlined above was here analyzed by employing Granger non-
66
Benefits of a pure public good are both nonexcludable and nonrival. See more e.g. Sandler-Hartley 1995, 4—
5. On further analysis, see Hummel-Lavoie 1990.
67
On this, see further Eloranta 2002a.
68
Roskill 1976, 177, 361—364; Eloranta 2001a; Smith 1959, 124—125.
28
causality tests between the military spending and economic growth variables for the selected 16 nations.
The results suggest that it was the economy that influenced the military spending levels rather than vice
versa, with additional support found for either two-way causation. Therefore, the implication of hegemonic struggle being harmful for the economy, at least without qualifying certain threshold levels, does
not seem supported by the evidence.
Yet, did a hegemonic struggle take place in the international system during this period? And, if so, how
did it affect the individual states? It certainly seems that the individual country responses varied, even
between the two dependent variables (military burden, defense share). Austria, Germany, Italy, and
Russia often responded to the same variables, whereas France and the United States formed another
elusive group. Furthermore, UK and Japan seemed to act in a less consistent manner. All in all, they all
displayed hegemonic tendencies in their military spending to varying degrees. And, to underline this
conclusion, income was not found to be significant explanatory variable for most of these states. Thus,
it was their external (systemic) ambitions that influenced their military spending behavior. However,
this conclusion needs to be checked against the performance of domestic political market variables,
similar as done in Eloranta (2002a) for the interwar period.
The third question was pursued with a system comprising the 16 selected nations. Systemic impacts,
including the possibility of hegemonic systemic leadership being one of the dominant forces in deciding
military spending patterns, were indeed ascertained and the results suggest that these nations indeed
responded to the declining leadership of the UK and the lack of leadership by the U.S. with an overall
military challenge. Moreover, the impact of the behavior of democracies seems to have motivated military spending decisions during the interwar period. Also, the inclusion of alliance dummies improved
the fits of the pooled equations, suggesting that more precise individual country variables should conceivably be included in the estimations. Thus, the impact of systemic changes in leadership has explanatory value, at least in the short run on the basis of the evidence here. Moreover, the notion that
military spending might be a significant force in determining economic development is quite clearly
rejected here.
This study suggests further challenges for the study of military spending among these states. The results
achieved here on the basis of the systems approach should be reviewed in connection with individual
29
country political markets, with pooling to be undertaken between the countries on the basis of coefficient equality tests. In addition, one should certainly attempt to include more precise individual country
threat and spillover effects. Moreover, the supply and demand factors should also include the impact of
domestic power structures and allocation patterns, as well as competition within the political markets.
REFERENCES:
ATOP (=Alliance Treaty Obligations and Provisions) Database [online] (2000). V. 1.0. Copyright Brett Ashley
Leeds with Jeffrey M. Ritter, Sara McLaughlin Mitchell, and Andrew G. Long. Available from:
<URL:http://garnet.acns.fsu.edu/~aleeds/atop.html>.
Autio, Jaakko (1992), Valuuttakurssit Suomessa 1864—1991. Katsaus ja tilastosarjat. Suomen Pankin keskustelualoitteita 1.
Banks, Arthur S. (1976), CROSS-NATIONAL TIME SERIES, 1815—1973 [Computer file]. ICPSR Ed. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [producer and distributor].
Blackaby, Frank & Thomas Ohlson (1987), ‘Military Expenditure and the Arms Trade: Problems of Data’. In
The Economic of Military Expenditures. Military Expenditures, Economic Growth and Fluctuations.
Proceedings of a Conference held by the International Economic Association in Paris, France. Ed. by
Christian Schmidt. London.
Bollen, Kenneth A. (1993), ’Political Democracy: Conceptual and Measurement Traps’. In On Measuring Democracy. Its Consequences and Concomitants. Edited by Alex Inkeles. New Brunswick.
Buzan, Barry, Ole Wæver & Jaap de Wilde (1998), Security. A New Framework for Analysis. Boulder, Colorado.
Carreras, Albert et al. (1989), Estadicas historicas de España: siglos XIX-XX (Historical Statistics of Spain: 19th
and 20th Centuries). Fundacion Banco Exterior. Barcelona.
Cars, Hans Christian (1987), ‘Negotiations to Reduce Military Expenditures — Problems and Possibilities’. In
The Economics of Military Expenditures. Military Expenditures, Economic Growth and Fluctuations. Proceedings of a Conference held by the International Economic Association in Paris, France. Ed. by Christian
Schmidt. London.
Chowdhury, Abdur R. (1991), ’A Causal Analysis of Defense Spending and Economic Growth’. Journal of Conflict Resolution, Vol. 35, No. 1.
Clement, Piet (2000), Government Consumption and Investment in Belgium 1830—1940. The Reconstruction of
a Database. Studies in Social and Economic History, Vol. 29. Leuven.
Conybeare, John A.C. & Todd Sandler (1990), ‘The Triple Entente and the Triple Alliance 1880—1914: A Collective Goods Approach’. The American Political Science Review, Vol. 84, No. 4.
Cornes, Richard & Todd Sandler (1996), The theory of externalities, public goods, and club goods. Second edition. New York.
Correlates of War Inter-State War Data [online], 1816-1997 (2000). [version 28.1.2000] Available in wwwformat: <URL: http://www.umich.edu/~cowproj/>.
Cullis, John G. & Philip R. Jones (1987), Microeconomics and the Public Economy: A Defence of the Leviathan.
Oxford.
de Groot, Gerard J. (2001), The First World War. Twentieth-Century Wars. Basingstoke.
Eichengreen, Barry (1990), Elusive stability. Essays in the history of international finance, 1919—1939. New
York.
Eloranta, Jari (1998), Different Needs, Different Solutions — The Importance of Economic Development and
Domestic Power Structures in Explaining Military Spending in Eight Western Democracies During the Interwar Period. A Licentiate Thesis in Economic History at the University of Jyväskylä.
Eloranta, Jari (2001a), ‘Military Competition between Friends? Hegemonic Development and Military Spending
among Eight Western Democracies, 1920—1938’. Essays in Economic & Business History, Vol. XIX.
30
Eloranta, Jari (2001b), ‘National Defense’. Forthcoming in The Oxford Encyclopedia of Economic History. Ed.
by Joel Mokyr.
Eloranta, Jari (2001c), ’OPPORTUNITIES AND CONSTRAINTS IN THE GAME FOR PUBLIC GOODS: The
Political Economy of Military Spending in Finland and Sweden, 1920—1938’. To be presented at the EHES
Summer School in Dublin, 2001.
Eloranta, Jari (2002a), THE DEMAND FOR EXTERNAL SECURITY BY DOMESTIC CHOICES: Military
Spending as an Impure Public Good among Eleven European States, 1920—1938. A Doctoral Dissertation,
European University Institute.
Eloranta, Jari (2002b), ‘“WEAK” EUROPEAN STATES IN THE INTERNATIONAL ARMS TRANSFER
SYSTEM OF THE INTERWAR PERIOD: The Impact of External Threats, Market Forces, and Domestic
Constraints’. Forthcoming in Scandinavian Economic History Review.
Fontvieille, Louis (1976), Evolution et croissance de l’Etat Français: 1815—1969. Economies et sociëtës. Institut de Sciences Mathematiques et Economiques Appliquees. Tome X. Octobre.
Geller, Daniel S. & J. David Singer (1998), Nations at War. A Scientific Study of International Conflict. Cambridge Studies in International Relations: 58. Cambridge, UK.
Gilpin, Robert (1977), ‘Economic Interdependence and National Security in Historical Perspective’. In Economic
Issues and National Security. Ed. by Klaus Knorr and Frank N. Trager. Lawrence.
Global
Financial
Data
[online]
(2000). [quoted 1.5.2001]. Available in www-format:
<URL:http://www.globalfindata.com>.
Gordon, Michael R. (1974), ‘Domestic Conflict and the Origins of the First World War: The British and the
German Cases’. The Journal of Modern History, Vol. 46, No. 2.
Gregory, Paul (1982), Russian National Income, 1885—1913. Cambridge.
Gurr, Ted Robert, Keith Jaggers & Will H. Moore (1993), ‘The Transformation of the Western State: The
Growth of Democracy, Autocracy, and State Power since 1800’. In On Measuring Democracy. Its Consequences and Concomitants. Edited by Alex Inkeles. New Brunswick.
Hardach, Gerd (1977), The First World War. London.
Harris, Richard (1995), Using Cointegration Analysis in Econometric Modelling. London.
Hawke, G.R. (1980), Economics for historians. Cambridge.
Herrera, Rémy (1994), Statistics on Military Expenditure in Developing Countries: Concepts, Methodological
Problems and Sources. Development Centre Documents. OECD, Paris.
Hewitt, Daniel P. (1996), ’Military Expenditures, 1972—1990: The Reasons behind the Post-1985 Fall in World
Military Spending’. Public Budgeting and Financial Management, Vol. 7, No. 4.
Hirvilahti, Jon (1993), Ensimmäisestä maailmansodasta toiseen maailmansotaan. Katsaus kelluvien valuuttakurssien ajanjaksoon vuosina 1914-1925. Suomen Pankin keskustelualoitteita 7. Helsinki.
Historical Statistics of the United States, Colonial Times to 1970 (1975). US Bureau of Census. Washington, D.
C.
Historische Statistik der Schweiz (1996). Under the supervision of Hansjörg Siegenthaler. Edited by Heiner
Ritzmann-Blickenstorfer. Zürich.
Historisk Statistikk 1978 (1978). Norges Officielle Statistikk XII 291. Oslo.
Hobson, John M. (1993), ‘The Military-Extraction Gap and the Wary Titan: The Fiscal-Sociology of British
Defence Policy 1870—1913’. The Journal of European Economic History, Vol. 22, No. 3.
Hodne, Fritz (1992), ‘New Perspectives in Economic History’. Scandinavian Economic History Review, Vol.
XL, No. 1.
Hummel, Jeffrey Rogers & Don Lavoie (1990), ”National Defense and the Public-Goods Problem”. In Arms,
Politics, and the Economy. Historical and Contemporary Perspectives. Ed. by Robert Higgs. Independent Institute. New York.
Huntington, Samuel P. (1997), The Clash of Civilizations and the Remaking of World Order. New York.
Il bilancio dello Stato italiano dal 1862 al 1967. Volume IV: Allegati statistici. Ministero del Tesoro, Ragioneria
generale dello Stato. Roma.
Johansen, Hans Christian (1985), Dansk historisk statistik 1814—1980. Danmarks historie, bind 9. Copenhagen.
Jones, Matthew T. & Maurice Obstfeld [online] (2000), Savings and Investment in 13 Countries (1850—1945)
[quoted 10.10.2000]. Available in www-format: <URL: http://www.nber.org/databases/jones-obstfeld/>.
31
Justino, David (1990), Precos e salaries em Portugal 1850—1912. História Económica 2. Lisboa.
Kaiser, David E. (1983), ’Germany and the Origins of the First World War’. The Journal of Modern History,
Vol. 55, No. 3.
Kennedy, Gavin (1983), Defense economics. London.
Kennedy, Paul (1989), The Rise and Fall of the Great Powers. Economic Change and Military Conflict from
1500 to 2000. London.
Kennedy, Paul (1991), The Rise and Fall of British Naval Mastery. Third Edition. London.
Kennedy, Paul (1994), Uuden vuosituhannen haasteet. Keuruu.
Keohane, Robert O. & Joseph S. Nye (1977), Power and Interdependence. World Politics in Transition. Boston.
Kindleberger, Charles (1973), The World in Depression 1929—1939. London.
Kindleberger, Charles (1988), ‘Hierarchy versus Inertial Cooperation’. In The International Economic Order.
Essays on Financial Crisis and International Public Goods. (Originally published in International Organization, vol. 40, no. 4, Autumn 1986). London.
Komlos, John (1992), ‘Interdisciplinary Approaches to Historical Analysis: The Uneasy Waltz of Economics and
History’. In Theory, Method, and Practice in Social and Cultural History. Ed. by Peter Karsten and John
Modell. New York.
Kraehe, Enno E. (1992), ’A Bipolar Balance of Power’. The American Historical Review, Vol. 97, No. 3.
Krantz, Olle (1987), Historiska nationalräkenskaper för Sverige: offentlig verksamhet 1800—1980. Skrifter
utgivna av Ekonomisk-historiska föreningen i Lund, vol. L. Lund.
Krantz, Olle (1997), ‘Swedish Historical National Accounts 1800—1990 — aggregated output series’. Umeå
(mimeo).
Maddison, Angus (1995), Monitoring the World Economy 1820—1992. OECD, Paris.
Mata, Eugénia (1993), As financas públicas portuguesas da regeneracão á primeira Guerra mundial. História
Económica 4. Banco de Portugal. Lisboa.
McCloskey, Donald N. (1987), Econometric History. Studies in Economic and Social History. London.
McNeill, William H. (1982), The Pursuit of Power. Technology, Armed Forces, and Society since A.D. 1000.
Chicago.
Milward, Alan S. (1965), The German Economy at War. London.
Milward, Alan S. (1970), The New Order and the French Economy. Oxford.
Milward, Alan S. (1977), War, Economy and Society 1939—1945. London.
Mintz, Alex & Chi Huang (1990), ‘Defense Expenditures, Economic Growth, and the “Peace Dividend”’. The
American Political Science Review, Vol. 84, No. 4.
Mitchell, B.R. (1992), International Historical Statistics: Europe 1750—1988. Basingstoke.
Mitchell, B.R. (1998b), International Historical Statistics: Europe 1750—1993. Fourth Edition. Basingstoke.
Mitchell, Brian R. (1998a), International Historical Statistics: The Americas 1750—1993. Fourth Edition. Basingstoke.
Mitchell, Brian R. (1998c), International Historical Statistics: Africa, Asia & Oceania 1750—1993. Third Edition. Basingstoke.
Modelski, George & William R. Thompson (1988), Seapower in Global Politics, 1494—1993. London.
Modelski, George & William R. Thompson (1996), Leading Sectors and World Powers. The Coevolution of
Global Politics and Economics. Studies in International Relations. Columbia, South Carolina.
Morgan, E.V. (1952), Studies in British Financial Policy, 1914—1925. London.
Morrow, James D. (1993), ‘Arms versus allies: trade-offs in the search for security’. International Organization,
Vol. 42, No. 2.
Murdoch, James C. & Todd Sandler (1986), ‘The Political Economy of Scandinavian Neutrality’. Scandinavian
Journal of Economics, Vol. 88, No. 4.
North, Robert C. (1990), War, Peace, Survival. Global Politics and Conceptual Synthesis. Boulder.
Pearton, Maurice (1982), The Knowledgeable State. Diplomacy, War and Technology since 1830. London.
Polity 3D Database [online] (2000). [quoted 20.12.2000]. Available in www-format: <URL:http://kgleditsch.socsci.gla.ac.uk/Polity.html>.
Pollins, Brian M. & Randall L. Schweller (1999), ‘Linking the Levels: The Long Wave and Shifts in U.S. Foreign Policy, 1790—1993’. American Journal of Political Science, Vol. 43, No. 2.
32
Prados de la Escosura, Leandro (1993), Spain's Gross Domestic Product, 1850-1990: a New Series. Universidad
Carlos III de Madrid. Dirección General de Planificación (Ministerio de Economia y Hacienda). Marzo.
Prados de la Escosura, Leandro (2000), ‘International Comparisons of Real Product, 1820—1990: An Alternative Data Set’. Explorations in Economic History, Vol. 37, No. 1.
Pryor, Frederick L. (1968), Public Expenditures in Communist and Capitalist Nations. London.
Rapkin, David P. (1990), ’The Contested Concept of Hegemonic Leadership’. In World Leadership and Hegemony. Ed. by David P. Rapkin. Boulder.
Rasler, Karen A. & William R. Thompson (1989), War and State Making. The Shaping of the Global Powers.
Studies in International Conflict, Vol. 2. Boston.
Rasler, Karen, William R. Thompson (1991), ’Technological Innovation, Capability Positional Shifts, and Systemic War’. Journal of Conflict Resolution, Vol. 35, No. 3.
Ray, James Lee & J. David Singer (1973), ‘Measuring the Concentration of Power in the International System’.
Sociological Methods & Research, Vol. 1, No. 4.
Rossi, Nicola, Andrea Sorgato & Gianni Toniolo (1993), ‘I conti economici italiani: una ricostruzione statistica,
1890—1990’. Rivista di Storia Economica, Vol. 10, No. 1.
Sandler, Todd & Keith Hartley (1995), The Economics of Defense. Cambridge Surveys of Economic Literature.
Cambridge, UK.
Singer, J. David & Melvin Small (1993), NATIONAL MATERIAL CAPABILITIES DATA, 1816—1985 [Computer file]. Ann Arbor, MI: J. David Singer, University of Michigan, and Detroit, MI: Melvin Small, Wayne
State University [producers], 1990. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor].
Singer, J. David (1979), ‘Introduction’. In The Correlates of War: I. Research Origins and Rationale. Ed. by J.
David Singer. New York.
Singer, J. David (1990), ‘Variables, Indicators, and Data. The Measurement Problem in Macropolitical Research’. In Measuring the Correlates of War. Ed. by J. David Singer and Paul F. Diehl. University of Michigan.
Smil, Vaclav (1994), Energy in World History. Essays in World History. Boulder, Colorado.
Smits, Jan-Pieter, Edwin Horlings & Jan Luiten van Zanden [online] (2000), Dutch GNP and Its Components,
1800—1913
[quoted
12.12.2000].
Available
in
pdf-format:
<URL:http://www.eco.rug.nl/ggdc/PUB/dutchgnp.pdf>.
Stiglitz, Joseph E. (1988), Economics of the Public Sector. New York.
Taylor, A.J.P. (1954), The Struggle for Mastery in Europe 1848—1918.
Taylor, A.J.P. (1961), The Origins of the Second World War. 2nd Edition. New York.
Taylor, Alan M. [online] (2000), A Century of Purchasing-Power Parity. NBER Working Paper 8012. Available
from: <URL:http://www.nber.org/papers/w8012>.
The economics of World War II (1998). Six great powers in international comparison. Ed. by Mark Harrison.
Cambridge.
The Rise of the Fiscal State in Europe c. 1200—1815. Edited by Richard Bonney. Oxford 1999.
Währungen der Welt I (1991). Europäische und nordamerikanische Devisenkurse (1777—1914), teilband II.
Beiträge zur wirtschafts- und sozialgeschichte, band 44, II. Herausgaben von Jürgen Schneider, Oskar
Schweitzer und Friedrich Zellfelder. Stuttgart.
Währungen der Welt II (1997). Europäische und nordamerikanische Devisenkurse (1914—1951). Beiträge zur
wirtschafts- und sozialgeschichte, band 45. Herausgaben von Jürgen Schneider, Oskar Schweitzer und Markus
A. Denzel. Stuttgart.
Watt, Donald Cameron (1977), ‘Some Aspects of A.J.P. Taylor’s Work as Diplomatic Historian’. The Journal of
Modern History, Vol. 49, No. 1.
Webber, Caroly & Aaron Wildavsky (1986), A History of Taxation and Expenditure in the Western World. New
York.
West, Robert L. (1993), ‘Background Note on Military Expenditure: Sources and Price Conversion Procedures’.
In Military Expenditure and Economic Development. A Symposium on Research Issues. Ed. by Geoffrey
Lamb with Valerian Kallab. World Bank, Washington D.C.
33
Wohlforth, William C. (1987), ‘The Perception of Power: Russia in the Pre-1914 Balance’. World Politics, Vol.
39, No. 3.
APPENDIX 1. Sources of Data and Data Solutions for the Individual Countries and on the Aggregate
COMMON DATA:
1870—1913:
Military personnel data (thousands) from Singer-Small 1993 (Norway: supplemented with Banks 1976). Iron and
steel production (thousands of tons) from Singer-Small 1993. Urban population (population living in cities with
population greater than 100,000 – in thousands) figures from Singer-Small 1993 (Norway: supplemented with
Banks 1976). Levels of democracy and autocracy (separate indices, scale from 0—10, with the score 6 or higher
in the democracy index taken as indicating a democracy, with the score 4 or higher in the autocracy index taken
as indicating an autocracy; these indices are constructed on the basis of the following variables: competitive participation in politics, regulation of participation, executive recruitment regulation, executive recruitment competition, executive recruitment openness, and executive constraints69), used to divide the military threat into either
democratic or authoritarian (neutral countries were excluded) aggregates, taken from Polity 3D 2000. Alliances
(defined very broadly as including any defensive, offensive, neutrality, nonaggression, or consultation obligations), used to construct alliance dummies for each country, by year from ATOP 2000. Exchange rates used in
various conversions from Global Financial Data 2000 (only partially for France, Germany, the UK, the USA),
otherwise from Währungen der Welt 1991, 1997 and Autio 1992.
AUSTRIA:
1870—1913:
Nominal military expenditures (ME) from Hobson 1993. Nominal GDP from Jones-Obstfeld 2000. Real GDP
from Maddison 1995, adjusted according to 1913 indirect PPP-level differences calculated in Prados de la Escosura 2000. Total population (including Hungary) from Singer-Small 1993. Nominal central government expenditures (CGE) from Mitchell 1998b. Regarding price data, only the wholesale price index (WPI) was utilized in
the deflation of ME, obtained from Mitchell 1998b (1910—1913 extrapolated with German WPI). Nominal ME
was then converted to 1913 prices and adjusted with the indirect PPP-converters found in Prados de la Escosura
2000 and the exchange rates mentioned above to come up with real ME in 1913 (quasi)-GBP.
BELGIUM:
1870—1913:
Nominal ME from Clement 2000 (except for 1913 from Singer-Small 1993, converted with the exchange rates
explained above). Nominal GNP from Clement 2000 for 1880—1913 (1870—1879 military burden extrapolated
using the defense share trend). Real GDP from Maddison 1995, adjusted as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b (except for 1913 from Banks 1976, converted
with the exchange rates explained above). Belgian CPI and WPI from Mitchell 1998b, combined with equal
weighting to come up with a deflator for ME. Real ME derived as in the case of Austria.
DENMARK:
1870—1913:
Nominal ME from Johansen 1985. Nominal GDP from Johansen 1985. Real GDP from Maddison 1995, adjusted
as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Johansen 1985. Regarding
69
For more on indices of democracy, see especially Bollen 1993, Gurr et al. 1993, Singer 1990. Criticism and
further evaluation of the use of such measures will be presented in a subsequent paper. Yearly lists of the 16
countries’ regime developments are available from the author by request.
34
price data, WPI (1870—1875 missing) and CPI from Mitchell 1998b, combined with equal weighting (for
1870—1875 only CPI taken) to come up with a deflator for ME. Real ME derived as in the case of Austria.
FRANCE:
1870—1913:
Nominal ME from Hobson 1993. Nominal GDP from Jones-Obstfeld 2000. Real GDP from Maddison 1995,
adjusted as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b.
French WPI from Mitchell 1998b and French CPI from Global Financial Data 2000, combined with equal
weighting to come up with a deflator for ME. Real ME derived as in the case of Austria.
GERMANY:
1870—1913:
Nominal ME from Hobson 1993. Nominal GDP from Jones-Obstfeld 2000. Real GDP from Maddison 1995,
adjusted as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b
(1870—1871 missing, military burden trend used to extrapolate defense share for these years). German WPI from
Mitchell 1998b and German CPI from Global Financial Data 2000, combined with equal weighting to come up
with a deflator for ME. Real ME derived as in the case of Austria.
ITALY:
1870—1913:
Nominal ME from Hobson 1993. Nominal GDP from Jones-Obstfeld 2000. Real GDP from Maddison 1995,
adjusted as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b.
Italian WPI and CPI from Mitchell 1998b, combined with equal weighting to come up with a deflator for ME.
Real ME derived as in the case of Austria.
JAPAN:
1870—1913:
Nominal ME from Hobson 1993. Nominal GDP for 1885—1913 from Jones-Obstfeld, for 1870—1884 extrapolated by forecasting with nominal CGE and nominal exports (and a constant), converted to yens, taken from
Banks 197670. Real GNP for 1885—1913 from Mitchell 1998c, for 1870—1884 extrapolated by forecasting with
nominal CGE, nominal exports, nominal imports, and nominal revenues (with a lag of one year) (and a constant),
converted to yens, taken from Banks 197671; adjusted to 1913 prices using the indirect PPP-converters found in
Prados de la Escosura 2000. Total population from Singer-Small 1993. Nominal CGE from Mitchell 1998c.
Regarding price data, Japanese WPI from Mitchell 1998c (the so-called Asahi Shinbun index) and CPI for
1900—1913 from Global Financial Data 2000, combined with equal weighting (for 1870—1899 only WPI
taken) to come up with a deflator for ME. Real ME derived as in the case of Austria.
THE NETHERLANDS:
1870—1913:
Nominal ME from Singer-Small 1993 (converted to NLG with the exchange rates outlined above). Nominal GDP
from Smits et al. 2000. Real GDP from Maddison 1995, adjusted as in the case of Austria. Total population from
Maddison 1995. Nominal CGE from Mitchell 1998b. Regarding price data, only the GDP-deflator introduced in
70
Performed with variables in logs. Adjusted R2 of the fit for 1885—1913 equal to 0,97. Chained together by
multiplying with the ratio (actual nominal GDP in 1885, billions of yen) 0,807 / (forecast nominal GDP in 1885,
billions of yen) 0,723. Passes the Breusch-Godfrey LM serial correlation test up to 5 lags.
71
Performed with variables in logs. Adjusted R2 of the fit for 1885—1913 equal to 0,97. Chained together by
multiplying with the ratio (actual real GNP in 1885, billions of yen) 3,852 / (forecast real GNP in 1885, billions
of yen) 3,967. Passes the Breusch-Godfrey LM serial correlation test up to 5 lags.
35
Smits et al. 2000 was available, which was then used as a deflator for ME. Real ME derived as in the case of
Austria.
NORWAY:
1870—1913:
Nominal ME for 1870—1904 from Banks 1976 (converted to NOK utilizing the exchange rates outline above),
for 1905—1913 from Singer-Small 1993 (converted to NOK utilizing the exchange rates outline above); the two
series were chained using the ratio of their difference in 1905. Nominal GDP from Jones-Obstfeld 2000. Real
GDP from Maddison 1995, adjusted as in the case of Austria. Total population from Maddison 1995. Nominal
CGE from Mitchell 1998b. Regarding price data, Norwegian WPI for 1891—1913 from Mitchell 1998b (extrapolated backwards to 1876 using Danish WPI), and Norwegian CPI for 1901—1913 from Mitchell 1998b
(extrapolated backwards to 1870 using Danish CPI)72, combined with equal weighting to come up with a deflator
for ME. Real ME derived as in the case of Austria.
PORTUGAL:
1870—1913:
Nominal ME from Mata 1993. Nominal National Product from Mata 1993. Real National Product derived by
using the price index in Mata 1993, adjusted to 1913 prices using the indirect PPP-converters found in Prados de
la Escosura 2000. Total population from Singer-Small 1993. Nominal CGE from Mata 1993. Regarding price
data, Portuguese WPI from Justino 1990, and the price index from Mata 1993 were used with equal weighting to
come up with a deflator for ME. Real ME derived as in the case of Austria.
RUSSIA/SOVIET UNION:
1870—1913:
Nominal ME from Hobson 1993. Nominal NNP for 1885—1913 from Gregory 1982, for 1870—1884 extrapolated by forecasting with nominal revenue, nominal imports, and a first-order autoregressive term (and a constant), exchange rates used in conversions to silver rubles, taken from Banks 1976, and nominal GDP derived by
raising the nominal NNP by 8,4 per cent73. Real NNP for 1885—1913 from Gregory 1982, for 1870—1884 extrapolated by forecasting with nominal exports (with a lag of one year), nominal imports, and nominal CGE (and
a constant), converted to silver rubles, taken from Banks 197674; adjusted to 1913 prices using the indirect PPPconverters found in Prados de la Escosura 2000. Total population from Singer-Small 1993. Nominal CGE from
Mitchell 1998b. Regarding price data, the deflator used to convert nominal ME to 1913 prices was the NNPdeflator derived from the nominal and real NNP series described above. Real ME derived as in the case of Austria.
SPAIN:
1870—1913:
Nominal ME from Carreras et al. 1989. Nominal GDP from Prados de la Escosura 1993. Real GDP from Prados
de la Escosura 1993, adjusted to 1913 prices using the indirect PPP-converters found in Prados de la Escosura
2000. Total population from Singer-Small 1993. Nominal CGE from Carreras et al. 1989. Regarding price data,
only the Spanish WPI was available for the period, taken from Mitchell 1998b, to be used as the deflator for the
nominal ME. Real ME derived as in the case of Austria.
72
These extrapolations could equally as well have been done using Swedish price data, since the Norwegian
prices moved similarly to the prices in these two countries.
73
Performed with variables in logs. Adjusted R2 of the fit for 1885—1913 equal to 0,96. Chained together by
multiplying with the ratio (actual nominal NNP in 1885, billions of rubles) 6,286 / (forecast nominal NNP in
1885, billions of rubles) 6,670. Passes the Breusch-Godfrey LM serial correlation test up to 5 lags.
74
Performed with variables in logs. Adjusted R2 of the fit for 1885—1913 equal to 0,94. Chained together by
multiplying with the ratio (actual real NNP in 1885, billions of rubles) 7,904 / (forecast real NNP in 1885,
billions of rubles) 8,474. Passes the Breusch-Godfrey LM serial correlation test up to 5 lags.
36
SWEDEN:
1870—1913:
Nominal ME from Krantz 1987. Nominal GDP from Krantz 1997. Real GDP from Maddison 1995, adjusted as
in the case of Austria. Total population from Maddison 1995. Nominal CGE for 1870—1879 from Banks 1976
(converted to SEK using the exchange rates described above), and for the rest of the period from Mitchell 1998b.
Swedish WPI and CPI from Mitchell 1998b, combined with equal weighting to come up with a deflator for ME.
Real ME derived as in the case of Austria.
SWITZERLAND:
1870—1913:
Nominal ME from Historische Statistik 1996. Nominal GDP from Historische Statistik 1996. Real GDP from
Historische Statistik 1996, adjusted to 1913 prices using the indirect PPP-converters found in Prados de la Escosura 2000. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b. Swiss WPI and CPI (CPI
for 1890—1913) from Mitchell 1998b, combined with equal weighting (only WPI included for 1870—1889) to
come up with a deflator for ME. Real ME derived as in the case of Austria.
THE UNITED KINGDOM:
1870—1913:
Nominal ME from Hobson 1993. Nominal GDP from Mitchell 1998b. Real GDP from Maddison 1995, adjusted
as in the case of Austria. Total population from Maddison 1995. Nominal CGE from Mitchell 1998b. British
WPI and CPI both from Mitchell 1998b, combined with equal weighting to come up with a deflator for ME. Real
ME derived by converting the nominal ME to 1913 prices, and subsequently corrected for its purchasing power
(relative of the U.S.) with the indirect PPPs of Prados de la Escosura 2000.
THE UNITED STATES:
1870—1913:
Nominal ME from Hobson 1993 (same series as in Historical Statistics 1976). Nominal GNP from Historical
Statistics 1976. Real GDP from Maddison 1995. Total population from Maddison 1995. Nominal CGE from
Historical Statistics 1976 (federal government outlays). American WPI and CPI both from Mitchell 1998a, combined with equal weighting to come up with a deflator for ME. Nominal ME was then deflated to 1913 prices and
converted to GBP using the exchange rates outlined above.
37
APPENDIX 2. DATA TABLES
Table 1. Country and Variable Abbreviations Used in the Text and Appendices:
ABBREVIATION
AR
ATC
AUT
BEL
CGE
CINC
CONCMILRES
CONCCINC
CV
DEM
DEN
DFSHARE
DUM
FRA
GDPCAP
GER
ITA
JAP
MA
ME
MILBUR
MILCINC
NED
NOR
POR
RUS/SOV
SPA
SWE
SWI
SYSTOTME
SYSTOTMECV
UK
USA
DEFINITION
AUTOREGRESSIVE ERROR TERM
AUTOCRACIES
AUSTRIA
BELGIUM
CENTRAL GOVERNMENT EXPENDITURES
COMPOSITE INDEX OF NATIONAL CAPABILITIES
CONCENTRATION OF MILITARY RESOURCES
CONCENTRATION OF CINC
COEFFICIENT OF VARIATION
DEMOCRACIES
DENMARK
DEFENSE SHARE (ME OF CGE, %)
DUMMY VARIABLE
FRANCE
GDP PER CAPITA
GERMANY
ITALY
JAPAN
MOVING AVERAGE TERM
MILITARY EXPENDITURES
MILITARY BURDEN (ME OF GDP, %)
MILITARY RESOURCE SHARE
THE NETHERLANDS
NORWAY
PORTUGAL
RUSSIA/THE SOVIET UNION
SPAIN
SWEDEN
SWITZERLAND
SYSTEM TOTAL ME
SYSTEM TOTAL ME, COEFFICIENT OF VARIATION
THE UNITED KINGDOM
THE UNITED STATES
Table 2. Granger Non-causality Relationships for 16 Countries, 1870—1913
DEPENDENT
VARIABLE
1. AUT:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
2. BEL:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
INDEPENDENT
VARIABLE
NUMBER OF LAGS
(FOR BEST p)
BEST
p-VALUE
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
7†
8
7†
1
1†
5
0,074
0,403
0,026
0,111
0,009
0,201
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
11
11
3
2
2†
0,098
0,105
0,029
0,045
0,041
38
Country GDP of tot. GDP
3. DEN:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
4. FRA:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
5. GER:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
6. ITA:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
7. JAP:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
8. NED:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
9. NOR:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
10. POR:
Defense share
Real GDP per capita
Military burden
Real GDP per capita
Country ME of tot. ME
Country GDP of tot. GDP
11. RUS:
Defense share
Real GDP per capita
Country ME of tot. ME
6
0,191
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
1
3
8†
4
11
13
0,053
0,093
0,064
0,080
0,406
0,110
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
2†
8†
9
8†
8†
6
0,000
0,067
0,293
0,050
0,063
0,092
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
1†
4†
1†
2†
12†
9
0,030
0,000
0,000
0,047
0,034
0,172
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
12
6
10
3†
1
3†
0,120
0,116
0,095
0,085
0,201
0,016
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
7
11
11
12
12†
14
0,193
0,334
0,190
0,070
0,050
0,167
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
3†
12
4†
3
13†
1
0,000
0,163
0,010
0,325
0,005
0,022
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
10†
8†
10†
1
2†
10
0,084
0,017
0,014
0,059
0,025
0,188
Real GDP per capita
Defense share
Real GDP per capita
Military burden
Country GDP of tot. GDP
Country ME of tot. ME
11
14
11†
13
2†
8
0,091
0,057
0,005
0,092
0,043
0,263
Real GDP per capita
Defense share
11
1
0,122
0,134
39
Military burden
Real GDP per capita
12†
0,030
Real GDP per capita
Military burden
3
0,242
Country ME of tot. ME
Country GDP of tot. GDP
11†
0,013
Country GDP of tot. GDP
Country ME of tot. ME
1
0,051
12. SPA:
Defense share
Real GDP per capita
5
0,070
Real GDP per capita
Defense share
5†
0,062
Military burden
Real GDP per capita
3
0,115
Real GDP per capita
Military burden
6
0,144
Country ME of tot. ME
Country GDP of tot. GDP
2†
0,011
Country GDP of tot. GDP
Country ME of tot. ME
10†
0,015
13. SWE:
Defense share
Real GDP per capita
5†
0,008
Real GDP per capita
Defense share
6†
0,000
Military burden
Real GDP per capita
13†
0,017
Real GDP per capita
Military burden
5†
0,000
Country ME of tot. ME
Country GDP of tot. GDP
2
0,143
Country GDP of tot. GDP
Country ME of tot. ME
5
0,571
14. SWI:
Defense share
Real GDP per capita
4†
0,013
Real GDP per capita
Defense share
5
0,035
Military burden
Real GDP per capita
7†
0,070
Real GDP per capita
Military burden
5†
0,003
Country ME of tot. ME
Country GDP of tot. GDP
11
0,226
Country GDP of tot. GDP
Country ME of tot. ME
5
0,164
15. UK:
Defense share
Real GDP per capita
1
0,084
Real GDP per capita
Defense share
9†
0,028
Military burden
Real GDP per capita
2†
0,016
Real GDP per capita
Military burden
10
0,122
Country ME of tot. ME
Country GDP of tot. GDP
13
0,073
Country GDP of tot. GDP
Country ME of tot. ME
2
0,162
16. USA:
Defense share
Real GDP per capita
6
0,445
Real GDP per capita
Defense share
9
0,069
Military burden
Real GDP per capita
12
0,194
Real GDP per capita
Military burden
10
0,116
Country ME of tot. ME
Country GDP of tot. GDP
1†
0,087
Country GDP of tot. GDP
Country ME of tot. ME
9†
0,031
Sources: see Appendix 1.
Note: All variables are in logarithmic form. AUT real GDP per capita, DEN real ME of total (16 countries) real
ME, ITA real GDP per capita, ITA military burden, JAP real GDP per capita, RUS real GDP per capita, RUS
real GDP of total (16 countries) real GDP, USA defense share, and USA military burden are I(1). † = null rejected at more than one lag.
40
Table 4. SUR (Seemingly Unrelated Regressions) Estimation of the Determinants of Pooled Defense Share and Military Burden in a 16-State System, 1870—1913
INDEPENDENT VARIABLE
COEFFICIENTS
(DEPENDENT
VARIABLE DFSHARE)
1.419***
0.097***
-0.253***
0.476***
0.271***
-0.192***
-0.041*** (t-1)
0.032***
0.190***
-0.089***
-0.112***
0.319***
0.062***
-0.042*
0.085***
-0.151**
-0.036**
-0.098**
-0.077***
†
1122.339
0.975
COEFFICIENTS
(DEPENDENT
VARIABLE MILBUR)
†
-0.118***
-0.268***
1.200***
0.502***
-0.093*** (t-1)
-0.136***
0.582***
-0.244***
0.565***
0.068*** (t-1)
-0.041***
0.129***
-0.175***
0.045***
-0.067***
-0.025**
0.842***
1344.605
0.958
CONSTANT
POOLED GDPCAP
POOLED CINC
POOLED MILRESSHARE
UKCINC
USACINC
UKMILRESSHARE
USAMILRESSHARE
SYSTOTME
SYSTOTMECV
CONCCINC
CONCMILRES
DEMCINC
DEMTOTME
ATCCINC
ATCTOTME
AUT ALLIANCE DUM
BEL ALLIANCE DUM
DEN ALLIANCE DUM
FRA ALLIANCE DUM
GER ALLIANCE DUM
ITA ALLIANCE DUM
JAP ALLIANCE DUM
NED ALLIANCE DUM
NOR ALLIANCE DUM
POR ALLIANCE DUM
RUS ALLIANCE DUM
SPA ALLIANCE DUM
SWE ALLIANCE DUM
SWI ALLIANCE DUM
UK ALLIANCE DUM
USA ALLIANCE DUM
AR(1)
LOG LIKELIHOOD
ADJUSTED R2
Sources: see Appendix 1.
Note: Variable definitions listed in Appendix 2, Table 1. Only statistically significant coefficients are listed. All
variables are in logarithmic form. AUT real GDP per capita, AUT total resource share (of 16 countries), ITA real
GDP per capita, ITA military burden, JAP real GDP per capita, RUS real GDP per capita, RUS real GDP of total
(16 countries) real GDP, USA defense share, USA military burden, and USA military resources share (of 16
countries) are I(1). † = cross-section specific, coefficients not listed here. * = null hypothesis of no correlation
rejected at 10 per cent level; ** = null rejected at 5 per cent level; *** = null rejected at 1 per cent level.
41
Table 5. SUR (Seemingly Unrelated Regressions) Estimation of the Determinants of Pooled Defense Share and Military Burden in a 13-State System, 1870—1913
INDEPENDENT VARIABLE
COEFFICIENTS
(DEPENDENT
VARIABLE DFSHARE)
1.238***
0.088***
-0.106***
0.305***
0.312***
-0.220***
-0.042** (t-1)
0.033***
0.201***
-0.072***
0.153**
0.058***
0.134***
0.079***
-0.058***
†
955.467
0.983
COEFFICIENTS
(DEPENDENT
VARIABLE MILBUR)
†
-0.275***
-0.299***
0.878***
0.732***
0.070*** (t-1)
0.055***
0.569***
-0.815***
1.234***
0.075**
0.144***
-0.226***
0.038***
-0.074**
0.839***
1030.734
0.941
CONSTANT
POOLED GDPCAP
POOLED CINC
POOLED MILRESSHARE
UKCINC
USACINC
UKMILRESSHARE
USAMILRESSHARE
SYSTOTME
SYSTOTMECV
CONCCINC
CONCMILRES
DEMCINC
DEMTOTME
ATCCINC
ATCTOTME
AUT ALLIANCE DUM
BEL ALLIANCE DUM
DEN ALLIANCE DUM
FRA ALLIANCE DUM
GER ALLIANCE DUM
ITA ALLIANCE DUM
JAP ALLIANCE DUM
NED ALLIANCE DUM
NOR ALLIANCE DUM
POR ALLIANCE DUM
RUS ALLIANCE DUM
SPA ALLIANCE DUM
SWE ALLIANCE DUM
SWI ALLIANCE DUM
UK ALLIANCE DUM
USA ALLIANCE DUM
AR(1)
LOG LIKELIHOOD
ADJUSTED R2
Sources: see Appendix 1.
Note: on variable specifications and definitions, see Appendix 2, Table 1. Only statistically significant coefficients are listed. Countries excluded in comparison with the 16-state system: JAP, NOR, and RUS. † = crosssection specific, coefficients not listed here. * = null hypothesis of no correlation rejected at 10 per cent level; **
= null rejected at 5 per cent level; *** = null rejected at 1 per cent level.