Abstract

Poverty and armed conflict are closely connected, so the poorest countries face the prospect of being caught in a “conflict trap” of poverty and recurring conflict (Collier and others, 2003). One group of countries particularly at risk for conflict are those that have recently emerged from conflict. It is particularly important to help these countries ensure a quick recovery from conflict and a return to sustainable development. This effort requires an understanding of the economic features of the conflict cycle, and this chapter seeks to contribute to this topic. In particular, this chapter looks at the evolution of economic performance and the role of macroeconomic policy and aid in a selection of 24 countries as they passed through civil conflict and through the first few years of postconflict recovery. The chapter offers three main findings leading to three policy implications.

Section I. Introduction

Poverty and armed conflict are closely connected, so the poorest countries face the prospect of being caught in a “conflict trap” of poverty and recurring conflict (Collier and others, 2003). One group of countries particularly at risk for conflict are those that have recently emerged from conflict. It is particularly important to help these countries ensure a quick recovery from conflict and a return to sustainable development. This effort requires an understanding of the economic features of the conflict cycle, and this chapter seeks to contribute to this topic. In particular, this chapter looks at the evolution of economic performance and the role of macroeconomic policy and aid in a selection of 24 countries as they passed through civil conflict and through the first few years of postconflict recovery. The chapter offers three main findings leading to three policy implications.

Main Findings

Shift in the key features of the conflict cycle. The first finding, developed in Section II, is that, although the economic performance of countries affected by conflict shares many features in common, the data point to a discernible and statistically significant shift in the key economic characteristics of the conflict cycle occurring at the start of the 1990s, where the conflict cycle is defined to end when real GDP per capita recovers to its preconflict level. Compared with earlier conflicts, those of the 1990s were shorter and associated with deeper economic contractions. In addition, whereas countries emerged from earlier conflicts following a prolonged period of recovery, they emerged from conflicts in the 1990s at a much earlier stage of the conflict cycle and faced significantly worse conditions. At the same time, compared with earlier conflicts, countries also generally came out of later conflicts with modestly higher growth in the first few years after conflict. This growth, although only modestly higher than after earlier conflicts, represented a significantly stronger rebound in growth from the low (negative) levels prevailing during conflict. In many cases, the conflict cycle to the recovery of output to preconflict levels remains incomplete. The chapter provides projections that suggest that the shift tended to reduce the overall length of the conflict cycle while redistributing the time spent in conflict in favor of the time spent in recovery. At the same time, it has probably also tended to reduce the overall economic cost of output foregone over the conflict cycle.

The role of macroeconomic policy. The second finding, discussed in Section III, relates to the linkages between the evolution of economic activity and macroeconomic policy. Conflict was typically accompanied by a deterioration and recovery in key macroeconomic policy variables that was much more pronounced in the conflicts of the 1990s. The changes observed over the past decade suggest that the stance of macroeconomic policy has also had a more discernible and statistically significant impact on economic activity in recent conflicts than in earlier years.1 The chapter provides estimates that the policy stance moderated the decline and also the initial recovery in output growth in earlier conflict cycles, but accentuated it in more recent episodes. In particular, the stronger macroeconomic stabilization effort, especially with respect to inflation, has been an important factor underlying the stronger postconflict recovery of growth observed in the 1990s. Another perhaps rather surprising result also emerges. Once other factors, including policy, are taken into account, the initial impact of the start and end of hostilities on output growth was quite similar for the two sets of conflict—suggesting that, despite the different economic profiles, the same underlying “conflict process” was at work.

The role of external assistance. The third finding, discussed in Section IV, is that these changes in part reflected a shift in donor practices following the end of the cold war that resulted in donors being less inclined to support countries during conflict but also more willing to provide assistance after conflict. Establishing linkages statistically is beyond the scope of this chapter, so the argument is only suggestive. Donors tended to increase financial assistance during pre-1990 conflicts but reduced assistance once conflict ended. In the conflicts of the 1990s, donors generally reduced assistance sharply during conflict, but also tended to increase assistance equally sharply after conflict. This difference may have contributed to more severe economic contractions and imbalances experienced by countries in these later conflicts and plausibly also contributed to their shorter duration. The greater donor willingness to provide support after recent conflicts has also contributed to stronger postconflict recoveries, which points to the potentially high productivity of aid targeted toward macroeconomic stabilization in the early postconflict recovery period.

Experience in countries receiving emergency postconflict assistance. These findings are buttressed by the experience, discussed in Section V, of six countries that have received emergency postconflict assistance (EPCA) from the International Monetary Fund since 1995.2 The experience of these countries was broadly similar to that of other countries in the 1990s, but their performance in the first two years after conflict was generally stronger. This was arguably because their stronger commitment to sound macroeconomic policies provided the basis for the international community to provide financial support soon after conflict. In this respect, there was an important virtuous cycle in operation: Sound policy attracted external assistance that made these policies easier to implement and more fruitful.

Experience in the Democratic Republic of the Congo (DRC). These findings are again buttressed by the experience, also discussed in Section V, of the DRC (which did not receive EPCA from the IMF) where stabilization and the start of economic recovery were made more difficult by delays in official external assistance. However, although the DRC’s initial postconflict growth performance was unfavorable, the government’s firm commitment to good policies was rewarded by an improvement in performance that was one of the strongest in the 1990s.

Policy Implications

The findings of this chapter point to three policy implications.

Postconflict policy priorities. First, if civil conflicts in the 1990s are representative, then compared with their earlier counterparts, countries emerging from recent conflicts face a more urgent need to restore macroeconomic stability, and the economic benefits of stabilization are also correspondingly larger. However, they also face competing political pressures that can be at odds with stabilization, including, for example, the urgent need to increase government spending to meet immediate social priorities. Nevertheless, there appears to be the need to assign a higher priority to postconflict macroeconomic stabilization than in the past.

Postconflict aid. There are possible implications for the timing and type of postconflict aid. Aid following recent conflicts has tended to peak immediately after conflict, but the ability of a country emerging from conflict to make use of aid is constrained by its political and administrative capacity. It has therefore been suggested that aid might be more effectively used if delayed until capacity was restored. This is perhaps particularly the case for project aid. In contrast, the evidence on recent conflict cases presented in this chapter indicates that the productivity of external assistance can be high in the initial postconflict period when the government is committed to following a sound macroeconomic strategy, particularly if assistance is provided to the budget in support of macroeconomic stabilization.

Aid during conflict. Finally, the findings suggest the intriguing possibility that the international community may be able, through its aid policies, to influence the economic profile of the conflict cycle, especially the trade-off between the duration and the economic severity of conflict. However, it is not clear what portion of the trade-off is preferable because, at first glance, the shift observed in the 1990s appears to have been accompanied by only a modest reduction in the short-term economic cost of conflict. Also, any evaluation would need to take into account the longer-term human and economic costs involved. This is obviously a complex issue that needs further attention.

Literature Review

Five areas of the literature of particular interest here relate to the causes of conflict, the length of conflict, the impact of conflict, the impact of macroeconomic policy, and the role of external assistance.3

Causes of conflict. Recent analyses have tended to downplay the traditional explanations of civil conflict revolving around the politics of grievance and have instead tended to highlight economic factors (Blomberg and Hess, 2002; Collier, 2000; Collier and Hoeffler, 1998, 2000, 2002a; Fearon and Laitin, 2002; Nafziger and Auvinen 2002). In particular, the propensity to civil conflict has been closely linked to economic stagnation and poverty, although the direction of causality is not altogether clear.

Length of conflict. The shortening of conflicts in the 1990s has been noted by Fearon (2002). No clear consensus has emerged on the factors underlying the length of conflict. It has been persuasively argued by Collier, Hoeffler, and Söderbom (2001) that the sort of factors typically used to explain the initiation of conflict have generally had little bearing on the duration of conflict. One regular feature is that, as well as being more prone to conflict, poorer countries also typically endure longer conflicts arguably because of their lower capacity to inflict damage. Fearon (2002) has also linked the length of conflict to the political nature of the conflict. For example, civil conflicts arising from coup attempts or popular revolutions or involving successful peripheral secessions have tended to be relatively brief, whereas conflicts revolving around land claims or natural resources tend to be relatively prolonged. Collier and Hoeffler (2000) have sought to explain the length of conflict by focusing on sources of financing for conflict, especially natural resources, as well as the balance of benefits to the parties involved, once conflict has started, to perpetuate conflict. The role of external assistance, especially related to military spending, in sustaining conflict has been noted in Michailof, Kostner, and Devictor (2002), and Elbadawi (2000) has also looked at the role of foreign interventions, especially in terminating conflicts.

Impact of conflict. A number of authors have looked at the cause of output losses in conflict, including Arunatilake, Jayasuriya, and Kelegama (2001), Caplan (2001), Collier (1999), Imai and Weinstein (2000), and Knight, Loayza, and Villanueva (1996). A broad consensus has emerged that civil conflict reduces annual real GDP growth by about 2 percentage points. Collier (1999) also found that the negative impact of conflict persisted long after conflict. As might be expected, this work links output losses to the geographical extent of the conflict and the destruction of the human and capital stock; the disruption of government capacity to collect revenues and provide essential services; and the general disruption of commerce. A promising line of inquiry pursued by Murdoch and Sandler (2001a, b) focuses on the spillover effects from conflicts in neighboring countries and the compounding of the damage when they are part of a broader set of regional conflicts. There has been limited work on the impact of conflict on other key macroeconomic indicators. Caplan (2001) found no discernible effect of civil conflict on inflation and only a limited tendency for government spending to increase relative to GDP. Gupta and others (2002b) provide more conclusive evidence that conflict led to higher inflation, higher government spending, and higher fiscal deficits.

Impact of macroeconomic policy. The impact of macroeconomic policy on output during conflict has received very limited attention. Gupta and others (2002b), looking at experience in the 1990s, found that growth was affected by changes in the composition of government spending and the reduction in social spending in favor of military spending. Helpfully, there are a number of studies across a broad spectrum of developing countries (not necessarily in conflict) with results that can also shed light on conflict situations, especially concerning postconflict macroeconomic stabilization.

There is now a consensus that healthy fiscal balances are generally good for economic growth, but there is less agreement on the short-term impact of fiscal consolidation. The standard Keynesian conclusion that fiscal consolidation reduces growth relies on the multiplier for government spending exceeding that for tax revenues. This conclusion finds support in a survey of the empirical literature on multipliers for government spending and factor input taxes by Gerson (1998). However, Gupta and others (2002a) have shown that fiscal consolidations in the 1990s have had a positive short-term impact on growth: a reduction of 1 percentage point in the fiscal deficit-to-GDP ratio led to a short-term increase in per capita output growth of 0.5 percentage point. The impact was larger when consolidation was based on current spending cuts rather than on revenue increases or capital spending cuts and when offset by reduced domestic rather than external financing.

The linkage between fiscal deficits and inflation also remains contested, but both Catao and Terrones (2001) and Fischer, Sahay, and Vegh (2002) found strong support for a linkage when inflation is high. A 1 percentage point reduction in the fiscal deficit-to-GDP ratio reduces inflation by up to 6 percentage points. There is now strong support that even moderate inflation can damage economic growth (Brauman, 2000; Ghosh and Philips, 1998) above a threshold in developing countries of about 10 percent (Khan and Senhadji, 2000). There is less agreement on the impact of disinflation. Disinflation has a positive impact on growth when inflation is very high (Fischer, Sahay, and Vegh 2002), but may have a contractionary effect if the inflation rate is already low or if the disinflation is too severe (Ghosh and Philips, 1998).

External assistance. Michailof, Kostner, and Devictor (2002) noted the role of aid in sustaining conflict during the cold war and the changes in aid patterns since the cold war. Otherwise, discussions of the role of aid have tended to focus on the postconflict recovery period. An important strand in the literature relates to the timing and type of aid. Collier and Hoeffler (2002b) and Collier and others (2003) noted that aid tends to peak immediately after conflict and argue that aid would be more effective if it peaked about three to five years after the end of conflict when absorptive capacity is at its highest. These papers also provide evidence that, relative to other countries, aid to conflict-affected countries in support of social priorities is relatively more effective than aid in support of economic reconstruction and macroeconomic stabilization, in part because of its impact in reducing the probability of renewed hostilities. Demekas, McHugh, and Kosma (2002) have also argued that, compared with humanitarian assistance, reconstruction aid supports longer-term capital accumulation and growth but at the expense of lower current consumption.

This chapter within the literature. The chapter seeks to extend the conflict-related literature in three directions. The literature has tended to focus on the bulk of civil conflicts that started before 1990, whereas this chapter argues that there are important features that do not carry over to the more recent conflicts. A second feature of the literature is its tendency to treat conflict as a single event, while the current chapter argues that the economic phase at which a country comes out of conflict has important implications for the nature of the postconflict recovery. Finally, this chapter emphasizes the role of macroeconomic policy, and indirectly of aid, as a determinant of growth during the conflict cycle.

The Data

The chapter looks at economic developments in 24 civil conflicts in 23 countries that have taken place since 1970 (Table A5.1). This set has been pared down from a much larger set, according to data availability and whether the conflict had a discernible macroeconomic impact.4 For example, the conflict in Vietnam was not included because of a lack of data, while India’s regional conflict in Kashmir was excluded because of a lack of a discernible economic impact on India as a whole. The 24 conflicts are divided by starting date into two groups: 10 that began before 1990 and 14 that commenced after 1990.5 In addition, performance was markedly different in four conflicts in the 1990s that arose out of dissolved federal entities (DFEs) in the Soviet Union and the Republic of Yugoslavia.

Setting beginning and end dates of conflict often requires judgment. This is especially true for internal conflicts, in which the descent into and emergence from conflict is often gradual and intermittent. For example, Sierra Leone experienced internal disturbances for several years before the coup in 1997—the date used here for this conflict. Decisions on the dates to use for analytical purposes are based on information from the Swedish International Peace Research Institute (SIPRI) and IMF staff reports.6 The chapter uses data up to 2002 from the IMF’s World Economic Outlook (WEO) database (2003). As might be expected, the quality of the data during and after conflict is weak. Data for several conflict cases are limited and are available for up to 2 years for 24 countries but are available up to 5 years for only 17.

Section II. Economic Performance over the Conflict Cycle

Profile of the Conflict Cycle

The economic cycle related to conflict is normally divided into three distinct phases that correspond to its political phases: a preconflict phase of economic deterioration, the period during conflict of reduced growth or contraction, and a postconflict phase of economic recovery. However, this perspective can be misleading if the political and economic phases of conflict are not in fact synchronized, or if there is a change in how they are synchronized. Precisely such a shift appears to have occurred in the 1990s, with important implications for the stage of the economic cycle at which countries emerged from conflict (Table 5.1, Figures 5.1 and 5.2).

Table 5.1.

Length of Contraction and Recovery During Conflict

(Years)

article image
Source: WEO.

Only includes contraction during conflict.

Figure 5.1.
Figure 5.1.

Real GDP

(Level, Index = 100 one year before conflict)

Source: WEO.
Figure 5.2.
Figure 5.2.

Real GDP Per Capita

(Level, Index = 100 one year before conflict)

Source: WEO.

Contraction. A distinguishing feature of civil conflicts, unlike cross-border conflicts, is that their onset is typically associated with a decline in economic performance.7 However, in many instances, this deterioration predated the conflict: real GDP growth began to fall as much as five years before conflict, and more than half the countries (mostly in post-1990 conflicts) entered conflict with reduced GDP per capita. With the onset of conflict, most countries experienced a contraction in output that averaged about three years for all conflicts, but that was generally longer for the pre-1990 conflicts. In terms of output per capita, the contractions tended to last about a year longer. More important, the end of the contraction and the start of the economic recovery did not necessarily coincide with the end of conflict.

Recovery. A major change in the 1990s was a sharp decline in the length of conflicts, organized by starting date, which fell from an average of 12 years for pre-1990 conflicts to only 4 years for those of the 1990s.8 In the pre-1990 conflicts, hostilities tended to end well after the start of the recovery so that there was a prolonged period of “in-conflict” recovery. In the shorter post-1990 conflicts, hostilities tended to end at about the same time as the end of the contraction (and occasionally before), and in these cases an “in-conflict” recovery period was typically absent.9 As a result, countries emerged from these conflicts at a much earlier phase of the economic cycle than did their earlier counterparts.

Phases of recovery. Where along the economic cycle the country emerged from conflict had an important bearing on the economic conditions it faced. It is useful to distinguish three phases: a stabilization phase in which major macroeconomic imbalances were corrected along with positive output growth; a reconstruction phase, during which the security and policy environment was normalized and recovery was fully supported by donor-funded reconstruction programs; and a final development phase leading to a return to normal growth. The timing of these phases relative to the end of conflict depended, in part, on the overall length of conflict and whether there was a period of in-conflict recovery. In the pre-1990 conflicts, which typically included a lengthy in-conflict recovery period, the modest stabilization required was mostly in place by the end of conflict. In the post-1990 conflicts, where the contraction typically continued to the end of conflict, the stabilization phase only began once conflict ended and in some cases was delayed for several years after conflict.

Economic Growth

The shortening of conflicts in the 1990s was also accompanied by a marked worsening of economic performance during conflict and consequently in the conditions that countries faced as they emerged from conflict. At the same time, there was a broad improvement in economic performance in the initial years after conflict (Table 5.2). For each set of conflicts, Table 5.3 shows the difference between growth during a period extending five years before and after conflict and growth in all other periods (i.e., under normal circumstances) as well as the difference in growth between the two sets of conflicts. It also shows the statistical significance of these differences.

Table 5.2.

Evolution of Real GDP and Real GDP per Capita

article image
Source: WEO.Note: DFE, dissolved federal entity.

The contraction period and trough are defined in terms of real GDP and real GDP per capita, respectively.

Not all post-1990 conflicts experienced an in-conflict recovery.

Data for 3, 4, and 5 years after conflict are available for 11, 7, and 7 of 14 post-1990 conflicts, respectively.

Table 5.3.

Evolution of Output over the Conflict Cycle: Differences in Growth During and Between Conflicts

article image
Source: WEO.Notes: Differences evaluated using regressions allowing for serial correlation. Levels of significance at the 1, 5, and 10 percent levels are indicated by ***, **, and *, respectively.

Pre-1990 conflicts. In the pre-1990 conflicts, the pace and depth of the contractions tended to be relatively modest with an average annual decline in real GDP per capita of about 4 percent—or about 5 percentage points below normal—and a cumulative decline in real GDP per capita at its trough to about 84 percent of its preconflict level. However, since their contractions were also typically followed by lengthy recovery periods during the conflict itself, these countries emerged from conflict with a level of real GDP per capita not far below the preconflict level and real GDP significantly higher than before the conflict. As a result, over the whole conflict, output actually increased by an annual average of 2 percent. Also, in these countries, the end of hostilities had no immediately discernible impact on growth.

Post-1990 conflicts. The experience in the far-shorter conflicts during the 1990s was altogether different. In these conflicts, the pace and depth of the contractions were much more severe: real GDP per capita declined by about 12 percent each year—or more than 14 percentage points below normal—to about 71 percent of its preconflict level, although this was biased downward by the particularly sharp contractions in the DFEs.10 Moreover, the contractions in the post-1990 conflicts typically continued to the end of conflict, and most countries emerged from conflict with output still far below the preconflict level. Once conflict ended, per capita output growth in the first five years was modestly higher for the later than for the earlier conflict countries, but this difference was not statistically significant. However, it is important to keep in mind that, for the later conflicts, growth after conflict was recovering from deep contractions during conflict, and the rebound was large and statistically significant. Moreover, performance in these later conflicts varied considerably because of the delayed recovery in some countries, especially in two DFEs.11 Once the DFEs are excluded, the superior performance of the remaining post-1990 conflict countries was particularly pronounced in the initial two years after conflict when average per capita output growth in these countries was several times higher that in the pre-1990 conflict countries.

Length of the Overall Conflict Cycle

To assess the impact of the shifts in the 1990s, it is assumed that the return of real GDP per capita to its level the year before conflict marks the point of recovery.12 By this measure, the recovery period after conflict was often longer than the time spent in conflict. In many instances, recovery remains incomplete. Recovery is even more prolonged, especially for conflicts in the 1990s, if the level of GDP per capita that prevailed before the preconflict deterioration is used as the benchmark. Of the 17 countries in the sample with sufficient data, only 5 had regained their preconflict level of GDP per capita within the first five years after conflict. From a different perspective, GDP per capita, which averaged 84 percent of its preconflict level at the end of the conflict, had risen to only 93 percent after five years.

Although the conflict cycles for most countries remain incomplete, WEO projections can be used to get some sense of their likely length. Using these projections, the recent developments have most probably lengthened postconflict recovery, especially in the DFEs (Table 5.4). For the most part, this is because the shift in the 1990s has redistributed the time spent in recovery from during conflict to after conflict, leaving the overall length of the conflict cycle broadly unchanged. However, once the idiosyncratic DFEs are excluded, the shift in the 1990s has probably tended to reduce the overall length of recovery, including the in-conflict recovery, and has consequently reduced the length of the conflict cycle from about 15 years to about 11 years.

Table 5.4.

Length of the Conflict Cycle Periods

(Years)

article image
Source: WEO.Notes: Estimates are based on WEO projections (Winter 2003) up to 2008 and extrapolated beyond. DFE, dissolved federal entity.

The contraction to the trough of per capita output during conflict, excluding contractions after conflict.

The conflict cycle is defined to end when real GDP per capita recovers to its preconflict level.

The Cost of Conflict

These projections can be used to estimate the overall economic cost of conflict. This can be measured by the net present value of the output foregone, again taking the preconflict level of output per capita as the benchmark (Table 5.5). These estimates are sensitive to the discount rate and also understate the cost to the extent that, in the absence of conflict, per capita output growth would have been positive, and also because the estimates do not capture the potentially large longer-term human and economic costs.

Table 5.5.

The Economic Cost of Conflict: Net Present Value in Months of Preconflict Output per Capita

article image
Sources: WEO and IMF staff estimates.Notes: The conflict cycle is defined to end when real GDP per capita recovers to its preconflict level. Estimates are based on WEO projections (Winter 2003) up to 2008 and extrapolated beyond. DFE, dissolved federal entities.

Subject to this qualification, the developments observed in the 1990s have reduced the average cost of the conflict period alone—the cost of the deeper contractions of the post-1990 conflicts has been more than offset by their greater brevity. However, these developments have also been accompanied by higher costs associated with the elongation of the postconflict recovery period noted above. Nevertheless, excluding the idiosyncratic DFEs, there has probably been a tendency for the overall cost of the whole conflict cycle to decline. Assuming a discount rate of 3 percent, the cost of conflict alone has declined from 13 to 6 months of preconflict economic activity, while the cost of the whole conflict cycle has declined from 18 to 15 months.13

Evolution of Macroeconomic Policy Indicators

The above developments have also been reflected in the evolution of policy indicators (Tables 5.6 and A5.2).

Table 5.6.

Macroeconomic Policy Indicators

(Period average unless otherwise indicated)

article image
Source: WEO.

Data for 3, 4, and 5 years after conflict are available for 11, 7, and 7 of 14 post-1990 conflicts, respectively.

Fiscal balances. Countries typically ran fiscal deficits (including grants) before conflict that deteriorated and then recovered as they passed through the conflict cycle. These fluctuations tended to be modest relative to GDP but were much more severe relative to revenues (including grants).14 Because of their deeper economic contractions, lower levels of external assistance, and absence of a prolonged recovery period during conflict, countries generally emerged from post-1990 conflicts with much larger fiscal deficits than was the case in earlier conflicts.15 However, in the initial postconflict period, the post-1990 conflict countries benefited from stronger economic growth as well as greater external assistance, so the improvement in their fiscal balances was generally more pronounced.

Revenues and expenditures. Both revenues and expenditures were initially compressed during conflict but, while revenues tended to remain low or decline, there were strong pressures to maintain or increase expenditures.16 In the pre-1990 conflict countries, increased fiscal deficits during conflict were generally accompanied by increased revenues and expenditures whether in real terms or relative to GDP. After conflict, fiscal consolidation tended to rely more on an adjustment in expenditures than in revenues, and real spending per capita tended to decline. In the post-1990 conflicts, however, both revenues and expenditures generally remained compressed in real terms during conflict, but the compression of revenues tended to be larger. After conflict, the fiscal adjustment tended to rely more on increased revenues to accommodate increased expenditures.

Inflation. Changes in the level and composition of assistance caused financing to evolve differently from the fiscal balances.17 In the pre-1990 conflicts domestic financing during conflict actually declined relative to GDP but increased following conflict. In the 1990s, domestic financing increased more sharply than the fiscal deficits during conflict, but also improved more sharply following conflict. The evolution of domestic financing was reflected in CPI inflation.18 The increase in inflation was more pronounced in the conflicts of the 1990s and, by the end of conflict, median inflation in the post-1990 conflicts (41 percent) was much higher than in the pre-1990 conflicts (12 percent).19 Once conflict ended, inflation generally declined, but the decline was uneven. In the pre-1990 conflict countries, where domestic financing initially increased, median inflation actually accelerated in the initial two years after conflict and only fell to single digits in the fourth year. In most of the post-1990 conflict countries, where domestic financing was sharply reduced, inflation declined to single digits within two years.

External sector. External current account balances (including transfers) generally deteriorated during the conflict mainly because of reduced official transfers. The role played by the trade balance during conflict was mixed, because there were several conflicting tendencies relating to trade volumes and on the effective terms of trade, but there was probably a larger tendency toward deterioration.20 The performance of the current account balance after conflict tended to diverge between the two sets of conflicts. The end of the pre-1990 conflicts followed a prolonged recovery that was already supported by large aid flows and robust exports and was accompanied by a surge in imports that was not supported by either stronger aid flows or export receipts so that current account balances worsened. The end of the post-1990 conflicts was also accompanied by a surge in imports, which, however, was supported by strong increases in aid transfers as well as export receipts so that current account balances in these countries initially improved.

External debt. Countries also emerged from conflict with increased external debt that averaged 117 percent of GDP and 743 percent of exports—which was not sustainable.21 Countries were therefore in urgent need of debt relief, and two-thirds of them subsequently became eligible for assistance under the Heavily Indebted Poor Countries Initiative. Debt-service obligations were also very high, and more than half the countries emerged from conflict with arrears, which, in some instances, posed a major hurdle to the provision of external assistance.

Section III. The Impact of Conflict and Macroeconomic Policy

The movements in output and economic indicators described above were a natural outcome of conflict. However, they were also responding to shifts in the stance of macroeconomic policy. For example, the tendency after pre-1990 conflicts for inflation to rise initially while fiscal balances were improved by reducing expenditures arguably reduced growth. In the post-1990 conflicts, the tendency for inflation to fall while fiscal balances were improved by increasing revenues to accommodate increased expenditures may have supported growth.

The Equations

These issues are assessed using equations (1) through (3), each of which contains real GDP per capita growth as the dependent variable. Equations (1) and (2) are each estimated separately for each set of conflicts while equation (3) is estimated using a panel of all conflicts.

The level of per capita real GDP growth:

Yit=β1,75.Yi75+β1o.do+kΣβ1ok.do.χkit+cΣβ1c.dc+ΣkΣcβ1ck.dc.χkit+ɛit(1)

Difference between growth during and outside the conflict cycle:

Yit=c+β2,75.Yi75+Σkβ2k.do.χkit+Σcβ2c.dc+ΣkΣcβ2ck.dc.Xkit+ɛit(2)

The difference between growth during the pre-1990 and post-1990 conflict cycles:

Yit=β3,75.Yi75+β3o.do+Σkβ3ok.do.χkit+Σcβ3c.dc.+ΣkΣcβ3ck.dc.χkit+β3po.dp.do+Σkβ3pok.dp.do.χkit+Σcβ3pc.dp.dc+ΣkΣcβ3pck.dp.dc.χkit+ɛit(3)

In these equations, Yit is real GDP per capita growth in country i in year t, Yi75 is the level of real GDP per capita in country i in 1975, χkit is the value of explanatory variable k for country i in year t, ɛit is the error term, c is the equation constant, do is the dummy for period outside the conflict cycle, the dc’s are dummies for the subperiods during the conflict cycle, dp is the dummy for post-1990 conflicts, and the β’s are the coefficients to be estimated. There are no individual country-specific dummies.

For each set of conflicts and for each period in the conflict cycle, equation (1) shows the estimated growth rate as a linear combination of an “underlying” growth rate, captured by the coefficient on the dummy, plus the unit contributions of the other explanatory variables. Because there are no cross-period linkages, this equation generates the same results as would a separate equation estimated for each period.

Equation (2) compares growth during the conflict cycle with growth outside the conflict cycle under “normal” circumstances. Growth is estimated as the normal growth outside the conflict cycle plus a difference attributable to conflict. For each set of conflicts, the difference between growth during the conflict cycle and normal is given by the variables cross-multiplied by the conflict cycle period dummies, dc.

Equation (3) compares growth in the two sets of conflicts. For each period, growth is given by growth in the pre-1990 conflicts plus a difference attributable to the post-1990 conflicts. For each period, the difference between the two sets of conflicts is captured by the variables cross-multiplied by the dummy for the post-1990 conflicts, dp.

For estimation purposes only, the conflict cycle is here defined, somewhat arbitrarily, to extend five years before and after conflict. All other periods are here taken to be outside the conflict cycle. The preconflict period is included to permit the estimates for the impact of conflict to take into account the deterioration in economic performance before conflict. Defining the conflict cycle to end after only five years is solely due to data availability.

To explore the evolution of output more closely, each equation is estimated with two variants of the dummies for the conflict cycle. Variant (A) divides the conflict cycle into the five-year period before conflict, the conflict itself, and the five-year period after conflict. Variant (B) divides the conflict period into its contractionary and in-conflict recovery periods and divides the five-year postconflict period into the initial two years and the subsequent three years.22 This permits a closer look at the more immediate impact of the start and end of hostilities on growth. For the post-1990 conflicts, only a small number of countries went through an in-conflict recovery and data availability falls off significantly in the three- to five-year period after conflict, so the results for these periods need to be interpreted cautiously.

Efforts to model output growth have typically used multiyear averaging to smooth out short-term fluctuations and have relied on a smorgasbord of explanatory variables, such as education rates, as proxies for supply-side variables. Multiyear averaging is precluded here by the focus on the short subperiods during the conflict cycle. Nor is it possible to make use of supply-side indicators, which tend to be surveyed infrequently during conflict, if at all.

Instead, output growth is modeled on the demand side and as a function of the key policy variables, which include growth in the terms of trade, growth in real per capita government spending, the change in the fiscal balance, and the CPI inflation rate.23 The fiscal balance is expressed as a ratio to revenues (both including grants), rather than as a ratio to GDP, so as to reduce the linkage to output. The change in the fiscal balance and government per capita spending are both included so as to differentiate between changes in the size of government and pure fiscal consolidation. The terms of trade capture external effects on the assumption that exporters are price takers. Finally, to capture possible convergence effects, due to different initial conditions, the equations also include per capita output in 1975.24

The equations are estimated using generalized least squares with cross-section weights (WGLS) with adjustments for serial correlation and heteroskedasticity. The results are shown in Tables 5.7 and 5.8. It is important to keep in mind that equations (2) and (3) refer to differences. To ease the presentation, the tables only show the coefficients for the difference components of these equations—tagged by the conflict cycle dummies dc in equation (2) and by the post-1990 conflict dummy dp in equation (3).

Table 5.7.

Real GDP per Capita Growth During the Conflict Cycle

(Results for equations (13) using variant (A) of dummies during the conflict cycle; for equations (2) and (3), only the coefficients on the difference terms are shown)

article image
Source: WEO.Notes: Levels of significance at the 1, 5, and 10 percent levels are indicated by ***, **, and *, respectively. All estimates done by generalized least squares with cross-section weights (WGLS). Gvt., government; rev, revenue; p.c., per capita; %chg, percentage change.All estimates done by WGLS.
Table 5.8.

Real GDP per Capita Growth During the Conflict Cycle

(Results for equations (13) using variant (B) of dummies during the conflict cycle; for equations (2) and (3), only the coefficients on the difference terms are shown)

article image
Source: WEO.Notes: Levels of significance at the 1, 5, and 10 percent levels are indicated by ***, **, and *, respectively. All estimates done by generalized least squares with cross-section weights (WGLS). Gvt., government; rev, revenue; p.c., per capita; %chg, percentage change. All estimates done by WGLS.

The contraction period is defined in terms of real GDP per capita.

The Impact of Conflict

The period dummy variables give an indication of the direct or underlying effect of the various episodes of the conflict cycle, abstracting from the impact of the policy and other variables. Table 5.9 extracts the regression results for the coefficients on these dummy variables and compares these against actual performance. For equations (2) and (3), the actual performance shown is the difference between growth during the conflict cycle and normal and the difference between the two sets of conflicts, respectively.

Table 5.9.

Impact of Conflict on Growth

(Actual growth and regression coefficients on period dummies only; for equations (2) and (3), only the coefficients on the difference terms are shown)

article image
Source: WEO.Note: Levels of significance at the 1, 5, and 10 percent levels are indicated by ***, **, and *, respectively.

The results support the conclusion that, compared with the earlier conflicts, countries passing through the later conflicts went through a deeper underlying deterioration before the conflict as well as a stronger underlying recovery three to five years after conflict. In addition, once allowance is made for other factors, the results for equation (2) point to a significant negative impact that persisted after the earlier, but not the later, conflicts.

The more interesting results relate to the impact of the start and end of conflict. Looking at the results for equations (1) and (2) using variant (A), the coefficients on the dummies for the conflict and for the one- to five-year postconflict period as a whole suggest that the direct negative impact of conflict on growth as well as the direct positive impact of the end of conflict were more pronounced in the later conflicts. This conclusion, however, is modified by a closer look at the subperiods of the conflict cycle using variant (B). The dummy coefficients for the contractionary period as well as for the first two years after conflict are close in magnitude for the two sets of conflicts (although, for the first two years after conflict, neither are statistically different from zero). Moreover, the results of equation (3) support the conclusion that these two coefficients are not statistically different between the two sets of conflicts.

These results using variant (B) are very much at odds with the different actual growth rates experienced by these two sets of countries. This invites the important and perhaps surprising conclusion that, once “other factors” are taken into account, the onset and ending of conflict had much the same impact on growth in both sets of conflicts. In other words, despite their different economic profiles, much the same conflict process was at work for both groups.

The Impact of Macroeconomic Policy

The conclusion that the direct impact of the onset and termination of conflict on growth was broadly similar in the two sets of conflicts leaves open the question why actual performance over the conflict cycle was so different. The results suggest that differences in macroeconomic policies may have been an important factor.

The impact of policy. Equation (1) shows how policy (and other variables) affected growth. The coefficients on government spending and inflation have the expected signs: higher government spending growth increases output growth, and higher inflation reduces growth. Although at times negative, the effect of the fiscal balance was generally positive when also statistically significant. The effect of changes in the terms of trade was generally positive outside the conflict cycle but negative during the conflict cycle.25 The results provide no support for the presence of convergence effects. The results for equation (2) suggest that over the course of the earlier, but not the later, conflict cycles there was a decline in the magnitude and statistical significance of the impact of the policy variables.26 The results for equation (3) suggest that once conflict started and into the postconflict period, the impact of the three policy variables was significantly larger in the later than in the earlier conflicts.

Contribution to growth. Table 5.10 shows the combined contribution made by the three policy variables to real GDP growth. A full decomposition of growth is shown in Table A5.3.27 The policy stance was generally a detriment to growth under normal circumstances and remained a negative factor through the conflict and after conflict. However, there were important differences in the extent of the negative impact and the improvement after conflict.

Table 5.10.

Policy Contribution to per Capita Growth

(Percentage points, average1)

article image
Source: WEO.Note: DFE, dissolved federal entities.

The sum of the contributions of the three policy variables is given by the generalized least squares with cross-section weights (WGLS) coefficients for equation (1) in Table 5.8 times the value of the variable.

In the pre-1990 conflict countries, the policy stance improved considerably with the onset of conflict—the impact turned mildly positive during the contractionary period—so that it helped moderate the reduction in growth. The policy stance deteriorated during the prolonged in-conflict recovery to become an important drag on growth in the initial two years after conflict. It was only three to five years after conflict that the policy stance was turned around to have a substantial positive impact on growth.

In the post-1990 conflict countries, the policy stance deteriorated substantially once conflict started, accentuating the reduction in growth. Although it continued to diminish growth during the initial two years after conflict, the negative impact was considerably reduced from the level during conflict and this improvement was therefore supportive of the recovery. The pace of improvement was sustained so that by three to five years after conflict, the policy stance had a positive impact on growth that was also larger than in the earlier conflicts. However, policy performance in the post-1990 conflict countries was uneven and particularly poor in the idiosyncratic DFEs. Excluding the DFEs, the postconflict improvement in the policy stance was much stronger, and its contribution to growth turned positive in the first two years after conflict.28

Robustness

Reverse causality. As is often the case in regressions with output growth as the dependent variable, the estimates suffer from the endogeneity of the explanatory variables. This is particularly the case for inflation and government spending growth, but there is little evidence that output growth caused movements in either the growth in the terms of trade or in the change in the government balance as a percentage of revenues. The seriousness of this problem depends on the dominant direction of causality between output growth and the explanatory variables.29 To account for possible reverse causality, equations (1) and (2) are estimated by weighted two-stage least squares (WTSLS) using the lagged growth in government spending and the lagged inflation rate as instruments.30 The results (Table A5.4) are broadly the same as before, so there is little reason to conclude that reverse causality is a significant concern.

Fixed effects. Another potential problem is the presence of unobserved country-specific effects. Depending on whether these effects are correlated with the explanatory variables, they can be estimated using a fixed-effects or a random-effects estimator. Only fixed effects can be addressed here because estimating random effects requires that the number of countries in the panel is larger than the number of coefficients, which is not the case. Because the fixed effects are not time specific and so replace the equation constant, they can only be estimated using equation (2). The results (Table A5.4) suggest little reason to conclude that there are significant unobserved country-specific effects.

Section IV. The Role of External Assistance

The lines of inquiry pursued by the conflict-related literature have been useful in accounting for key features of specific groups of conflicts. For example, the focus on the role of natural resources provides valuable insights into the conflicts in countries such as Sierra Leone and the DRC where illicit diamonds have been an important factor. Similarly, the role of spillover effects is particularly pertinent to the set of regional conflicts in West Africa and the Great Lakes region. However, they do not adequately address the sort of systemic changes in the economic profile of conflicts observed in the 1990s as outlined above.

The timing of the developments described above obviously suggests a linkage to the end of the cold war and to changes in the role played by the international community in low-income countries. However, establishing a statistical linkage is beyond the scope of this chapter, and the argument is therefore only suggestive.

From the early 1990s on, external assistance flows to low-income countries as a whole, and to sub-Saharan African countries in particular, declined significantly (O’Connell and Soludo, 2001). This decline was accompanied by a significant change in the profile of assistance to conflict-affected countries. On balance, the role played by the international community in the 1990s was less encumbered by geopolitical considerations and moved toward being less supportive of conflict in favor of being more supportive of the postconflict recovery effort (Tables 5.11 and A5.5).

Table 5.11.

Net Official Resource Inflows

(Percent of GDP)

article image
Source: WEO.

Once conflict started there were important differences in the level and phasing of assistance, measured by net official resource inflows.31 For the pre-1990 conflict countries, assistance increased in both real terms and relative to GDP (but not relative to population) and also tended to be front loaded, frequently spiking upward just before the conflict. The reverse occurred in the 1990s—and once conflict started the level of assistance declined, it also tended to be back loaded, frequently spiking up toward the end of conflict.

This change in the profile of assistance arguably affected both the duration and economic impact of conflict either directly through the level of material assistance provided or indirectly through the impact on macroeconomic stability.32 By this account, the higher level and front loading of assistance during the earlier conflicts provided governments with material support to conduct hostilities, as well as support for economic activity, and helped support macroeconomic stability (including low inflation). This arguably also mitigated the extent of the economic contraction—but at the expense of prolonging the conflict. Conversely, the reduced assistance during conflict in the 1990s may have accentuated the severity of the economic contraction and, together with the back loading of assistance to the end of conflict, also helped bring about a quicker termination of conflict.33

Once conflict ended, the pattern of assistance again differed between the two sets of conflicts. In the earlier conflicts, resource flows declined compared with their conflict levels whether in real terms or adjusted for population and GDP. Countries emerging from conflict were therefore under pressure to restrain and reduce government spending and to resort to inflationary financing, with negative implications for the recovery. In more recent conflicts, resource flows have tended to increase substantially from conflict levels, again whether in real terms or adjusted for population and GDP. Moreover, at least once adjusted for population or GDP, aid levels to these countries surpassed those to countries coming out of earlier conflicts. The higher assistance to these countries in the 1990s provided support for their stabilization efforts and permitted them to increase government spending while reducing their reliance on inflationary financing with positive implications for growth.

Section V. Performance in EPCA Countries and the DRC

Performance in the EPCA Countries

The arguments of the preceding sections are buttressed by the experience of six post-1990 conflict countries that received external financial assistance, including EPCA from the IMF, soon after the end of their conflicts.34,35,36

The performance of these six EPCA countries during the conflict period had many similarities to that of other post-1990 countries (Table A5.6). Like the other post-1990 conflicts, the EPCA country conflicts were generally short and accompanied by sharp contractions, so they also emerged from conflict with severe economic imbalances and needing macroeconomic stabilization. In the initial two years after conflict, the overall performance of the EPCA countries was generally stronger than that in other countries, although performance was varied and particularly weak in the Republic of Congo. Overall, real GDP per capita growth averaged 4 percent, and the inflation rate declined to single digits in all the EPCA countries (except Tajikistan). Their recovery effort in this initial period was supported by large increases in net resource inflows that exceeded inflows to other countries.

The eight other conflict countries in the 1990s (non-EPCA countries) also generally emerged from conflict with a more urgent need for macroeconomic stabilization, especially with respect to inflation. These countries also received external financial assistance after conflict, but total external assistance was generally less than to the EPCA recipients.37 Despite substantial progress, stabilization was not reached in these countries within two years after conflict. For example, the median inflation remained high at about 50 percent, while average per capita real GDP declined 1 percent.38 It was not until three to five years after conflict that these countries were able to reduce inflation and increase output growth to a level comparable to that of the EPCA countries.

Table 5.12 shows how the postconflict stabilization efforts in these countries affected growth (a full decomposition of growth is provided in Table A5.3).39 In both sets of countries, policy deteriorated during conflict, accentuating the negative impact of conflict. However, after conflict, the EPCA countries’ policy performance improved sufficiently to make a positive 2.5 percentage point contribution to growth. In the non-EPCA countries, policy performance deteriorated further and diminished growth by –5.6 in the first two years after conflict.

Table 5.12.

Contribution of Policy to per Capita Output Growth

(Percentage points, average1)

article image
Source: WEO.Note: …, not available.

The sum of the contributions of the three policy variables given by the generalized least squares with cross-section weights (WGLS) coefficients for equation (1) in Table 5.8 times the value of the variable.

The relatively good performance in the EPCA countries was likely the result of a combination of factors. The circumstances of these six countries permitted the authorities to address their difficulties with greater commitment to sound macroeconomic policies. This in turn provided the basis for the international community to provide financial support to help these countries meet their policy objectives. In this respect, there was an important virtuous cycle in operation: sound policy attracted external assistance soon after conflict, which made these policies easier to implement and more fruitful and therefore also more politically acceptable.40

Performance in the DRC

The arguments of the preceding sections are also buttressed by the experience of the DRC, where stabilization and the start of economic recovery was made more difficult by delays in official external assistance. The DRC emerged from conflict in 1999, but the provision of external assistance to the government, including a comprehensive financial arrangement with the IMF, was delayed until mid-2002 following the conclusion of a peace agreement and until the DRC could clear external arrears.

The DRC emerged from conflict with real GDP per capita reduced to 80 percent of its preconflict level and hyperinflation that was above 500 percent. In 2000 inflation remained high, and the economy continued to contract. In mid-2001, the new Kabila government introduced its enhanced interim program to stabilize the economy. Despite the lack of external financial support, end-year inflation was reduced to about 135 percent in 2001, but output continued to contract. It was only in 2002 that end-year inflation was reduced to the moderate level of 16 percent and that output growth resumed.

Table 5.12 provides estimates of policy’s contribution to per capita growth over the DRC’s conflict cycle, and Table 5.13 provides greater detail for 1999–2002.41 The policy stance (especially inflation) detracted significantly from growth until 2002, but the negative impact was significantly reduced by the adjustment efforts initiated under the enhanced interim program. The improvement in the policy stance added more than 15 percentage points to growth from 2000 to 2002. However, the full impact of these efforts was not felt until 2002 when policy started to make a large positive contribution to growth.

Table 5.13.

Democratic Republic of the Congo: Decomposition of Output Growth

(Percentage point contribution to per capita growth1)

article image
Source: WEO.Notes: p.c., per capita; Gvt., government; rev, revenue, p.c., per capita; %chg, percentage change.

The contribution of each variable given by the generalized least squares with cross-section weights (WGLS) coefficients for equation (1) in Table 5.8 times the value of the variable.

The DRC’s output performance in the first few postconflict years was one of the weakest in the 1990s. Conversely, compared with the conflict period, the DRC government’s commitment to good policies after conflict was rewarded by improvement in performance that was also one of the strongest and also significantly stronger than for the post-1990 conflict or sub-Saharan Africa countries as a whole. Moreover, by the end of the postconflict period, the contribution made by policy in the DRC was significantly larger than in these other groups of countries. The policy stance added 3.5 percentage points to growth in the DRC in its third year after conflict versus not more than 1 percentage point in the post-1990 conflict countries or the sub-Saharan Africa conflict countries (Tables 5.10 and 5.12).

Section VI. Conclusions

The chapter argues that there was a shift in the key economic characteristics of conflict during the 1990s. The underlying conflict process at work has remained much the same, and this shift has probably been reflective of differences in the stance of macroeconomic policy over the conflict cycle that appear to be related to changes in donor practices toward countries affected by conflict since the end of the cold war. These findings would seem to have important implications for policy and aid priorities during and after conflict.

The chapter leaves a number of important questions unanswered. The argument linking aid patterns to these shifts has been only suggestive and needs to be explored more thoroughly. The chapter also suggests an important role for direct budgetary support in the postconflict recovery period, and further work needs to be done in exploring the effectiveness of such aid.

Appendix Tables

Table A5.1.

Conflict Dates and Length

article image
Sources: SIPRI; and IMF staff reports.Note: DFE, dissolved federal entities.
Table A5.2.

Macroeconomic Policy Indicators

(Period average unless otherwise indicated)

article image
article image
Source: WEO.

Data for 3, 4, and 5 years after conflict are available for 11, 7, and 7 of 14 post-1990 conflicts, respectively.

Table A5.3.

Decomposition of Real GDP per Capita Growth

(Contribution to real GDP per capita growth, percentage points, average)1

article image
article image
Source: WEO.Note: …, not available; p.c., per capita; %chg, percentage change.

For each variable, the contribution in each period is equal to the average of the generalized least squares with cross-section weights (WGLS) regression coefficients for equation (1) from Table 5.8 times the value of the policy variable.

The results for all the post-1990 subgroups make use of the same common sample coefficient estimates for the full set of post-1990 conflicts.

The contraction period is defined in terms of real GDP per capita.

Table A5.4.

Real GDP per Capita Growth During the Conflict Cycle

(Equations (1) and (2), adjusting for reverse causality and fixed effects; for equation (2), only the coefficients on the difference terms are shown)

article image
article image
Source: WEO.Notes: Levels of significance at the 1, 5, and 10 percent levels are indicated by ***, **, and *, respectively. p.c., per capita; Gvt., government; rev, revenue; %chg, percentage change.

The contraction period is defined in terms of real GDP per capita.

Estimates done by weighted two-stage least squares (WTSLS) using the lagged growth in government spending and lagged inflation as instruments.

Estimates done by generalized least squares with cross-section weights (WGLS).