Although removed from direct consideration in the context of American diplomacy, the subject under study bears at least obliquely on U.S. policies. Professor Smaldone, a political scientist, reaches somewhat surprising counterintuitive conclusions based on considerable, if (as he notes) preliminary, research. -- Ed.
This paper aims to shed light on whether African military spending during the 1990s purchased its intended commodity, security - conceived herein as making a difference to African states experience of war and peace. Specifically, it seeks to discover whether ME-related factors were associated with patterns of war and peace or the duration of major armed conflicts in sub-Saharan Africa (SSA) during 1989-1999. The paper is a preliminary, rudimentary, and generalized analysis subject to significant questions, qualifications, qualms, and quibbles. That said, we proceed.
African military budget data are notoriously opaque, incomplete, and deceptive. This is not likely to change dramatically any time soon, although one can hope for improvement. Since 1981 there has been a global ME reporting mechanism, the United Nations Instrument for Reporting Military Expenditures. More than 110 states have submitted reports at least once, and participation has increased substantially in recent years. Throughout the 1980s and 1990s annual submissions ranged between twenty and thirty-five. However, in 2001 national submissions surged to sixty-one, and then in 2002 jumped to eighty-two. Regrettably, African participation remains low (recent reports are accessible on the UN website). SIPRIs project on "Budgeting for the Military Sector in Africa: The Process and Mechanisms of Control" may also promote increased transparency and hence more and better data. Until then, we are left with the standard annual sources of worldwide ME data: SIPRI, IISS, and the U.S. Department of State (its World Military Expenditures and Arms Transfers [WMEAT] series is used here).
According to WMEAT 1990, military spending by all African states less Egypt was only 1.8% of worldwide ME in 1979, and 1.5% in 1989. In 1999 it was still only 2.4%, with a few states DROC, Nigeria, Ethiopia and Eritrea accounting for most of the growth in the late nineties (WMEAT 1999-2000). That Africas fifty countries, representing 29% of the nations covered by WMEAT, have spent less than 2% of global defense outlays, is often overlooked. As for war and peace, Africa has gotten a bum rap also. Widespread perceptions that Africa is a region in perpetual turmoil, and that conflicts have become even more rampant in the 1990s, are exaggerated. Africas share of worldwide major armed conflicts was lower than 25% during most of the period, though it did rise to about two-fifths in 1998-99 (SIPRI Yearbook 2000). Moreover, taking into account the large number of states on the continent compared to other major regions, Africas relative conflict profile diminishes further. Within the continent itself, the average number of states in conflict during 1989-1999 was eight per year (16%) not good, but not as bad as typical characterizations.
Military Spending and Major Armed Conflict
According to SIPRI, sixteen SSA states experienced one or more major armed conflicts (MAC) during 1989-1999 (actually seventeen, but Eritrea is not included here because it was not independent for the entire time span). In other words, near two-fifths of SSA states were afflicted by at least one MAC for at least one year. Conversely, twenty-six other states did not experience any MAC during the period. SIPRI defines MAC as a prolonged conflict between the armed forces of two or more governments, or between a government and at least one organized armed group, involving at least 1,000 battle-related deaths over the course of the conflict; both internal and international conflicts are included. (Note: it is possible for a given state to experience multiple MACs simultaneously; our analysis is concerned with states in conflict, not conflicts in states.)
Table 1 shows the incidence of MAC among these sixteen states for the chosen period. Two years of MAC were recorded in Congo (1997, 1999), Guinea-Bissau (1998-99), and Senegal (1997-98). Burundi and DROC registered MAC-level political violence in the last three years of our selected period (1997-1999). All other states suffered one or more continuous or discontinuous MACs for at least four years, with Sudan being the only one wracked by such intensive warfare over the entire eleven-year period.
As noted above, the research questions examined here are whether military spending levels and other ME-related factors were associated with patterns of war and peace in SSA, or with the duration of armed conflict in states that experienced MAC. Our elementary analytical models make no explicit assumptions about the direction of causality. Indeed, the data and relationships between MAC and ME can be interpreted in various ways to get at the questions: Do patterns of war/peace explain patterns of ME? . . . or do patterns of ME explain patterns of war/peace? However, the results obtained in the following series of analyses show pronounced patterns that allow us to make reasonable inferences and judgments.
Table 1. Major Armed Conflicts in Sub-Saharan Africa, 1989-1999
Source: Stockholm International Peace Research Institute, SIPRI Yearbook 1990 2000;
Major Armed Conflict (MAC): a prolonged conflict between the armed forces of two or more governments, or between a government and at least one organized armed group, involving at least 1,000 battle-related deaths over the course of the conflict; both internal and international armed conflicts are included.
Notes: (1) This table lists states in which MACs occurred each year, not necessarily all MACs that occurred in these states. By SIPRIs definition, a government could be involved in multiple MACs simultaneously. (2) Eritrea is not included because it was not independent during the entire period of study.
National security is traditionally considered to have two dimensions: national defense against external aggressors and internal security against domestic enemies. Both of these are covered by SIPRIs MAC data. Arguably, ME-related indicators can serve, inter alia, as proxies for a states ability to deter or defeat armed opposition (external or internal). Obviously, peace is not purely a function of defense or deterrence, and numerous factors enter into the determination of success or failure of deterrence and of military action against enemies of the state, but this analysis will zero in only on ME levels and derivative variables. A state that remained at peace by avoiding MAC is considered to have successfully deterred it, whereas the occurrence of MAC can indicate failure of deterrence. Among states that experienced MACs, we will be interested to discover if ME-related variables were associated with conflict duration. Although the level of aggregation in this paper does not permit us to make judgments whether ME-related factors affected military success or failure, one could hypothesize that states that are able to terminate wars earlier were more "successful" than those that endured extended conflicts.
ME data can be used in a number of ways to examine war-peace relationships. The actual (current) value of annual ME levels expressed in national currencies or even US dollar equivalents are not suitable for time series or comparative analyses. Fortunately, WMEAT computes ME and other economic indicators at constant US dollar values (constant 1999 US dollars for the period 1989-99). Even so, the enormous international disparities in ME across diverse nations around the world make comparisons of absolute ME levels of limited if not dubious utility.
However, ME data can be standardized, transformed, and combined with other indicators to facilitate analysis. For example, WMEAT calculates ME/GNP and ME/GCE for each nation to determine "military burden" relative to national wealth and central government expenditures respectively, and computes military spending per member of the armed forces (ME/AF) as a rough measure of military capability. This paper explores patterns of relationships between the incidence and duration of MAC on the one hand, and ME, ME/CGE, ME/AF, and a composite index of these three, on the other. Specifically, it employs WMEATs global rank order data on these indicators for forty-two SSA states for the years 1989, 1994, and 1999, averaging the rankings for these three years to serve as a single, admittedly rough measure, of each indicator for the eleven-year period. Global rankings for 1989 are available in WMEAT 1990 (144 countries), for 1994 in WMEAT 1995 (172 countries), and for 1991 in WMEAT 1999-2000 (172 countries). The composite index was computed by simply averaging the average rankings on all three indicators ME, ME/CGE, and ME/AF for the three specified years. Since this composite score combines relative military spending levels with relative military fiscal burden and relative military capability measures, it is termed the Military Effort Index (MEI).
War, Peace, and Military Spending
Table 2 shows the results of cross-tabulating MAC incidence (yes or no) with high or low average global rankings on ME levels (1989, 1994, 1999). Remarkably, twenty-four of the twenty-six states that remained at peace were low spenders. Nearly as stunning is fact that eleven of the sixteen MAC-afflicted states (about 70%) were also relatively cheap on defense outlays. Put differently, the overwhelming majority of SSA states (thirty-five of forty-two, or 83%) opted to keep military spending low during the 1990s. For most of this subset (about 70%), low ME levels did not diminish whatever deterrence value was achieved by defense spending. Even the great majority of states that suffered MACs were able to keep their ME levels low despite wartime exigencies.
Table 2. War, Peace, and Military Spending in Sub-Saharan Africa, 1989-1999
Table 3 shows the cross-tabulation of MAC incidence (yes or no) with high or low average global rankings on military fiscal burden (ME/CGE) in 1989 and 1999. (WMEAT did not calculate ME/CGE rankings in 1994 because of date reliability and availability issues.) Among the twenty-six SSA states enjoying peace throughout the period, fifteen had low ME/CGE scores and eleven scored high. This dichotomous distribution suggests that peace (avoidance or deterrence of MAC) can be maintained using either "low-budget" or "gold plated" defense spending strategies relative to the governments overall fiscal resources, the former being slightly favored.
By contrast, thirteen (81%) of the sixteen states that suffered MAC exhibited high ME/CGE scores. Putting aside the chicken-and-egg issues, this strikingly pervasive association of high military fiscal burdens and MAC underscores the deleterious effects of war on other government priorities and programs.
Table 3. War, Peace, and Military Fiscal Burden in Sub-Saharan Africa, 1989-1999
If ME/AF is even crude approximation of military capability, Table 4 reveals some particularly interesting results. A total of thirty-five of the forty-two SSA states (83%) ranked low on ME/AF. Cross-tabulation of war/peace and low/high average global ME/AF rankings (1989, 1994, 1999) shows that twenty-one of the twenty-six SSA states (81%) that remained at peace during this period maintained low military capability scores. As in the case of ME, this suggests that for most SSA states, deterrence was not achieved by maintaining overweening military strength. Even more remarkably, fourteen (88%) of the sixteen conflicted states also had low ME/AF scores.
It is relevant to note here that the computation of average global rankings for these tables induced a small bias in favor of high ranking (lower numbers). In 1989 there were 144 countries in WMEAT tables, compared to 172 countries in 1994 and 1999. The average global rankings used in these tables are stated in terms of 172 countries, but each countrys average ranking was "pulled upwards" slightly by the smaller number of countries recorded in 1989. In Table 4, Cameroon (83), Kenya (85), and Angola (82) are among the states with high ME/AF average global rankings, but if they scored 87 or higher, they would be deemed to have low military capability rankings. If the calculation of average global rankings took account of the changing size of the worlds population of states, one or more of these three states might have been assigned to the respective low-ranking quadrants, thus magnifying the highly skewed results already observed.
Table 4. War, Peace, and Military Capability in Sub-Saharan Africa, 1989-1999
After examining patterns of relationships between ME, ME/CGE and ME/AF on the one hand and peace/war occurrences on the other, does the composite Military Effort Index (MEI) produce different results? In Table 5 all SSA states are classified according to whether or not they experienced MAC and whether they scored low or high on average global MEI rankings in 1989, 1994, and 1999. As in the case of our military capability measure (ME/AF), there are high concentrations of both peaceful and MAC-afflicted states in the low MEI cells. Notable too, as in the case of ME/AF, five of the nine high-ranking states had scores close enough to "low" that their assigned quadrant could have been affected by database size anomalies. If so, it would produce an even more pronounced concentration of states in the low military capability quadrants.
Table 5. War, Peace, and Military Effort in Sub-Saharan Africa, 1989-1999
Conflict Duration and Military Spending
Table 6 assigns the sixteen SSA states that experienced MAC into quadrants depending on the duration of their conflicts and average global ME rankings. Ten conflicts were short, six long. Interestingly, the modal outcome, reflected in seven (44%) of the cases, was a conjunction of low ME ranking and shorter wars. One might be tempted to state the obvious, that short wars do not cost as much as long ones; however, two-thirds of the states that suffered protracted MACs also had low ME rankings. Among the five states that ranked high on ME levels, three experienced short wars and two had long ones; the small number of states in these two cells makes any further judgment inadvisable.
Table 6. Conflict Duration and Military Spending in Sub-Saharan Africa, 1989-1999
In Table 7 we examine the distribution of MAC-affected states according to conflict duration and average global rankings on military fiscal burden (ME/CGE). The thirteen states that registered high ME/CGE ranks were about evenly split between short and long wars. Thus, when deterrence failed and war broke out, higher defense outlays relative to government budgetary resources was the price few could avoid. Equally apparent is that making greater sacrifices by increasing the proportion of budget allocations to the military did not appreciably affect the duration of conflicts.
Table 7. Conflict Duration and Military Fiscal Burden in Sub-Saharan Africa, 1989-1999
Table 8 reveals a reverse pattern when the average global rankings on military capability (ME/AF) for the sixteen SSA states with MACs are mapped against war duration. Nearly all states in conflict had low military capability scores, regardless of length of war, but the majority of cases (9 56%) experienced shorter wars; for five other (31%) low-spending states, longer wars prevailed. If, as was surmised, states with greater relative military capabilities might be able to terminate conflicts in a shorter time than states with lesser military capabilities, the SSA data do not afford an opportunity to test that hypothesis: there were only two states that ranked high on military capability; one had a long war and the other a short one.
Table 8. Conflict Duration and Military Capability in Sub-Saharan Africa, 1989-1999
Finally, when all three ME-related indicators are taken into account by using the average global ranking on the composite military effort index (MEI), the numerical distribution of cases is identical to that obtained by using military spending levels (ME) alone and very close to that of military capability. In Table 9 the cell with the largest number of states (44%) is that with seven countries characterized by both shorter MACs and low levels of overall military effort. Four other states with low MEI rankings experienced longer wars.
Table 9. Conflict Duration and Military Effort in Sub-Saharan Africa, 1989-1999
Summary of Findings
However, relative military fiscal burden (ME/CGE) did turn out to be a significant differentiator. Among the twenty-six peaceful states, there was a notable split between those that were cheap on defense and those that invested rather more heavily in it. Among the sixteen MAC-afflicted states, a large majority carried high military fiscal burdens.
Regarding the duration of armed conflict, ten of the sixteen states with MACs experienced short wars, and the modal or majority groupings of states were those with low military spending (ME), low military capability (ME/AF), low overall military effort (MEI), and short wars. In other words, there appears to be a modest association between lower ranking ME-related factors and short conflicts.
However, unlike the findings regarding relative military fiscal burden and the incidence of conflict, ME/CGE bore no clear relationship to the duration of war. Most conflicted states labored under high military burdens, but they were almost evenly divided between short and long wars.
Further Observations and Thoughts
Only one measure military fiscal burden (ME/CGE) stood out among our selected ME-related indicators as a discriminator. Curiously, the twenty-six peaceful states employed either low or high military fiscal burden strategies in nearly equal measure. It would be interesting and perhaps instructive to compare these two categories of states to determine the reasons for their divergent approaches to military burden bearing.
On the other hand, for most of the sixteen war-prone SSA states, the price of conflict was a higher proportion of national treasure allocated to the defense sector. But unfortunately, such greater sacrifices did not make any apparent difference in the duration of their wars: the thirteen states with the highest ME/GCE scores were almost evenly divided between those that experienced short conflicts and those that suffered protracted wars.
That ME/CGE should turn out to be the only discriminating factor to emerge in this study is significant for both analytical and policy-related reasons. Analytically, our finding corroborates the crowding-out effect of increased military spending on government social expenditures found to be almost universal among Third World states in conflicts (Stewart & FitzGerald 2001: 83-89, 231). Policy-wise, by underscoring the difficult choices and real tradeoffs made by governments at war, it highlights the responsibility they, their allies, and their armed adversaries share for the consequences of falling into the war trap. Not only do they suffer heavy social and economic costs and developmental regression, but even the increased burdens of defense do not reduce the duration of war.
Finally, two questions bear further reflection and investigation. First, regarding war and peace: how should we interpret the curious finding that the overwhelming majority of both peaceful and war-prone states had a common military profile: low rankings on most ME-related indicators? What does this say of the debates about "peace through strength" and realist theories of conflict and deterrence? True, such theories are about international relations and conflict, and most of Africas wars have been internal. But does this mean that the military power of states (even weak ones) is irrelevant to their internal security? For SSA states in the 1990s it certainly seems so.
Or does it? Since a substantial majority of the SSA states that suffered MACs had low ME-related scores, does this imply that military weakness was a prelude to or correlate of conflict? Not in the face of the finding that more than four-fifths of the states that enjoyed uninterrupted peace also had low military capability scores. The puzzle remains.
The second question concerns the duration of armed conflict: how should we interpret the evident but modest consistent association between lower ranking ME-related factors (ME, ME/AF, and MEI) and short conflicts? At first blush, such patterns seem counterintuitive: one might logically expect states with lower levels of military spending, military capability, and overall military effort to be unable to terminate their conflicts sooner than more militarily robust states. Here we would suggest that other factors were probably at work, combining with low ME-related indicators to produce this peculiar outcome. As acknowledged earlier, the simple research design of this paper is not well suited to using conflict duration as a measure of military success or failure.
In the larger picture, this paper has only scratched the surface of a complex set of issues and relationships. Hopefully, it will stimulate further and more advanced research on the security-related aspects of military spending in Africa and elsewhere.
Mohammed, Nadir A. L. 1999. "Civil Wars and Military Expenditures: A Note." Draft paper at http://econ.worldbank.org/files/13215_Mohammed.pdf.
Omitoogun, Wuyi. 2000. "Military Expenditure in Africa." Appendix 5D in SIPRI Yearbook 2000. Oxford: Oxford University Press: 281-298.
________. 2001. "Military Expenditure and Conflict in Africa." DPMN Bulletin 8/1: 10 pp. [Leiden: Development Policy Management Network, African Studies Centre].
Stewart, Frances, Valpy FitzGerald and associates. 2001. War and Underdevelopment. 2 vols. Volume 1: The Economic and Social Consequences of Conflict. Oxford: Oxford University Press.
Stockholm International Peace Research Institute. SIPRI Yearbook: World Armaments and Disarmament. Annual.
U.S. Department of State. World Military Expenditures and Arms Transfers. Annual.
This paper was prepared for presentation on the panel "Managing Conflict: The Military-Diplomatic Dimension" during the 46th Annual Meeting of the African Studies Association, Boston, MA, October 30 November 2, 2003. Its original title was "War and Peace in Sub-Saharan Africa, 1989-1999: Does Military Spending Matter?"