Some Economic Consequences of the Transition from Civil War to Peace Jean-Paul Azam David Bevan Paul Collier Stefan Dercon Jan Gzennixg Sanjay Pradhan TheWorld Bank Policy h e a r c h Department ]PublicEconomicsDivision December1994 Summary findings Drawing on evidence from Africa - especially Ethiopia being financeable. The transition to peace is primarily and Uganda -the authors of this volume draw the transition from fear and the defensive responses that conclusions about economic policy in the aftermath of became ingrained in wartime. The peace dividend comes civil war. A sample of conclusions follows. as a gradual recovery of confidence induces repatriation Civil wars differ from international wars. They are of financial and human capital. informal, often have no clear beginning and end, weaken Such confidence can be boosted by the early rather than strengthen the authority of the state, and sequencing of investment-sensitive policy reforms and by leave two unreconciled armies to be demobilized within preserving low inflation through direct consumer price one territory. Civil wars erode the institutions of civil index targeting. Lack of confidence can be compensated society, leading to a decline in the stock of social capital, for by temporary undervaluation of the exchange rate, or which takes some time to restore. Private investment and by temporary tax incentives for investment which, government revenue are slow to recover, and military however, may prove more difficult to make credibly expenditures are not easily reduced. As a result, there is time-bound. Finally, aid can permit accelerated little or no peace dividend in the short run. rehabilitation of the infrastructure (especially transport The period of transition to peace is a particularly networks) needed to return to a market economy. suitable time for radical policy reform, despite the high Contrary to the study's hypothesis, the authors found degree of polarization typical in countries engaged in that demobilization - at least in Uganda- did not lead civil war. And speedy reform, far from increasing to a significant upsurge in insecurity. In the short term, uncertainty, is likely to reduce it. After a civil war, demobilization significantly reduced crime, unless the private agents are fearful both of each other and of the demobilized lacked access to land. If the demobilized government. This, perhaps even more than physical returned to their home areas and were given some damage to infrastructure, hinders private-sector-led assistance, with identifiable exceptions they were able to recovery, as irreversible investment is delayed despite find income-earning opportunities. This paper-aproductof the PublicEconomicsDivision,Policy ResearchDepartment -is partof alargereffort inthedepartment toexaminecomponentsof publicspending (suchas militaryexpenditures) and the interface between the publicand privatesectors. The study was funded by the Bank's ResearchSupport Budget under the research project "Some Economic Consequences of the Transition from Civil War to Peace"(RPO677-31)and by the Africa RegionalOffice. Copies of this paper are availablefree from the WorldBank, 1818 H StreetNW,Washington, DC20433. Pleasecontact CarlinaJones, roornN10-063, extension37699 (139 pages). December 1994. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues.An objective of the series is to get the findingsout quickly, even if the presentations are less than fully polished.The papers caw the names of the authors and should be used and cited accordingly. The findings, interpretations, and conciusions are the authors' own and should not be attributed to the World Bank, its Executive Board of ~irectors,or any of its member countries. Produced by the Policy Research Dissemination Center POLICY RESEARCH WORKING PAPER Some Economic Consequences of the Transition from Civil War to Peace Jean-Paul Azam, David Bevan, Paul Collier, Stefan Dercon, Jan Gunning and Sanjay Pradhan The World Bank Policy Research Department Public Economics Division Contributors Jean-Paul Azarn CERDI, University of Auvergne David Bevan Centre for the Study of African Economies, Oxford University Paul Collier Centre for the Study of African Economies, Oxford University Stefan Dercon Centre for the Study of African Economies, Oxford University Jan WillernGunning CEPR, Free University Amsterdam SanjayPradhan Policy Research Department, World Bank Chapter 5 Demobilization and Insecurity.A Study in the Economics of the Transition from War to Peace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 Paul Collier 1.Introduction............................................................... 39 2. Demobilization and Micro-insecurity ........................................... 39 3. Demobilization and Macro-insecurity .......................................... 44 4. Conclusion ............................................................... 45 Chapter 6 Food Markets. Liberalization and Peace in Ethiopia: an econometric analysis ........... 47 Stefan Dercon 1. Introduction ............................................................... 47 2 The effects of the quota system. trade controls and the war on grain prices ............. 48 . 3. Teff Prices and Margins in Ethiopia ............................................ 50 4. Analyzing market integration and the effects of liberalization and peace .............. 53 5.Results ................................................................... 60 6 Conclusion ............................................................... 64 . Chapter 7 Fiscal Aspects of the Transition from War to Peace: with Illustrations from Uganda and Ethiopia .......................................................... 87 David Bevan and Sanjay Pradhan 1. Introduction ...............................................................87 2. Fiscal Characteristics of Transition ............................................ 87 3. The Revenue Transition ..................................................... 90 4. Expenditure Patterns ........................................................93 5. Decentralization ........................................................... 95 6. Conclusion ...............................................................96 Chapter 8 Economic Aspects of the Ethiopian Transition to Peace ........................... 109 David Bevan 1. Introduction ..............................................................109 2. The Rural Economy ....................................................... 110 3. The Macroeconomy .......................................................113 4. Government Structures ..................................................... 116 5. Conclusion ............................................................. 117 Chapter 9 Economic Aspects of the Ugandan Transition to Peace . . . . . . . . . . . . . . . . . . . . . .119 Paul Collier and Sanjay Pradhan 1. Introduction: The Economic Meaning of 'War' . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 2. Aggregate Consequences of War and Peace: Production. Expenditure and Their Composition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 3. Private Responses to Peace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125 4. Government Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .128 5. Donor Responses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 6. Conclusion: Commitment and Adjustment in a Frightened Society . . . . . . . . . . . . . . . 133 References ....................................................................... 135 List of Tables Table 1.1. The Composition of Ugandan GDP by War-Vulnerability ........................... 5 Table 3.1. A Taxonomy of Policy Uncertainty ............................................ 27 Table 3.2. Effects of a Devaluation on the Incentive to Repatriate ............................. 27 Table 5.1. Regression Results for Crime and Demobilization ................................42 Table 6.1: Margins between Addis Ababa Real Teff Prices and specific markets ................ 52 Table 6.2. Price differential among regional towns in Ethiopia ...............................54 Table 6.3. Cointegrating Relationships .................................................61 Table 6.4. Short-run integration (F(6,58)) ............................................... 63 Table 7.1. Ethiopia: fiscal accounts ..................................................... 99 Table 7.2. Ethiopia: Public Expenditures Budget ......................................... 100 Table 7.3. Ethiopia--Composition of Public Expenditures .................................. 101 Table 7.4. Uganda--Composition of Recurrent Expenditures ............................... 102 Table 7.5. Uganda--Recurrent Expenditures ............................................ 103 Table 7.6. Uganda - Central Government Operations (As a % of GDP) . . . . . . . . . . . . . . . . . . . . . . . 104 Table 7.7. Economic Characteristics of Infrastructure Projects in Uganda .................... 105 Table 7.8. World Bank Projects Uganda, 1984--1989 .....................................106 Table 7.9. World Bank Projects Uganda, 1984- 1989 ..................................... 107 Table 9.1. The Composition of GDP by War-Vulnerability ................................ 121 Table 9.2. Index of Components of GDP by War-Vulnerability ............................. 122 Table 9.3. Relative Price Changes by War-Vulnerability .................................. 124 Chapter 1 Introduction Paul Collier 1.Introduction This study attempts to draw some general long wars, the post-war recoveries have only been economic inferencesabout economic policy in the partial. There appears to be an initial boost to aftermath of civil war. It draws upon evidence output: in Uganda, in the first two post-war years from Africa, particularly Uganda and Ethiopia, but (1987 and 1988) the economy averaged 8.3% the emphasis is analytic rather than empirical. growth, and in Ethiopia in the first post-war year Civil wars differ from international wars: they are (1992) growth was 7%. These are high growth informal,often having no clear beginningand end, rates both historically and by African standards but they weaken ratherthan strengthen the authorityof left both economies well below plausible the state, and they leavetwo irreconciled armies to counterfactuals.In Uganda, even eight years after be demobilized within the one territory. My main peace the economy is still far below its pre-war argument is that civil warserode the institutionsof peak. Civil wars also lead to a loss of exchange civil society, leading to a decline in the stock of efficiencyand rapid recovery is a potentialpart of social capital. One consequence, discussed in the the peace dividend. However, a study of the next section, is that there is little or no instant reintegration of the Ethiopian grain market peace dividend. Section 3 discusses the gradual (Dercon, Chapter 6) finds that the contributionof restoration of security. Sections 4 and 5 consider peace was modest. Neither the private nor the inter-sectoral and inter-temporal resource public sectors behave as though in receipt of a reallocations induced by civil war, and the scope prospective windfall. Collier and Gunning for policy to assist in the reversal of these (Chapters 2 and 3) show that the private sector responses during the post-war recovery.Section 6 remained wary of recovery.It neither invested nor concludes. dissaved as might have been expected had it anticipated recovery with confidence. The fiscal 2. Is there a Peace Dividend? implicationsof peace are also discouraging(Bevan and Pradhan, Chapter 7). Because of their Civil wars vary greatly in terms of economic informality,civil wars generally do not end in the damage. The relatively short civil war in Nigeria same decisive manner as internationalwars and so appears fromthe NationalAccountsto havecaused militaryexpenditure cannotbe veryeasily reduced. only a fairly small output loss which was fully In Uganda it actually rose 40% in the early years recovered within three years of peace. By contrast, of peace. Further, revenue may riot be buoyant. In in Uganda, a conservative counterfactual is that Ethiopia during wartime the state used coercive withoutthe long and intermittentcivil war, by the measures for revenue-raising which had to be arrival of more peaceful conditions GDP would abandoned during peace, causing a revenue have been double its actual level. Similarly declirie. In Uganda, where revenue had collapsed dramatic magnitudes would apply to Mozambique during the war, it was very slow to recover, for and Angola. reasohs discussed below. Finally, during the last However, while the potential for a peace years of a civil war the fiscal position may dividend may thereby seem to be larger in these deteriorateto a position which is unsustainable. In effect. the post-war government inherits an cant and distinctive impact effect in the three imperative need for fiscal retrenchment. In months following the demobilization: soldiers who summary, there should be no expectation of a had reported no access to land were one hundred substantial 'peace dividend' either for the economy times more likely than the Ugandan average to as a whole or for the government. commit crime, while those with access to land significantly reduced crime. The latter effect was 3. Re-establishing Security substantiated by interview evidence from commu- nity leaders who claimed that the presence of men Civil wars generate two types of insecurity. Micro- with military training had discouraged existing insecurity is the fear of personal violence and theft criminals. In Uganda most soldiers had access to which is present in any society but is usually land and so the latter effect outweighed the former: heightened in the context of civil war by the crime fell by 7% folIowing demobilization. Nei- breakdown of policing. Macro-insecurity is the ther of these effects was sustained in the following fear that the state will be used in a partisan way nine months: there appeared to be no relationship against some social groups. casting doubt over all between demobilization and the level of criminal- aspects of economic life which depend upon the ity. This may be because both types of soldiers impartiality of state services, such as the enforce- rapidly civilianised. Alternatively, it might be ment of property rights. Evidently, during a civil because demobilized soldiers relocated from the war the state is partial, and the social groups which home district to which they were discharged, will in future control the state is contingent. Hence. although there is some evidence that this did not both opponents and supporters of the current happen on a significant scale. The Ugandan demo- government face macro-insecurity. Both types of bilization was well-managed and indeed long- insecurity can be expected to have economic delayed after the end of the war, but could have consequences, these being analyzed in Section 5. been improved upon. Given the high propensity of The ending of a civil war does not usually substan- the landless to commit crime and that only 12% of tially resolve these concerns. Micro-insecurity the army described themselves as landless, access persists because the reconstruction of peacetime to land could have been used as a criterion for security services takes time. Indeed, criminality selecting those to be demobilized. may increasedue to demobilization. Macro-insecu- Macro-insecurity is less amenabIe to investiga- rity is diminished, but there is a continuing risk tion. Azam (Chapter 4) uses a simple game-theo- that the war will be resumed and the state over- retic framework in which peace can be maintained thrown. Civil war starts in societies with underly- by a combination of military expenditure and re- ing causes for conflict. and the war itself increases distributive expenditures in favor of the losers. The bitterness. Hence, when a war stops there is a conversion of swords into plough-shares directly significant risk that it will resume, as happened in increases macro-insecurity and needs to be offset Angola and, intermittently, in Uganda. by a visible redistribution of plough-shares to the The study investigated whether demobilization potential enemy. The winner is obviously regarded worsens micro-insecurity, using data on the de- as partisan. Since this perception motivates the rnobilizatio~lof 20,000 Ugandan soldiers in late potential threat to the government, public gestures 1992 (Collier. Chapter 5). The approach was to of redistributiveexpenditure are needed to counter determine whether changes in crime in the 38 it. On this analysis, the Angolan government made districts of Uganda could be related to the district- the mistake of reducing military expenditures level incidence of demobilization. The composi- without pre-committing to high redistributive tion of the army is also somewhat distinctive as a expenditures. Demobilization can increase macro- result of a civil war. In Uganda much of the army insecurity not only by reducing the capacity of the had been recmited in the bush as a guerilla force government to defend itself, but by furnishing while very young and so had few non-military opponents or others with potential recruits. The skills and little education. Almost all demobilized demobilization of the Ethiopian army may have soldiers would need to earn their livelihoods in contributed in this direct way to the insurrection in peasant agriculture. The results showed a signifi- Oromo,and the Ugandan demobilization may have inputs. Marketed agricultural produce has this contributed to the civil war in Rwanda. characteristic. The most vulnerable activities are The implication is that military expenditures those which depend upon the market for both can be reduced without increasing either micro- inputs and outputs. Formal sector manufacturingis insecurityor macro-insecurity,but that the govern- the main instance of such an activity. ment may need to have offsetting increases in That war jeopardizes assets follows as a spe- expenditure in the form of compensation to losers. cial, but important,case of the more general decay of social capital. Assets depend upon enforceable 4. Civil War and Inter-Sectoral Substitution rights. For example, in Uganda, when the Asians lost property the new owners held theirclaims in a The breakdown of security is not just an aggregate most insecureform. Not only might theyanticipate shock to the economy. Rather it induces both inter- that at some stagethe property might be restored to temporal and inter-sectoral resource reallocations its Asian owners, but of more immediate concern, which the return to peace gradually reverses. I now they lacked clear title and so could neither sell not consider the sectoral implications of social disor- borrow against the asset, and risked having it der. The argument is that social disruption jeopar- reassigned to some other 'owner' through the dizes transactions and assets and that this has same arbitrary process by which they themselves differential consequencesas between sectors of the had acquired it. The illegitimate possessionof an economy. asset creates powerful incentives to strip it. Since An environment in which transactions can be possession is likely only to be temporary it is conducted cheaply requires considerable social safest to transform it into an invisible form: invisi- capital. The range of social capital required is bility is the best substitute for legitimacy other rather wide. In the early 20th century informal than the possession of overwhelming force. This agricultural marketing was greatly eased by the transformation is worthwhile even if in the process introductionofstandardizedweightsand measures, a substantial part of the value of the asset is forfeit. which requires a continuous process of public Invisible assets take various forms. The least checks (Ensminger (1992)). The provision of a visible is an asset held abroad. Cash may also be legal system enables the parties to a contract to easy to conceal. Finally, crops which can be kept enforce it relatively cheaply. The provision of a underground (tubers) are safer than those which communications system reduces the costs of must be stored above ground (grains). In addition information, and a transport system reduces the to invisibility, immobility is desirable in an asset. costs of movement. Many aspects of a transaction Immobility will not constitute a defense against cannot be specified in a contract because not all large scale predationsuch as the arbitrary power of contingenciescan be anticipated. An environment a ruler, but it is a defenseagainstthe micro-insecu- of cooperation, achieved either by the internaliza- rity of casual theft. An immobile asset is only tion of activities within an organization (the inte- forfeit if anotheragent is able to secure the space grated fim) or a high expectation of repeat trans- on a long-term basis. Obvious immobileassets are actions. is necessary to reduce the incidence of land and buildings. There are no assets which are opportunistic behavior. both immobile and invisible, but there are some Activities vary considerably in their transac- which are both mobileand visible,such as vehicles tions intensity. The least transactions-intensive and consumer durables. activity is subsistence production. However, it Just as activities vary in their transactions should be noted that although subsistence produc- intensity, so they vary in their intensity in visible tion does not, by definition, involve the marketing and mobile assets. Manufacturing is again highly of output, it will still usually involve the purchase vulnerable since both its inputs and its outputs of some inputs.Hence, it is not immunefrom a rise must be stored at the site of production and are by in the costof transactions. An intermediatestage in their nature mobile. Subsistence is in one respect the hierarchyof transactions intensity are activities more asset-vulnerablethan production for market which depend upon the market for the sale of in that the latter can be sold as it is harvested, output, but are not very dependent upon it for whereas the former cannot usually be eaten as it is harvested (exceptfor some tubers)and so must be forces took Kampala. although low-intensity stored on site. Livestock is a disastrously asset- fighting continued until 1990. Hence, unlike an vulnerableactivity in that it iscapital-intensiveand international war, a civil war tends to have some- the entire capital is fairly visible and highly mo- what imprecise dates. On economic criteria it bile. Theservicesector is relativelyinvulnerablein seems reasonableto date the commencementfrom that its inputs are largely people and buildings. 1972 and the ending around 1986. So far I have considered how sectors differ A measure of the relative performance of a according to whether they use visible or mobile sector is how the quantity of its value-added assets, and whether their production process is changed relative to GDP (see Table 1.1). The intensive in transactions.A particularlyvulnerable manufacturing sector, which I have suggested is sector on these criteria suffersa large cost-shock. both transactions intensive and asset-vulnerable, However, additionally,somesectors produceeither halved relative to GDP (which itself halved rela- assets or transactions. These sectors suffer a tive toa reasonablecounterfactual).Manufacturing demand collapse because war operates like a tax recovered post-1986, although even by 1993194it on their output. In African civil war contexts the was still far below its 1971 share of GDP. The asset-producing sectors are usually construction sector which contracted most severely during the and livestock. Transactions production covers war was the asset-providing sector, construction. transport, trade, and the financial services sector. This was the most successful sector post-1986. To give an illustrative parody of the above analy- More limited evidence on livestock suggeststhat it sis, the worst possible activity in which to be followeda qualitativelysimilar pattern toconstruc- engaged might be the manufactureof cash-regis- tion. From 1982-86 there was a rapid absolute ters. The production process is vulnerable because decline in the sector, which was reversed post- manufacturing uses mobile assets and has a high 1986. However, since the livestock is concentrated ratio of transactions to value-added.The demand in the North, which was the last part of the country for output collapses because it is a mobile capital to be pacified, the recovery would be expected to good and one used only for processing transac- be limited. The other sector which suffered se- tions. At the other end of the spectrum would be verelyduringthe war period was the transactions- growing cassava for own-consumption. providing sector, commerce and transport. Again To summarize, I have distinguishedactivities this grew more rapidly than GDP post-war, but according to whether they are transactions-inten- remained much smaller relative to GDP than its sive, vulnerable asset-intensive, transaction-pro- pre-war level.Thesector which did relatively well viding, and asset-providing. I now show how this out of the war wassubsistenceagriculture. Subsis- is consistent with the remarkable changes in the tence grew relative to GDP during the war and composition of Ugandan GDP both during and contracted relatively thereafter. However, it is after the period of war. noteworthythat even subsistenceoutput grew only In 1972 Amin declared "economic war" slowly between 1971 and 1986: its performanceis against the Asian community. Although violence only good in relative terms. In effect, the substitu- did not erupt until later, from thisdate the business tion of resources into subsistenceas they became community could no longer regard the state as a less productive in other sectors, was offset by the neutral provider of services and the economic negativeeffectsof the war through the disruption manifestationsassociated with civil war appeared: of inputsand assetssuch as stored grain. Post-war, an exodus of human and physical capital, falling subsistence activities have grown least (25%), aggregateoutput, and a shift in the composition of though only around half of the retreat in subsis- output. The subsequent periodic warfare was tence has been reversed. substantiallyresolved in 1986 when the NRA Table 1.1. The Composition of Ugandan GDP by War-Vulnerability (% share of GDP at 1991 constant prices) 1971 1986 1993194 Transaction and asset intensity: high: manufacturing 8.8 4.4 6.0 medium: marketed agriculture 22.6 24.5 22.7 low: subsistence 20.5 36.0 32.1 Transaction-providing: transportand commerce 21.2 16.1 17.2 Asset-providing:construction 12.5 3.5 5.5 Unassigned activities 14.4 15.5 16.5 Note: The National Accounts provide data at 1966 prices for 1963-85 and at 1991 prices from 1982-93194. 1982 was selected as the year to be used for conversion from 1966 to 1991 prices. Since output changes 1971-92 are only measured at 1966 relative prices, the conversion of 1971 output to 1991 prices is only approximate. Sector i in 1971 at 1991 prices is approximated =:[(sector i in 1971 at 1966 prices)l(sectori in 1982 at 1966 prices):l.[sector i in 1982 at 1991 prices]. This has the advantage that since 1982-91 GDP was calculated on a consistent set of definitions of sectors, changes in definitions between the 1966 series and the 1991 series only lead to a mis-estimate to the extent that the alter the growth rate of the sector between 1971-82. Total GDP in 1971 at 1991 prices was then calculated as the sum of the sectoral outputs so revaiued. Note that this will differ from a direct adjustment of total GDP in 1971 by the factor [(total GDP in 1982 at 1991 prices)l(total GDP in 1982 at 1966 prices)]. Sector shares in 1971 at 1991 prices are then sector outputfGDP. The post-warrecoverythus involvesa return to incentive problems which this would normally the market and a restoration of confidence in induce. Implicit taxation of transactions occurs assets. The latter is taken up in the next section, through the inflation tax since domestic currency here the policy implications of the former are is held for market-derived income but not for considered. subsistence income. The government should The return to the market is important both therefore set the inflation rate below the revenue- because market-based activities are normally more maximizing rate. The revenue-maximizing infla- productiveand because they provide opportunities tion rate for Uganda is not known, but that for for risk-reduction. The latter gives rise to an Kenya is in the range 10-15% (Adam (1992)) so externality arising from the substitutionof activi- that the appropriate policy response might be to ties from subsistence back into the market: as aim for inflation somewhat below this. markets become denserthey becomemore reliable, A second policy implication is that the return so that society has an interest in subsidizing the on the rehabilitation of transport infrastructure is return to the market. likely to be high. Collier and Pradhan (Chapter 9) One way in which this can be done is by the reviewall the road transport projects undertaken in government setting both explicit and implicit Uganda since 1986 and find an average rate of taxation of transactions low in the early post-war return of 39%, clearlya very high rate of return by years. Explicit taxation ratesarecommonly high in the standards of public expenditure. This is sup- war economies as the government searches for ported by a rural household survey conducted in revenue and uses coercion to offset the dis- 1990 which asked which government actions had been most appreciated since 1986 (Bigsten and The economics of the transition to peace can Kayizzi-Mugerwa).Road improvements were the therefore be thought of in part as the process of most popular choice. shifting private portfolios from foreign liquid assets to domestic liquid assets. An equivalent 5. Civil War and Inter-Temporal Substitu- gradual repatriation takes place with respect to tion human capital,since civil wars generate an exodus of the skilled. If these repatriations can be induced During civil war private profits can be high. As they deliver a genuine 'peace dividend' of poten- markets become disrupted they become lesscom- tiallysubstantial proportions.For example, private petitiveand so marketing margins widen. Taken to transfersto Uganda are estimated to have grownat the limit, disruptioneliminates trade and so elimi- 35% per annum (in current dollars) in the past four nates profits, but at lesssevere levels profits might years(1990191-93194).By 1992193 they exceeded be sustained or even enhanced as reduced volumes exports. Net emigration of Ugandan citizens fell of trade are offset by wider margins. Collier and from massive proportionsduring the civil war era Gunning (Chapter 2) cite evidence from Mozarn- to a few thousand a year in the late 1980s and a bique and Somalia, and Keen (1994) analysesthe few hundred a year in the early 1990s, with net Sudan. in each case wartime conditions being immigration of non-residentcitizens. Here I focus associated with high profits. There are also oppor- upon financial repatriation, although some of the tunities to acquire assets by means that would argument may extend to return migration of the normally be illegitimatebut which becomefeasible skilled. once the state has abandoned its role of impartial The uncertainties which gave rise to liquidity defender of property rights. Those who make are not automatically removed by the end of civil substantial trading profits or acquire assets in war. A central propositionof the paper is that both questionableways havean incentive to havea high microand macro-insecuritycontinue at high levels savings rate. The circumstanceswhich have gener- in the post-war environment,so that the realization ated these opportunitiesare unlikely to persist, and of a peace dividend is contingent upon the recov- luxuryconsumptionis generallydifficultand even ery of privatesector confidence.The policy prob- dangerous in civil war conditions. Hence, those lem is therefore to allow for the initial lack of whodo well out of civil war face portfoliochoices. confidence and to assist in its restoration. Collier Civil war is not an environment conducive to and Gunning(Chapters 2 and 3) explorethe impli- investment (see Sen (1991)): there is too much cations of a highly liquid private sector which is uncertainty and, as discussed below, physical too fearful to make irreversibleinvestments.They assetsare vulnerable. The corollary is that savings advance four policy prescriptions, two of which are held in liquid form since during periodsof high have as the objective the recovery of private uncertainty liquid assets cany a premium in the investment and two of which are concerned with form of an option value. Normally, during civil government macroeconomicmanagement. war domesticfinancial assets are subject to a high They argue that the recovery of private invest- and variable inflation tax so that the only liquid ment can be encouraged by a combination of financial assets available are foreign. The capital reassurance and subsidy. A post-civil war govern- flightcommon during civil wars might thereforein ment lacks reputation and so its promises lack part bea reflectionof increased demand for liquid- value. Yet typically, during wartime the economy ity. An indication of this is that in Ethiopia, where will have become heavily regulated and so one of unusuallycivil war did not jeopardizeprice stabil- the post-war tasks is policy liberalization. The ity, there was a substantial increase in domestic most reassuringaction that such a governmentcan liquidity. During the intensive phase of the war, do with respect to economic reform is to imple- 1986-91,the ratioof currencyto GDP increased by ment it. Since not all reformscan be implemented 93%. Hence. by the onset of peace. a significant at once due to bottlenecks in the legislative and group of private agents are highly liquid, usually policy formulation processes, prospectivereforms holding foreign assets. should be sequenced according to their impact upon investment.The most urgent reforms concern those which impact directly upon the capital values whereas in Uganda a recoveryof confidence would of prospective investment. For example, civil wars include a switch into domestic currency and so be leave a legacy of contested property rights, and it counter-inflationary. Since changes in confidence is importantthat these should be resolved rapidly. are unpredictable, the high liquidity makes the Even with efficient sequencing of reforms, demand for domestic currency unusually volatile there is a high irreducible element of investor risk. so that monetary targeting would produce price Since private investment is likely to have substan- volatility. In current formal models of the costs of tial externalities whereas much of the risk which inflation, inflation shocks have no cost and so the private investors consider should not be taken into inflation tax becomes the ideal buffer policy. account by society, there is a case for the govern- However, the transfers to which such shocks give ment to subsidize the act of commitment. One rise can be disruptive, provoking bankruptcy, and policy instrument which can be used for subsidiz- will accentuate the already severe problem of low ing investment is the exchange rate. Typically in confidence. If these costs are sufficiently severe Africa the exchange rate has been overvalued, and then the stability of the price level becomes a this has subsidized the purchase of imported legitimate objective. Since monetary targeting will capital goods purchased with domestic currency. not achieve this objective, the CPI must itself be By contrast, in post-war conditions liquid assets directly targeted, with some combination of the are in foreign currency, while the least reversible, fiscal deficit and the reserves as control variables. and hence most risky, investments are not im- The liquidity of the private sector and its lack ported capital goods (such as transport equipment) of confidence in the government also has implica- but non-tradable capital goods such as buildings. tions for the government's scope for domestic Such capital goods are location-specific and funding of its deficit. Government liabilities, such depreciate slowly, and so are more sensitive to as bonds and treasury bills, have the wrong risk uncertainty.An under-valued exchange rate subsi- properties, whereas the large stock of fairly safe dizes the transfer of foreign currency assets into real assets such as residential property held by the non-tradableassets. Combining the two points, for governments in Ethiopia, Uganda and Angola the same reason that the exchange rate can subsi- might be purchased by private agents without dize the repatriation for investment decision, it being so heavily discounted. In Uganda the gov- directly affects the asset value of investment, and ernment borrowed through three month Treasury so it is an urgent reform. Bills at a real interest rate in excess of 20%. In Undervaluation is not the only policy instru- 1992 it attempted to sell longer, nine month matu- ment which can encourage repatriation into irre- rities but could find no purchasers. In 1993 it was versible investment, but it has the advantage of able to find takers, but only at real interest rates of being unsustainable. A manifestly temporary around 40% which it was not prepared to pay. subsidy is more effective in bringing forward the Clearly, the act of borrowing at such high interest act of commitment. Tax benefits may be harder to rates would itself have signalled that the govern- make credibly temporary and are administratively ment had little confidence in its continued exis- more cumbersome, but have the advantage that tence. In 1994, shortly after the elections for the they can be confined to productive investment Constituent Assembly, nhich were successful for whereas undervaluation also subsidizes the pur- the government in both their conduct and their chase of residential property. result, real interest rates fell swiftly to single The high level of private liquidity also has figures and longer term borrowing became a implicationsfor macroeconomic management: the possibility. That the government faces a steep yield demand for domestic currency might alter substan- curve may be a common feature of the post-civil tially, and the government has some scope for war legacy, and the gradual flattening of the curve selling assets. In Ethiopia part of the excess liquid- is a way of conceptualizing the return to peacetime ity is in domestic currency, whereas in Uganda it levels of confidence. has been in foreign currency. Thus, in Ethiopia a Finally, in a post-civil war environment the recoveryof confidence would induce a switch out private sector is not only suspicious of the state, it of money into real assets. and so be inflationary, is afraid of it. The power of the state has been used in a partisan manner. This provides an argument 6. Conclusion for restoring the tax base only gradually. Due to the decay of the institutions and conventions of To summarize, during a prolonged civil war the civil society the state has lost most of the mecha- institutions and trust through which civil society nisms needed for compliance with tax-gathering assists the economy are eroded. The economy systems. The private sector has learned how to loses some of its 'social capita!'. Like other types evadethe state, by corruption and a failureto keep of capital, this takes time to restore: in the after- records, and these practices can only be changed math of a civil war private agents are fearful both slowly. For example, in the absence of audited of each other and of the government. This, perhaps business records, the tax authorities essentially even more that physical damage to infiastmcture, bargain with enterprises over lump sums rather is the obstacle to a private-led recovery as irrevers- than tax in proportion to activity. In this situation ible investment is delayed despite being finan- a rapid attempt to increase revenue requires the cable. The peace dividend comes not from a swift state to intensify just the arbitrary actions which resumption of activities directly disrupted by the have defined the decay of civil institutions and war, such an affect being modest, but rather from procedures. In Uganda government expenditure a gradual recovery of confidence which induces has recovered its pre-warshareof GDP long before repatriation of financial and human capital. Public- other components of GDP have recovered. It is led recoveryfinancedthroughaggressive increases arguablethat at the margin, the discouragementto in taxation risksconfirming the fears of the private the return to the market-sector involved in fairly sector. The policy environment can, however, be arbitrary taxation slows recovery by more that the made conducive to private recovery. Confidence extra government expenditure induces it. In rural can be assisted through the early sequencing of Uganda thesamesurveythat found road infrastruc- investment-sensitivereforms and by the preserva- ture had been most appreciated found a high h r e tion of low inflation using direct CPI targeting. priority was the reform of the authority of local The lack of confidence can be compensated by government, since the institutions did not have temporary undervaluation of the exchange rate or clearly defined and well-understood powers. In temporary tax incentives for investment. Finally, Ethiopia, one of the first post-war actions was the aid can permit eccelerated rehabilitation of infra- grass-roots dismantling of the institutions of structure needed for the return to a market econ- government rural authority. A period of low gov- omy, most notablythe transport network. ernment revenue might thus be an investment in the revival of private sector confidence. Chapter 2 War, Peace and Private Portfolios1 Paul Collier and Jan WillemGunning Abstract During civil wars trading is profitable as marketsfvagment. Pro$ts may be saved in liquid form, because investment is too risky. In a successful economic transition to peace these liquid assets are switched into investment. However, continuing fears of insecurity may keep portfolios liquid. We consider three policy consequences. The unpredictable return of confidence causes erratic changes in the demand for money, complicating monetav targeting. Government liabilities become unmarketable except at prohibitive interest rates, though the state may be able to sell real assets. The government can subsidize the act ojinvestment commitment by temporarily undervaluing the exchange rate. 1. Introduction In parts of Africa prolonged civil wars have now will be a resumption of warfare. The conjunction ended (e.g. in Ethiopia and in Uganda) and there of uncertainty and liquidity can arise in other are prospects for peace in several of the current circumstances, but it is particularly poignant in the African conflicts. The analysis of the economic aftermath of civil war and has important policy consequences of peace tends to focus on three implications. Since private agents are atypically issues: demobilization, the rebuilding of war- liquid at the start of the transition, governments damaged infrastructure and the reallocation of can benefit from private portfolio adjustment. At resources previously allocated to war, the "peace the same time the scope for government policy is dividend". In this paper we focus on a different limited as uncertainty makes private agents issue which has received relatively little attention. unwilling to commit themselves to irreversible This is how the end of war affects the portfolio investments and unwilling to hold government choices of private agents and the implications of liabilities except at high real interest rates. those choices for government policy. In this paper we argue that in these Peace finds private agents with portfolios circumstances three policies are appropriate: a which reflect war circumstances and are therefore monetary policy targeted on the price level rather no longer appropriate. In particular, the legacy of than on any monetary aggregate; the sale of war is likely to be reflected in private agents government assets such as housing and land as a holdingan unusually large part of their portfolio as substitute for government borrowing; and a liquid assets (either domestic or foreign). In subsidy on fixed investment through a policy of addition the war-peace transition is typically undervaluing the exchange rate. characterized by considerable uncertainty, both as The structure of the paper is as follows. In the to the extent and the timing of the lifting of war- next section we consider the consequences of a time economic controls and as to whether there war-peace transition for private portfolio choices. We then consider three aspects of economic that rather than collapsing, as implied in the use of policy: in section 3 the implications for short run the term 'crisis', the economy of Southern Somalia monetary management, in section 4 the medium has gone underground. and many herders and run policy towardsthe government's own portfolio, traders have benefited from the growth in and in section 5 the policy towards the promotion unofficial exports' (p. 1 19). of private fixed investment. The agents who benefit from high income opportunities during wartime are likely to have 2. Private Portfolio Responses to War unusually high savings rates. First, they are - unlikely to be established wealthy persons: the African civil wars, especially in the context of an traditional occupations of economic elites are, attempt by the government to maintain an likely to be adversely affected during African civil economic control regime, create remarkable wars. In Ethiopia, Uganda and Angola ethnic opportunities for profit. The environment of war minorities were disproportionately represented in and illegality causes the disintegration of the established economic elite and in each case competitive markets and so creates haphazard but there was a voluntary exodus'. Hence the agents substantial rents for a whole new class of agent. who benefitfrom war circumstances will wish, like the illegal trader. The circumstances which in any other nouveau riches, to accumulate wealth. Britain had by 1920 produced the 'men who had Secondly, they have an additional incentive in that done well out of the war,' and which in America the circumstances which provide their own during the 1930s produced bootleg millionaires opportunity to acquire rents are unlikely to persist. likethe Kennedys, are combined. By their nature, Hence, during wartime illegal traders have a these opportunities for high income are not easy to powerful incentive to build up assets. observe. Nor is profit monotonically increasing in If traders have high savings rates they must market disruption: profits are zero both when the make portfolio choices. Investment in fixed real market is 'perfect' and when transactions costs are assets is likely to be unattractive. First, as long as so high that trade is eliminated. It is the the duration of war is uncertain there is a premium intermediate range in which markets are on liquidity: activities which are profitable during sufficiently fragmented by civil war to widen war may cease to be profitable during peace. When trading margins, but in which trade volumes are investment is irreversible this creates option value not massivelycurtailed, which is conducive to high (Dixit (1989)): rather than investing now the profits. We give three examples from African civil entrepreneur remains liquid until uncertainty is war environments. Azam (1993) shows that in resolved. Secondly, real assets are vulnerable in Ethiopia the onset of peace reduced geographic wartime. They may be damaged but in addition price dispersion in the grain market, which would war makes property rights insecure: thefts are imply reduced profit margins for trading. Little more common and the legal system and other (1992) shows that in Mozambique between 1988 supporting conventions of ownership may break and 1991 peri-urban market trading became less down. For example, in a household survey of post- profitable as it became less subject to predation: war Uganda which included long recall questions 'Most traders appreciate the ability to transact trade on assets, Bigsten and Kayizzi-Mugwenva (1992) without fear of government reprisal or sanctions'. found that peace brought a particularly marked However, 'trade was more profitable in increase in land transactions. They argue that this 1988...when there was not so much competition' reflected the pent-up need for asset transactions (p. 7). Finally, a study of the Somali cattle trade which were infeasible during the period of social (Little and Coloane (1993)) concluded that 'the breakdown. Somali materials show that, even under the so- Hence, because war leads to a premium on called 'crisis' conditions, certain traders and liquidity and to asset vulnerability, illegal traders producers do quite well' (p. 118). 'Because of wanting to accumulate assets are likely to prefer internal warfare and limited agricultural resources, financia! assets. They must choose between few countries better typify the African 'crisis' than domestic and foreign financial assets and between does Somalia. The analysis presented here shows currency and interest-bearing instruments. The safest and highest return financial asset is probably phase of the war, 1986187-1991/92, the ratio of a foreign interest-bearing claim such as a bond or currency outside banks to Gross Domestic bank deposit. However, holding such an asset Expenditure increased by 93% (Collier and involves the inconvenience of access, both for Gunning (1992)). deposits and withdrawals since all transactions By contrast, in Uganda during the highly must be conducted externally. By contrast, disturbed times from the mid-1970s until 1986, transactions can be conducted locally for the other there was a decline in domestic currency relative to three financial assets: foreign currency, domestic expenditure. Given the history of high and variable currency and domestic interest-bearing assets inflation and outright currency default this is not (such as bank deposits or bonds). Even foreign surprising. Our hypothesis is that Ugandan trading currency can be bought or sold in exchange for profits were disproportionately held in foreign domestic currency on local illegal foreign currencies. While this is intrinsically unobservable, exchange markets. This difference in transactions during 1992 there was an unofficial private capital costs reduces the liquidity of foreign interest- inflow of around $200m, some of which is likely bearing assets relative to the other financial assets. to have reflected previous capital outflows. Foreign currency may offer a higher return than To summarize, at the onset of peace, a group domestic financial assets and yet still not dominate of private agents, those engaged in illegal trading the latter in portfolios. This is analyzed by Dowd in wartime, is likely to be holding substantial and Greenaway (1993) who show that despite poor liquidity in a range of foreign and domestic returns the transactions demand for domestic financial assets. We now turn to how portfolio currency may persist because of the high costs of choices change in the early stages of the return to coordinated switching. Hence, portfolios are likely peace. Civil war economies will have been starved to include domestic currency as well as foreign of private investment even if there are high return assets unless the penalty from holding domestic opportunities as agents preserve their options by financial assets is very high. In the context of holding liquidity. Although it is not possible to get African civil wars, domestic bank deposits were direct evidence on the rate of return on private largely a claim on the government and so liable to investment in the aftermath of civil war, indirect default, while being highly visible to the evidence is provided by a study of public authorities. The offsetting advantage, that they investment in Uganda (Collier and Pradhan, paid interest, was modest since interest rates were Chapter 9). During the early post-war years, the set low. Hence, they offered little advantage and evaluation reports on the major category of public particular costs compared to domestic currency. infrastructure, roads, showed an average real rate The use of the inflation tax in wartorn Africa has of return of 39%. Clearly, this is an exceptionally been highly variable. Uganda and Angola had high rate of return for public investment and is at bouts of hyper-inflation. By contrast, Ethiopia and least suggestive that the return to private Chad have had fairly stable price levels. In investment was also high. Thus, a distinctive Ethiopia, over the entire period 1960-91, the feature of Africa's post-civil war economies may depreciation of the parallel exchange rate against be that important groups of entrepreneurial private the dollar (which is the pertinent rate for the choice agents face high return opportunities without being between the two currencies) was only 4% per constrained by lack of finance. We now consider annum. Since the acquisition of domestic financial whether they will use these opportunities. assets is in aggregate observable3, unlike that of The ending of a civil war does not necessarily foreign assets, this affords a testable proposition: reestablish economic security. We distinguish if the conjunction of civil war and a government three types of insecurity which, though not unique economiccontrol regime leads to the accumulation to post-civil-war economies are present in a of significant financial assets, then in Ethiopia it heightened form. First, micro-level insecurity will should be visible as a continuous build-up of be a concern because civil warfare leaves a legacy domestic currency relative to expenditure. of an armed population desensitized to violence. Throughout the Ethiopian civil war this has been Somalia, Sudan, Uganda, Ethiopia, Rwanda, the case. Most spectacularly, during the intensive Burundi, Angola and Liberia all now have heavily armed civilian populations.A high level of micro- fluctuated in the range +230% to -I%, and in insecurity discourages the acquisition of visible Ethiopiaafter years of low and stable inflation the assets. For example, interview evidence suggests rate jumped to 40% and then fell to zero. Such that in Ethiopia lack of personal security was the variable inflation discourages investment. main obstacle to investment in commercial In a war-peace transition all three forms of agriculture during the immediate post-war period uncertaintyare likely to betemporarilyhigh so that 1992193.4 there is an incentive to avoid irreversible The second risk heightened in the aftermath of investment: the premium on liquidity which civil war is macro-insecurity.Whether the conflict existed during war is maintained during peace. has been resolved by military victory (Uganda, Hence, despite the onset of peace, potential Ethiopia) or a brokered peace agreement (Angola, investors may continue to hold their assets in Mozambique, Rwanda and Burundi) there are liquid forms. Further, to the extent that considerable risks that the government will not consumption has been repressed during the war, survive. This risk may be exacerbated by there may be a switch out of liquidity into demobilization since demobilized soldiers consumption rather than investment. In Ethiopia, constitute a potential recruiting ground for wheresome liquid assets are visible becausethere subsequent rebellions. There is now clear is greaterconfidence in domestic financial assets, econometric evidence that political instability it is possibleto observe the behavior of liquidity in directly discourages private investment (Alesina the post-war environment. The banking system and Perotti (1993),Alesina et al. (1992))and such remained highly liquid, with banks reporting a a link is scarcely surprising. Interview evidence shortage of investment projectsand implying that from commercial banks and private potential any consumption spree was modest (Collier and investors in Ethiopia during 1992 suggested that Gunning (1992)). the primary explanationfor the lack of investment An alternative to remaining liquid is to make was concern about the forthcoming Eritrean investments which are relatively safe. The referendum, since this might conceivably lead to riskinessof a capital good is related to itsdegree of circumstancesin which warfarewould be resumed. sector-specificity, duration, and country- In Uganda, where the government has made specificity. Country-specific capital goods are sustained efforts to reassurethe Asian community, clearly most prone to macro-insecuritysince they interviews conducted in 1993 revealed irreducible cannot be removed from the area of jurisdiction. fears that the complexion of the government might The least desirabletype of investment is therefore change after elections. one which cannot be removed from the country, A third type of risk arises because the cannot be redeployed between sectors, and transition to peace creates large fiscal shocks. depreciates only slowly. For example, an Peace reduces some military expenditure (the investment in worker housing for a tea plantation publicsector "peacedividend"), but increasesother would have all these characteristics, whereas expenditures. It may permit the government to investment in a vehicle would have none of them. increase some components of revenue collection We would thereforepredict that, to the extent that such as import duty (Uganda), but wartime asset holders venture out of liquidity, they will methods of expropriation and coercion become favor investmentssuch as vehiclesover thosesuch unacceptable, reducing other components as tea estates, even if (abstracting from risk) the (agricultural taxation in Ethiopia). Each of these return on the latter is considerably higher. More effects is large and unpredictable, the net effect generally, investment in agriculturetends to be less being that the fiscal deficit is subject to extreme easily reversible than in other sectors. fluctuation (Bevan (1993, 1993a)). Since the This is indeed reflected in the pattern of post- government has little or no recourse to domestic war investmentin Ugandaand Ethiopia.In Uganda debt instruments (as private agents fear that the only3% of private investment is being directedto government may default), these shocks must be agric~lture.~Analysis of private investment in monetized, resultingin large variationsin inflation. Ethiopia reveals a similarly extreme skew away In Uganda, the annualized inflation rate has from the agricultural sector6. Yet in both economies, not only is agriculture by far the retreated into subsistence so that peace induces largest sector, but it is the dominantexport activity remonetization. In Ethiopia, more systematic and the sector favored by policy changes. The coercion of the peasantry gave rise to involuntary investment that does take place is largely in peasant integration into the market, so that peace transport equipment and machinery for light may even increase the size of the subsistence manufacturing, both of which are highly mobile. sector. Thirdly, the volatility of inflation will lead Currently even Mozambique is able to attract to switches into and out of domestic currency. In textile investment into export processingzones, in Angola, the stabilization of the price level induced spite of the high degree of country-specific a portfolioswitch out of durable goods into money uncertainty.' This reflects the relative ease with (Azam et al.(1994)). In Uganda, during the yearof which such investment can be reversed: investors a falling price level (1992193) the demand for real are able to take advantage of specialtax incentives money balances rose by 20%. without having to take a view of the long term. One implication of this volatility in money In summary, private portfoliosat the onset of demand and supply is that monetary targeting is peace are highly liquid, agents have an incentiveto impractical. If donor conditions include monetary maintain this liquidity, and to the extent that they targets (as they do in Uganda) they are liable to be do switch from liquid assets to fixed investment breached unless the government matches its they have an incentive to avoid the type of expenditure to its revenue on a high-frequency investment which is most needed. We now turn to basis. More importantly, monetary targeting is three public policy problems generated by this under these circumstances an inappropriate private behavior. instrument for avoiding monetary shocks, since shocksoriginating on thedemand side now need to 3. Implications For Short Term Monetary be neutralized by changes in supply. Money Policy demand shocks may potentially affect either the price level or real output since money demand is As discussed above, during the war-peace not stable. For example, when money demand falls transition the fiscaldeficit will be volatile. Further, (as would happenwhen investorconfidencereturns the inflationary consequences of a given deficit in Ethiopia) then monetary targeting would be become unpredictable because the demand for counterproductive: keeping the money supply money is also liable to be subject to unpredictable constant when demand falls would be inflationary. shifts. First, to the extent that investorconfidence By contrast, in Uganda the recent rise in real returns, there will be a portfolio switch out of moneydemand permitteda non-inflationary breach liquid assets into real assets. This will have of a monetary target. Had money supply not different implications for the demand for domestic accommodated this demand then either prices money depending upon which liquid assets are would have fallen or output contracted. being reduced. In Uganda and Angola, since One alternative to monetary targeting is fiscal holdings of domestic money were so depleted, a targeting, whether of expenditure or of the budget switch into real assets must be financed from deficit. However,the government must then accept foreign assets. Such an increase in real the resulting high degree of volatility in inflation. expenditures would actually increase the demand A corollary is that private agents will continually for domestic money since some of the new be wrong-footedin their inflation expectationsand transactions will require it. By contrast, in Ethiopia so domestic financial liberalization should be an increase in investor confidence is inflationary delayed. Otherwise, domestic financial since the asset demand for domestic money would intermediation will generate large transfers decline and this would be offset only partly by the between agentsorthogonalto economic efficiency. increase in transactions demand associated with For example, during the extraordinarily rapid the rise in investment expenditure. Secondly, the Ugandan disinflation of 1992, expost real interest demand for money is likely to be affected by rates rose to around 70%, rewarding depositors and switches back into(or out of) the monetized part of penalizing lenders. Fortunately, there was so little the economy. In Uganda, the war-torn economy financial intermediation that these transfers were not noticeably disruptive. only three months and being rediscountable at the An alternative to fiscal targeting is to targetthe Central Bank.'' They are therefore not as distinct price level directly. Rather than setting a target for from money as would be desirable for non- a monetary or a fiscal aggregate the government inflationary financing. Second, despite their short would set a target for the price level itself. Fiscal maturity,they remain domestic currency li~bilities instruments would be used but only in response to of the government and as such can be vif ~.edas a deviation of the price level from the target. Such rather risky. The local Treasury Departm ,it of a a strategy would imply that shocks originating multinational company in Uganda explained that it from the demand for money would be borne by did not hold TBs in spite of an attractive yield: fiscal responses rather than by the price leveL8To were there to be a default corporate headquarters implementsuch a strategy of direct targeting of the would penalize local executives for an event which price level requires an accurate and prompt would be regarded as predictable given prior consumer price index. The Ugandan government history." By mid-1993 the real interest rate on gave this a high priority, undertaking a household Ugandan TBs was still around 23% despite the expenditure survey in 1989, some two years after modest level of sales: the total outstanding stock of the establishment of peace, and two years before a government domestic currency debt was only one census. The Ethiopian government is currently percent of GDP.Attempts by the Central Bank to reversing these priorities, planning a census for sell debt of longer maturity, nine months instead of 1994, two-and-a-half years post peace, with no three months, found no buyers in 1992, and in firm plans for an expenditure survey. Its existing 1993 could only find buyers at a real interest rate price index is too unreliable to be used as a guide of 40%. This extreme steepening of the yield curve to fiscal policy: the government might find itself is a symptom of the doubts about the viability of tightening fiscal policy in response to a statistical the state which persist after the restoration of artifact. peace. The scopefor deficit financing from sales of To summarize, the transition to peace is likely government debt is therefore very limited: were to cause large and difficult to predict changes in the government to use it to any extent, the interest money demand so that monetary targeting will fail burden would seriously inflate future government to keep the price level constant. In these expenditure. Even with modest sales, real interest circumstances fiscal rather than monetary policy rates at 23% should have been regarded as should be used, with fiscal policy being tightened prohibitive for government borrowing. or relaxed in response to observed deviations of While post-war governments are in an the price level from a target. unusually weak position to borrow domestically, they own assets (public enterprises, commercial 4. Medium Term PortfolioPolicy and residential property) which private agents may prefer over government debt. In Angola and During wartime African governments will Ethiopiathe government owns all urban residential obviously not find willing buyers for their property and all urban land. In Uganda public financial instruments. Since such instrumentsoffer ownership is less extensive, but still includes large the government the prospect of non-inflationary quantities of housing. The government can sell financing, it is natural that post-war governments such assets provided its ownership is should attempt to develop the market as soon as uncontested.'* possible. The Ethiopian government indeed plans Whether the government can raise money to start a market in Treasury Bills (TB) at the time more easily by selling such assets than by selling of writing,some two years after the end of the war. government debt depends on how private agents The Ugandan government has been operating one compare the associated risks. First, all such assets for two years. However, the Ugandan market in are less vulnerable to predation by the government TBs is still extremely thin. The only buyers are than are government liabilities. While the four banks which between them hold around $ government may confiscate them, and so they are 30m.' The sale of TBs poses two problems. First, not entirely secure, the acts of confiscation and these assets are very liquid, having a maturity of default are not equally easy: omission is easier than commission. Secondly, equity in privatized Currently, in Africa's post-war economies, the public enterprises is more risky than the ownership government is often holding a substantial part of of real estate, since the future profits stream of an its portfolio in real estate. a low risk. low yielding enterprise is more vulnerable to policy change than asset, while attempting to fund its deficit through is the rental stream from residential (and to a lesser sales of liabilities at high real interest rates. To extent commercial) property. Real estate is illustrate, the Ugandan government owns some therefore likely to be considered an attractive asset. 4,000 houses with a market value of perhaps $ This is confirmed by the Ugandan The typical rental return on the Kampala phenomenon of 'dollar houses', luxury residences housing stock is around 8%,14 so that if the constructed in the immediate post-war years by government rented this asset out at commercial private investors for rental to expatriates. Such rates it would receive an income of $ 8m. By investment is country-specific, slow depreciating contrast, it has only around $30m of domestic and in the export sector. However, unlike currency debt on which. as noted above, it pays investment in, for example, the rehabilitation of a 23% real interest, or $7m. The government should tea estate which shares these characteristics, it is substitute in its portfolio away from safe, low relatively invulnerable to policy change since not yielding assets which the private sector is willing being a production process it is not input- to hold, and from liabilities which the private dependent, payment for the services can be made sector is unwilling to hold, towards development by the expatriate from a foreign bank account into expenditures which the private sector will not a foreig bank account without being controllable undertake. A further advantage of selling by the government, and its services can always be government property is that because it is less liquid sold on the domestic market, so that the investment than TBs it sterilizes money more effectively. A is mobile between the export and non-tradable dollar of real asset sales may therefore be more sectors. Finally, because house rental is not a skill- counter-inflationary than a dollar of TI3 sales. intensive process, the asset is saleable on a much Ethiopia currently has neither a market in wider market than is a tea estate. The number of government liabilities nor a market in public investors in Ugandan tea estates is so limited that, assets. The government owns a very substantial even if profitable, the investor must presume that part of the urban housing stock and here there is it is unmarketable, whereas houses are marketable the same scope for a profitable portfolio switch as in all but catastrophic circumstances. If investment in Uganda. In addition all land is government in "dollar houses" was sufficiently attractive in the owned. Land sales therefore offer enormous scope immediate post-war years to induce new for an improvement in the public portfolio. Until construction, at least part of the existing stock of early 1993 the government effectively rented out public residential property is probably marketable. land for free and administrative allocation systems In the earliest stages of peace private agents rationed the resulting excess demand. An will probably not be willing to purchase either important rationing mechanism involves approval government liabilities or public real assets. of an investment project by the Investment Office. However, as peace is sustained, they are likely to While such approval does not guarantee access to accept, because of the differences in risks, a land, it makes it considerably more likely: when a substantially lower implicit rate of return on project is approved the Investment Office writes a unconfiscated residential property and land than letter to the regional government in support of the upon government liabilities. In effect, the risks of investor's application for land. Since access to land subsequent loss of property rights recede more is highly profitable a substantial part of investment rapidly than the risks of a government implicit or proposals are probably bogus. For example, there explicit default on liabilities. Further, since the are numerous applications for the construction of housing market has a naturally broader base than hotels. Commercial bankers think that the most that for government liabilities, there should be likely explanation for this is that such applications considerably more scope to expand this asset offer a good prospect of prime commercial sites market without forcing up implicit interest rates which can then be retained by the partial, or if than there is for sales of government liabilities. necessary complete, construction of a building the use of which can remain flexible." In March 1993 sector-specific and more particularly country- the Addis Ababa regional governmentinstituted a specific. Non-tradable capital goods must be 'service charge' and thereby de facto started to sell constructed on-site and are ex post location- urban land. It charged around Birr 20,000 (about $ specific. By contrast, tradable capital goods are 4,000) per acre for prime sites. This was levied as intrinsicallyex ante not country-specific,and are a charge on the transaction which would be often mobile expost and therefore less subject to repeated if the land were to be resold. This country-risk, but non-tradable capital inherently approach to a land market has several weaknesses. contains country-risk. Further, sector-risk is First, it is a transactions tax rather than a sale and largely country-specific: the predominant reason so discourages resale. Second, the price charged is why the return in a sector might decline in the far below the market-clearing price to the extent African context is if government policy turns that it can be determined. The price at which prime against the sector. Hence, capital goods which are sites informallychange hands is reportedly around country-mobile largely avoid sector-risk even if Birr 800,000 (around $ 150,000) per acre. There they are sector-specific. If the government of was therefore a strong incentive to acquire land, Uganda removes protectionfrom the textilesector, even given the new service charge, but the charge there is some scope for relocating machinery in then encouraged owners to retain the land even if neighboring countries. Non-tradablecapital goods they had no use for it. Third, the public beneficiary are therefore not only inherently more prone to of urban land charges was the regionalgovernment country-risk than tradable capital, they are also of Addis, which is easilythe richestof the regions. likelyto be more prone to sector-risk.The neglect Hence, the policy in effect permitted the richest of of the agricultural export sector during wartime the regional governments to dispose of national implies that the non-tradablecapital stock in which public assets at a small fraction of their worth.16 the sector is normally fairly intensive(trees, feeder Hence in post-war circumstances placing roads, storagefacilities)is in need of replacement. government liabilities is an expensive form of Yet the high degree of macro-insecurity and governmentfinance as governmentdefault is still policy-risk in post-war economies discourages considered a non-negligible risk so that private preciselythistypeof investment. This isconfirmed agents require high real interest rates. At the same by the experience of Uganda and Ethiopia noted time the government is likely to own asset.suchas above: in both countries investment in agriculture reaI estate and land which private agentsconsider is extremely low. more secure and are therefore willing to hold at To summarize the Uganda experience, the lower rates of return than government liabilities. economy has been starved of private investment, especially in export agriculture. The post-war 5. Long Term Policy For Stimulating government introduces a legal environmentsuited Exports to private investmentwith an InvestmentAuthority to aid the process. What it gets is dollar houses and Civil war disproportionately damages the light manufacturing. This is not because these agricultural export sector of the economy: wartime investments offer the highest return but because governmentsover-tax it given their extraordinary they are substantially less risky than the need for revenue, and rural transportationof goods investments which are most needed. becomes vulnerable. Since the sector is the major Many of the factors which make non-tradable source both of foreign exchange and government investment risky for private agents should be revenue, there is a social premium upon private discounted by society because they are generated investment in it. by the society. Ideally, the risks should be Even if currently profitable, investment in the removed, but some of them are irreducible in the sector is highly risky because it happens to be aftermath of civil war. A possible strategy is for atypically irreversible(sincedepreciation rates for the government itself to make the investment in such investment are very low) and process- non-tradable capital, leasing it to private dependent. Agricultural investments are largely entrepreneurson short, but renewable,tenure. This would shift the risk involved in the irreversible commitment associated with ownership from the investment can be removed once the tax holiday private entrepreneurto the government. Even if the ends." government is able to financesuch investment. the In the past most African governments more specific the non-tradable capital purchased implicitly subsidized investment through over- by the government the more severely does the valued exchange rates. The over-valuation government encounter the problem of 'backing subsidized all imports purchased with domestic winners', namely that it lacks the knowledge on resources, but rationing gave priority to capital which to make good choices. Further, much of the goods. This was, therefore, not a general subsidy problem is the fear of government on the part of on investment, but a subsidy on imported capital private investors. Although leasing avoids an goods purchased with domestic resources. The irreversiblecommitment in non-tradablecapital by subsidy needed in Africa's post-war economies is the private investor, it does so at the cost of precisely the converse of this: there is a case for increasing the direct relationship between the usingthe exchange rate tosubsidizeinvestment but private agent and the government: the government now in the form of an under-valuation. becomes the landlord. This actually increases the This will have two effects. First, it will induce scope for the government to be predatory and so substitution of non-tradable for tradable capital. may make the entrepreneur wary of other Secondly, as noted, private agents hold large irreversible non-marketable investments such as amounts of liquid assets and part of this is held reputation, without which activitycan only remain abroad. For example, at present Ethiopian limited. businessmen living abroad hold very large If the government does not take on the amounts of foreign assets which they may want to investment itself then in effect, it needs some bring into the country to finance domestic mechanism for subsidizing the act of commitment investment. Such foreign capital inflows will be of private investors, becauseotherwise irreversible encouraged by a policy of undervaluation. An investment will be socially sub-optimal. What is additional administrative advantage of working needed is a simple mechanism for implementing through the exchange rate is that a firm only such a subsidy. receivesthe subsidy if it actually makes the desired However, strategies of subsidy can be expenditure. problematic. The African record of 'backing During exchange rate under-valuation the winners' is so poor and the political pressures to government will be temporarily accumulating misappropriatetargeted subsidies are so acute that reserves. Under-valuation cannot be a long-term the subsidizing of particular investment projects is policy because it would imply ever-rising reserves. not likely to be desirable. However, this is itself quite a desirable feature At present the Ugandan government is since smart private agents should therefore subsidizing investment by precommiting itself to recognize that the subsidy is temporary. This five year tax holidays. This policy has several recognition should both limit the tendency to drawbacks. First, a tax holiday is not a credible act misallocate irreversible factors to activities which of commitment since the government has several are only profitable because of the under-valuation explicit and implicit tax instruments in addition to and further induce the purchase of non-tradable corporationtax. Most notably, it could revert to an capital goods since delay is expected to reduce the over-valued exchange rate, extracting any desired subsidy. rate of taxation on investments while leaving the Such a policy will certainly not be sufficient to corporate tax rate at zero. Second, a tax holiday attract investmentcommitments in the absence of obviously provides no insurance against the complementary public investments in collapse of profits due to a reversion to civil infrastructure. Nor should investment be the unrest. Third, it permits the government to impose primary objective of exchange rate policy. high corporate tax rates once the tax holiday period However, to the extent that undervaluationindeed is over. Knowingthis, investorswill favor projects implicitly subsidizes commitment, which is itself which generate profits only in the short term, or sociallydesirable, there is a case for including it as install only e-rpostmobile capital goods so that the a consideration in the setting of economic policies. 6. Conclusion peace transition situations. Since money demand is highly unstable the government should target the African economies in post-warenvironments. such price level (changing its fiscal stance in response as Ethiopia and Uganda, are likely to have some to deviations of the price level from some target) common characteristics. Some private agents. directly rather than attempting to control inflation whether illegal traders or expatriated ethnic through monetary or fiscal targeting. It should minority business communities, are financially secure its medium term financing needs through liquid and so in a position to afford investment. the sale of property and land rather than through The economy has been extremely short of private attempts to establish a domestic market in its investment for many years and so the return is liabilities since as long as default is considered likely to be high. Yet private agents are resistant to likely debt financing will be unnecessarily costly. non-tradablecapital goods investments of the type Finally, it should subsidize the act of commitment that the export sector needs for recovery. We have implied by investment in irreversible capital proposed three policies as appropriate in these through temporarily undervaluing the exchange circumstances which are fairly distinctive to war- rate. Notes 1. The authors would like to thank two anonymous referees for extremely useful comments on an earlier version of this paper. 2. Obviously, in Uganda there was also an involuntary exodus. 3. The exception to this is the Franc Zone member countries such as Chad since there is no national' currency. 4. Interviews conducted by the authors during 1993. 5. Figures from the Uganda Investment Authority, covering all planned and implemented private investment (other than very small scale informal activities) since its inception in 1991 to mid-1993. 6. Source: Ethiopian Investment Authority, 1993. 7. Authors' interview with Zimbabwean textile entrepreneurs. June 1993. 8. An advantage of this policy is that it is conducive to the revenue-maximizing amount of seigniorage. Recent analyses of the demand for money in Africa (Adam (1992), Adam, Ndulu and Sowa (1993)) suggest that fairly low and stable inflation rates is revenue-maximizing. 9. TBs have been supplemented by Promissory Notes, issued to suppliers with whom the government has fallen into serious arrears in its payments. They cany a maturity of one year and pay interest at 15% and are not rediscountable at the central bank. They are therefore forced sales of government paper. However, since they can serve as collateral for credit they are in effect rediscountable at the commercial banks. The banks must then hold them to maturity but they are clearly close substitutes in the banks' portfolios for TBs. 10. In Zambia in a post-Kaunda transition not dissimilar from those considered here, the maturity of TBs has been shortened to only 28 days. 11. Authors' interview. 12. Government ownership is likely to be contested when it results from confiscation. In Uganda and Ethiopia part of public property is owned by the present government following confiscation from private owners by a previous government. Since the post-war government has come to power by overthrowing its predecessor, it is under no obligation to regard these actions as legitimate and so restitution is at least potentially on the political agenda. The Ugandan government committed itself to restitution, but with a time limit on claims. If the government is at some stage to raise revenue from asset disposals such a limitation is necessary because until that date the government cannot establish its ownership of properties for disposal. The Ethiopian government has resisted general restitution of property. 13. There is no current valuation of the publicly owned housing stock. However, there are believed to be 4.000 properties. Most of these are in very poor condition but some of them are on very valuable sites. The market price for newly constructed housing of similar size to the average of the public housing stock is $ 50.000. The estimate in the text halves this to allow for poor condition and the depression in the market from the greatly increased volume of sales implied by disposal of the public housing stock. However, it is probably a conservative figure because some government properties are extremely valuable because of their locations. 14. Data on house prices and rentals were supplied by Mr. Kasekende, General Manager, National Housing and Construction Corporation. 15. There is also a hotel boom in Uganda, where again there is evidence in the form of uncompleted buildings which serve as security for loans, that investors favor hotels because of the high content of sector-unspecific fixed capital combined with an explicit activity which is attractive both to government and bankers. 16. The central government consequently declared the Addis user fee system to be illegal and established a study group to consider land policy. In August 1993 the study group participated in a televised debate with the business community, as the government edged towards making its major asset marketable. 17. The government of Botswanaencouraged manufacturing investment through tax holidays, but has found that at the end of the holiday period the investment is footloose. Chapter 3 Policy Uncertainty, Repatriation and Investment Paul Collier and Jan WillemGunning' 1. Introduction In 'War, Peace and Private Portfolios' we argued a governmentwhich intends to break its promises.' that in a post-civil-war economy private agents Alternatively, credibilitymay be achieved through would hold considerableforeignfinancial assets. A institutional arrangements which make policy second feature of war economies is that they are reversal extremely ~ostly.~Here we take a different usually heavily regulated. We considera post-war and indeed perhaps more obvious approach. The economy such as Ethiopia in which the typical prospectivereform package includesa wide government inherits a 'control regime' during range of policy intentions. These policy intentions which there has been considerable capital flight. vary both with respect to the credibility with which As in our previous paper, the policy objective is to they are viewed and the impact they will have if induce repatriation for purposes of private implemented. We abstract from fears of policy investment. However, while in that paperour focus reversal: the government can resolve the was on compensatingfor irreducible risk, here we uncertainty surrounding a prospective reform by consider the sensitivity of repatriation to the implementingif but becauseof administrativeand control regimeand its reform. The policy problem politicalconstraints, implementing one policy now arises because policy reform is a slow process due has the opportunity cost that another must be to legislativeand political obstacles so that it must deferred. The pace of reform is given, but the be spread over several years. While it is common govenunenthasscope forchoice in determiningits for governments in this position to pre-announce sequence. We thereforeinvestigatewhether, given their agenda in broad terms as a statement of different credibility ratings of different policies, intent, a post-war government is at best an there are more and less efficient sequences of unknown quantity (more likely during the reform. We distinguish between two types of insurgency phase it will have created a frightening doubt: those which concern the timing of a impression),and so some of the promised reforms promised reform and those which concern whether will inevitably be less than fully credible. Doubts a promised reform will in fact completelydrop of will arise both as to whether the government will the agenda. We refer to theseas timing doubtsand stick to its timing, and more fundamentally fundamental doubts. Some componenrts of the whethersome proposed reforms will happen at all reform package are subject to each of these types or, even if implemented, be reversed. In the worst of doubt and the government can only resolve scenario the government may be so little trusted them by implementation. We further distinguish that potential repatriatorswill not repatriate until policies by their impact if implemented. Some many years after all reforms have been made. policies directly affect the price or worth of an To date, the main focusof the literature which asset, whereasothers do so only indirectlythrough addresses this problem has been on how the altering the return upon it. government might increase the credibility of its Policy incredibility may have consequences promises to undertake or maintain reforms. The for various aspects of private behavior. A government may enhance its credibility by particularfocus of the literaturehas been the effect adoptingsignalling rules which distinguish it from on private investment. It may reduce investment because capital once installed is immobile. A reasonable expectation would be that land rights broken policy promise may change the relative would definitely be settled but over an uncertain profitability of sectors (Rodrik, 1989), or reduce period depending upon policy choice. This was the return on domestic investment below other therefore a timing doubt and the policy impact was assets (van Wijnbergen, 1985; Rodrik, 199 1). directly on the worth of the asset. The acquisition Additionally, policy uncertainty might reduce of good title to urban land would not of itself investment due to its effect upon savings (van confer the right to sell it at the market-clearing Wijnbergen, 1992). Here our focus is specifically price.Given the ideological background of the new upon private investment financed by repatriation. government, it was uncertain whether it could There are two reasons for this narrower focus. permit the emergence of a competitive land First, arguably it is these funds which are most market, as its program hinted, or whether it would policy-sensitive.In the economies most in need of continue to control prices. As with property rights private investment due to a long history of its the impact of policy was directly on the worth of discouragement, such as Ethiopia and Uganda, an asset, but doubt was now fundamental. Inherited private foreign assetsare probably large relative to labor laws reduced the return on investment by domesticassets. Second, since the government can conceding to workers many of the powers which attach a low weight to the non-domestic would normally be reserved for management.This expenditure of unrepatriated funds, investment included restrictions upon dismissal which had led financed by repatriation has a lower social to substantial overmanning. These laws were opportunitycost than domestic savings. naturally a sensitive issue for the new government Recent policy debates in Ethiopia can serve to because existing employees wished to preserve illustrate the distinctions we will be making. Our there entrenched rights. There was therefore again taxonomy of policy doubts and policy impact fundamental doubt as to whether policy reform in generatesfour possible cases: doubtscan be timing this area would be other than cosmetic, but the or fundamentaland the impact can affecteither the impact was on the return on capital. Finally, the return on capital or directly affect asset values. rate of return was also reduced by corporate taxes Four policy reforms which were under which were virtually confiscatory in the Mengistu consideration by the Ethiopian governmentduring period. It was apparent that the new government 1992 fall into each of these categories,as depicted would reduce corporate taxes as part of its in Table 3.1. First consider property rights. As in promulgation of an investment code. However, it other post-warand post-communistsituationsthere was unclear what priority the government would was considerable uncertainty as to the ownership assign to the new code given that it had so many of property. Assets currently operated by the state other commitments on its legislative capacity. could at some stage be subject to rival property Hence, doubtconcernedtimingand the impactwas claims from previous owners who had been on the return on capital. expropriated.Hence, the purchase of non-tradable The first question we address is whether, a capital goods (like buildings) carried the risk that priori, there are grounds for the government to title might subsequently be contested. The prioritize the urgency of making these reforms, acquisitionof land for industrialor commercial use given that the policy and legislative process is slow carried the same risk but in a more extreme form so that not all changes can be undertaken since all that was available was the acquisition of simultaneously. We consider only otherwise a right of usage rather than of ownership. The commensurate policies. By commensurate we government was aware of the problem posed by mean policy stances which, if sustained with the lack of clarity over property rights but faced a certainty, have the same effect upon permanent choice. It could, as in Uganda, recognize and invite income. That is, we are not investigating whether claims from previous owners. Alternatively, it uncertaintyabout labor laws is necessarily more or could declare such claims invalid ab initio,vesting less costly than that about property rights, but ownership in itself. The former course implied a rather the relative urgency of reform were the much longer period of uncertaintyon the validity effect of permanent unreformed labor laws to of title for a repatriator than the latter. Hence, a depress the return on capital by the same extent as a permanent failure to clarify property rights. agent has to decide whether or not to repatriate, he Section 2 sets out the analysis of prioritising this is uncertain as to which policy will prevail in that topology of policies. period. The decision problem is in this case Our second question concerns exchange rate extremely simple: since it is certain that the policy. In many developing countries the domestic rate of return will exceed the foreign one overvaluation of exchange rates was probably the (rr > r*) for all later periods, the investor will single most serious policy failure of the early choose to remain liquid for at most one period. He 1980s and it has consequently been the central will do so if the return over the infinite horizon is focus of reform. On the argument of our previous higher than if he repatriates immediately: paper. exchange rate reform is yet more important in transition to peace economies: the rate needs not only to be corrected from over-valuation, but switched to under-valuation in order to create an Clearly, only the first period returns need to be implicit subsidy on repatriation into irreversible compared: the investor will repatriate if the domestic investment. Whereas it is straightforward expected value of the return on domestic to classify the four Ethiopian policies discussed investment in that period exceeds r*. From (1) above into those which affect the return on capital repatriation will be postponed if: and those which directly affect its price, the classification of the exchange rate is more complex. In Section 3 we show that the exchange rate may have very different effects depending This defines a critical probability p,. If delay of upon other aspects of the control regime, so that reform is sufficiently likely (p > p,) repatriation the urgency of reform differs accordingly. will be postponed. In the second case there is fundamental doubt. 2. Ranking the Consequences of Policy The reform is either implemented in period 1 (with Uncertainty probability I -p), or it is completely discounted. If the investor remains liquid in period 1, he will Consider an investor who has to decide whether to repatriate in the second period only if reform has repatriate assets held abroad. The domestic rate of then taken place. Hence he will decide to keep his return is r, if policies are unchanged and rr after assets abroad during period 1 (that is to remain reform. The return on foreign assets is r*. All rates liquid for the time being) if: of return are finite and we assume rr> r* > ru.The returns are measured in foreign currency and in the case of domestic investment profits are measured after being remitted a b r ~ a d . ~ The investor is risk-neutral; he maximizes the Note that after the first period he then has the expected value of the return on his assets over an option to repatriate, which he will exercise if infinite horizon, discounting at a fixed rate r,. reform has then taken place. In this case there is a Foreign assets are liquid, but domestic investment premium on liquidity: even though the expected is irreversible. If reform is implemented it is return on domestic investment in period 1 may be sustained indefinitely. We consider, on the basisof much higher than the foreign rate of return, it may the distinctions between fundamental and timing still be advantageous (since r* > r,) to hold assets doubts and between rate of return and capital value abroad during the first period to avoid being impacts, three types of policy uncertainty. locked in should reform not take place. This, of In the first case there is timing doubt. Reform course, is the option value of waiting. will take place either in period 1 or in period 2. Condition (3) defines a critical probability p,; Viewed from the start of period 1, the domestic if reform is sufficiently unlikely (p > p,) then the rate of return will rise from r,,to r, with probability repatriation decision will be postponed. From (2) I-p in period 1 and with probability p in period 2. and (3) the critical probabilities in the two cases Hence. at the start of the first period, when the satisfy: very low values of p (p < p,) the investor will repatriate in the first period in zll three cases. In and since r, > r* > r, this implies p, > p,. The between these extremes, the combination of timing critical probability is lower than in the first case doubt and a policy impact directly on the asset because investment is irreversible: in the first case value (the third case) is most damaging for repatriation involves a risk of a return lower than repatriation (in the sense that the critical on the foreign asset (r, < r*), but only for the first probability is lowest), fundamental doubt period; in the second case the risk applies to all combined with a policy impact on either the asset future income. value or its return is intermediate, and the Finally, in our third case reform applies to the combination of timing doubt and a policy impact value of capital (unlike in the first case) and there upon the return on capital is least damaging to is timing doubt (unlike the second case). The agent repatriation (in the sense that the critical must decide whether to repatriate at the start of probability, p,, is relatively high). period 1, not knowing whether policy will be The conclusion from this is that even at modest changed for period 1or only for period 2. If policy probabilities of postponement or non-reform, is not reformed until period 2, a rival claimant for policy uncertainty can induce investment deferral his newly purchased asset may come forward if it relates to timing and the policy impact is on during period 1. asset values so the government needs to resolve it It is convenient to establish a baseline case. by giving the needed legislation high priority. By Suppose that should policy reform be delayed until contrast if uncertainty relates to timing but the period 2 the risk of asset loss during period 1 is impact is on the return on capital, reform can be ru/r, and that allowing for this risk, the expected postponed until later in the legislative program. rate of return is ru. Hence, in the event of the Applying the results to the previous example of reform not being delayed, the return will be r,. In four policies in the prospective Ethiopian reform this case the expected return on domestic program, the implication is that the most urgent investment is [(I -p)r,+ prJ( l+I/r,), just as in case refom would be to establish unchallengeable 2, in spite of the difference between fundamental ownership claims. The least urgent reform would and timing doubt. However, there is a distinction be to amend profits taxation. The other two between the two cases. If the investor remains reforms, the resolution of whether urban land liquid in period 1 (earning r*), he will thereafter markets should be permitted, and whether the labor earn r, indefinitely in case 3 (since reform is code should be substantially revised, are of an certain5) but only (I-p)r, + pr* in case 2 (because equal and intermediate urgency. then there may be no reform). Hence postponing Our results differ in several ways from those of repatriation is even more attractive in this case Rodrik (1991). In our formulation uncertainty is than in case 2. The investor will choose liquidity in resolved after the first period. In the timing doubt the first period if: case of condition (I), at the end of period 1 either reform has taken place or the agent knows with r* + rjr, > [(1-p)rr+ pru](l+l/rd) (5) certainty that it will occur in period 2. Alternatively, in the fundamental doubt case, Comparing (3) and (5) the critical probabilities reform has either taken place in period 1 (and it must satisfy: will then be sustained) or it has not (and one then knows that it will not occur). In Rodrik's model the reform is introduced in period 1 and in each subsequent period there is a (constant) prcbability which implies that p, > p,, as expected. (p) of policy reversal. Hence even though Rodrik considers policies which affect the return on Our results imply 1 > p, > p, > p3> 0.Hence, as capital rather directly impact upon capital values, the probability, p, of delay or non-reform and uncertainty relates to timing, his result is much approaches unity (p > p, ) then in all three cases closer to our second and third cases than to our the investor will prefer liquidity. Conversely, for first, apparently eq$valent, case. Indeed, he concludes that even low probabilities of policy price effect depends on the composition of reversalmay deter investment.This simply reflects investment. that in Rodrik'smodel policy reversal is possible in The official exchange rate may also affect the every period so that policy uncertainty affecting rate of return on capital. It can do so through two the return on capital is analytically identical to routes, the return in domestic currency and the uncertaintyaffectingasset values. An option value mapping to remitted profits. Consider, first, the cannot arise in his model. If the agent decides to return in domestic currency. In the limiting case a remain liquid initially he will never have to change in the official exchange rate has no effect reconsider that decision since matters cannot upon the rate of return. Again let the price of non- improve: if the reform is sustained for one more tradables be unaffected by the devaluation, so the period the probability of a future reversal is limiting case holds if all inputs(other than capital) unaffected. Hence when the agent reconsiders the and outputs of the investment project are non- profitabilityof investmentafter a period of waiting tradable. If the characteristics of the project are the problem is either still the same as before or the reversed, so that all non-factor inputs and outputs reform has collapsed in the mean time so that the are tradable and their prices reflect the official investor will decide to remain liquid. In our case a exchange rate both before and after the reform may be adopted (rather than reversed) and devaluation, then the rate of return rises in hencethe option value argumentfor staying liquid proportionto it. This positiveeffect may be seen as does apply (in case 2). the removal of an implicit tax, the overvaluationof the domestic currency. 3. Exchange Rate Uncertainty and Now consider the return as measured by Repatriation remitted profits, taking the domestic currency return as given. Again there is a limiting case in We now consider how uncertainty attached to the which a change in the oficial exchange rate has no official exchange rate should be classified effect. Suppose, analogously to our assumption accordingto the aboveschema. The questionarises about non-tradable goods prices, that the because the official exchange rate may, depending devaluationof the oficial rate has no impact upon upon detailsof economic structure, directly affect the parallel exchange rate, and that the devaluation eitheror both of the price of capital and its rate of is sufficient to achieve convergence at that rate. return viewed from the perspective of the The limiting case now holds if, prior to reform, repatriator. profits were remitted at the parallel rate. Note that The official exchange rate may directly affect this is not a curiosum. In the more extreme control the price of capital. Suppose that it is devalued regimes the remittance of profits was illegal. The from e, to e,. If domestic prices of tradable capital usual private response was to remit profits by the goods rise in proportion, then, if all capital goods device of over-invoicing of imports. Bevan et involved in the investment project are tradable, the a1.(1990) establish that the implicit exchange rate price of capital to the repatriator is unaffected by at which such remittances take place varies with the devaluation. This will be the case if tradable the details of the control regime, but that in the capital goodsare imported at the official exchange most common case it is the parallel rate. While rate both beforeand after the reform. Since capital this is a limiting case, in less severe control goods imports tend to be highly visible (and regimes profits of fcreign firms could be remitted therefore more difficult to smuggle) this at the official rate. In this case the devaluation assumption may not be unrealistic. However, reduces the profitabilityof domestic investment in prices of non-tradable capital goods are unlikely to proportion (i.e. by a factor eJe,). While the former rise proportionately.We considerthe limitingcase effect on the return on capital can be interpreted as (which may be applicable in the short run) in the removal of a tax, this effect amounts to the which these prices are unaffected by devaluatiom6 removal of a subsidy: before the reform foreign In this case, if the asset purchased by the investors were implicitly subsidized through the repatriatoris non-tradable,the devaluation lowers overvaluationof the exchange rate. the price of capital by the factor e,/e,.' Hence the The above effects generate the taxonomy follows from the analysis of Section 2. Recall that shown in Table 3.2.The 'capital' column indicates for fundamental doubt there is no distinction whether the capital goods acquired by the investor between the impact upon the price of capital and are tradable (T) or non-tradable (N). The next its return. The four cases are thus similarly columnsimilarlyclassifies(non-capital)inputsand affected by fundamental doubt. By contrast, for outputs. The third column classifies according to timing doubt, effects on the rate of return ::re less the pre-reform remittance regime, a P indicating serious than those which operate directlq %jnthe remittance at the parallel rate and an 0 at the price of capital. The four cases in which official rate. The 'effects' column summarizes devaluation alters the incentive to repatriate span which of the three effects (if any) a devaluation these routes, since effects I1and 111are on the rate has: a change in the price of capital (I), in the rate of return whereas effect I is on the price of capital. of return in domestic currency (11) or in the Table 3.2 showsthat in one of the fourcases, 7, the exchange rate for remitted profits (111). These only route by which devaluation induces effects are aggregated and signed in the last repatriation is through a price of capital effect, column. implying that with this economic structure In the third case none of the three effects devaluation is urgent. In another, case 1, the only operates: all capital is tradable, so the quantity of route is through a rate of return effect, implying a investment is not affected, all other inputs and lower degree of urgency. The remaining cases, 5 outputs are non-tradableso that the domestic rate and 6, are intermediate since both routes operate. of return does not change and the exchange rate relevantfor sending profits abroad does not change 4. Conclusion since prior to the reform the parallel rate applied. In two other cases two effects operate and they We have considered a post-war economy in which offset each other precisely. In case 2 the rate of the government has the typical inheritance of an return rises in proportion to the devaluation, but economic control regime with private this is exactly offset by the rise in the official entrepreneurs holding substantial foreign assets. exchange rate at which profits are remitted. The policy objective is to induce repatriation for Similarly, in the last case the price of capital falls, purposesof private investment, but policy reform but this is precisely offset by the removal of the is a slow process due to legislative and political subsidy implicit in the exchange rate at which obstacles so that it must be spread over several profits were remitted. In one case, 4, the net effect years. The government pre-announces its agenda, is negative. Here the removal of the subsidy is the but some of the promised reforms are less than only effect of the devaluation. These cases fully credible because the new government has no demonstrate that there can be no general reputation. Doubts arise both as to whether the presumption that devaluation increases the ' government will stick to its timing, and more incentive to repatriate. Afortiori, in thesecases the fundamentally whether some proposed reforms degree of uncertainty related to a prospective will happen at all. The question then becomes devaluation is irrelevant for repatriation. Since whether, apriori, some aspects of uncertaintyare under certainty the effect is either zero or negative a more severe deterrent to repatriation than others, and since without reform the investor prefers to and thus which reforms are most urgent and which keep his assets abroad, an uncertain prospect of can be postponed with relatively little damage. If devaluation can neither induce nor deter it chooses the most appropriate sequence then repatriation. On the criterion of inducing repatriation for investment will begin prior to the repatriation it would therefore be prioritized later completion of the reform program, whereas if it than other reforms. chooses the least appropriate repatriation will be However, in the remainingfour cases, the net delayed until after program completion. effect is positive. Since in these cases a We have shown that some types of policy devaluation induces repatriation, uncertaintywith uncertainty are more detrimental than others. respect to the course of exchange rate policy is a Three types of uncertainty need to be deterrent. The magnitude of the deterrent now distinguished,dependingupon whetherthe price of capital is affected directly, or only indirectly The implications for the order of structural through the return on capital. and upon whether the adjustment measures were first illustrated with reform itself is in doubt or merely its timing. respect to four prospective Ethiopian policy Intuitively it might seem that it is more important reforms each of which could be classified into one to resolve doubts as to whether a reform will be or other of the categories distinguished. We then implemented than doubts as to timing, so that the turned to the most important economic policy first round of reforms should be those which reform during the transition to peace, namely the repatriators would otherwise doubt might be exchange rate. Whereas some policies are readily undertaken at all. This intuition is shown to be classifiable, we showed that the implications of incorrect. Uncertainty about the timing of a reform exchange rate uncertainty for repatriation depend is found to be more serious than doubts about upon details of economic structure. We identified whether a reform will be implemented if the those structures in which exchange rate uncertainty impact of the reform is directly on the price of was irrelevant to repatriation, and those in which it capital. would fit into each of our categories. Table 3.1. A Taxonomy of Policy Uncertainty Impact Type of Doubt Fundamental Timing Value of Capital Land Prices Property Rights Return on Capital Labor Laws Profits Taxation Table 3.2. Effects of a Devaluationon the Incentiveto Repatriate Initial rate for Case Capital Inputs/outputs repatriation Effects Net effect 1 T T P I1 + 3 T N P none 0 Notes 1. The authors are grateful to Robert Bates and Mohsin Khan for comments on an earlier version of this paper. 2. A bizarre case involves a government imposing trade restrictions to convince private agents that it will eventually, once its credibility has been fully established, adopt free trade (Engel and Kletzer, 1991). 3. Elsewhere we have argued that the credibility of African trade reform could be enhanced were the EC and African customs unionsto establish reciprocalfree trade. Reversal of trade reform would then lead to loss of free accessto the EC market (Collier and Gunning, 1993). 4. Hence the rise from rr to r, may reflect a change in the profitability (in domestic currency) but also a change in the rate at which profits remitted abroad are taxed or in the exchange rate applicable to remittances. 5. Hence in case 3 there is no option value. The investor may decide to wait but if he does so his decision at the end of period 1 is a foregone conclusion: he will repatriate, because by then the rate of return will have risen to rr. 6. In some contexts this is a reasonableassumption. For example, the very large Ethiopiandevaluation of 1992 had no impact upon eitherthe parallelexchangerate or the domestic price level, and the similarly large Nigerian devaluation of 1986 had a negligible effect. 7. This can have a powerful effect on the incentive to repatriate. For example, in Ethiopia e)eUwas about 3. Chapter 4 How to Pay for the Peace? A Theoretical Framework with References to African Countries Jean-Paul Azam Abstract This chapter analyses a simple game-theoretic model to highlight the choice of the government between raising its defense expenditures or giving away some "gzjls'90his opponents, as a means to defend his position in power. Ifthe government is a Cournot-Nash player, then there is no gift in equilibrium, and any increase in the budget will lead to more ineflcient defense expenditures. However, if the government is a Stackelberg-leader, then he will use the "gift" as a tool in his policy for staying in power. 1. Introduction The outbreak of a civil war is the worst failure of countries, especially during the Cold War. The a peace-keeping policy, or the dreadful result of Lybian leader Khadafi has often been held the lack of it. Mostcountries in the world are made responsible for various rebellions, whereas the of a heterogenouspopulation, divided by ethnicity, formertwo Super-powers, the USA and the USSR, religion, language, ideology, etc. It generally takes played a part in supporting civil wars in various some conscious efforts by the government for a parts of the continent, includingthe Honlof Africa state of peace to be maintained, with some clear (Africa Watch, 1991, Doornbos et al., 1992, impact on public finances. Wubneh and Abate. 1988). In many parts of the world, at various points in However, there are havens of peace in this time, open war between countries, or within the violent continent. The examples of countries like boundaries of various countries, was the normal Cdte d'lvoire or SCnCgalprove that it is possible to state of life. This was the case in Europe for most live in peace in Africa. Cdte d'lvoire offers a of its history,and it is now ravaging many parts of striking example of political stability, as the late Africa, from Liberia to Somalia, from Chad to President Houphouet-Boigny was ruling the Angola, etc. A recurrent cause of conflict in country even before independence in 1961, up to African countries is ethnic diversity, and his death in December 1993. He was a Cabinet "tribalism" is often a cause of civil war. The Minister in the French Govenunent in the 1950s. perennial rivalry between the Utus and the Tutsis Nevertheless, this is a clear example of a country in Rwanda and Burundi are cases in point. with a population divided between different ethnic Religion is often a cause of conflict as well. groups, whose territories extend largely over the However, foreign intervention has often played a neighboring countries (Azam, 1994). While the role in igniting violent conflicts in African largest ethnic group, the Akan, is related to the Ashanti from Ghana, the Krou group in the south have been heavily taxed for funding, among other west is part of the largest ethnic group in things, a lot of public investment in the poorer neighboring Liberia. The Senoufo in the north are north (Azam, 1994). Hence, there is a public more closely related, ethnically, to most people in finance aspect of a peace-keeping policy, as well Burkina-Faso, than to the southern Ivorians. as an "attitudinal" one (Brown, Hassan and Fole, Moreover, immigration has been going on during 1992). the recent decades, and it is estimated that 50% of This example sheds some light on the contrast the population in Abidjan, the capital-city, and between the policies pursued in Uganda by Obote, 25% of the whole population, are of foreign origin. after the fall of Idi Amin's dictatorship, and by Many come from Burkina-Faso and Mali in the Museveni, after the overthrow of Obote. Whereas north. Therefore, this example shows that ethnical the former tried to favor systematically his own division can be reconciled with internal peace in ethnic group and its close allies, the latter tried on Africa. Moreover, the north is predominantly the contrary to open his government to Muslim, while the south is Christian or animist, representatives from all the groups (Mutibwa, showing that religious diversity can also be 1992). Obote's attitude has triggered a violent accommodated peacefully. But this country had opposition with a strong ethnical character, only one political party, the PDCI, up to 1990. On whereas the latter mainly got some fairly irrational the contrary, SCnCgal was long regarded as the opposition, with a definite religious or magical almost unique example of Africandemocracy, with content, as was the case with Alice Lakwena regular elections being held with competitive (Behrend, 1991). As a result, Museveni's Uganda candidates, and limited violence. Therefore, may be regarded as almost completely peaceful, multipartism can exist together with peace, and a whereas Obote's Uganda witnessed one of the single party is not necessary, as the case of CGte worst slaughters ever, sometimes compared to that dlIvoiremight suggest, for avoiding conflict. of the Pol Pot regime in Kampuchea (Mutibwa, Although it is not possible to build a 1992). convincing counterfactual analysis showing Similarly, one can interpret as a case of whether this country would have experienced more voluntary redistribution in favor of potential political violence had it pursued a different set of opponents some aspects of the demobilization policies, it is interesting to use the case of CGte policy pursued currently in Ethiopia. For example, d'Ivoire for illustrating two important ingredients whereas some of the veterans of the defeated of a peace-keeping policy. First, the late President Mengistu army have been maintained in the had not spent much effort fighting his potential military, or, having been demobilized, have kept opponents, whose opposition he overturned by their ranks and their pension rights, a large number other means: with an extremely small number of of the victorious soldiers have been simply asked exceptions, he had always managed to share some to go back to their homeland with no particular of the spoiIs of the state power with the rising personal advantage.' politicians, even if they initially seemed to be The aim of the present chapter is to provide a confronting him (Azam, 1994). They have theoretical framework for discussing these issues. generally been attracted by the President's offers, In a simple game-theoretic setting, we discuss the and subsequently involved in the government and appropriate relative weights to be given to (defense the ruling party, in prominent positions. Second, an expenditure and to redistribution of the state important share of the government budget was money within a peace-keeping policy. The model used for redistributing between regions some of the presented highlights the trade off that faces the fruit of the economic miracle of the two decades government between spending budgetary resources following independence. The coffee and cocoa for fighting the opponent on the one hand, and growing farmers from the south, including the giving up some resources for reducing the BaoulC farmers of the President's ethnic group, opponent's willingness to seize power, by sharing some of the benefits of being in power, on the following. Assume that this probability of A other hand. This game may result in an inefficient staying in power is an increasing function of the equilibrium, where the two contenders are ratio DIF, given by: spending too much in military expenditures. This case is somewhat akin to the type of competitive 0 i p=p(D/F)i 1, pl>O,p" O. V" G. Define x=- pffD/Fpfand assume, 3. The "No Gift" Equilibrium for the moment, that x < I. Then B's reaction function can be written: It is intuitively obvious that a Cournot-Nash government will not give away,anythingas a "gift" F=F(T, G, D), FT>0, FGO. (6) to his opponent. As the level of G does not enter directly in the function p(-), and oniy acts Moreover, it can be shown thatd log F/a log indirectly as an incentivefor B to reduce his forces D = (I-x)/(2-x) < I. This can be seen intuitively F, it cannot affectA's behavior under the Cournot- by examining (5). If D and F are increased Nash assumption. proportionately,so that D/F does not change, then This can be shown formally by defining A.20 the right-hand side does not move, while the left- as the Lagrange multiplier attached to constraint hand side increases. Therefore, in order for (5) to (3), with complementary slackness. Then, hold, D/F must increase when F increases. maximizing (2) under (3) yields the followingfirst If the condition x = - pffD/Fpf< I does not order conditions: hold, with x ~ J 12[, then pf(D/F) D/F becomes , a decreasing function of D/F, and a F/a D < 0. As a function of D/F, x is first increasing, and then decreasingfor large values of D/F, because of tile and assumed concavity and boundedness of the function p(-) [0, I]. It is thus reasonable to E .' conclude that F is first increasing, and then decreasing in D, as D 6 F, where 6 is such Noticethe formal analogy between (7) and (5). that x(6) = I. We can safely neglect the case Becauseof complementaryslackness, we have A. G where x 22 , which is liable to give rise to a = 0. Thiscannot be consistentwith (8) and G > 0, multiplicityof equilibria. unless either - Hence, in this model, F is not a monotonic functionof D. There is a very high level of defense (i) T- and lim U'(T-G) = 0 as T-G - -,or - expenditures by the government where the opponentis discouraged from increasing his forces (ii) F- and D/F 0, andp(0) = 0 andpf(0) - as a response to the government increase in - - . defense expenditures. and reduces them instead. There is no ambiguity regardingthe two other Hence, under realistic assumptions, the effects, provided 2 x . The positive impact of T 2 government does not give any positive G to B in is not surprising, and captures the effect of the the Cournot-Nash equilibrium, as (8) implies that A > o . the intersection between the nvo will necessarily This "No Gift" equilibrium is not generally move upwards. Hence, in this case, an increase in Pareto-efficient.In order to analyze its properties, the government budget necessarily entails an let us draw the reaction functions of the two increase in the opponent's forces. It will plausibly agents. We already know from the previous entail as well an increase in the government section the main properties of the function F(-) defense expenditures. But this is not necessarily describing B's reactions. It is represented as the the case, as the new equilibrium point can lineFFin figure4.1. The government behaviorcan graphically be located to the north-westof pointE. be derived from (7), by putting G = 0. It thus But this will only occur in the unlikely case where reads: the marginal utility of S is very small for the government,at S = T, so that the DD curve shifts very little. It can be shown that the equilibrium response of D to an increase in T is positive It is straightforwardto find that the chosen D provided x > o/fo+3 , where o is the elasticity (-) is an increasing function of T. This is like an of V(-) with respect to Tor G (assumed constant), income effect: as the budget increases, the and E is the elasticity of U(-) with respect to S. government will consume more, and spend Similarly, it can be shown that the equilibrium more on defense. This may be checked by response of F to an increase in T is positive differentiating (7') to find a D/a T= U'() pr(D/F) provided x < (~+o)/o, which is automatically D/Fp > 0.Aspr(-) is a decreasing function of D/F, satisfied in the case of figure 4.1. we find easily that the chosen D(-) is an ambiguousfunction of F, with a log D/a log F < 1. Figure 4.1 For if D and F are increased proportionately, so that D/F does not change, then the right-hand side of (7') does not move while the left-hand side increases.Aspr(-) is a decreasingfunction of D/F, D must grow in fact less than proportionately to the increasein F, or even decrease, in order for (7') to hold while F increases. Using the notation previously introduced, it can be shown that A's chosen functionD(F, g is such that d log D/a log F = (x - I)/x < I. Comparing this to the results found when discussing (6) above, we find that the slopesof the reaction functions of A and B must be of opposite signs when they intersect (for x <2). The function DfF,g is represented in figure 4.1 as the DD line. Hence, in the most plausible case, the "no gift" In figure 4.1, we restrict the analysis to the equilibrium is such that an increase in the case where the equilibrium point lies above the F government budget, broughtabout for example by = (I/b) D locus, i.e. where the DD and FF foreign aid, will in fact result in the build up of intersect in the zone where x < I. The other case forces, F for the opponent, and D for the can be analyzed analogously, with no special government. This equilibrium thus supports the difficulty. view expressed by Deger and Sen (1992) that aid Now, if for some reason, like an increase in may well end up funding indirectly some increase foreign aid or any other financial support, T is in useless military expenditures.' increased. then the DD curve shifts rightwards, The Pareto-inefficiency of this Nash while the FFcurve shifts upwards. It follows that equilibrium can be seen graphically by drawing a ray through the origin and E. Then, moving account. In this case. using (2) and (6), one finds: inwardsalong this ray, by reducing proportionately D and F, would leave the political equilibrium un- changed, with p(-) constant, while leaving more resources to be shared by the hvo players for consumption purposes. The following section Therefore, a positive term has been added to - shows that a government which adopts the role of p(-) U'(-)< 0, which is only taken into account in a Stackelberg leader may improve on the "no gift" the Cournot-Nash case. The term between square equilibrium, by pursuing a more peaceful policy, brackets is negative, and makes a positive using the instrumentof the gift to the opponent, in contribution to a W/a G as it has a tilinus sign addition to the tool of defense expenditures, in before it. Hence, the Stackelberg-leader order to try and remain in power. government values more the "gift" than the As a step in this direction, let us look at the Cournot-Nash one. Conversely, it discounts the impact on the "no gift" equilibrium of an marginal contribution of defense expenditures to exogenous increase in G. We know from (6) that its welfare by taking into account their positive the FFcurve shifts downwards. Regarding the DD impact on the opponent's forces. Using again (2) curve, its impact is like that of a cut in the budget and (6) we thus find: T. Hence, DD shifts to the left. Consequently, an exogenous increase in G leads normally to a cut in military spending by both the government and its ~pponent.~But it will in general entail a change in D/F, and hence inp(-), so that we cannot conclude Now, the term in square brackets is normally that it implies necessarily a Pareto-improvement. positive, at least in the case of figure 4.1, and it is subtracted from the term pf(-)I;(-)/F- I, which is 4. The Stackelberg-Leader Government only taken into account in the Cournot-Nash case. Therefore, the fact of taking account of the It is intuitively obvious that a Stackelberg-leader opponent's reactions to the government policy government is more likely than a Cournot-Nash (or tools entails a reduction in the value attached by Stackelberg-follower)government to use the "gift" the latter to defense expenditures, and an increase as a tool in a peace-keeping policy. By taking into in the value of the gift. However, although this fact account the negative impact of a gift on the level opens the way to the existence of a Stackelberg of the opponent's forces, the government now equilibrium with a positive level of the gift, this realizes that it can increase the probability of does not necessarily lead to it. We need a new remaining in power not only by increasing its condition to be fulfilled, in order to guarantee that defense expenditures, but by giving something to the change in weights given to D and G, as its challenger. It is thus in a position to trade off embodied in (9) and (lo), results in an actual move more gift for less defense expenditures. From a away from the "no gift" equilibrium. For this to social point of view, it is clear that the former is happen, there must exist a value of G > 0 such preferable to the latter, as it is an argument in B's that a W/a G as given in (9) is 2 0. Intuitively, this utility function, whereas D only affects the requires that the negative impact on F must be probability of staying in power, which can be strongenough, and with a strong enough impact on regarded, roughly speaking, as a distributional the probability of staying in power, to offset the issue. The aim of the present section is to give utility loss of giving away G to the opponent some precise analysis of this intuitive argument.' instead of consuming it. Some tedious First, one can check that the marginal calculations show that this occurs if 0 2c(2- contribution of G to the government welfare W is 3,where 0 = pf(D/F) D/Fp(-), and c = (T-G) enhanced when the impact on F is taken into U'(-)/U(-) > 0. In other words, the existence of a "positive- aid. T is increased, we cannot be sure that this gift" Stackelberg equilibrium is more likely, the results in an increase in D and F like in the "no more responsive to the gift is the opponent's level gift" equilibrium. Now, the FF curves shift of forces, the more sensitive is the probability of upwards, and the WWcurves shift upwards and to remaining in power to a change in the balance of the right. The result on point S is thus ambiguous. forces. and the less elastic is the government's Hence dealing with a Stackelberg-leader utility function to a cut in consumption. government rather than with a Cournot-Nash one Figure 4.2 illustrates such an equilibrium. The is a necessary, but not a sufficient, condition for DD and FF curves are drawn as in figure 4.1, and aid to be used in a peaceful way. The appendix representthe Cournot-Nash reaction curves of the gives a more precise discussion. two players. The YFV curve is an indifference curve of the government; it is horizontal when it 5. Conclusion cuts the DD curve, by definition of the latter. Then, as the FF curve has a positive slope at E, it follows In this paper, we have analyzed a simple game- that a Stackelberg government gets more welfare theoretic model to highlight the choice of the than it would get in the "no gift" equilibrium E. government between raising its defense Then, provided the condition spelt out above for expenditures or giving away some "gifts" to his the existence of a "positive-gift" equilibrium is opponents, as a means to defend his position in satisfied, then it should look like point S in figure power. If the government is a Cournot-Nash 4.2. WrW' is a government indifference curve, player, then there is no gift in equilibrium. This corresponding to a higher welfare level than E, and "no gift" equilibrium is quite inefficient, as it is it is tangent to the curve FrF',a reaction function possible to reduce proportionately the military of Brs, corresponding to a higher level of G than expenditures of the two players without changing FF. Provided that E and Sare located above the F the political balance, while leaving more resources =(I/b)D locus (i.e.x < I), this entails a lower level to be shared between the players for consumption of D. purposes. This equilibrium has as well the unpleasant property that any increase in the budget, brought about for example by an increase in aid, will lead to more inefficient defense Figure 4.2 expenditures by both players. However, if the government is a Stackelberg- leader, then he will use the "gift" as a tool in his policy for staying in power. Hence this model illustrates how the use of redistribution as a tool in a peace-keeping policy can be related to some behavioral characteristics, or some attitudes, of the government. It is certainly the mark of a more sophisticated government than a fairly primitive Cournot-Nash one. As in Azam (1993), it is thus shown here that some expenditures without any apparent economic value, or some forgone consumption opportunities. can in fact play a positive part in a politico-economic equilibrium, yLf3 D and hence have a positive economic return after all. Therefore, it can reasonably be said that a government may invest in s redistribution policy, Now, if for some reason. like an increase in whose returns are made of some enhanced chances of peace being maintained in the country. of redistribution on military expenditures depends However, it is not claimed that redistributive on various parameters, nor in the real world, where policies are able to ensure peace in all cases. This ideological or ethnical opposition are sometimes is not so in the model presented, where the impact too strong to result in anything but the outbreak of violence. Appendix We give here a more formal analysis of the Comparing the "pseudo-reduced-form" Stackelberg equilibrium than in section 4. The equations (A.3) and (A.4) to (7) or (T),one can see government seeks to solve the following problem: that the former may be derived f r o ~the ~ i latter by subtracting the term in square brackets. The latter is positive provided x < I. Therefore, except in pathological cases, we should get lower F and lowerD in the Stackelberg equilibrium than in the "no-gift" equilibrium. But, from (A.2) and (AS) From the first-order conditions we can write we can see that G depends only on Tand D/F, so the following equations: that a cut in D and Fdoes not entail ~lecessarilyan increase in G. F=p'U- [(I -x)pU'/VID/F], (A.3) D = Up'D/F - [(I -x)pU'/V'], (A-4) Acknowledgements Stimulating discussions with David Bevan, Paul Collier, and Jan Willem Gunning, as well as useful comments by an anonymous referee are also acknowledged, without implicating. Notes 1. I got this detail from a personal discussion with the Chief of Staff at the Ministry of Defense in Addis Ababa. 2. Notice that this effect would disappear if it was clear that aid would be cut if the government was overthrown. Then, in a sense, foreign aid is more likely to fuel the build up of forces the more "politically neutral"is the donor, in a "no gift" equilibrium. In this sense, ideologically-orientedbilateral aid might be more conducive to peace than neutral multilateral aid. 3. This raises the issue of political conditionality,analyzed by Palda (1993). 4. A more detailed mathematical analysis is provided in the appendix. Chapter 5 Demobilization and Insecurity A Study in the Economics of the Transition from War to Peace Paul Collier 1. Introduction 2, Demobilization and Micro-insecurity I Demobilization is an integral part of the transition There are two reasons to expect that demobiliza- I from war to peace in Africa's civil wars. Ethiopia tion might increase crime. Demobilized soldiers l and Uganda have both recently undertaken large are not placed into employment and so start their I scale demobilizations. In Ethiopia some 400,000 civilian life as unemployed. It is likely that the I soldiers of the former government were demobi- lack of an income source increases the propensity I I I lized during late 1991 and early 1992 and the to commit crimes. Additionally, soldiers tend to be I demobilization of part of the victorious army is unskilled except in the use of weapons and so I currently being planned. In Uganda 20,000 soldiers might have a comparative advantage in criminal were demobilized at the end of 1992, and there are activities. I plans to demobilize a further 30,000 shortly. This The skill composition of the Ugandan and paper does not provide a general review of the Ethiopian demobilized armies was somewhat I demobilization process, but focuses upon one different, reflecting their different origins. Of the aspect, the relationship between demobilization two, the composition of the Ethiopian army was and security. better suited for demobilization. The Ugandan The hypothesis to be investigated is that army was largely composed of those recruited as demobilization constitutes a potential threat to guerillas in rural areas during the war. Recruits security. The motivation for the study is that were typically very young at the time of recruit- insecurity is likely to have substantial economic ment, and had little education. The Ethiopian army consequences. Two types of insecurity are distin- of 400,000 which was demobilized during 1992, guished. Micro-insecurity is the fear that the was recruited by the former government, largely individual will be the victim of crime. Macro- by conscription, and so was closer to being a cross- insecurity is the fear that the state will be over- section of the male population. A survey of the thrown by insurrection. Micro-insecurity, it is characteristics of the Ugandan army just prior to hypothesized, does not just constitute a set of demobilization (MSEC, 1993a) found that over a involuntary transfer payments. but inhibits asset third had no education, while a majority had had formation and disrupts transactions. Macro-insecu- no economic activity prior to recruitment, and so rity may reduce the credibility of government had no work experience either as wage earners or commitments and the propensity to make irrevers- in self-employment. Nor had the army proved to be ible investments. If demobilization reduces secu- a training ground for non-military slcills: only 3% rity, there is therefore a trade-off between the had acquired such skills during their time as direct economic gains from reducing a component soldiers. In both countries most soldliers had been of government expenditure and the indirect eco- in the army for several years and so their connec- nomic costs. Potentially, demobilization affects tions with their original areas had weakened. This both micro-insecurityand macro-insecurityand the reduced their access to land and possibly also their paper considers them in turn. interest in farming as a career. However, most soldiers had acquired dependents, two thirds on crime and the synchronized nature of the de having families. Hence, the need for an income mobilization precluded time series econometric wasthat for a substantial household rather than just analysis. However, the spatial pattern of the demo- an individual. In both countries demobilized bilization provided the basis for a cross-section soldiers were not allowed to keep their weapons, approach. but guns were widely available. A demobilized Soldierswere demobilized to their districts of soldier who chose to become a criminal would origin and the Veterans Assistance Board kindly have had little difficulty in gaining access to a provided data on the numbersof demobilized sent weapon. to each district. We therefore compare the district In assessing the link between demobilization crime rates before and after the event using demo- and crime three approaches were used. First, bilization as an explanatory variable. Normally, interviews were conducted with senior police this would not be valid since demobilized soldiers officials and private security services in both might choose not to return to their districts, or countries. Second, data was gatheredon the district subsequently migrate from the initial place of level pattern of crime as reported to policestations demobilization. However, the government of before and after demobilization,and matched with Uganda went to considerable efforts to get soldiers the district-level incidence of demobilization. to return to and to stay in their originalhome areas. Third, a follow-up survey of demobilized soldiers All demobilized soldiers were transported to their and community leaders was analyzed. The two home areas and received there. Further, three latter approaches were only feasible for Uganda. installments of payments were made to bank In Ethiopia the policeconsidered that demobi- accounts which soldiers were required to open in lization had initially caused an increase in crime, their home areas: payments could only be made particularlyrural banditry. However, by early 1993 into these accounts. Migration therefore incurred the incidenceof crime was quite modest in almest the need to return to the home area to claim these all areas, comparingfavorablywith Kenya accord- payments. Finally, we have evidenceon the extent ing to the Police Commissioner. In Uganda the of subsequent migration from the follow-up sam- police perception was that crime had risen as a ple survey of 300 demobilizedsoldiers conducted consequenceof demobilization,and this was more during April and May 1993 (MSEC, 1993b). This emphatically the perception of the expatriate found that less than 2% of the demobilized had at security services. that stage migrated from the district to which they We now turn to the analysis of Ugandan had been transported. statistics on crime and demobilization. Crime Immediately prior to demobilization, a census statistics in Uganda are not computerized and can was conducted of those to be demobilized, estab- only be collated with reliability back to the begin- lishing their skills and prospects. As discussed ning of 1991. The CID kindly collated data on the above, skill levels outside the militarysphere were total numberof reported crimesfor 1991,1992 and very low. However, a potentially differentiating 1993, the last distinguishing between the first characteristic concerning prospects was access to quarter and the rest of the year, for each of the 38 land. Although most of those to be demobilized districtsof Uganda. The demobilizationtook place reported that they had access to land, some 12% in late December 1992 and January 1993. The claimed to be landless. Because this was a census national data on reported crime show that the rather than a sample, the data can reliably be used demobilization coincided with a decrease in the when disaggregated to the district level. Further, crime rate. During 1992 there were 20% more landlessness is concentrated: nine of the 38 dis- reported crimes than in 1991 (which may simply tricts account for virtually all of the landless reflect a higher rate of reporting as the country demobilized, so that the district-level incidence of grows more accustomed to civilian government). the demobilization of the landless is very different Yet during the first quarter of 1993, immediately from that ofthose with land. Hence, we distinguish following the demobilization, crime fell by 7% in the subsequent analysis between the landless compared with the fourth quarter of 1992. While demobilized and other demobilized. this is suggestive, the lack of adequatetime series Both crimeand demobilization by district were We distinguish between the impact effect of expressed as rates relative to the district population demobilization and its longer-term effect. The as given in the 1991 population census.'There is impact effect is that in the first quarter of 1993, substantialvariation in both the crime rate and the immediately after the demobilization. During this demobilization rate. On average, during the first period the demobilized were in receipt of modest quarter of 1993 there were 0.56 reported crimes lump sum pay-offs. The longer term effect is per thousand inhabitants, with a range between gauged by the remaining three quarters of 1993 districtsfrom 1.20 to 0.11 and a standard deviation (the most recent data point). During this period of 0.31. The incidence of demobilization was on pay-offs had ceased, but the demobilized had by average 1.30 soldiers per thousand inhabitants, then had greater opportunity to generate incomes with a range between 3.59 and 0.05 and a standard from agricultural self-employmentor from finding deviation of 1.OO (reflecting the previous regional other non-criminal sources. skew in recruitment). Hence, there is sufficient The dependent variables were therefore the variation in both rates for a test of the relationship crime rates during the first quarter of 1993 and the between demobilization and crime by means of remaining three quarters of 1993. These were cross-section data to have some credence. If one regressed on the crime rates for 1991 and 1992, tenth of the soldiers were involved in crime then thereby controlling for fixed effects and trends, its incidence would rise by around a quarter on and the incidence of demobilization,distinguishing average, more than doubling in the district with the between the landless and others. The results are highest incidence of demobilization and barely reported in Table 5.1. changing in the region with the lowest incidence. Table 5.1. Regression Results for Crime and Demobilization Dependent variables Crime93(1) Crime93(2,3,4) coeff. se t coeff. se t constant Crime91 Crime92 Landless Others Crime91 = number of crimes in district during 199l/population of district in 1991; sources: CID and 1991 Population Census. Crime92 = number of crimes in district during 1992/population of district in 1991; sources as above. Crime93(1) = number of crimes in district during the first quarter of 1993lpopulation in 1991; sources as above. Crime93(2,3,4) = number of crimes in district during the second, third and fourth quarters of 1993lpopulation of district in 1991; sources as above. Landless = number of those demobilized into the district in December 1992 who reported no access to landlpopulation of the district in 1991; sources: MSEC (1993b) Table 12, and population census. Others = number of those demobilized into the district in December 1992 who reported access to land/population of the district in 1991;sources: MSEC (1993b) Table 12, and population census. The results reveal that in the short term there was because the demobilization of those with is a significant relationship between demobiliza- access to land, who were the large majority, signif- tion and the crime rate. The demobilization of the icantly reduced crime. Our discussion of the landless significantly increases crime. Applying sample survey of community leaders, below, will the coefficient in the Table, the 2,346 landless shed some light on why this might be the case. demobilized accounted for 140 crimes, an inci- Applying the coefficient reported in the Table, the dence of criminality of 6%. This incidence is one 17,668 demobilized with access to land reduced hundred times higher than that for the Ugandan the number of crimes by 1,060. Hence, the net population as a whole during the period. However, effect of demobilization is predicted to have been the short term effect of the Ugandan demobiliza- a short run reduction in crime of 920. Between tion was not to increase crime but to reduce it. This 1991 and 1993 (the only period for which there are data) there was an increase in reported crime regression results. The low involvement in crime averaging 2.8% per quarter, implying an expected was corroborated by the evidence of the survey on increase between the fourth quarter of 1992 and the economic reintegration of the demobilized. A the first quarter of 1993 of 303 crimes. Instead, substantial majority had been taken up peasant there was a reduction in crime of 757. The shortfall agriculture, with a further 20% engaged in small of actual crime on expected crime of 1060 is non-agricultural enterprises. Virtuaiiy none had remarkably close to the predicted reduction due to entered the labor market, reflecting both the small demobilization of 920. Hence, the coefficients size of that market in Uganda, and the very low derived from cross-section data on district-level skill and education levels of most of the demobi- variations in crime cohere well with the observed lized. Because the survey was conducted only coincidence of demobilization a sharp drop in the some four months after demobilization, it bears aggregate crime rate. most directly upon the regression results for the During the first quarter of 1993, the consensus first quarter of 1993. Interestingly, the consultants in the expatriate security community was that concluded from their experience that whereas demobilization had already increased crime and overall the demobilization had progressed success- that this would get much worse once the demobi- fully, assistance should have been targeted upon lized had exhausted their pay-offs.' However, the landless. This is consistent with our short-run according to our data, both of the above effects of results which suggest that a minority of the land- demobilization proved to be ephemeral. The crime less faced unusual pressures which through crime rate for the remainder of 1993 was unrelated to inflicted social costs upon the local population. either type of demobilization, both the coefficients However, these negative effects appear to have and the t-statistics being negligible. Hence, what been temporary so that the impact upon criminality appears to have happened is that the landless, after would not justify major public expenditure. The some initial desperation, settled into their commu- costs of crime in Uganda are not known. However, nities, while the landed majority rapidly lost the to give an illustrative calculation, were crime to military attributes which had provided a deterrence cost 5% of GDP then a reduction of one reported to crime in their localities. Both groups soon crime would be worth around $6,000. Hence, were became normal civilians. it possible to prevent the excess criminality of the We now turn from the regression analysis to a demobilized landless, the reduction of 140 crimes third data source, a sample survey of demobilized would be worth $840,000, justifying per capita soldiers and community leaders conducted in May expenditure targeted on the landless of around 1993 by a consultancy company (MSEC, 1993b). $360. Through interviews with a sample of 300 demobi- The conclusion from theabove evidence is that lized soldiers this survey investigated the initial the Ugandan demobilization has not to date given economic activities of the demobilized. Addition- rise to significant micro-insecurity. The Ugandan ally, through interviews with community leaders, demobilization was still in its early stages, but the the survey established the number of cases of absence of any relationship between demobiliza- crimes in which demobilized soldiers had been tion and the crime rate in the subsequent 4-12 convicted and the perceptions of the community months is certainly encouraging. By early in this leaders as to the impact of demobilization on period the modest pay-offs would have been security. The survey found that the rate of crime exhausted, and if the demobilized were at even this identified with convictions of demobilized soldiers stage sufficiently established to maintain them- was very low, only 0.2% of soldiers being in- selves without atypical resort to criminality, it volved. Further. this was the perception of commu- would seem unlikely that there would be a later nity leaders. Community leaders reported either deterioration. Recall that in Ethiopia, where there that demobilization had not affected the security is a longer post-demobilization experience, the situation or that it had been beneficial as a result of police perception was that although the security the presence of those with military experience situation initially deteriorated it then improved. acting as a deterrent, this being consistent with our 3. Demobilization and Macro-insecurity Toconclude, theconcern about demobilization weakening the securityof the state must be taken The consequences of demobilization for macro- seriously.The relationship between disaffection of insecurity are more difficult to observe. The the demobilized and the scope for insurrection possibility that the state will be overthrown pointsto the importanceof successful reintegration through insurrection is a genuine concern in both into the economy. Policy should therefore be countries. The fear is not fanciful because in both designedso asto achievea cost-effectivereduction such an event has recently happened, indeed, in in disaffection. In Uganda this was part of the Uganda it has happened repeatedly.Rather likethe rationale for the establishment of the Veterans Pesoeffect, this type of fear is unobserved and it is Assistance Board. We now turn to the design of a hard to find evidence for it, let alone quantify it. policy with this objective. An example which demonstrates the pervasive To date, in both countries poIicies designed to presence of a sense of macro-insecurity is there- assist the demobilized have combined thinly fore appropriate. Recently, a Ugandan District spread financial and material assistance to all the Commissioner was observed to be leaving his demobilized with small job creationschemes from home in a hasty manner. The response among which a few have benefited substantially. Experi- observers was panic and flight, because they ence with the latter in both Uganda and Ethiopia interpreted the behavior as indicating theapproach has been very poor. Generally, the jobs created of rebels. In fact, the Commissionerhad just heard have been at high cost, newly established enter- that the President was paying a surprise visit.3If prises have not been viable without continued private agents attach non-negligible probabilities subsidy, and only a trivial number of jobs have to the overthrow of the state then this is likely to been generated. A hypothesisof the study was that reduce the propensity to make irreversible invest- theactivity with the most scope for employing the ments, and the credibilityof governmentcommit- demobilized was construction. However, inter- ments. It is therefore not a phenomenon to be views with construction firms and with experts on dismissed. labor-intensive public works established that this Demobilizationmight affect macro-insecurity was not the case. In urban areas, constructionwas through two routes. First, demobilized soldiers fairlyskill-intensive.Firms wished to recruit either poorly integrated into the economy and therefore those with experience or those with some educa- disaffected, constitutea pool of potential recruits tion. Although rural construction could be made for a subsequent military challenge to the state. In highly labor-intensive, it was not well-suited to the both countriesthere is evidence for a link between employmentof the demobilized. The technique of demobilization and insurrection. In Ethiopia, the rural infrastructure building using labor intensive demobilization is believed by the authorities to methods involved the deployment of a small core havecontributed to the insurrection by the Oromo of skilled, well-paid labor, which travelled with the Liberation Front during mid-1992, which at one project (for example, a road) and otherwise relied point involved a substantial pitched battle. In upon the recruitment of unskilled labor from the Uganda, insurrections fromthe Northare known to locality. The advantage of this was that the un- involve former soldiers of Amin's army. The skilled labor had its own accommodation and second route by which demobilizationmayweaken sources of food. The use of demobilized labor macro-securityis that beyond a point, the smaller would have necessitated travelling labor gangs is the army the more prone is the state to be chal- which would have considerably increased the lenged. Although there is clearly some size below expense. which thearmycannot be reduced withoutencoun- Not only are employment-creation projects tering this danger, the size of army needed for the high cost, they are unnecessary. As noted above, victorious insurrections which both achieved is the follow-up survey in Uganda found that the evidentlyconsiderably larger than that needed for majority had successfully integrated into peasant deterrence, hence, this should not be a significant agriculture or non-farm self-employment. The consideration in the planned demobilizations. modest generalizedassistancein the form of seeds. agricultural implements and roofing sheets for housing, was appreciated. although since delivery deliver are liable to be counter-productive. While mechanisms had frequently proved inadequate it employment-creation schemes are costly and was a source of friction. Presumably because of the ineffective, it is appropriate to target additional successful integration, the survey also found a support to the landless in land scarce areas, and to largely positive attitude on the part of those demo- arrange long term support for the severely disabled bilized to the demobilization experience. The two which is likelyto mean retaining them in army-run main exceptions were landlessness and the treat- facilities. ment of the disabled. Because of its localized nature, landlessness is predictable and amenable to 4. Conclusion policy intervention. For example, the government could have a lower rate of demobilization for Contrary to the prior hypothesis of the study, we soldiers from land-scarce districts, or land might have found that demobilization, at least as con- be purchased for them. The problem of disabilityis ducted in Uganda, does not lead to a significant substantial in the armies of Uganda and Ethiopia. upsurge in insecurity. Indeed, we have found that No special provisions were made in the Ugandan in the short term demobilization significantly demobilization for the disabled and this has caused reduces crime unless the demobilized lack access intense resentment, not just among the disabled, to land. The demobilized, if returned to their home through the perception that the government has areas and given some assistance, are with identifi- discarded onto communities incapable of handling able exceptions able to find income-earning oppor- the problem those who made the greatest sacrifices tunities. As this perception is transmitted from the in its cause. The government in Ethiopia faced first wave of demobilized, subsequent demobiliza- with the same problem, is planning special facili- tion becomes easier as more soldiers volunteer (as ties for this group. To conclude, the design of a is currently the case in Uganda). Conversely, the policy to minimize disaffection has some rationale attempt to link demobilization to special employ- in macro-security over-and-above its direct bene- ment schemes appears to be ill-advised. While fits to the recipients. Generalized assistance seems there may be a good case for expanding the con- to be needed only on a modest scale, and promises struction sector, this should be delinked from the beyond the competence of the government to demobilization exercise. Notes 1.Source: Government of Uganda (1993), Table 18. 2. I am indebted to the UNDP for an opportunity to attend a regular meeting of the heads of expatriate security services and for their cooperation in answering my questions. 3. The incident was reported in the New Vision, for June 21, 1993. Chapter 6 Food Markets, Liberalization and Peace in Ethiopia: an econometric analysis Stefan Derconl 1. Introduction During the 1980s Ethiopia has gone through 8,000 in towns (Holmberg (1977)). Regional turbulent times. Politically, the government of markets may well have been reasonably integrated President Mengistu faced persistent armed during that period, with food prices even during secessionist activity, especially in the Northern the 1973-74 famine period not rising by more than regions of Eritrea and Tigray. It culminated in May 20 percent (Holmberg (1977))3. With price 199 1 in the take-over by a Tigrayan-led coalition controls and licensing of trade came also the rule of guerrilla movements which form the present that traders would have to supply a percentage of Transitional Government, and the secession of their grain turnover (usually 50 percent) to the Eritrea. Economically, stringent controls on trade AMC at the quota price plus a margin of 4 to 5 birr and production repressed both rural and urban - leaving the traders' profits considerably lower areas, forcing the government to start a reform than for grain sold as part of arbitrage between program in 1990, which was continued after the districts and regions. Enforcement turned quickly accession to power of the rebels. In 1992, an IMF- brutal: in 1976 traders were executed for allegedly sponsored structural adjustment program was hoarding and evading price controls. Since then started. and throughout the 1980s, numerous cases of Food production and marketing has been at the forced requisition of produce, and execution and foreground of both the crisis and the reforms'. torture of traders have been documented. Land reform after the 1974 revolution provided for especially in the war zones (Africa Watch (1991)). many rural households for the first time a hopeful Licenses were randomly taken away, and all trade break with centuries old feudal system. Around the of any scale was banned from 1982183 in Gojam same time, initially with World Bank support, a and from the beginning of the harvest in 1985 in new agency was set up in 1976 to participate in Arssi. Movement of grain up 100 kg was usually grain trading and sufficiently strong to influence allowed in principle, but grain could always be grain price formation. the Agricultural Marketing requisitioned if a specific trading purpose was Corporation (AMC). Traders had to obtain licenses suspected. In 1988 it was announced that as long as and price controls were established. Pan-territorial 50 percent of the grain would be delivered to the pricing was used from 1980181 onwards. From that AMC, traders would be issued with licenses in all marketing year onwards. farmers were required to areas. but the implementation bf this ruling was deliver between 50 to 100 percent of their reportedly quite different across regions (Franzel production as a quota to the AMC at fixed prices. et al., 1989). well below the market prices. In a tragic irony, No estimates of the number of traders which peasants only recently freed from a demanding remained active are available, but their number is feudal system of fixed grain deliveries to landlords likely to have diminished considerably in the were subjected to a new exploitative system of 1980s. At the same time, beyond the measures forced deliveries, executed by the state. specifically aimed against traders, the war resulted Before these reforms, up to 25,000 grain in a substantial destruction of infrastructure, merchants had been active in rural areas and up to especially in the North. The intensification of the fighting from around 1988 limited further the delivery of a fixed quota will be discussed. Since ability to move grain, whether legitimately or it has been dealt with by others (e.g. Lemma Merid smuggled. The controlled distribution system (1986), Alemayehu Lirenso (1987), Franzel et al. could however increasingly less supply sufficient (1989), Legesse Dadi et al. (1992), Azam (1992)), grain to consumers at the controlled prices, partly we can be brief about this. Azam has shown that if due to a large diversion of grain to the military the household consumes both manufacturinggoods operations, and urban areas became more and an own-produced cereal, then under general dependent on the free market supplies. With the conditions a forced delivery of cereals at a price, increased politicaland military pressures placed on lower than the market price, can be viewed as a the government to salvage the regime after the lump-sum tax equal to the margin between market gradual demise of its main sponsors, the Soviet and quota pricestimes the quota, so that the impact Union, a liberalization of grain markets was of the quota system on the net sale by the farmer commenced in March 1990, which established the works only through an income effect. freedom to move grain, and quota's were Consequently, the net supply of grain is an abolished. With the demise of the Mengistu regime increasing function of the level of the quota and a in May 1991, the main remaining monopoly decreasing function of the quota price. sources of grain for the AMC, the producer co- Liberalization should then have reduced the net operativesand state f m s , began to be dismantled. supply of grain to the market, since on-farm The AMC is now supposed to compete with consumption will have increased, provided both private traders in the free market. cereals and other goods are normal. This paper deals with the consequences for Azam (1992)'s analysis deals only with the food markets and prices of the changes from a total net supply by the farmer, irrespective of controlled system to free trade, from war to peace. whether it went to the open market or the AMC. In particular,we want to establish, first, what were To derive the supply to the open grain markets, we the consequences for the level of consumerand should focus on the supply, net of the quota. Let H producer prices and for the marketing margin be the quantity of grain harvested and S, the net between surplus and deficit prices and secondly, sale of grain. is equal to H minus C, with C the on- can we concludeanything aboutthe functioningof farm consumption.The quota Q has to be delivered food markets, ie. has there been any change in the at a price Po, which is below the market price P. extent of market integration before and after the With other goods consumed denoted by M, we can changes? write the budget constraint faced as Methodologically, the first question will be addressed using a theoretical arbitrage model (section 2), after which data on the main staple, in which the right hand side is total income Y and teff, will be used to infer whether the predictions (P-PQ).Q= T(P,PQ,Q),the lump-sum tax T. The net of the model can be observed (section 3). These sale to the open market, F, equals S minus Q. The results will be further checked using an effect of a marginal increase in Q on F is positive econometric approach which will also be used to if address the issue of the functioning of markets (section 5). Before that, in section 4, a specific econometric methodologyis developed to analyze market liberalization, usingcointegration and error Using (I), this will be the case if correction mechanisms. This methodology tries to improve upon the classic Ravallion (1986) model for market integration. 2. The effects of the quota system, trade controls and the war on grain prices in which E, and b, is the income elasticity of grain First, the effect on grain producers of the forced and tlie budget share of grain in total expenditure. regional autarky the market price p, is srnaller than Since b, is smaller than one and since the right p,, in other words there is a larger demand relative hand side of the inequality is larger than one, the to supply in j than in i. For simplicity, let us call i income elasticity of grain has to be'sufficiently a 'surplus' region and j a 'deficit' area, even though larger than one for an increase in the quota to this may misleadingly and wrongly suggest the result in an increase the supply to the free market. presence of rationing. If the transport cost to move In other words, under usual circumstances, the one unit from i to j is t , and constant, then quota system will have repressed the supply of arbitrage with competitive markets suggests that grain to the open market. Liberalization should the margin between p, and p, will be reduced to tij have resulted in larger supplies reaching the open via exports from i to j. As discussed before, under market, despite the reduction in total marketed quite acceptable circumstances, the quota system production. will have reduced the net supply to the free market The second issue has to do with urban moving both p, and p, up while the supply of consumer demand. During the 1980s the AMC rations in the main consumer markets will have sold substantial amounts of cereals in Addis Ababa reduced prices both in i and j. Because of the at low, controlled prices via Kebele shops, even possibility of arbitrage, this effect is independent though most people had to supplement the of whether rations where distributed in all regions quantities obtained with purchases in the open or only in particular areas such as Addis Ababa, market. After liberalization, the quantities sold by although this would have affected tlie actual levels the AMC in this way barely reached 10 percent of of prices relative to the free market situation. the previous quantities. In a simplified way, the The introduction of restrictions on traders will effect of the change in this 'ration' can be seen in affect the margins between i and j. Suppose a figure 1. Suppose the household consumes cereals, trader can only obtain a license if a fraction q of all denoted C, and other goods, denoted NC. A ration purchases are delivered to the AMC at a price p,, R, of cereals is supplied at a lower price than the which is considerably less than the price in j. If the market cereals price, so that the budget curve is transport costs related to the delivery to the AMC kinked. If this ration is not sufficient, then some are ti, then arbitrage between i and j will take place extra cereals will be purchased to reach the as long as optimum at A, where total cereal consumption is C, and other consumption is NC,. A reduction of the ration to R?will result in an inward movement Reworking this expression, we find that the margin of the budget curve. This is a pure income effect, between prices in i and j will reduce after arbitrage so that as long as the other goods. NC, are a to composite normal good, consumption of both cereals and other goods will decline, to C2 and NC,, at the optimum B. The important point is that It implies that, under the obvious circumstance that the reduction in cereal consumption is smaller than the difference between the price in jand the price the reduction in the ration, in other words, the paid by the AMC to traders is larger than the reduction in the ration will have increased the difference in transport costs related to delivery in demand for cereals in the open market. The net j and to the AMC, the margin betwken prices in the effect from the liberalization on market prices is region i and j will have increased after the traders' consequently not very clear: increased open market restrictions. Prices in 'surplus' regions will have supply, but increased demand as well. This experienced a downward pressure and prices in conclusion ignores however the effects on trade of 'deficit' areas are likely to have become higher as the liberalization and the spatial distribution of a consequence. Trade will be discouraged and these effects. The basic spatial arbitrage model as margins will not be brought down to transport in Takayama and Judge (1974) can be used to costs. discuss this further. Liberalization could have a further effect on Let us consider two regions i and j with marketing margins if licensing limited the number differing demand and supply and assume that in of traders participating in moving grain before the reforms. Costs of licenses were reportedly high prices in 'surplus' areas and results in an upward before the reforms. Some routes may have been movement of prices in the main demand arias. If controlled by very few traders, who could pay the the return to peace also means that both security licensing costs andlor the necessary bribes. To the and the basic transport infrastructure is restored extent that new traders enter food trading after then peace can have further beneficial effects on liberalization, the rents to these traders in the form producers in the main surplus areas as well as on of supernormal profits could be eroded through consumer prices in urban areas. increased competition. The result would be that marketing margins move close to the actual 3. Teff Prices and Margins in Ethiopia marketing costs. Observing reduced marketing margins could therefore both be related to the In this section a discussion is presented of the removal of trade taxes via quota's and to increased evolution of price and marketing margins since the competition after liberalization. 1980s. The analysis focuses on teff, the main Finally, it is possible to discuss the effects of staple in the Ethiopian highlands. This valuable war and insecurity on cereal markets. Besides the and relatively drought resistant but very labor physical losses of cereal production in war zones, intensive cereal crop forms the basis for most the inability to move grain because the breakdown meals in rural and urban Ethiopia. Monthly data on of infrastructure and security risks will affect open market prices in urban areas have been prices considerably. Introducing these constraints collected systematically by the Agricultural on movement we can rewrite the arbitrage model Marketing Corporation for a few dozen markets under trader quota's as the maximization by traders starting in the mid-1980s. The Relief and of all profits x on moving quantities x from i to j Rehabilitation Commission (RRC) has also subject to a movement constraint, or collected data in up to 80 towns as part of its famine early warning system since the early 1980s. Both data sources are used in this section. The Central Statistical Authority (CSA) has published data on rural and urban prices, but important gaps s.2. x.. =AT..) exist in the published data, so this source was not rl !I used in the analysis. Some discrepancies existed T.. i T,;. between the RRC and the AMC data, but since the !I AMC data contained systematically fewer gaps, the latter source was more relied upon. All the data in which n is total number of markets and T is the were deflated by a non-food Consumer Price Index transport available and T' denotes the constraint on (CPI) obtained by reweighing the Addis Ababa moving grain because of the insecurity and the CPI excluding all food using CSA data. By no breakdown of infrastructure. Traded quantities x,j means is this a correct deflator for all the different have to be a function f of the available transport. towns, at least it will give us some indication of Denoting ltijas the shadow price of the transport the evolution of food prices relative to other infrastructure we can find that traders will move commodities. Unlessothenvise stated, prices are in grain between i and j until real January 1988 prices per quintal, and data were available for some of the main markets until September 1993. First, figure 2 shows the evolution of prices for white teff in Addis Ababa in real terms since If the transport constraint is binding, this implies August 1979. No clear trend can be observed for that the margin after arbitrage trading will be the entire period. The most striking relative price larger than simply transport costs corrected for the change occurred during the famine period 1984-85. government restrictions on traders. If insecurity is Starting in February 1984 after the Meher harvest, a strong element then this shadow price can be peaking in the middle of 1985, after which prices very large. War has therefore a negative effect on only slowly declined to reach pre-1984 prices after the Belg harvest of 1986. The effects of the famine Consistent data for most other towns in in Wollo and Tigray must clearly have been felt in Ethiopia are not available for the first part of the Addis Ababa as well. Low prices were then 1980s. Figure 3 and 4 show data since September observed during 1987 to 1989, with the post-Meher 1987. which allow us to build up a more complete harvest prices in 1989190 close to the lowest in the picture of the consequences of liberalization and entire period. Liberalization occurred in March peace. First. it can be seen that the price in Addis 1990, but no clear effect can be observed. except Ababa has consistently remained higher than in for an apparent seasonal effect. The last few most of the other towns before and after months of the war showed a rapid increase of liberalization: Addis Ababa clearly exercises a prices in the beginning of 1991, despite a post- considerable demand-pull to grain supplies from harvest period. Insecurity must have affected the virtually all directions. Only in Dire Dawa are distribution of the 1990191 Meher harvest price higher. a town which is quite far removed considerably. Since then, prices have declined from any supplying regions. In fact, from the gradually, even though prices were around the middle of 1993, prices in some towns East and beginning of the 1992193 Meher harvest still Southeast of Addis Ababa (Debre Zeit, Nazreth, higher than during the period 1987-90, but not Shashemene) became (probably temporarily) higher than in the pre-1984 period. During 1993 higher than in Addis Ababa, suggesting even a teff prices declined however further, to levels lower flow, away from Addis Ababa towards Dire Dawa. than at any point in the last 14 years, after a In general however, margins between Addis Ababa successful Belg harvest. Even the official data are and the rest of these important urban centers are not yet available, prices are said to have positive. Secondly, a clear effect of the marginally declined further, but are at present liberalization can be seen: margins between the (February 1994) increasing again, after a poor most important demand center, Addis Ababa and Meher harvest and expectations of a poor Belg all the other towns decreased after liberalization, as with rains being too late. From these data, it is not was predicted by the model. A clear increase in the possible to find therefore a clear result from the grain flows as part of arbitrage must have taken liberalization. In any case, some time after the place. By the end of the period, prices in some EPRDF take-over, prices have declined to towns were higher than before liberalization: e.g. historically low levels. Open market supply Debre Markos, Ambo, Hosaenna, Shashemene in - increases seem to have been able to more than fact these are close to surplus teff producing areas; compensate the demand increases in the open in others, such as Addis Ababa and Dire Dawa - market in Addis Ababa. Filtering the effects of a important deficit and demand centers, prices were (delayed) response to liberalization. a genuine lower. These results are consistent with the response to peace and security or the effect of very predictions of increased arbitrage after good Meher and Belg harvest, is quite impossible. liberalization in the previous section: more grain The fact that a good harvest can flood the Addis will shipped from surplus regions to deficit regions Ababa market and does not get wasted in the after liberalization, since incentives for arbitrage feeding of several large armies is likely to be have increased. another factor. Table 6.1: Margins between Addis Ababa Real Teff Prices and specific markets1 mid-86lend-89 mid-901Sep.93 mid-91lSep.93 result post-liberalization Shoa Nazreth 18 10 10 down Debre Zeit 12 3 2 down Wolliso 31 28 27 equal Ambo down Hosaenna down Ziway down Shashemene 31 12 10 down Kefa Jimma 32 31 28 equal Gojjam Bahar Dar* 45 32 30 down Debre Markos 59 26 26 down Arssi Assela* 37 17 17 down Dessie Kombolcha 28 7 7 down Dessie* -0 9 1 1 UP Gonder Gonder* Harmghe Harar -33 -26 -28 down? Dire Dawa -26 -22 -21 down? Ifubabor Mettu* 35 36 39 equal? Weffega Nekempte* 42 32 32 down Source: calculated from AMC data. 'towns are ordered according to regions as they were in use until the end of the 1980s. *data up to the December 1992 only. To interpret these results better. the margins obtained using the (longer) RRC data on 51 market were calculated and filtered for temporary erratic relations. Liberalization has had somewhat more movements using six-months averages. These mixed effects within the regions. even though averages were used to obtain an average margin mainly reducing the differentials. The take-over before and after liberalization, and after the did not affect the patterns set by liberalization. EPRDF take-over. Some of the results were To summarize the main effects of the sensitive to which data source was used. but liberalization, it was shown that the level of prices overall the picture was consistent. Table 1 gives after liberalization in main 'deficit' areas - big the difference between real teff prices in Addis urban centerssuch as Addis Ababa and Dire Dawa Ababa and in some of the main urban centers in - was, if anything, lower than before. Producers (or Ethiopia for which data were available. Margins at least traders) in the main surplus areasare likely declined in Shoa in six out of seven markets by 30 to have benefited considerably, since generally percentor more. In other regions, margins declined price differentials between trading areas were in 7 out of 11 markets mentioned, and increased brought down by arbitrage to lower levels than only in Dessie and Gonder. The decline is before in a majority of instances. Increased especially notable in Arssi and in Gojjam. the two reliance on open market supplies by urban big producing areas which faced banned private consumers has been more than compensated by trade before liberalization.The result for Dessie is increasedopen market supplies, and especially by mainly the consequence of the (erratic) effects of better incentives to trade. We cannot find clear food aid distribution in that area during the period evidence on the existence of a peace dividend: since the famine, depressing prices considerably in lately, prices have declined in the major consumer some periods.For the interpretation of these results areas, but only after a strong price increase around one also ought to keep in mind the flows of teff the ascent of the new government. No evidence within the country as depicted in figure 16, which can be found of increased arbitrage in food means that some of the markets for which markets: increasedsecurityshould have meant that differences with Addis Ababa prices are given are the constraints on moving commodities should only very indirectly linked with Addis Ababa.Still, have been reduced, and margins to have come the result is striking: liberalizationclearly reduced further down due to the reduction of the shadow price differentialsin the country, moving all prices premium on transporting. One should however closer to Addis Ababa levels. This result is notethat most data presented in the tables are from confirmed by using data on 44 more peripheral southern or central regions: data from the most marketsfor which the RRC collected data, with 38 affected war-zones, such as Tigray, Northern showingdeclines after liberalization. No effect can Gonder and Eritrea, are not available for the however be observed from the cessation of the obvious reason that a war was going on. war: in table 1 is no significantdifference between the post-take-over period and the rest of the 4. Analyzing market integration and the liberalization period. Relative security in the effects of liberalization and peace country has failed to deliver its dividend in the form reduced price differentialsin the country. Thus far we have analyzed the effects of Table 6.2 gives data on the price differentials liberalizationand peace on the levels of prices, but between some towns which may potentially trade one of the implicit aims of liberalization has to do with each other due to proximity. Of the 21 with the improvedjirncrioningof markets. In some presented, the pricedifferentialwent down in 10of respects, without sliding into a semantic the comparisonsand remained roughly the same in discussion,the analysis of the changes in margins 6 others before and after liberalization. In 5 cases across towns has to do with increased arbitrage. did the differentialgo up. A comparableresultwas Table 6.2. Price differential among regional towns in Ethiopia mid-86lend- mid-90lend- mid- result post- 89 92 91lend-92 liberalization Shashemene - Hoseanna equal Arba Minch - Hosaenna down Nazret - Asela down Dire Dawa - Nazret down Nazret - Harar down Debre Zeit - Navet equal Debre Zeit - Ziway down Debre Markos - Dangila UP Addis Zemen - Gonder down Gonder - Kola Diba equal Addis Zemen - Nefas Mechwa equal Bahar Dar - Mekane Selam UP Kombolcha - Dessie down Dessie - Haik UP Dessie - Woldiya down Dessie - Korem down Harar - Dire Dawa equal Bahar Dar - Gonder down Bahar Dar - Debre Markos down Assela - Robe UP Gimbi - Metu equal However, the effect was predicted from a integration can just as well be interpreted as static model of removal of implicit taxation on evidence for a competitive model as for an traders, and not from any effect of the increased oligopolistic model with base-point pricing. competitive forces within the market or a faster Market integration analysis provides however an transmission of information and response to interesting indication of the extent to which and trading opportunities. It is unlikely that the speed with which price signals in one market competitiveness of a market can be analyzed with are transmitted to another market. If market anything but detailed microeconomic information liberalizing measures result in an improvement of of traders and their cost structure and constraints4. this transmission and therefore increasing market Standard market integrationanalysis such as in integration, then a positive effect on the Ravallion (1986), using only information on functioning of markets from liberalization has been prices, cannot conclude anything about detected. Somewhat surprisingly, the methodology competitiveness. Indeed, as Faminow and Benson of market integration has thus far (as far as I know) (1991) have shown, Ravallion-type of market not been applied to liberalization even though its applicability would seem obvious. implies Granger-causality in at least one direction. In this section we will first explain the This result has to be put in context properly. methodology for such analysis. Given the progress One could think of market relationships in econometric methodology since Ravallion's developing as some continuum. Markets which are classic (1986) paper. we will first restate the not at all related - no trading takes place for central elements of his approach in the context of example - will show no cointegration at all. When the more recent theory of cointegration and error- market functioning improves, a point will come correction5. Some parts of this approach have when markets will reflect all shocks which occur appeared in the literature. notably by Goodwin and to the other: markets are integrated in the long-run Schroeder (1991) and by Palaskas and Harriss and showing statistical cointegration. Market (1993)6.As will be seen, the present paper can be operations are observed as part of a process of seen as an extension of their approach. First, the capturing profits from arbitrage. At some point. methodology is explained as if there is no markets are processing information so fast that all structural change (such as liberalization) to be information is immediately absorbed into the considered. prices, and become 'informationally efficient'. If two markets are integrated one states in fact showing no cointegration at all. CIearly, common that some long run relationship exists. in that sense and economic observation has to guide what changes in one market will transpire in the prices interpretation should be given to the absence of of the other markets in due course. Short run cointegration. Since many markets in developed market integration implies then that this long run countries. including stock markets, are found to be relationship is very quickly obtained - virtually informationally inefficient (e.g Taylor and Tonks immediately. Commodity prices are often non- (1989)), finding non-cointegration is unlikely to be stationary, and if so, standard inference is not a reflection of informational efficiency. valid. Cointegration analysis allows researchers to The various steps of the analysis are described make from this failure a strength. Simple tests below and expanded upon only in case they have exist to establish whether there is a stable long run not been dealt with in the literature on market relationship between different non-stationary integration. TheJirst step is to establish the order series, and the Granger representation theorem of integration of the different price series, using allows the dynamics of cointegrated series to be unit root tests. If the series are I(0) then the specified as an error-correction mechanism (Engle standard Ravallion approach can be followed. If and Granger (1987)). The existence of a not, as is often the case. then one should proceed. cointegrating relationship between prices from as the second step, with the estimation of the regionally different markets can then be seen as cointegrating regression. For two non-stationary evidence of long-run market integration (as in prices R and P for the same commodity in different Goodwin and Schroeder (1991)), while the locations, the long-run relationship to be tested can dynamic specification can be used to establish the be represented as: extent of short-run integration (as in Ravallion (1986), Palaskas and Harriss (1993)). This analysis needs to be distinguished from Following the spatial equilibrium literature the 'efficient market hypothesis' literature, with its (Takayama and Judge, (197I)), equation (9) emphasis of informationally efficient markets assumes the existence of a marketing margin, (Malkiel (1987)). The latter will state that two specified as a linear function of the price R, in markets are informationally efficient if past prices particular: from one market do not help to predict present prices from the other market, beyond information contained in past prices from the latter market. In This specification nests more common fact, statistically this means the absence of specifications, such as constant or proportional Granger-causality in any direction (Granger and marketing margins (ie. a = 0 and B = 0, Escribano (1987)). But this implies that the price respectively), while contrary to e.g. Ravallion series will not be cointegrated, since cointegration (1986) a is not required to be equal to 1'. Testing can be done using the Cointegrating Durbin and thereforehelps with the dynamicspecification. Watson (CRDW), the Dickey-Fuller test, the It is therefore included as the fourth step in the adjusted Dickey-Fuller test. Engle and Granger procedure. The test implemented here will use the (1987) proposedseveral more potential tests,while dynamic specification, and therefore the test will the Johansen maximum-likelihood procedure be for dynamic Granger-causality,and not simply allows the testing of multiple cointegratingvectors in the long-run relationship. Misspecificat:on can (Johansen, 1988). Evidence of cointegration however bias Granger-causality tests, so it is provides evidence of long-run market integration, appropriate to use a dynamic error-correction and the estimated values of u and 6 would provide formulation to implement this test. The point is an indication of the nature of the marketing that the Granger-representation theorem allows margins. Providing there exists additional every cointegrating relationshipto be represented information on the nature of marginal marketing as a vector-autoregressionwith an error-correction costs, it is possibleto interpret the estimated values term.The dynamic error correction model (ECM) in the light of efficiencyor competitivenessof the related to equation (9) for the testing of Granger- market. Otherwise, no welfare economic causality can be specified as interpretationcan be attached to these results. A P ,=B,,AP,, ... O,,AP,, + The interpretation of the cointegrating regression is complicated because of the small sample properties of the estimators involved. While the OLS estimatorsof the parameters of a cointegrating relationshipare superconsistent, they The error correction term (P,, - ah-,6) is equal - are biased in small samples, and this bias may be to the lagged error term from the cointegrating quite large (Stock (1987)), it is a biased estimate. relationship and the estimated residual of the In small samples, this bias may be quite large cointegrating regression can be used to estimate (Banerjee, et al. (1986)). Inference within the (11)according to the 'two-step' procedure (Engle cointegratingregression is also not possible with and Granger (1987)). An appropriate alternative the usual tables (Banerjee, et a1.(1986), Stock ECM specification with AR, as the dependent (1987). In applied work, a procedure can be variable can be derived in the same way. The proposed which in practicecould establishwhether number of lags, n, has to be chosen to make sure one should suspect the bias to be very large. The that the residuals behave according to the usual procedure to use is to estimate besides the conditions for correct inference. If P and R are I(1) cointegrating vector also an Autoregressive but cointegrated then all terms in equation (11)are Distributed Lag model (i.e. regress P, on lagged P stationary. As a consequence, inference using and on presentand past R). As the third step in the standard t and F-tables is possible. A Granger- analysis, one ought to calculate the long-run causality test for market integration to establish the solution of the coefficient on R in the ADL and relevant direction of the flow of price information check whether this value is significantly different from the estimate of the coefficient on R,in the can then be considered.Granger-causality from R to P can be implied from the rejection of the joint cointegrating regression(see Doornik and Hendry hypothesis: (1993) and Adam (1993)). This is not a formal test procedure, but it is very useful to warn the researcher against wrong conclusions on the existence of a particular long-run relationship and a similar test can be derived for Granger- between prices. causality from P to R. Failure to establish any The existence of a cointegrating relationship Granger-causality should alert the researcher that between two variables has been shown to imply there may be problems with the cointegrating that Granger-causality exist in at least one relationshipeven if the tests under the second step direction (Engle and Granger (1987)). Knowledge suggestedcointegration.Granger-causality can be about Granger-causalityallows the determination interpreted as providing evidence on the main of the direction of the flows of price information, direction of flows of information between markets. using weakly exogenous variables. In practice. Equation (11)is in the market integration literature lagged values of a R and lagged values from all referred to as the (dynamic) reduced form equation markets which can be thought of to influence AR of the structural model in which the present can be used. If the null of independence is rejected, difference of reference market prices ,Ah, at the instrumental variables or equivalent estimation right hand side of the equation has been solved for. techniques ought to be used: otherwise OLS would To do tests of short-run market integration one seem appropriate to estimate the dynamic ought to be able to estimate a 'structural' version of structural model. the model, with AR, at the right hand side. In The sixth step involves the estimation of this particular, one would like to estimate the model: dynamic structural ECM model. The estimates can be used for further testing and interpretation. First, it is possible to test the hypothesis of short run market integration, which means that a price increase in the reference market will be immediately passed on to the dependent market. The null of short-run integration implies testing of the joint hypothesis: To use OLS, this requires however weak- exogeneity of AR, relative to inference on its coefficient in the model, which may not be true in interdependent markets. Under the absence of exogeneity, alternative estimation methods such as In particular, under the null, none of the lagged instrumental variables estimation ought to be used variables are significant while any change in R is instead. These methods involve however a loss of transmitted according to the long-run relationship, efficiency if the bias caused by using OLS in the and any deviation in the last period is immediately presence of endogeneity is relatively limited corrected. In fact, under this null, the correct short- (Nakamura et al., 1981). The Wu-test based on run model is simply a stochastic re-specification of Hausman (1978), as discussed in Nakumura et al. the cointegrating relationship in (9). All the (1981), tests whether the explanatory variables are variables in equation (13) are stationary if P and R statistically independent of the errors of the model. are cointegrated, so the standard F-tables can be For example, to test whether in (13) this used for the test in (14). independence is violated through the inclusion of A problem exists with this test. The null- AR,one would estimate an auxiliary regression of hypothesis under (14) includes testing whether the process generating AR, and enter the fitted is equal to a, in which a is the coefficient on R, in values of this regression in (13). The OLS the cointegrating relationship. As wa:j mentioned estimator on this fitted value in (13) would under before, this coefficient is biased in small samples. the null of independence tend to zero in probability If we have good (economic) reasons to believe that and standard t-tables can be used. If the null is the 'true' a is a particular value, then we could use rejected, OLS is not appropriate. It should be it, both in the error correction term in (9) as in the stressed that this not a direct test of weak- test under (10). For example, Palaskas and Harriss exogeneity, but rather of the bias involved if OLS (1993), who propose a short-run market integration would be used despite the absence of weak- test not dissimilar to the present approach, suggest exogeneity (Maddala (1993)). Following Urbain a model that only deals with constant marketing (1992), we introduce the error-correction terms margins. so that a equals one. Alternatively, and into the auxiliary regression, since their consistent with the two-step procedure, one could significance suggests that exogeneity is unlikely to use - at least in a large enough sLmple- the hold8. estimated value for a. and proceed with the test. TheJifth step in the analysis is to implement Some loss of power of the test will be involved, this test. The auxiliary regression for the but no clear alternative exists. explanatory variable (AR,) needs to be specified Additional tests of long-run integration are check for autocorrelation, heteroscedasticity, non- possible as well at this stage. The non-significance normality, misspecification, etc., it is crucially of the coefficient of the error correction term, y, important to evaluate the stability of the estimates should make the researcher wary of the correctness over the sample. In particular, the type of market of the specification and/or the presence of a long- integration models considered here assume a run relationship, since the error-correction constant degree of market integration over the formulation is following the Granger entire sample period. In particular, they ignore the Representationtheorem a correct representation of consequences of structural changes in the the cointegrating relationship. A more formal test economy, such as new infrastructure, changes in of long-run market integration involves testing the market intervention, liberalization, relative presence of the same long-run relationship in the marketing cost changes, etc. Parameter stability short-run specification. If the markets are tests (e.g. using different Chow tests or recursive cointegrated then the following restriction should least squares) are required to check whether there not be rejected in equation (13) is any evidence of such structural changes, and if there is evidence of instability, the model and its conclusion on market integration are invalidated. This last step provides a way to adapt this approach for the study of the effects for example since the long-run solution to (13) is improved security or liberalization. One way to go about it would be to check whether over the entire period and check for structural breaks, and if they are absent it could be concluded that the policy or other changes did not significantly affect the long- so that the test under (15) simply an alternative run relationship nor the dynamic adjustment path. form is of testing whether the coefficient on AR is The alternative is to implement the above analysis in the long-run equal to the estimated value of a in the cointegrating regression - which it ought to be for the relevant sub-samples and compare the estimated results before and after the change. The if the earlier tests were passed. latter approach will be taken here, even though, These tests are equivalent to the tests proposed given data availability, some problems with small by Ravallion (1986) on market segmentation (the sub-samples arise, reducing the power of the tests significance of y), short-run and long-run market implemented. In particular, emphasis will be on integration (see (14) and (15)). Similarly, the index the period before liberalization (pre-March 1990), of market connection (IMC) proposed by Timmer the period since liberalization until the take-over (1987) can also be calculated in this equation. (May 1991) and the aftermath of the take-over. Without loss of generality these tests are now First, we will focus on the cointegrating proposed to be implemented in a dynamic regressions, which give an indication of changes in (differences) model, while the long-run the marketing margins. relationship is tested in this analysis in levels in a As was described in section 2, liberalization statistically more acceptable way using and peace are likely to have affected the long-run cointegration. Also, the tests above are more relationship between prices through an increase in general than those proposed by Ravallion since arbitrage. This implies that equation (9) will give they allow for market integration under a more different estimated values in the period before and general structure of the marketing margins, as after these occurrences. In particular, if there are described in (5) and does not require for a to be three states during the period under consideration, equal to one. The seventhstep is model evaluation - equally and assuming that any change in the marketing margin is proportionalto the prices in the reference important to estimation of the dynamic market, then the long-run relationship to be tested relationship, but easily ignored in applied work. Besides evaluating diagnostics on the residuals - to can be written as9: If appropriate. testing over sub-periods separately is possible as well. The structural equation, to be used for the testing short-run integration, can be written as: with dummy, is 1 in the period before liberalization and 0 afterwards, dummyz is 1 throughout and dummy, is 0 until 'peace' was achieved and 1 afterwardslo. This specification implies that the 'long-run' coefficient on the reference market price before liberalization equals (a,+az), between liberalization and peace az,and after peace (az+as). which includes the present difference in prices in Equation (17) reflects differences in the the reference market. appropriately interacted with marketing margins between different periods, and the different dummies. To establish the bias resultsconsistent with the 'static' theoretical results involved in using OLS, it is again appropriate to from section 2 can be expected. Given the results run a Wu-test as described before with respect to in section 3, it is likely that for most markets the the present difference in the reference market estimated coefficients will results in larger margins prices. before than after liberalization, but little difference After estimating equation (20), it will be after the take-over. Direct inference on these possibleto implement tests to establish any change values is however not possible, so we will have to in short-run integration. For each sub-period, it can employ both our judgement and the other tests on be tested whether short-run integration holds. In long-run integration mentioned above to check the particular,adjusting (14), this test becomes for the validity of any conclusions, including by first sub-period: estimating the ADL for the model. The dynamic analysis will form the basis of testing for any change in the functioning of the markets. The relevant equation for the Granger- causality tests becomes while for the next two periods this becomes: and in which EC,., are the lagged residuals from the estimated long-run relationship in (17)". A test for Granger-causality over the entire period involves the rejection of the joint hypothesis (see (12)): If stronger or new evidence can be found for short- run integration, this can be interpreted as statistical evidence for improved functioning of the markets, since the information will be transmitted faster. A further indicatorof changed functioning of markets will be to look at any change in adjustment speed to errors, as reflected in the parameters on the relevant error-correction terms. The closer to one, In the appendix all the econometric results are the faster errors relative to the long-run solution given. First, the order of integration of each series will be corrected, suggesting an improvement in (data between July 1987 until September 199312). arbitrage relative to periods of slower adjustment. All level seriesare I(1) so that their differencesare Finally, it is possible to test in this short-run stationary (appendix table 1). This allows us to dynamic equation whether the long-run solutions proceed to the second step, the estimation and implied by the cointegratingrelationshipare at all testingof cointegration between markets. Since we consistent with the evidence from the short-run want to be able to distinguish changes both in equation. The relevant restrictionsto be tested for long-run and short-run relationshipsin the period each sub-period are up to liberalization, compared to the aftermathand also the period after the take-over, a (general) cointegrating regression was estimated as described in equation (17), with the pre- liberalization period defined as between July 1987 and February 1990, and the post-'peace' period as starting in June 1991. Consequently, the parameters a,and a,give theadditional proportion of the Addis Ababa price being reflected in the particular market price. Table 6.3 shows the results. Most market prices are clearly cointegrated with Addis Ababa prices, rejectingthe absence of cointegrationat, at least, 5 percent. Three markets are problematic. For Kombolcha, the absence of cointegration cannot be rejected at 5 percent using an Adjusted 5. Results Dickey-Fullertest with four lags. Closer inspection of the result suggested that one lag was sufficient The methodologydescribed above was applied to to avoid autocorrelationin the test-regression. An some of the most important grain routes in the ADF-test with one lag implied that the absence of country.The most important markets, in some way cointegration could be rejected at 5 percent. connected with the main consumer deficit area, Kombolcha was therefore retained for further AddisAbaba, have been considered. In all possible analy~is'~.Cointegration is more clearly rejected directions, some of the main towns along the roads for Debre Markos, despite its location in a big going to Addis Ababa have been considered, surplus region, Gojam. Recall however that in this correspondingdirectly to the radial structureof the region traders' activity until the liberalization market as is often considered in the literature. period. This does not mean that long-run Giventhe sizeof the Addis Ababa market, itwould integration was restored after liberalization: seem fully appropriate to use it as the reference cointegration tests on the sub-period startingfrom market. In particular, the ten towns considered are the date of liberalization did still not reject the Debre Markos (Gojam), Kombolcha (Wollo), absence of cointegration at 5 percent. Debre Nazret, Ambo, Wolliso, Hosaenna, Shashemene Markos was therefor?excludedfrom the rest of the and Debre Zeit (Shoa), Jimma (Keffa) and Dire analysis. Even in the long-run, Debre Markos is Dawa (Hararghe). Debre Zeit, Debre Markos, not integrated with the Addis Ababa market14.The Shashemene and Hosaenna are in main surplus Jimma results suggest also no integration at all, areas, Nazret is along the road from Arssi, another and was furtherexcluded. main surplus area. Table 6.3. Cointegrating Relationships Price, =CONSTANT+ a ,.dummy,.R,+ az.dummyl.R,+ a,.dummy,.R, + e, with R, = price in reference market (Addis Ababa) dummy, = 1 from 7/87 until 2/90, then 0 dummy, = 1 from 7/87 until 9/93 dummy, = 1 from 6/91 until 9/93, before 0 ............ADDIS ABABA........... cointegrationtest-statistics CONSTANT ADDIS ABABA Cointegration 1:est-statistics a1 a2 a3 DW DF ADF (4 lags) WOLLISO 7.54 0.01 0.68 0.05 ].I]** -5.12** -3.50* (0.76) (0.33) (8.09) (1.98) DEBRE ZEIT 26.14 -0.03 0.7 0.05 1.50** -6.34** -3.54* (3.43) (-1.80) (11.31) (2.63) NAZRET 37.52 -0.05 0.62 -0.00 1 1.14** - -3.65* (4.68) (-2.97) (9.09) (-0.06) 5.09** JIMMA 69.3 1 0.03 0.12 0.11 0.44 -3.01 -2.27 (5.39) (0.91) (1.07) (3.60) AMBO 19.67 -0.07 0.65 0.05 1.12** -5.17** -3.21* (1.82) (-2.89) (7.02) (2.07) DEBRE MARKOS 40.16 -0.21 0.40 0.07 0.45 -3.08- 3.44* (3.68) (-8.46) (4.31) (2.63) DIRE DAWA 27.49 0.02 0.97 -0.0 1 1.13** -5.18** -3.65* (2.16) (0.55) (8.91) (-0.34) KOMBOLCHA 17.58 -0.14 0.78 0.02 0.80* -3.91* -2.86 (1.92) (-6.72) (10.07) (0.84) HOSEANNA 33.85 -0.12 0.43 0.07 1.60** -5.42** -5.83** (3.58) (-5.45) (5.39) (3.18) SHASHEMENE 45.44 -0.10 0.48 0.08 1.21** -5..34** -5.00** (4.64 (-4.65) (5.77) (3.46) ZIWAY 24.09 -0.13 0.72 0.05 1.35** -5..90** -3.45* (2.49) (-5.91) (8.77) (2.13) t-value in brackets ** =absenceof cointegration rejected at 1 %(DF and ADF only) Sourcefor critical values cointegration tests: Engle and Yoo (1987) Even though no correct inference can be made absence of Granger-causality could be rejected in in the cointegrating regression, a further virtually all cases. In two cases, Wolliso and interpretation of the results can still be attempted. Kombolcha, the rejection is only possible at 12 and The estimates for a, and for the constant term have 19 percent respectively. These markets are, large t-values, lending some support to the however. relatively peripheral, so this weak constant-cum-proportional marketing margin evidence seemed not sufficient to drop these specification. The estimates for a , and a, are markets from the rest of the analysis. In general, a broadly consistent with expectationsL5. Before flow of information from Addis Ababa to all the liberalization, prices in 8 out of the 11 towns were different markets seems acceptable, and therefore a smaller proportion of the Addis Ababa price. a specification with Addis Ababa as the reference implying in each case higher marketing margins. market will be maintained. For some, such as Debre Markos, Kombolcha, Next, the Wu-(Hausman)-test for endogeneity Ziway, Hosaenna and Shashemene, a, is -0.10 or was implemented using a fitted value for the more. Liberalization allowed prices in, for difference in Addis Ababa prices, explained by two example, Debre Markos and Hosaenna to increase lags of the dependent variable and by two lags of by 37 percent and 17 percent respectively. After the differences in prices for all the markets used in the EPRDF take-over, the proportion of Addis the analysis and including the error-correction Lg Ababa prices passed on to each market price was terms for each market . The auxiliary regression roughly the same or less than 5 percent higher in 8 and the Wu-test statistics are reported in appendix out of 11 cases, suggesting no or little effect from tables 4 and 5. The error-correction terms were not the peacei6. In four cases, however, (Jimma, significant in the auxiliary regression. Further, it Shashemene, Hosaenna, and Debre Markos), we turned out that the null of independence could not can observe a clear decline in the estimated be rejected at 5 percent in any of the equations, margin, with a, equal to 0.07 to 0.11. Since Debre suggesting that OLS would not result in an Markos, Shashemene and Hosaenna are close to important bias in the estimations of the structural surplus areas, peace may have had a positive effect equations. on prices in certain surplus regions through a Consequently, OLS was used to estimate the reduced marketing margin: prices in Debre Markos structural equations for each market, using and Hosaenna for example went up by about 13 equation (20). Differences in prices in each market percent. Since it is likely that at least some part of were regressed on two lags of the dependent this increase was passed on to producers, it is variable, present and up to two lags of changes in probable that they have been important gainers Addis Ababa prices, including different terms for from the liberalization and the peace". the period up to liberalization and the period from An autoregressive-distributed lag (ADL) the change of regime in 1991 to measure any model with four lags was estimated to check differences in the dynamic effects of price whether we should suspect a bias in the estimates changes. Finally, error-correction terms, the in the cointegrating regression. The results are in residuals of the cointegrating regression, are table 2 in the appendix. In general, the long-run included, again with additional terms for the two solutions obtained are consistent with the resultsof sub-periods as above. Two lags were found to be the cointegrating regressions. In some cases both sufficient to exclude diagnostic problems with the the constant and the estimate for a2 seem to be residuals. The specification is very general, and not somewhat different (e.g. Debre Markos and Debre surprisingly, many terms are not significant. The Zeit), but without affecting the conclusions estimated equations are shown in the appendix in reached thus far". tables 7 (a,b, and c). In preparation for the dynamic analysis, One indicator for changes in the functioning of equation (18) was estimated. The estimates were the markets would be changes in the adjustment used for testing Granger-causality from Addis speed to errors relative to the long-run relationship, Ababa prices to the other prices. The results are given by the coefficient on the error correction reported in appendix table 3. As expected, the term. In two markets, Hosaenna and Dire Dawa, this coefficient is significantly different for each of and vibrant market town in the South. From all the the sub-periods. In the case of Hosaenna, the other towns. only Nazret. close to Debre Zeit, was findings do not imply a change in the adjustment also integrated in the short run with Addis Ababa. speed, since in each sub-period the coefficient is as Of the remaining towns, in five towns (Wolliso, close to minus one as in the others (even though at Ambo, Dire Dawa, Kombolcha. Ziway) short-run different sides of minus one, so that the adjustment integration was obtained after libera~lization:it path is different but not necessarily the speed). The could not be rejected at 5 percent, even though in Dire Dawa results suggest a lowar adjustment Kombolcha and Ziway it could be rejected at 10 speed in the post-take-over period. For all other percent. In Hosaenna. the return to peace was markets, adjustment to errors in long-run instrumental to obtain short-run integration with equilibrium has not changed at all. Addis Ababa, while in Ziway short-run integration Next, short-run market integration was tested was now not rejected, even at 10 percent. In two in this equation for each sub-period, as was markets, Nazret and Dire Dawa, short-run described in (24), (25) and (26). Important short- integration was reversed after the EPRDF take- run integration changes were found (table6.4). over. First, only Shashemene and Debre Zeit have In appendix table 8, long-run integration tests consistently remained integrated in the short run are reported. They test the consistency of the with the Addis Ababa market. Short-run cointegrating regressions with the long-run results integration in this case means that changes in the implied in the short-run dynamics. In all but three monthly average price in one market are reflected cases the long-run relationships implied are not in the average price of the same month in the other rejected in the short-run equation. The exceptions market. Debre Zeit is the closest market to Addis are Dire Dawa before and after liberalization, and Ababa, while Shashemene is the most important Table 6.4. Short-run integration (F(6,58)) Addis Ababa to: Wolliso Debre Zeit Nazret Ambo Dire Dawa Kombolcha Hosaenna Shashemene Ziway -=rejected at 10 percent *=rejected at 5 percent **=rejected at 1 percent Hosaenna after liberalization. Some of the short- run and long-run results are surprising, and some 63 tests were done on the robustness of the results. analysis was applied to teff markets in Ethiopia. Tests on the residuals of the dynamic equations The quota-system and repression of private revealed no problems with the specification. The trade have had important consequences on both the short- and long-run tests were also repeated in a level of prices and the functioning of markets. model with up to 4 lags (instead of 2). None of the Liberalization has been successful in correcting the short-run results were changed, but the taxation effects through increasing prices, inconsistency in the Hosaenna dynamic especially in the main producing areas, without specification did not show up in the long-run raising prices in the main consuming areas. integration tests, while for Dire Dawa long-run Liberalization also has had a positive effect on the integration was only rejected at I0 percent. This functioning of markets: five markets became suggests that the long-run test results may not be integrated with Addis Ababa in the period after the cause for too much concern. policy change. The short-run test results show that The fall of the Mengistu regime has had liberalization clearly improved the functioning of relatively little effect oz the level of prices, except markets: an increasing number of markets became for a temporary effect. The return to security may in the short-run linked to the price movements in well have been important in some specific Addis Ababa. For some markets, such Hosaenna, peripheral markets or in war-zones, in most of the short-run integration was only achieved after the country the effect on prices has not been very EPRDF take-over, suggesting a positive influence large. There were some exceptions: in at least three from the return to peace and security. markets prices moved closer to the Addis Ababa Finally, both Dire Dawa and Nazret (on the prices after the take-over, and in one, Hosaenna, it main road linking Addis Ababa and Dire Dawa) allowed a further increase in the integration of showed a decline in short-run integration with markets with the center, Addis Ababa. Addis Ababa after the take-over. The former may This does not mean that after reform and the be reflection of the temporary insecurities along end of civil war, all markets are well integrated. In that route, following the EPRDF take-over and the particular, Debre Markos - the capital of Gojjam, problems with another movement, the OLF, and a main cereals growing area - was not integrated in more general bandit activity, which was reported the long-run with Addis Ababa. Using only the there. The Navet result is more striking: giving its more recent data, no evidence could be found for location, a clear integration with Addis Ababa integration with Addis Ababa either, even though would be expected. A close inspection of figure 1 prices in Debre Markos were able to rise shows, however, that the Nazret prices became considerably following liberalization and peace. during 1993 higher than the Addis Ababa prices, The lack of integration is despite increased which suggests a reversal of flows relative to security, a direct road connection (about 12 hours Addis Ababa. A possible explanation may be that by truck maximum) and proper communications. given the relative low prices in Addis Ababa, Recall, however, that in Gojjam grain trade was traders activity has switched during 1993 more banned for considerable time. The repression of towards Dire Dawa. The model is not designed to trade for more than a decade can seemingly not be allow for this possibility and this may result in an resolved in a few years. The re-emergence of an apparent failure of short run integration testsz0. effective number of grain traders seems a slow process; traders need capital and skills while 6. Conclusion experienced traders may well have switched to other profitable activities during the repression. This paper presented means of analyzing the Furthermore, the short-run market integration consequences of liberalization and peace. Standard found for several markets is based on monthly analysis focuses on static results of liberalization- data, meaning that the gains f r ~ marbitrage are levels of prices and margins before and after eroded within one month, which is far longer than reform. In this paper we extended and adjusted needed to pass on information on prices and to market integration analysis to allow conclusions move grain. More refined data, on a weekly or about the functioning of markets as well. The daily basis may well show the failure of short-run integration across these markets. run arbitrage fails to be perfect is not addressed. The analysis is able to pick up changes in and To reach strong policy conclusions, it is necessary problems with the relationships between markets. to supplement this analysis with microeconomic However, it can only notice these changes or studies of market structure and traders behavior. problemswith trade routes. Why long-run or short- Notes 1. This paper was started while I was a Visiting Lecturer at the Economics Department, Addis Ababa University. I am grateful to the staff at the department for facilitating the research. 2. For an analysis of the evolution of grain marketing in Ethiopia, see Alemayehu Lirenso (1987) 3. This famine has been identified by Sen (1981) as resulting from a dramatic entitlement failure by peasants and pastoralists, even though this view is not without controversy (see e.g. Devereux (1988)). 4 Examples of such studies are Bryceson (1992) or Santorum et al. (1992) on Tanzania. 5. It should be stressed that Ravallion's approach is still a valid way of approaching the problem if prices are stationary. If not, it could not yet use the body of literature on testing long-run relationship in non- stationary series (cointegration), while if the model was estimated in levels non-stationarity could invalidate the inference. 6. Some aspects of the modelling methodology have appeared in Alexander and Wyeth (1994), but in their attempt to 'simplify' the approach, several fundamental econometric mistakes (see for a discussion, Dercon (1993)). 7. Alternatively, the relationship could be specified in logarithms if this seems a more appropriate way of considering transport costs. 8. Pesaran and Smith [I9901discuss this type of orthogonality tests. Urbain [I9921questions however the power of these tests in error-correction models. In particular, the outcome seems very sensitive to the specification of the auxiliary model. He takes up Engle and Granger's [I9871suggestion to introduce the error-correction term from the model of interest (ie. with AP, as dependent variable) into the auxiliary regression for A&. The significance of this term would suggest some feedback from P to R, and is likely to suggest a failure of weak-exogeneity of R relative to P. This may be the justification for the test suggested by Alexander and Wyeth [1994],even though it is not a standard nor formal way of testing exogeneity. The appropriate route would then seem to be to introduce the error-correction term into the auxiliary regression and to combine a check of its significance with the standard Wu-test as suggested above. Note however that the tests remains indirect, since no direct and formal tests for exogeneity exist [Banerjee,et al., 19931. 9. The latter assumption is in fact not necessary in general: differences could also be found in the constant 6 between periods, and this extension is easily dealt with. In the actual empirical analysis, it was however found that these extra dummy-variables did not become significant in the long-run relationship, so the present specification is used in this section, simplifying the exposition. Some of the main anomalies in the results were re-checked using the more general specification, but it turned out that none of the conclusions reached were affected. 10. This equation allows tests whether over the entire period a cointegrating relationship exists between the different price series, but with different values of a and 6 in the different sub-periods. If cointegration is rejected, then it would be appropriate to check whether there has been a change, for example after liberalization compared to before: cointegration may only have started afterwards. This tests cannot be nested in an estimated equation over the entire period, but will require to take the sub- periods separately. 11. Note the absence of interaction terms on lagged vaiues of AP. There is no reason to expect a change in which lagged prices within the same market location affect present prices after liberalization of the flows between markets. 12. The analysis is in real prices defined as before. In this way we avoid apparent cointegration between series because of a common inflationary trend without any direct interdependence between thle markets. 13. The limited significance of the tests for Kombolcha may be related to its location, just South of some of the most famine-vulnerable areas in the countries. Throughout the 1980s, and even later, large amounts of food aid were distributed in Wollo. Kombolcha was one of the main storage points for NGO food aid. It is very likely that the food aid would have found its way through to the open markets, at least partially. This may well have resulted in distortions in the market for teff for considerable periods of time, affecting its market relations with Addis Ababa. The problems were not obviously caused by the specification. All the tests were repeated with dummies for each sub-period to allow for changes in the constant in (16). The absence of cointegration was again only rejected at 5 percent using the different tests, and the short-run integration tests gave similar results as the results described below. 14. Cointegration tests using the Debre Markos prices were also implemented allowing for changes in the constant in (1 6), and for sub-periods of the sample. Even for the period since the end of civil war, the absence of cointegration could not satisfactorily rejected. Even though the sample period becomes very short for cointegration testing (25 observations), long-run integration can still not be observed in the more recent periods, justifying the exclusion of Debre Markos for the short-run integration tests. 15. The interpretation of the coefficients in the cointegrating regressions needs some care. Recall that the marketing margin is defined as in equation (6): R,- P,= -6 + (1-a).&. For all but one market (Dire Dawa), the margins with Addis Ababa are positive, so that a higher margin corresponds to a lower value for a. Margins are also a function of &.The higher &, the higher a needs to be to keep the level of the margin the same. The latter may confuse the conclusions somewhat. In particular. the mean Addis Ababa price before liberalization was about 126 birr per quintal; after liberalization but before 'peace' 119 birr per quintal and afterwards 124 birr. Consequently, the same level of the margin before or after liberalization (or peace) would actually require a , (or a,), given the estimated values for a?to be (up to) +0.05 (or +0.04 for a,). In other words, to conclude safely that a lower long-run margin pertained before liberalization (or after peace), would require the estimated values of a , (a,) to reach values well above 5 percent (or 4 percent for a,). 16. See the previous note. 17. Given the failure of cointegration tests, the interpretation of the results for Debre Markos can only be suggestive. 18. Also. little difference could be found between the residuals from the long-run solution to the ADL- model and those of the cointegrating regression. The dynamic analysis was later repeated with the estimated residuals from the ADL-model as the error-correction term, but, as expected. this did not affect any of the conclusions. 19. Implicitly we therefore allowed that all markets are endogenous, but that all interaction between different markets will be happening through the Addis Ababa market. 20. This is not to say that the Nazret market has fulIy adjusted towards price movements in Dire Dawa. First, multivariate cointegration tests showed that adding Dire Dawa prices to cointegrating regression with Nazret prices explained by Addis Ababa prices, gave insignificant results for the Dire Dawa price terms. Also, a bivariate dynamic model was estimated with the change in Navet pricesexplained by the appropriateterms of Dire Dawa prices and an error correction term from a cointegrating regression of Nazret on Dire Dawa prices. This dynamic model had substantial explanatory power, but short run integration was rejected in all three sub-periods. If there has been a reversal of flows in which teff is moved from Nazret to Dire Dawa instead of to Addis Ababa, this does not (yet) imply that Nazret .-ad Dire Dawa are integrated in the short run.Finally, the breakdown of short run integration between (~ddis Ababa and Nazret after the EPRDFtake-over was also found to pre-date the reversal in the relative level of prices: short-run integration was also rejected in the period between the 'peace' and the end of May 1993, before Nazret prices became higher than Addis Ababa prices. The conclusion is that some adjustment in flows is likely to be happening, but short-run arbitrage seems not to have been fully restored. For a systematic treatment of the spatial relationship between two marketswith potentially reversing trade flows, see Dahlgran and Blank [1992]. Figure i Retail price White Teff Addis Ababa in 1988/1 real birr per quintal REAL 1 988 /I TEFF PRICES SELEC~EDMARKETS iPRDF TAKE-OVER 1 1 I-AODlSABABA- NAZRETH +WOLLlSO *AMBO i DEBRE MARKOS DEBRE ZElT i Figure 3 REAL 1988 / 1 TEFF PRICES SELECTED MARKETS ( 2 ) 1-ADDIS ABABA I - JIMMA -+DIREDAWA I +-KOMBOLCHA HOSAENNA SHASHEMENIE I Figure 4 Appendix Table 1. Unit Root Tests for All Variables 1. Levels (I(1) vs. I(0)) DW DF ADF (4 lags) Addis Ababa Nazret Wolliso Ambo Dire Dawa Debre Markos Kombolcha Debre Zeit Hoseanna Shashemene Ziway 2. Differences (I(2) vs. I(1)) DW DF ADF (4 lags) OAddis Ababa 1.9 -7.895** -4.959** ONazret 2.1 -8.511** -3.984** OWolliso 2.1 -8.621** -4.650** OAmbo 2.5 -10.450** -4.717** ODire Dawa 2.5 - 10.560** -4.404** ODebre Markos 1.7 -7.270** -3.970** OKombolcha 1.7 -7.141** -4.817** ODebre Zeit 2.3 -9.727** -4.71 1** OHoseanna 2.4 -10.690** -5.210** OShashemene 2.3 -9.873** -5.625** OZiway 1.9 -8.4 17** -4.662** ** = rejecting H at 1 percent (in case of DF and ADF). Table 2. Long-run Solution to Autoregressive Distributed Lag Model (4 lags for dependent and independent variables) Price =Constant + a,.dummy,.R + a,.dummy,.R + a,.dummy,.R with R = price in reference market (Addis Ababa) dummy, = 1 from 7/87 until 2/90, then 0 dummy, = 1 from 7/87 until 9/93 . dummy, = 1 from 6/91 until 9/93, before 0 Addis Ababa Constant El a? Q3 Wolliso 11.14 -0.55 Debre Zeit 51.19 -2.95 Nazret 34.03 -2.56 Ambo 34.31 -1.14 Debre Markos 8.42 -0.27 Dire Dawa 24.87 0.02 0.98 -0.03 -0.95 (0.32) (4.44) (-0.55) Kombolcha 48.87 -0.16 0.55 -0.00 -1.38 (-3.08) (1.91) (-0.05) Hoseanna 52.6 -0.12 0.28 0.08 -5.21 (-6.62) (3.32) (4.07) Shashemene 54.18 -0.13 0.42 0.07 -5.58 (-7.19) (5.17) (3.44) Ziway 39.18 -0.15 0.60 0.04 -3.05 (-6.70) (5.55) (1.88) Table 3. Granger-causality Test: Do Addis Ababa Prices Granger-cause Others? Wald test for linear restrictions (4 lags for dependent and independent variables) 1. QWOLLISO: 2. QDEBREZEIT: 3. ONAZRET: 4. OAMBO: 5. ODIRE DAWA: 6. OKOMBOLCHA: 7. QHOSAENNA: 8. OSHASHEMENE: 9. OZIWAY: - = significant at 10 % * significantat 5 % = ** = significant at 1% Table 4. Auxiliary Regression (for Wu-test of Endogeneity) Modelling OAA by OLS (1987 (10)to 1993 (9)) Variable Coefficient Std.Error t-value PartRZ Variable Coefficient Std.Error t-value PartRZ Variable Coefficient Std.Error t-value PartR2 R2=0.962858 a =5.56255 DW = 1.90 * Some variables had to be excluded since singularity of matrix for inversion during estimation. Table 5. Endogeneity Tests (two lags for dependent and independent variables) First test: add fitted value for hAA from auxiliary regression, as well as hAA interacted with the dummy, (pre-liberalisation) and dummy, (post-peace). Test is joint significance of three tenns. Second test: add fitted value for hAA from auxiliary regression. Test is significance of this term. (All tests are Wald-tests for linear restrictions.) 1. hWOL: (1) LinRes F( 3,55) = 2.30 [0.09]- (2) LinRes F( 1,57) = 2.54 [O. 121 2. hDZ: (1) LinRes F( 3,55) = 1.21 [0.31] (2) LinRes F( 1,57) = 2.50 [0.11] 3. hNAZ: (1) LinRes F( 3, 55) = 0.79 [0.511 (2) LinRes F( 1, 57) = 0.5 1 [0.48] 4. OAM: (1) LinRes F( 3, 55) = 1.36 [0.27] (2) LinRes F( 1, 57) = 0.44 [0.5 11 5. ODD: (1) LinRes F( 3,55)= 0.51 [0.68] (2) LinRes F( 1, 57) = 0.98 [0.33] 6. hKO: (1) LinRes F( 3, 55) = 0.98 [0.41] (2) LinRes F( 1, 57) = 0.11 [0.74] 7. hHO: (1) LinRes F( 3, 55) = 0.30 [0.83] (2) LinRes F( 1, 57) = 0.1 1 [0.75:1 8. hSH: (1) LinRes F( 3, 55) = 0.47 [0.70] (2) LinRes F( 1, 57) = 0.5 1 [0.48] 9. OZI: (1) LinRes F( 3, 55) = 0.55 [0.65] (2) LinRes F( 1,57)= 1.12 10.291 - = rejected at 10 % Table 6. Variable Names AA = Addis Ababa WOL = Wolliso DZ = Debre Zeit NAZ = Nazret AM = Ambo DM = Debre Markos DD = Dire Dawa KO = Kombolcha HO = Hosaenna SH = Shashemene ZI - - Ziway ...pea = variable multiplied by dummy,, implying variable zero except for period 6/91 until 9/93. ...pre = variable multiplied by dummy,, implying variable zero except for period 7/87 until 2/91. Res = residual from cointegrating relationship. Table 7. Short Run Dynamic Equations (1)-(3) EQ(1) Modelling AWOLby OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value A WOL- 1 0.066159 0.1284 0.515 AAA AAA-1 A AA-2 AAAprelib AAApreli-l AAApeace AAApeace-1 A AA peace-2 ResWOLPre-1 0.28122 0.29169 0.964 ResWOLPea-1 0.30637 0.367 0.835 R2 = 0.477365 a 6.84989 DW = 1.95 = EQ(2) Modelling ADZ by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value ADZ-] -0.21067 0.14732 -1.43 AAA AAA- 11 AAAprelib AAApreli-1 AAApeace -0.11203 0.22638 -0.495 EQ(3) Modelling ANAZ by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value ANAZ-1 0.11342 0.13345 0.85 AAA AAA-I AAAprelib -0.13838 0.20192 -0.685 A AApeace -0.013274 0.22175 -0.06 ResNAZpe- 1 -0.043357 0.37055 -0.117 R2 =0.408536 a = 5.24222 DW =2.05 EQ(4) Modelling AAM by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value AAM-1 -0.22743 0.15845 -1.435 A AM-2 AAA AAA-1 AAAprelib A AApreli-1 A AApeace A AApeace-1 A AApeace-2 ResAMpre- 1 -0.033189 0.27531 -0.121 ResAMpea-1 -0.068001 0.37507 -0.181 R2 = 0.457627 a =8.47496 DW = 1.98 EQ(5) Modelling ADD by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value ADD-1 -0.011755 0.1371 -0.086 ADD-2 -0.027053 0.11621 -0.233 AAA 0.83479 0.26394 3.163 AAA-1 -0.0010809 0.29107 -0.004 AAA-2 -0.47285 0.27607 -1.713 AAAprelib -0.39964 0.35857 -1.115 AAApreli-1 -0.24116 0.37289 -0.647 AAApreli-2 0.25563 0.33462 0.764 AAApeace 0.14771 0.44284 0.334 AAApeace-1 0.057207 0.40933 0.14 AAApeace-2 0.72371 0.41038 1.764 ResDD-1 -1.2867 0.48336 -2.662 ResDDpre-1 0.61206 0.49873 1.227 ResDDpea-1 0.93496 0.5201 1.798 R2 =0.498357 a =8.97905 DW = 1.97 EQ(6) Modelling AKO by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value AKO-1 0.1895 0.14651 1.293 A KO-2 0.0691 ' A AA 0.48395 AAA-I 0.078851 AAA-2 0.069392 AAAprelib -0.19535 ~AApreli-1 -0.058647 AAApreli-2 -0.087047 AAApeace 0.2482 AAApeace- 1 -0.23279 AAApeace-2 0.25593 ResKOM-1 -0.61393 ResKOMpr- 1 0.24363 ResKOMpe- 1 0.054779 R2 =0.388707 a =6.81589 DW = 1.95 EQ(7) Modelling AH0 by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value AHO-1 0.26488 0.1351 1 1.961 AAA 0.30788 0.20145 1.528 AAApeace -0.019429 0.3097 -0.063 EQ:~) Modelling ASH by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient ,, Std.Error t-value ASH-1 0.12985 0.12537 1.036 ASH-2 0.048348 0.10656 0.454 AAA 0.29797 0.16828 1.771 AAApeace 0.15894 0.28071 0.566 EQ(9) Modelling AZI by OLS The present sample is: 1987 (10) to 1993 (9) Variable Coefficient Std.Error t-value AZI-1 0.25585 0.1247 2.052 Az1-2 0.032221 0.10348 0.31 1 AAA 0.43942 0.15388 2.856 AAApeace AAApeace-l Table 8. Long-run Integration (F(1,58)) (in model with hvo lags for dependent and independent variables) Addis Ababa to: 7187-2/90 3190-5191 6191-7/93 Wolliso 0.34 0.13 0.17 Debre Zeit 0.42 2.58 1.18 Nazret 0.43 0.58 0 Dire Dawa 7.52** 2.92- 0.23 Kombolcha 0.4 0.03 0.58 Hosaenna 0.22 5.41* 0.02 Shashemene 0.1 1 0.08 0.25 Ziway 0.07 0.19 2.22 *=rejected at 5 percent. **=rejected at 1 percent. Chapter 7 Fiscal Aspects of the Transition from War to Peace: with Illustrations from Uganda and Ethiopia David Bevan and Sanjay Pradlian 1. Introduction When a country emerges from a lengthy war, it is what would have been possible had the war contin- inevitable that the transition to a peacetime econ- ued. omy should have profound budgetary con- To make the point clear, we may contrast two sequences. A number of African countries have hypothetical transitions. In the first, the country embarked on this transition in recent years, and it had been engaged in a long-running border dispute, is to be hoped that others will soon follow suit. In had never lost control of territory, and had reached these circumstances, it is natural to enquire a military stalemate and fiscal steady state. The whether the fiscal challenges and opportunities war was costly but sustainable, and the end of it accompanying the transition have common fea- does release resources which can be slhifted into tures, and whether the experience of countries beneficial uses. In the second example, the country which began the process relatively early may be had been engaged in an exhausting civil war. useful in the formulation of policy in the others. Neither of the warring factions had been able to This paper considers this question, using the maintain control of national territory, and the war experiences of Uganda and Ethiopia to illustrate. led to the steady erosion of the country's infrastruc- Section 2 identifies a number of potentially sys- ture, its productive base, and the government's tematic features, and later sections look at these in revenue base. Eventually, the war became unsus- more detail. Section 3 discusses the revenue tainable as a result of economic and fiscal exhaus- characteristics of the transition and Section 4 tion, and ended. Since it would not have been considers expenditure patterns. Section 5 looks feasible to continue military spending at its previ- briefly at the impetus to decentralize, and Section ous inflated level, the end of the war does not 6 concludes. releases resources for civil purposes. Far from there being any peace dividend in the conventional 2. Fiscal Characteristics of Transition sense, the economic and fiscal collapse which accompanies the transition may mean that the In one sense the fiscal implications of the cessation government's resources are falling faster than its of hostilities must be benign. At the very least, commitments. Without an increase in external there will be the possibility of a substantial redirec- finance, the short run outlook for civil expendi- tion of government spending from military to civil tures would then be contractionary rather than and, hopefully, socially productive purposes; expansionary. Of course, these circumstances also alternatively, the resources can be released from offer the prospect of accelerated change in the public to privatecontrol, permitting a reduction in medium or long run as the lost ground is recov- government financing. Unfortunately, it does not ered. follow that there is a substantial "peace dividend" While these two cases are stylized, they serve for distribution in the short run, relative to the to demonstrate how very different the fiscal conse- period immediately before the cease-fire. The quences of transition may be, depending on the potential redirection or reduction is relative only to antecedent circumstances. The overwhelming majority of military disputes in Africa involve civil typically triggered by the ending of hostilities. This war, so that very substantial damage to the infra- carries with it three distinct fiscal complications. structure and to the economy in general is routine. First. since the level of aid is typically very sub- Both Uganda and Ethiopia fall into this category. stantial and must be expected to decline in the The immediate post war period is therefore charac- medium term, there is an important issue concern- terized by unusually high expenditure requirements ing the feasibility of substitution of domestic for rehabilitation. Whether the government's revenue andlor contraction of expenditure in the revenue base (relative to the inevitably diminished medium term. Second. since the aid necessarily GDP) has also been damaged is more various; in accrues as foreign exchange, whereas the govern- Uganda, for example, it had been massively con- ment's rehabilitation spending could well be tracted for some time before the peace, while in largely nontradable. the real exchange rate will be Ethiopia it remained at a very high level. In conse- affected. Third, aid flows are subject to condition- quence, the immediate revenue prospects were ality and a variety of release procedures which quite different in the two countries. In Uganda, means the flow of revenue is volatile and unpre- there was every reason to suppose that revenue dictable. These issues are not unique to economies effort could usefully be increased, though it re- undergoing war to peace transitions; indeed very mained unclear how fast revenue actually collected similar circumstances arise in countries like Zam- could be raised. In Ethiopia, the very high wartime bia where the trigger for serious increases in aid is collections reflected an unacceptably authoritarian the combination of thoroughgoing political and approach, rather than an efficient structure of economic reform. However, it is certainly likely to taxation. Domestic revenue would inevitably fall be a systematic and important feature of war to sharply in this case, prior to a probably slow and peace transitions. Finally, one feature of enhanced partial recovery as a reformed structure is put in aid flows in the transition is their potential role in place. providing a signal of international donor confi- Even the view that military expenditure must dence in the government's overall economic pro- necessarily fall is mistaken. In a civil war, there gram. Given uncertainty and apprehension in the are (at least) two armies; during the fighting, only aftermath of war, donors' effective underwriting of one of these is on-budget. It may not be thought the economic program could become an important prudent immediately to release the defeated army factor in allaying some of the private sector's because of the dangers of banditry or insurrection, hesitancy in making irreversible investments. and it may be felt necessary to reward the victori- The discussion so far has focused on a fairly ous guerilla contingents by regularizing them. In narrow interpretation of fiscal impact; in effect, we the short run, the new government may end up have assumed that the underlying view of the role financing two armies. To the extent that the mili- appropriate to the state is little changed. The war tary hardware was supplied off-budget by foreign may have devastated the economy, it may have governments, there may be little scope for major severely damaged the public capital stock, it may reductions in material military spending either. In have eroded the institutions of government them- consequence, the compressibility of the military selves. In consequence, it may have distorted the budget in the short run may be quite limited or shape and scale of government for a time, but in even perverse. Both Ethiopia and Uganda, how- due course it is intended to revert to the original ever, have managed to compress their military pattern. This is a perfectly plausible outcome, budgets to some extent, though the process took particularly when the incumbent government is longer in Uganda and resulted in initially higher victorious. However, there is an alternative possi- levels of spending. bility, which is more likely when the government To summarize the argument, in African transi- is overthrown, so that the peacetime administration tions there need not necessarily be any substantial, marks a break from that which fought the war as domestically generated peace dividend in the short government. In this case, it is possible that the run. Expenditure needs may be temporarily high transition will be more thoroughgoing, with a and revenue temporarily low. The real peace general reconsideration of the purposes and mech- dividend lies in the enhanced flow of aid which is anisms of public activity. There are examples of both types of transition in Africa. but recently territorial gains are made but do not include the cases where the incumbent is overthrown have administrative institutions of government. In these predominated; both Uganda and Ethiopia fall into circumstances, the leaders of a guerrilla army may this category. have to undertake at least a limited delivery of There are a number of reasons why an incom- public services outwith the usual institutional ing administration may wish to break not only with channels. Good examples of this process are selected policies but with the entire policy regime provided by the organs of the National Resistance of its defeated predecessor. One follows naturally Movement in Uganda, and by those of Ithe Eritrean if the war was fought partly on ideological People's Liberation Front which had effectively grounds. Curiously enough, this does not well barred the institutions of the Ethiopian government describe recent events in either Uganda or Ethio- from operating in Eritrea for some years before the pia. In both countries, a government of predomi- end of the war. These institutional improvisations nantly socialist - or at least interventionist - lean- may prove quite durable after the end of hostilities. ings was defeated by forces of rather similar First, they constitute an important part of the economic outlook. While in both cases the transi- apparatus which brought the new regime to power, tion has been marked by a profound shift in the may still provide a significant power base subse- role that is perceived to be appropriate for the quently, and may also engage many loyalties. state, that was not a consequence in any profound Second, they may be perceived to have been way of the incoming government's prior economic effective vehicles for the delivery of local services philosophy. and other functions, and this against a background A second reason is that the policies of the where the more routine institutions of government defeated regime may have become discredited, are perceived to have failed. These circumstances either by simple association, or more comprehen- are calculated to induce a relatively experimental sively as a contributory factor in the defeat. In this approach to institution building by the incoming case extensive changes may be attractive, even in regime. the absence of any ideological commitment, There is a fourth reason, linked to the second simply because what went before is perceived not and third; it is that ousting the previous regime to have worked. In fact, substantial changes in required a coalition to be assembled, of groups economic regime may have been inevitable even in which had been previously excluded from power. the absence of the switch in political regime; the This process is likely to bring into the successor latter may simply permit the former to take place regime ethnic and other groups which are suspi- more rapidly and comprehensively. An interesting cious of centralization. Indeed, in the limiting case, illustration is provided by Ethiopia; as already the war may have been fought on secessionist noted, the economic philosophy of the incoming grounds. More generally, the transition may in- regime differed little from that which had charac- volve a powerful impetus towards devolution. The terized their predecessors. Indeed, inexorable secession of Eritrea is an example of the former economic decline had persuaded those predeces- extreme, but there is ample evidence in both sors to adopt a substantial package of liberalizing Uganda and the remainder of Ethiopia of the reforms; in the first months of the transition it strength of the latter. The circumstance in which a appeared possible that the new government might government is overthrown also means that the old attempt to put the clock back, and refuse to main- constitution is effectively suspended; while the tain the impetus to reform. In the event, this para- new regime could in principlesimply adopt the old doxical outcome did not materialize. However, it constitution, in practice it typically does not. The is clear that important components in the economic actions of the Transitional Government of Ethiopia policy switch predate the end of the war and the provide a case in point, with an extended process associated change of government. of constitutional redefinition still under way as A third reason, not wholly unrelated to the well as an accompanying redrafting of legislation. second, is that the incoming regime typically has Finally, there is the fact that, for the reasons no experience of government, or experience of an outlined earlier, the successor government will unusual kind, for example that obtained when both be in need of, and able (conditionally) to obtain. enhanced donor financing. The nature of initial positions. both as regards explicit tax and the conditionality will vary only in minor respects, non-tax revenue, and as regards the availability of and will largely reflect the international commu- seigniorage. nity's current view of good practice. This is not Ethiopia has had a remarkably varied fiscal only understandable,it is in many respects admira- history, culminating in the later years of the Derg ble. However,conventional wisdom is itself some- in a very high rate of resource extractior! by the what fickle, and not ideology free. When an in- government, albeit with a still higher rate o ipublic coming government does not have a deeply en- spending. Since the overthrow of the Derg in trenched position in respect of some public activ- 1991, there has been a substantial fall in revenue ity, and is in need both of funds and guidance, it is collection. not only vulnerable to external pressure, it may The data for the 1980s exhibit a number of even welcome some aspects of it. In these circum- striking features.(Bevan, 1992) Tax revenue itself stances, part of a policy shift may simply reflectan was not unusually high by African standards, external agenda, whether this has been imposed or averaging 17 per cent of GDP over the decade, accepted by the domestic government. with relatively little trend. Non-tax revenue, by Several of these reasons tend to reinforce each contrast, was both substantial, and broadly rising, other in shaping the type of policy shift that oc- so that total revenue averaged 24 per cent over the curs. Since the ousted regime was often highly decade (22 per cent in the first five years, 27 per centralized, interventionist, and perceived to lack cent in the second). This non-tax revenue was both legitimacy and competence, and since the comprised largely of the residual surplus of public international community has been pressing for enterprises. Despite the relatively high level of liberalization and deregulation, there is a conver- total revenue, it fell very substantially short of total gence of forces toward markets and away from expenditure, which climbed rapidly at the begin- controls; toward devolution and away from the ning of the 1980s, to settle thereafter at a fairly center; toward expenditure on social services and trendless level around 39 per cent of GDP. In away from (non-infrastructural) economic ser- consequence, the overall deficit between 198213 vices; and toward greater fiscal prudence. and 1989190 only twice dipped below 10 per cent These various strands may be summarized, for of GDP, and averaged 13 per cent. A little over a representative case, as follows. There need not half of this was externally financed (7 per cent), necessarily be a real prospect of a domestic peace with the domestically financed component being dividend in the short run; on the contrary, the more or less entirely financed by bank borrowing. immediate prospect may be one of low or falling It has been a remarkable, and much discussed. domestic revenues and elevated expenditure feature of the Ethiopian economy that such a high requirements. A key issue is whether countries rate of monetization of the deficit should have will be able to reduce their military budgets in the been associated with so little inflation especially aftermath of war. More significantly, there is the given the lack of real growth. The Ethiopian likelihood of a substantial temporary increase in government's record is often described as one of aid. The success in the expenditure transition fiscal prudence, but this can only be justified by depends upon whether countries are able to make the fact that it did not engender a high rate of major changes in ,public expenditure patterns inflation. Approaching the matter from the other towards the civilian sector, and within the civilian end, an overall deficit of 13% of GDP, half fi- sector, towards economic infrastructure and social nanced from the banking system, looks anything spending. However, the phasing of these shifts is but prudent. The rapid monetary deepening that uncertain, and the process may be slow. made this fiscal stance consistent with so little inflation may have been a transitional process 3. The Revenue Transition leading to a permanent shift in the economy, or it may have been a temporary and reversible process As the preceding discussion stressed, the revenue which will be unwound in the near future. How- positions inherited by new regimes are extremely ever, it seems inconceivable that this process of varied. Ethiopia and Uganda had very contrasting deepening could continue. Equally, it is not feasi- ble for Ethiopia's external debt to be much in- much provide a fiscal dividend as avert a fiscal creased, except on the softest of terms. In conse- catastrophe. quence, a very substantial reduction in budget In any event, the early part of the transition has deficit had become a necessity towards the end of been accompanied by a sharp fall in domestic the war as a matter of financing imperatives if revenue, from 25% of GDP in 1989190 to 20% in accelerating inflationor excessive debt service was 1990191 and around 16% in 1991/92. Of this, tax to be avoided. revenue amounts to barely 1I%, down from around Two further aspects of the inherited situation 17%; this fall partly reflects the operation of a are important. First, despite the very substantial relatively primitive and as yet somewhat unre- scale of the fiscal intrusion into the economy that formed tax system when excessive coercion is is revealed by these data, they significantly under- removed. One feature that has been much dis- estimate the real weight of the government's fiscal cussed is the virtual disappearance of revenuefrom operations. In particular, the direct taxation of the explicit taxation of agriculture. This was partly agriculture embodied in the reported figures is due to the collapse of the institutions which acted only a small part of the total, since there was as tax collectors, partly to the fall in the world heavy implicit taxation via compulsory quotas coffee price and the increase in smuggling, partly levied on producersat artificially depressed prices to the relatively tenuous control which the central as well as a plethora of forced contributions in government was initially able to exert in the kind, in cash and in labor services. regions of agricultural surplus, and partly to the Secondly, and notwithstanding this substantial decentralizing thmst of its fiscal thinking. degree of under-reporting of fiscal impact, the data Non-tax revenue - mainly from state-owned do make it clear that, by any conventional stan- enterprises - continued to provide substantial dards, the Ethiopian government had become too resources to central government, though not at the big relative to the economy. It would probablynot previous scale. Similarly, financing from the have been feasible and would certainly not have domestic banking system fell over the period from been desirable to attempt to close the excessively around 10% of GDP to less than 4%. Apart from a large budget deficit by increasing revenue; the bulk brief surge in prices as the war came to a close, of the adjustment inevitably had to take place on inflation has remained low and the level of sei- the side of expenditure. (World Bank, 1990) Even gniorage revenue high. (For example, seigniorage if the presentshort run revenue problems are put to of nearly 3% of GDP will be available if inflation one side, the longer run perspective must be that remains at its current level of around lo%, real the fiscal stance of the 1980s was unsustainable, growth is at the level assumed for planning pur- and that massive expenditure reductions would poses of 5.5%, and the main money to GDP ratios relatively soon have become inevitable. are stable.) Finally, foreign financing has risen To put the matter somewhat differently, the from about 10% of GDP in 1989196)to 13% in direct fiscal costs of the war were being carried 1992193, in each case very roughly evenly split either off-budget (as already noted) or by an between grants and loans. unsustainable increase in the deficit. Indeed, The aggregate fiscal experience of these four ostensible, on-budget defense expenditure aver- years, two before and two after the end of the war, aged a little under 10per cent of GDP over 198012 may be summarized as follows. Domestic revenue - 8718, compared to the average budget deficit over fell by 8 percentage points, from 25% to 17%; total this period of a little under 12 per cent of GDP. In expenditure fell by 12 percentage points, from consequence, there was virtually no scope for 46% to 34%; in consequence the overall deficit substituting peaceful government expenditure for narrowed from 21% to 17%. The contribution of the reduced military spending; in the absence of foreign financing rose from 10% to 13%,and that peace, any continuation of military expenditure at of domestic financing (overwhelmingly from the its previous level would have required both contin- banking system) fell from 11% to 4%. The in- ued (dejacto)grants of equipment from the East- crease in external finance was therefore less than em bloc and a reduction in non-military spending. sufficient to match the much needed contraction in In this sense, the transition to peace did not so domestic borrowing, and spending had to fall Ethiopian case, defense spending grew in the in Table 7.7 which shows that donors financed initial years of the transition, reflecting problems 85% of the rehabilitation costs. The potential in financial management and provision of uniforms concern that donor financing of infrastructure and basic supplies for the victorious army. In the rehabilitation may have a Dutch Disease impact first two years of the transition, defense spending was noted earlier. As it turns out, this problem may grew by 40% per annum in real terms, (World not materialize in the aftermath of war: given the Bank,1991) and its share grew to 46% of recurrent collapse of the construction sector and the lack of expenditures and 3% of GDP in 1987188. How- availability of materials and equipment, the bulk of ever, since attention began to focus on this growth, project costs consist of imports, mitigating the on the severe underfunding of the economic and Dutch Disease impact. As Table 7.7 shows, the social sectors, and the consequent unsatisfactory average foreign exchange component of the infra- provision of public services, the share of defense structure rehabilitation projects amounted to 85% spending has been reduced substantially; by of total costs! Moreover, given the constraints on 1992193 it had fallen to 28% of recurrent expendi- implementation capacity in the aftermath of war, tures and 1.5% of GDP (see Tables 7.4 and 7.5). there was a heavy reliance on importing manage- The Government has also embarked upon a care- rial services and using foreign contractors. fully designed demobilization program, funded The contribution of the resulting infrastructure through external assistance. At the same time, rehabilitation in Uganda can be seen in the very reflecting increased concerns about poor health high estimated economic rates of return, which and education indicators, the share of social sectors average close to 40%. The value of infrastructure has increased from 27% to 33% of recurrent rehabilitation is also seen in the results of a 1990 expenditures. survey of rural households in Uganda conducted Overall, the increase in external assistance by Bigsten and Kayizzi-Mugerwa. In this survey, combined with expenditure restructuring has respondents were asked to name the single biggest permitted Uganda also to begin the process of improvement in the area brought about by the redirecting expenditures towards priority sectors. government since 1985; 48% of the respondents A central feature of Uganda's expenditure transi- listed improvements to the road network. tion has been the emphasis placed upon and suc- While donor financing of infrastructure reha- cess achieved in the rehabilitation of physical bilitation has made an important contribution, it infrastructure. War inflicts very substantial also highlights likely medium- to long-term prob- damage to infrastructure, and its rehabilitation is lems of sustainability. Indeed, while the major vital for restoring economic activity, fostering trunk roads have been rehabilitated, there is a market integration, and inducing the private sector massive maintenance crisis looming over the to make irreversible investments. Indeed, in the rehabilitated roads. This will call for considerably first few years of the transition, all major trunk enhanced and sustained recurrent expenditures, roads in Uganda were rehabilitated and this has and will necessitate that over time domestic re- been a key factor in the revival of economic activ- source mobilization substitute for external assis- ity and initial high levels of growth. Given the tance. constraints on domestic resource mobilization Another critical characteristic of a budget so stressed above, external financing played the dependent on external finance, particularly in the dominant role in this rehabilitation. As the at- context of very low monetization, was discussed in tached map of donor participation in the road the opening section of the paper. It is the poten- sector indicates, donors effectively carved up tially great vulnerability of macroeconomic stabil- major regions in Uganda as their principal domain ity to the vagaries of donor finance, in particular to of responsibility. relatively minor lags in the release of funds. The In order to examine the economic implications Ugandan experience of 1991/92 illustrates the of this infrastructure rehabilitation, a detailed point well. There was a shortfall in import support survey was canied out for all the major infrastruc- receipts early in the fiscal year, due to implementa- ture rehabilitation projects financed by each donor tion delays. Expenditure continued at programmed during the transition. The results are summarized levels, before being cut back later in the year. The consequent deficit, which was financed through the effect were among the first actions of the Transi- domestic banking system, barely exceeded 1% of tional Government. While it would have been GDP. Given the excess sensitivity noted earlier, possibleto pursuethis aim by devolution of power this small monetary shock triggered a serious within a unitary state. the present intention is to go inflationary surge, at an annual rate of 70% (with further than this and adopt a federal constitution. the monthly rate reaching 10%). The response of Fourteen regions were declared in the transitional the government to this unwelcome demonstration charter (subsequently reduced to ten) and a series of the vulnerability of its program to relatively of proclamations have been issued which have minor aid shocks was to adopt monthly cash progressively defined regional policy. Elected budgeting. If the path of revenue is volatile, then councils are now in place in all regions. Two the path of expenditure is made equally so. If complications with the new structure should be properlyexecuted, this system does indeed prevent noted. First, the regional boundaries have been unprogrammed deficits from materializing, but at determined to a great extent by ethnic rather than the cost of a disruption in the delivery of services administrative or economic considerations. Conse- and in efficient public activity. Once again, there quently, they vary enormously in nature, size and is nothing in this story which fixes it uniquely into viability, and will pose difficult challenges to the the war peace framework; indeed, a more or less designers of the federal constitution. Particularly identical story could be told for Zambia. However, acute difficulties will be involved in the design of war peace transitions are more likely than most to the revenue sharing arrangements. Second, the new be characterized by the combination of heavy regions have been inserted into an already existing externaldependence and inflationsensitivity which local government structure, and it is far from clear makes this outcome likely. how the various parts of the system will relate to each other. 5. Decentralization While the question of regionalisation is im- mensely pervasive and there have been a number While emergence from a war is neither a necessary of policy statements, it remains unclear what the nor a sufficient condition for an impetus towards net effects will be, not least in the fiscal sphere. decentralization,there is evidence for a strong link, (Bevan, 1993) It is very difficult to assess what the at least when the incumbent regime loses a civil partition of revenue will be, either between center war. There is certainly such an impetus in both and regions in aggregate, or amongst the regions Ethiopia and Uganda, which appears to be partly themselves. This partly reflects uncertainties over rooted in the guerrilla origins of the resistance,and the eventual ownership of state assets, partly the partly a response to the imperialistic behavior absence to date of any public statement concerning attributed to the defeated center. By way of con- the rules for partitioning joint revenues, and partly trast, there appears to be no decentralizing impetus the lack of a comprehensive fiscal data base. of comparable strength in countries undergoing However, it appears inevitable that all the regions thoroughgoing but peaceful transitions, such as will require heavy subsidization, given the scale of Zambia. the obligations devolved to them. This suggests Ethiopia has a long history of centralized that the central government not only possesses the government, a feature carried over from the impe- means to maintain macroeconomic control (via rial period and maintained after the revolution. In borrowing restrictions) but, informally, will have consequence, political institutions outside of the the bargaining power to control the composition of center are rudimentary, and there is little adminis- expenditure as well. trative experience or capacity at the regional and There have also been extensive statements district level. This high degree of past centraliza- setting out the powers and duties of central and tion is perceived as being intimately connected regional governments. The most interesting feature with a history of unsatisfactory and authoritarian of these atrangements is the dual accountability government. (Eshetu Chole, 1992) There is now a that is implicit in them, with a regional bureau songpolitical commitment to createa substantial typically responsible both to the regional govern- clegree of decentralization, and statements to this ment of which it is an executive agency, and to the Ministry which formulated the policy it must this appears all too likely. However. in the short implement. This construction seems certain to run. the sheer lack of appropriately qualified and cause considerable tension both between the experienced personnel should prevent this out- central and regional tiers and within each tier, come, always provided that reasonabi.: tight between the executive and budgetary branches. control of recruitment is maintained. Indeed it may prove an unworkable hybrid be- Third, will the central government's c,\:ltrol of tween administrative decentralization and a more macroeconomic management be compromised? In thoroughgoing federalism. the case of Uganda, the district budgets will appar- The administrative aspects of the new arrange- ently be considered within the central budgetary ments are also far from clear. There have been process, so there should be no additional problem. movements of personnel between central and In Ethiopia, the central government is likely de regional administrations, and some between re- facto to retain more direct control of general gions. However, in view of the general lack of government outcomes than was intended, because capacity with which the regions started, most of of the practical difficulty of devolution in the short them remain well short of being capable of execut- run; but the new arrangements appear in any case ing their new responsibilities. So far it has been to give adequate control, provided that proper necessaryfor the central government to continue to budgeting and accounting procedures can be fulfil functions which have theoretically been operated. A serious danger in the medium run may devolved; in some ways, this may have postponed be the difficulty of maintaining macroeconomic some of the potential efficiency losses. control in a federal system if many of the constitu- While in other respects the Ugandan transition ent governments are incapable of producing accu- is further advanced than that in Ethiopia, this is not rate and timely management information. true of its decentralization program. As in Ethio- Fourth, will the central government be able to pia, the eventual shape of the institutional arrange- maintain control over the composition of public ments and their fiscal consequences remain ob- expenditure? Here again, the retention of central scure. Two major differences are the more highly control over the national policy agenda should in disaggregated structure in Uganda (39 districts), theory ensure this. The two caveats are first, and what appears to be a less thoroughgoing whether the informational base will be adequate to degree of devolution. However, primary education permit effective control, second whether the new and health are now the responsibility of the district arrangements are politically workable. In particu- councils, as of the Ethiopian regions, and there lar, will the lower tier governments go on being seem to be similar ambiguities as regards the chain content to accept policy from the center, and what of command and the financing arrangements which action will the center be prepared to take if its will permit these responsibilities to be executed. In policies are not implemented? both countries, policy in this area is being made "on the hoof'. 6. Conclusion Drawingthese various considerations together, what can be said about the likely fiscal conse- The three preceding sections have summarized quences of these developments? First, will decen- aspects of the two countries' fiscal experience that tralization reduce the efficiency with which the bear on the issues raised in the introduction. rest of the transition is managed? The main danger Despite remarkably different initial conditions, here is probably the potential diversion of scarce they provide ample evidence that while a reduction political and administrative resources. So far, in military spending is feasible to some extent in though the relatively small cadre of policy makers the transition, it offers only limited possibilities of in both countries is very stretched, there is no a peace dividend. The increased spending on social evidence that the devolution issue has impacted services and infrastructurethat is certainly required adversely on other policy areas. must therefore be found by reducing other catego- Second, will expenditure rise as a consequence ries of civil expenditure, which appears dificult to of duplication? In the long run, as in other coun- execute though the process has begun in each tries which have undertaken extensive devolution. country, or by increased flows of aid, which appear easier to achieve. There is an added. important that donor assistance has also been used to finance benefit from external assistance in the transition. high return infrastructure rehabilitation projects, Donor assistance. especially from the IMF and the the bulk of which consists of foreign exchange World Bank. effectively underwrites the overall requirements, the Dutch Disease concern is likely economic program. and provides an important to be mitigated. positive signal to a cautious and apprehensive There are, however. three other problems with private sector in the aftermath of war. this dependence on aid. For one. the volatility and One danger of going this route is that it may potential inflationary leverage of aid, Uganda induce a form of Dutch Disease, inhibiting the provides a cautionary note. For the other two, the structural recovery of the economy. This is most jury is still out. Whether it will prove possible to likely if the government uses import support to substitute domestic revenue for aid rapidly and finance nontradable activities. However, if it is durably enough as aid tapers down, and whether also liberalizing the economy. and correcting for governments can reacquire effective capacity for an overvalued exchange rate, the effect may only public service provision remains to be seen. On a be to slow the reallocation of resources. not to more positive note and despite these uncertainties, reverse it. In any case. the alternative is a radically the difficult process of expenditure restructuring -- smaller level of government activity, which is redirecting expenditures towards economic infra- likely to constitute a still more serious brake on structure and social spending -- has begun in both economic rehabilitation. Moreover. to the extent countries, although much more remains to be done. UGANDA Donor Financing for Road Rehabditatiomaintenance ADB 1YE UGANDA CLASSIFIED ROAD NEMBRK Table 7.1. Ethiopia: fiscal accounts ( O hof GDP) FYS7 FYSS FY89 FY90 FY91 FY92 FY93 FY94 Variables act. act. act. act. act. act. prlm. proj. Tax Revenue 18.4 19.6 19.1 17.4 15.5 11.3 12.2 16.1 Non-tax Revenue 7.3 9.7 12.3 7.5 4.3 4.3 5.4 2.8 Total Revenue 25.7 29.3 31.4 24.9 19.8 15.6 17.6 18.9 Grants 2.8 5.4 6.4 3.7 3.5 3.3 5.8 9.5 Revenue and Grants 28.5 34.7 37.8 28.6 23.3 18.9 23.4 28.4 Recurrent Expenditures 23.0 28.9 31.0 34.0 27.4 22.5 22.1 22.5 Capital Expenditure 12.1 12.3 15.6 11.2 8.9 6.7 12.2 15.2 Total Expenditure 35.1 41.2 46.6 45.2 36.3 29.2 34.3 37.7 Deficit before Grants 9.4 11.9 15.2 20.3 16.5 13.6 16.7 18.8 Deficit after Grants 6.6 6.5 8.8 16.6 13.0 10.3 10.9 9.3 Net External Financing 3.5 3.4 5.4 5.5 3.3 1.8 6.5 7.6 Domestic Financing 3.1 3.1 3.4 11.1 9.7 8.5 4.4 1.7 Sources: MOF Table 7.2. Ethiopia: Public Expenditures Budget (%of GDP) Variables 1987 1988 1989 1993 1994 General Administration 3.12 2.88 2.82 3.18 3.19 Defense 8.8 1 10.7 13.5 4.89 3.40 Economic Infrastructure 0.58 0.52 0.43 0.53 0.6 1 Education 3.27 3.12 3.13 3.96 3.87 Health 1.04 1.01 0.99 1.33 1.46 Other Economic Services 0.97 1.12 1.24 1.53 1.63 Other Social Services Othersa Total 11. Ca~italBudget Economic Development Economic Infrastructure Education Health Other Social Development Others Total 111. Sector Budget Economic Infrastructure Health and Education Total Note: pension, debt servicing, subsidies, safety net, and external assistance a Sources: MOF Table 7.3. Ethiopia-Compositionof Public Expenditures FYI987to FYI994 A. Recurrent Budget Sector 1. General Administration 2. Defense 3. Economic Infrastructure 4. Education 5. Health 6. Other economic services 7. Other social services 8. Others(1) TOTAL B. Capital Budget 1. Economic Development 2. Economic Infrastructure 3. Health 4. Education 5. Other Social Development 6. Others TOTAL NOTE: Figures in per cent of total. For absolute expenditure levels see annex tables 2.1A, and 2.1B. (1) These include pensions, debt servicing, subsidies, safety net. and external assistance. SOURCE: Ministry of Finance Table 7.4. Uganda--Composition of Recurrent Expenditures (% Recurrent Expenditures) UGANDA 1993194 1992193 1991/92 1990191 1989190 1988190 Percentage of Total Recurrent Expen- Budget Actuals Actuals Actuals Actuals Actuals diture Economic Services Agriculture Infrastructure Other Social Services Education Health Local Government Other Defensd Public Administration TOTAL 'I Excludes Section 1 from the Ministry of Defense. Table 7.5. Uganda--Recurrent Expenditures (% of GDP) 1993194 1992193 1991192 1990191 1989190 1988189 Budget Actuals Actuals Actuals Actuals Actuals GDP at MP - U Sh millions 4,501,000 3,888,676 2,7 18,870 1,821,485 1,395,073 901,936 As Percentage of GDP at MP Economic Services Agriculture Infrastructure Other Social Services Education Health Local Government Other Defense/ Public Administration TOTAL 7.1% 5.4% 7.6% 5.6% 6.3% 5.7% Excludes Section 1 from the Ministry of Defense , 'Pable 7.6. llganda -Central Government Operations (As a %of CDP) Current Expenditure ** ** Repayment 1.8 2.2 1.7 I .O 1.3 1.5 2.0 3.2 4.6 3.2 2.7 2.0 ** ** Debt Relief 0.0 0.0 0.0 0.0 0.6 1.2 3.2 1.5 4.9 1.9 0.6 0.6 Domestic 2.8 3.1 2.2 3.4 2.0 1.5 -1.6 0.5 2.1 -0.6 -0.1 -0.8 ** tt ** +1. Bank -0.3 3.2 1.4 3.8 2.0 1.4 -1.4 0.2 1.9 -0.4 -0.3 -0.9 ** ** Non-Bank 3.1 -0.1 0.9 -0.3 0.0 0.1 -0.2 0.3 0.2 -0.2 0.3 0.1 Memorandum Item: Revenue and Grants 11.7 9.7 8.0 5.1 7.4 6.9 8.2 11.4 14.0 15.3 15.7 14.1 14.6 15.1 Overall Deficit plus Grants -2.1 -4.0 -3.3 -3.7 -3.9 -3.3 -4.3 -3.4 -7.4 -3.2 -3.3 -3.5 -2.8 -2.6 Source: Ministry of Finance, INF, and staff estimates Tablc 7.7. Economic Characteristics o f Infrastructure Projects i n llgrrndn (blillion o f USS) Project Name TIME 'TC FOREIGN LOCAL 0 I ) A COVER'T I'A/CONSU~.T ERR(%) WB: THIRD HIGHWAY PROJECT\ 84-87 74 58 16 58 16 7.12 58% WB: FOURTH HIGI~WAY PROJECT" 7.7 7.7 6.1 38% U K : ROADMAINTENANCE PROJECT^ 1.18 1.18 1.46 40% UK: ROADMAINTENANCEPROJECT IID 1.30 1.30 0.36 40% EC: NORTHCORRIDOR ROAD I~IPROVEMENT PROJECT N.A. N.A. 1.82 28.8% EC: NORTH CORRIDOR ROADIMPROVEMENTPROJECTI1 N.A. N.A. 4.94 16% EC: KAMPALA-MASAKA ROAD A+B+C PROJECT 3.36 3.36 N.A. 38% EC: KAMPALA-MASAKA ROAD D PROJEC~ N.A N.A N.A 46% EC: KABALE-RATUNA PROJECT 1.11 1.11 0.53 50% GERMANY:HIGHWAYSREHABILITATION PROJECT 3.56 3.56 3.34 N.A. GERMANY:HIGHWAYSREHABILITATION PROJECT I1 2.94 2.94 2.52 N.A. GERMANY:HIGHWAYSREHABILITATION OST-UGANDA PROJECI 5.6 5.6 3.08 N.A WB: RAILWAYS PROJECT^^ W B : WATER & SANITATION PROJECT WB: SECONDPOWER PROJECT NOTE: A. REQLJIRE400 MAN-~IONTHCONSULTANTSERVICES s.REQUIRE 515 MAN-MONTHCONSULTANT SERVICES C. WITH PERFORMANCEINCENTIVE US$0.05 MILI.ION D. WITHSTAFF EMOLUMENT US$O. 12 LIII.LION E. REQUIRE48 MAN-MONTH CONSULTANTSERVICES F. NEW UNITS WERE SET UPTOOPl?RATETllEPROJECT SOCIALSECTOR HEALTHRECOVERY EDUCATIONIV STRUC-TLIREAWUS'I'MENT ECONOMIC RECOVERY I ECONOMICRECOVERY H AGRICULTURE AGRICULTUREDEVELOPMENT Sour11W e nAGRICULTURE REHAHILITATION TABLE7.9. WORLD BANK PROJECTS UCASDA, 1984-1989 PROJECT SECTOR # OF PROJECT TC(MILLIONF U S ) O %OF PROJECT %OF TC INFRASTRUCTURE 7 197.2 39 69.6 SOCIALSECTOR 2 64.5 1 1 13.3 STRUCTURE ADJUSTMENT 2 115.7 1 1 24 AGRICULTURE 4 57.9 22 12 INDUSTRY 1 I5 5.6 3.1 TECHNICAL ASSISTANCE 2 33 1 1 6.8 TOTAL 18 483.3 100 100 Chapter 8 Economic Aspects of the Ethiopian Transition to Peace David Bevan 1. Introduction The war in Ethiopia is usually regarded as having delinking the objectives of the war(s) very largely continued for the thirty years from September from broader political or economic issues. 1961, when the first armed engagement between Third. prior to the lengthy set of wars under the Government and the Eritrean Liberation Front consideration here, the area presently covered by took place, until the fall of the Dergue in May Ethiopia and Eritrea had seen a very long history 1991, so that by any standards it was a lengthy and of territorial conflict of a brutal and feudal kind, very damaging conflict. However, while sheer with armies of both sides living off the land and duration is probably its most striking characteris- the victors being rewarded with grants of occupied tic, there are three other features which are central land, coupled with the services of the occupiers. to any analysis of the problems posed by the Sincethe prosecution of this traditionalltype of war transition to peace. required any army to sequester the agrarian sur- First, there was really not one war, but a set of plus, there was an alternative strategy to that of wars which took place in parallel: the Eritrean war direct military engagement, namely to pursue a lasted for the full thirty years, while the others scorched earth policy which denied the opposing were of varying but shorter durations. For most of army the means of sustenance. while incidentally the period, there was fighting on more than one inflicting great hardship on the civil population. front: what was common to each of these subsid- Thesetraditions were largely carried over into the iary wars was that they (almost invariably) in- more recent conflicts, which meant that the dam- volved the Government as a participant, not that age to productive activity was substantially greater they involved a coordinated or cohesive opposi- than would necessarily have followed from the tion. The intensity of fighting shifted between scale of the military actions themselves. Since the different fronts over time, as did the identity of the bulk of the action took place in the rural hinter- opposing groups in particular regions. This feature land, the consequences were particularlysevere for had two important consequences, one in splitting the rural populations and for agricultural activity the Government's forces, dividing its attention and and development. making its ultimate failure more probable; the Whileevents have moved rapidly since theend other in reinforcing the centrifugal tendencies of the war, not least in the secession of Eritrea, the which were in any case likely to be important transition is still at an early stage of a necessarily during the peace. protracted process. The longer run characteristics Second, the identity of the Government itself of the successor states are uncertain, so that both changed dramatically in the middle of the conflict the destination and the path toward it are difficult (and arguably partly in response to it), when the to discern. However the idiosyncratic features of Emperor was overthrown in the coup of 1974. the conflict itself are bound to loom large in any While the ideological orientation of the govern- likely transition. First. the evolution of agriculture ment and many of its policies were radically has been powerfully affected; provided the restora- altered,the powerful centralizing thrust of govern- tion of peace is accompanied by appropriate ment was unaffected.This had the consequence of policies, there must be scope for a very substantial and extended growth in production. Second. the to the various insurgents. In consequence peasant defeat of a distant, predatory and highly central- agriculture was subject to massive interventions, - ized government by a subset of a widely disaf- both by means of taxes, levies and confiscation of fected set of regions has pushed the likely succes- produce, and by means of actions to destroy live- sor state (in Ethiopia) towards a smaller. more stock, standing crops and productive assets These devolved and possibly federalist political structure. interventionswere compounded after 197 by the Interestingly, Ethiopia has escaped one typical revolutionarygovernment's attempts to caryrrout a inheritance of so extensive and protracted a war, sociaIist transformation of agriculture. While the namely extreme macroeconomic imbalance result- original impetus for these policies is not attribut- ing in a collapse in confidence and very high able to the war, the energy with which they were inflation. pursued, and the shape they took, were profoundly This paper examines salient features of the influenced by it. Ethiopian transition; it makes reference to the The rural economy of Ethiopia, particularly in companion papers which discuss particular themes the Northern Highlands, is characterized by a very in this transition, but does not summarizeor para- uneven degree of local self-sufficiency.In conse- phrase those discussions. It is structured as fol- quence inter-regional sales of grain from surplus to lows. Sections 2,3 and 4 examine the issues raised deficit areas, and migration of labor in the con- in the preceding paragraph in more detail, and verse direction have been crucial to sustaining trace the effects of the war and implications of adequate consumption in normal times and in peace for agriculture and the rural economy, for permitting survival in bad ones. The government the macroeconomy and for the structures of gov- did not content itself with directly attacking the ernment respectively. Section 5 considers the means of rural production, the producers and the possible responsesof variousagentsto the changed product in insurgent areas: it also acted to stifle situation further, and concludes. marketsand to prevent traditionalmigration. This was achieved both by legislation in respect of 2. The Rural Economy trading and the hiring of labor, and by direct military means such as bombing markets and According to the 1984 census, nearly 90 per cent vehicles.(Africa Watch, 1991). The effect predict- of Ethiopia's population was then rural, the over- ably was to severely damage market integration, whelming majorityof them peasants. Nomadicand reduce the rural economy's capacity to handle semi-nomadic pastoralists accounted for only 10 shocks and to create the conditions which turned per centof the population,though they ranged over the 1984 drought into a major human catastrophe. 60 per cent of the land area, and non-peasant Closely related to the war itself and to its agricultural production (formerly private estates, consequencesarefoursetsof policies; land reform, later state farms) engaged only 1 or 2 per cent of agricultural marketing, resettlement and villagis the population (Clapham, 1988). Since more than ation. Land reform was the central component in three-quarters of all Ethiopians are peasants, the the revolutionary government's plans to transform fortunes of the whole economy are necessarily rural agricultureand was instituted in April 1975, heavily dependent on what happens in peasant six months or so after the fall of the Emperor. It agriculture. It is conceivable for a civil war to be abolished all existing forms of tenure, making all conducted in a fashion that impacts relatively little rural land the collective property of the Ethiopian on the peasant economy, or where the main impact people; restricted the maximum size of holding; is to restrict the access of peasants to market prohibitedthe hiringof labor to work the land; and interactions, leaving them free to substitute into established peasants' associations, whose most relatively undisturbed subsistence activities. This important task was the allocation of land, both was very far from being the pattern in Ethiopia. As immediately after the reform and every two or noted in the introduction, the government viewed threeyears thereafter.(Their other major role was the peasant population as central to the conduct of to have been to induce a move towards voluntary the war, both as a source of resources for itself, cooperativisation, but in this they met with very and as a potential source of support and resources restricted success). Given the complexity of the previous tenure systems and the prevalence of tive was established, and its powers were consoli- litigation they had induced, the reform was carried dated and extended over the following several out remarkably successfully. In those areas under years. While there was always a parallel system of the government's control, (and these comprised by private traders, their freedom of action became far the greater part of cultivated land)the reorgani- steadily more circumscribed. They not only had to zation was rapid, effective and relatively complete be licensed and operate subject to AMC restric- (Clapham, 1988). In Eritrea and Tigray only a tions, which restricted grain movements between proportion of households was covered and this regions, but they had to deliver a proportion of varied substantially with the territorial ebb and their purchases to the AMC at official (low) prices. flow of the war. The identity of the gainers and This proportion was raised over time until it losers and the size of the gains and losses also covered all purchases in many cases. Meanwhile, varied substantially across the country. Where the a system of quotas was established in a hierarchi- pre-revolution peasants had owned most of the cal arrangement running from the regional admin- land they worked, they had little to gain from the istrations right down to the individual household, abolition of tenancy, and their plots were some- under the aegis of the peasants' association. The times smaller after the reform than they had been AMC was not in general able to achieve its quotas, before: but in areas which had been taken over by but the pressure placed on peasants was often settler landlords, as in much of the south. the bulk intolerable, and led to them having to make pur- of the population had been tenants, wage laborers chases at the (substantially higher) open market and landless. For them the gains were potentially price to meet their quota. The overall impact of substantial, either in obtaining access to adequately these arrangements was twofold. First, they rein- sized plots, or in being relieved of the exactions of stated the system whereby a large share of peasant the pre-revolutionary landlord (this relief was production reverted to non-peasants. However, frequently worth a quarter of the crop, sometimes - these exactions were now universal, and not lim- as in coffee-growing Kaffa - more like a half of it). ited as before to tenants; and in place of landlords, Since there had been relatively few large the beneficiaries were now the urban population, private estates prior to the revolution, and the especially the inhabitants of Addis Ababa, and the largest of these had been taken over intact as state army, or more generally the constituency favored farms, the reform did not increase the total land by the government. Second, they powerfully available to peasants appreciably, unlike land reinforced the impact of the other developments reform in some other countries. In addition, the noted above in ensuring the progressive fragmenta- very process which guaranteed peasants access to tion of Ethiopia into a set of insulated local mar- land in their own area denied them access to land kets. elsewhere, and halted the tradition of landless What the policies had not achieved was any peasants migrating south in search of land. Cou- very significant growth in voluntary co- pled with low rates of pay on state farms, the operativisation. The government's response was to prohibition on private hiring of agricultural labor, abandon the emphasis on voluntarism, and embark and restrictions on movements to the towns, these on a program of villagisation. A major national policies impeded the operation of factor markets campaign was operated between 1984 and 1990, just as those discussed earlier impeded the opera- but there had been prior experience, both in the tion of goods markets. In the context of a difficult south and in the north, of using villagisation as part climate and difficult terrain, the impact was to lock of counter-insurgency operations (involving the peasants into an initial configuration which be- construction of fortified "strategic hamlets"). came progressively less adapted to circumstances. Remarkably, a very substantial proportion of the The leaders of the peasants' associations also rural population was successfully villagised, at became progressively more divorced from their least in the south, despite the very considerable constituents, as the associations became instru- initial costs of compliance and the very uncertain ments of central government, rather than agencies and generally negative long run consequences. The for self administration. latter include increased difficulties in obtaining In 1976 the Agricultural Marketing Coopera- water and firewood, increased distances between dwelling and fields, and severe problems of live- reform had not provided peasants with any guaran- stock management,which sometimes enforced the tee of retainedcontrol of any specific piece of land abandonmentof traditionalmixed farminglherding (as opposed to rights to an allocation in the area). in favor of wholly sedentary agriculture. These In consequence the incentive to undertake long difficulties were exacerbated by the strategic term investments in terracing, planting trees and agenda in the program, which dictated that village the like was much weakened. Indeed, given the locations should permit rapid access (often along political control over allocation, energetic im- roads constructed using compulsory labor levied provement of land might actually shorten the from the villagers). Against this incomplete list of household's tenure of it. Far from encouraging costs, it is dificult to set any clear benefits, at least investment, the system was better designed to for the peasants. From the government's perspec- induce asset stripping and disinvestment. These tive, villagisation may well have increased the effects were magnified by the asset losses occa- possibilityof efficientextraction of theagricultural sioned by the war, both directly from military surplus, even if its main effect was to reduce the action, and in consequenceof the famines, which size of the surplus. The lack of violent resistance, forced massive slaughter and sales of livestock, particularly in the south, possibly reflected the including ploughing oxen. The role of the state in long history (dating back to the imperial cam- inducing or making investment in the sector was paigns of the nineteenth century) of severe repri- directed at the state farms and producer coopera- sals if the government's writ were not closely tives which yielded very low returns. followed. In any event, by 1988 the government Given this background, even the very lack was reporting that more than 12 million individu- lustre performance exhibited by peasant agricul- als had been villagised, or fully half of the rural ture over the lastdecade or two is quite creditable. population in the areas it then controlled. While yieldsdid not increase, at least they did not The other major policy of moving peasants fall, which could easily have been the consequence was the much longer range one of resettlement, of SO sustained an "economic war" on the peas- largely from the north to thesouth. There has been antry. However, since the area under cultivation much debate about the motives, necessity and was also stationary - and probably had been since consequences of this policy. In one perception, it the 1960s (World Bank, 1990) - while the popula- was a device for movingthe population away from tion was growing rapidly, per capita food produc- the control of the insurgents; in another, it was an tion trended strongly downwards. This required attempt to achieve the type of internal migration increased food imports with no increase in the which would have taken place atomistically and volume of export crops (and a sharp deterioration automatically if land reform and related policies in the terms of trade) to pay for them. In conse- had not inhibited it; in another, related to the last, quence, there was a growing macroeconomic it was a necessary response to the emergence of imbalance, as well as a deterioration in average sustainedfood deficitsandenvironmental degrada- nutritionaladequacy and an increasedvulnerability tion in the north. Whatever the rationale, there is to adverse shocks. no doubt that extensivecoercion was involved in Partly in response to the evident failure of recruiting individual settlers, ranging from the these policies, many of them were beginningto be conditional use of food aid, through quotas oper- abandoned or reversed,even prior to thefall of the ated by the peasant associations, to people being Dergue. Liberalization of internal trade had been rounded up at gun-point. The whole exercise was implementedsystematically in the areas under the very costly, both in material and human terms, control of the EPRDF from 1989, and the Aovern- with very high mortality rates amongst the reset- ment's own position shifted decisively rn 1990. tled (Africa Watch, 1991). This shift was in one sense overtaken by events, It was inevitable that this immensely disrup- with the defeat of the regime the following year, tive compound of policies and events would but since in many respects the incoming govern- severely damage the functioning of the rural ment pursueda similar reforming trajectoryto that economy, paralyse its capacity for adaptation, and belatedly initiated by the Dergue, there was more greatly inhibit innovation and investment. Land continuity in economic than in political matters from 1990. lation of the freight system and importation of new It remains to consider the likely impact of ve'hicles and replacement parts can be achieved these reforms. Preliminary studies of agricultural relatively quickly, but construction and rehabilita- supply response in Ethiopia (Soares, 1992) have tion of an always inadequate and now highly estimated supply elasticities of the type usually degraded rural road system cannot. found in such studies, positive but well below It seems clear that there is scope for very unity. It can be argued that the most significant substantial growth in agricultural output, if only short run impactof liberalization is not to be found because of the history of stagnation, low invest- in aggregate supply changes, but in the reintegra- ment and lack of innovation. Since this lack of tion of the regional markets. Studies have found a progressiveness is all too easily explained by the marked reduction in the regional price dispersion self-reinforcing mix of warfare and perverse for grain, and a reduction in the average trading intervention, there need be no inference that the margin, consistent with prior expectation (Azam, Ethiopian peasantry (who are in any case ex- 1992; Dercon 1993). tremely heterogeneous both ethnically and in Other aspectsof the transition are lesssatisfac- agricultural practices) are constitutionally lacking tory. M i l e much of the dismantling of institutions in dynamism. However, advancing from such a is inevitableand often desirable (as with the hated low base. and with a history so inimical to initia- producer cooperatives),some institutionsof value tive, it would be unreasonableto expect the growth have been damaged too. For example, the service to be very rapid. cooperatives had played a central role in the distribution of inputs and provision of services to 3. The Macroeconomy peasant farmers, and they, or similar successor agencies, will have a major and expanded role in The macroeconomic evolution of Ethiopia is the economic transformationof peasant agriculture notable in two ways. First, and unsurprisingly, which is vital to the longer run prospects of the aggregate growth has been erratic and on average Ethiopian economy. However, in the north, many very slow; since 1974 per capita income has of these cooperatives and their assets were effec- declined at roughly 0.5 per cent per annum. tively destroyed during the war, and most of those Though other sectors have grown more rapidly and in the rest of the country were incapacitated soon consistently than agriculture, the latter is dominant after the fall of the Dergue. This was a resultof the and its poor performance hasseriously retarded the power vacuum which was created in most locali- whole economy. Second, severe macro imbalances ties, and the consequent uncertainties, which emerged, but, surprisingly, these did not lead to permitted and encouraged asset stripping even of rapid inflation or to a rapid depreciation of the socially useful institutions (Kello, 1992). parallel exchangerate. Inflation rarely exceeded 10 Most serious of all has been the refusal of the per cent per annum, and the parallel rate depreci- TransitionalGovernmentto resolve the issueof the ated against the dollar at an average annual rate of ownership of rural land. The initial policy guide- only 4 per cent over the entire 30 years of the war lines made it clear that this would have to await a (Collier and Gunning, 1993). referendum following the eventual transition to It is clear that the budget was a major de- democratic government. If anything, this has stabilising influence in the macroeconomy for increased uncertaintyabout security of tenure and many years (Bevan, 1992, 1993a, 1993b). The the appropriability of the returns to individual interesting question is why this did not engender investments. While the government has success- more inflation than it did. The rapid monetary fully addressed some of the problemsconstraining deepening that made this possible may have been short run recovery in the sector, it has not done so a transitional process involving a permanent shift in respect of this crucial constraint on long run in the private sector's behavior, or it may have recovery and growth. Other requirements for the been a temporary and reversibleprocesswhich will long run are in hand, but inevitably some will be unwound in the future. For this to have hap- require many years to achieve. A major example is pened, events in the 1980s must have had an the provision of adequate transport, where deregu- asymmetric impact on savings and investment behavior, with investment becoming temporarily ing government comprised an economy in which infeasible or unattractiveto a much greater extent growth of output had failed to match population than was the case for savings. growth for nearly a decade; where aggregate It is extremely difficult to obtain a reliable investment had probably fallen below replacement picture of either investmentor saving in the Ethio- levels,and private investment all but disappeared; pian case. Government capital formation figures where private savings were very low and public are presently available only up to 1987/88,data on sector savings negligible; where the overall budget public enterpriseare incomplete,and those on the deficit was typically in excess of 10 per cent of privatesector are more or less non-existent.How- GDP, and substantially financed by borrowing ever, for what they are worth. the available figures from domestic banks; where there was severe suggest that in the mid 1980s,private investment import compression and a significantly over- accounted for only 2 percentage points out of a valued exchange rate. As remarked above, this national investment total running at the low level catalogue of problems had not yet engendered a of 14 per cent of GDP.While this may well be an runaway inflation, but the position was clearly not underestimate, it does imply an immensely inhib- sustainable. After a period of hesitation, the new ited rate of provision for the hture. Since these governmentdevalued the currency (from 2.07 birr figures are gross, it is hard to avoid the conclusion tothe US dollar to 5 birrto the dollar). Thisdid not that the private sector was dis-investing in real set off the inflationaryresponse which the govern- assets at a fairly rapid rate. This is not implausible, ment had feared, since the domestic price level given the options available. These did not include already reflected the scarcity of foreign exchmge. real estate to any great extent, since it was forbid- It also continued with the program of reform and den to own more than one modest house, and all market liberalization which had been initiated land, both rural and urban, was vested in the state. belatedly by its predecessor. This program in- Investment in productive assets was rendered cluded sweeping reforms of some aspects of the difficult or unattractive by a number of factors; regulatory structure impeding private initiative, these includedthewar itself(a major concentration such as the labor code and grain marketing ar- of industry had been located in Eritrea), the im- rangements, and a statementof its intention gradu- mense difficultyof negotiatingfor a site, the labor ally to limit its role in productive economicactivi- and other codes which virtually prohibited flexible ties in favor of the private sector. operation, the difficulty of obtaining inputs, and While the end of the war saw a large reduction the general.hostility of government to private in military expenditures, it also saw a collapse in enterprise. Taken together, these conditions must revenues (Bevan, op. cit.). Even with the signifi- inevitably have greatly dampened the enthusiasm cant increase in aid inflows which has taken place, of potential investors. if domestic financing is to be held within prudent Private saving figures are probably even less limits, it is only possiblefor total civil spending to reliable, but the observed monetary deepening be maintained at its war-time level, not increased. itself suggests that saving held up better than Whatever the long run implicationsof the end of investment in real assets, and whether or not there the war may be, there is certainly no short run was substantial capita1 flight, the domestically fiscal peace dividend in the sense of increased retained share was also substantial. This presum- resources for the government's social and eco- ably means that even though the confidence was nomic activities. lacking for real investment in the short run, there The previous point referred to the govern- was no comparable loss of confidence in the ment's resourcing in terms of flows; it remains to currency, or in the longer run possibilities. It was consider whether there is scope for temporarily acceptable to go liquid in the domestic currency relaxing the resource constraint by selling assets and wait on events (Since a sizeable share of these and generating capital revenues. This is particu- liquid funds were probably generated in parallel larly attractive in a context where the end of war, exchange dealings, the option of capital flight and the concomitant need for wholesale recon- would have been readily available, if desired). struction and rehabilitation, is combined with a The macroeconomic inheritanceof the incom- transitionfrom socialism,where the redefinition of the scope appropriate to the state includes a sub- the recovery in private investment. Since the land stantial reduction in the type and quantity of assets is now vested in the regional governments, not in it should hold. In Ethiopia, the state currently owns the central government, these revenues will not all land, both rural and urban; it also owns a high accrue directly to the central budget. However, the proportion of the urban housing stock. the bulk of regions are likely to be in receipt of grant aid from the country's industrial assets, and a great deal of the center for the foreseeable future, so this may military hardware, in addition to the economic make little difference in practice. infrastructure which is state owned, as in most Sale of government owned housing is actively countries. Leaving the last category aside, there under consideration, but is unlikeb to generate might appear to be extensive scope for realizing substantial revenues in the short nln. Current rents asset values, and generating revenue, while shift- are very much below market rates, the bulk of the ing to a more appropriate portfolio. Unfortunately, occupiers have very low incomes (and many are there are difficulties with each of these assets. not on the public sector payroll, so could not be It is essential that some redefinition of rural compensated for rental rises), and could not afford land rights be made to increase peasants' actual and to buy: sales would have to be at very deep dis- perceived security of tenure and provide assurance counts or to third parties, with the danger of subse- that they will be able to appropriate the returns to quent eviction. Privatization of productive para their own efforts and investments. In general, statals is also in process, but as elsewhere in much the simplest and most reliable way of achiev Africa, this will inevitably be a lengthy business, ing this is to vest ownership of small holdings in and will not yield large revenues in the short run. the peasants themselves. However, achieving this Finally, sale of military hardware as scrap is in Ethiopia would probably be a politically fraught, unlikely to generate large revenhes, and sale in acrimonious and lengthy process, partly because it working order raises difficult mcrral and political would not involve a return to a clearly defined issues. original position, partly because of the large All in all there are a range of political, social amount of involuntary locational shifting that has and economic constraints which mean that the taken place over the last 15 years. In any case, desirable switch in the government's portfolio is vesting the land in peasants as poor as these is not unlikely significantly to ameliorqte its resourcing 3 mechanism for raising substantial cash revenues. constraints. Equally, it leaves the government with If it were seriously intended to sell government a problem if success in creating a climate for owned rural land, rather than merely transfer title, private investment leads to a swkch out of liquid then this would have to be to third parties, and assets as private portfolios are switched into real would re-create a class of landlords; it does not assets. It must either create an attr~ctivelonger run appear conceivable or desirable that the govern- government liability (which would also have the ment should obtain capital revenue by auctioning perverse effect of partially crowding out the pri- re-created rental rights over peasants. For whatever vate investment recovery) or run a domestic sur- reason, the government has refused to address the plus to correct the domestic finanking balance. To issue of ownership to date, so this is not on the the extent that it can engineer rapid economic agenda as a revenue device. growth, and/or an increase in the private propen- The circumstances governing urban land are sity to save, these difficulties will be correspond- rather different; while land occupied by dwellings ingly reduced. carries a similar political charge, and may have In the longer run, the government should be somewhat similar properties to those just discussed able to engineer some recovery ih domestic reve- (occupiers do not currently reimburse the state for nue, partly by reform of the relatively antiquated the implicit rental; however they may still end up tax system, partly as economic recovery shifts paying it if they are in the private sector), land for activity into a more tradable, and hence more industrial use does not. While outright sales of taxable, configuration. However, the share of freehold are not on the agenda, sales of long leases revenue in GDP will not recover to the levels certainly are. These may generate considerable reached under the previous regime, which involved revenues, depending on the strength and speed of a politically unacceptable and economically dam- aging degree of coercion. In the plausible circum- imposed by force after the fall of the Dergue, and stance that donor preparedness to close the fiscal the EPRDF, in its new incarnation as the Transi- gap does not remain indefinitely at its present high tional Government of Ethiopia, adopted a strategy level (around 13 per cent of GDP), it may be of recognizing ethnic differences within a loosely difficult to maintain spending at its present level defined federal structure. The secession of Eritrea (around 34 per cent of GDP). While the present was bitterly resented by those who still subscribed macroeconomic positionappears to be comfortably to the concept of an inclusive and centralized under control, the medium term will require care- Ethiopian state, but attempting to block this seces- ful managpent. A tapering of aid inflows, cou- sion was never really an option for the EPRDF. pled with a sluggish revenue response to tax Given its political base in Tigray,a small, poor and reforms and an unwillingness to grasp the nettle of distant part of the country, and given that it had expenditure control would lead to a rapid re-emer- itself no secessionist aspirations, the EPRDF faced gence of high domestic deficits in circumstances a problem. If Ethiopiacould no longer be governed where they had ceased to be financeable. by military force, least of all from the periphery, it would be necessary to build some form of ruling 4. Government Structures coalition. However, this requires different skills from those appropriate to waging a guerilla war. Changes as momentous as those that have taken Equally serious, if a coalition has to be forged place in Ethiopia are bound to have major conse- out of the country's different nationalities and quences for the nature of government and its regions, is the disparity betweenthese. Tigray is in structures. This section focuses on three of these. many respects atypical; it is coherent, with a strong Both the Dergue and the imperial regime which sense of regional identity, and no serious problem preceded it were highly centralized and highly of internal minorities (it has not been subject to authoritarian; in addition, until almost the very end inward migration). It has been an integral part of of its period in power, the Dergue was strongly Ethiopia from the beginning, aspires to a substan- committed to a socialized scheme of ownership tial degree of autonomy within it, but stands to and control, and to a rejection of market processes. gain substantially from transfers from the center. The military defeat of the Dergue, the natureof the Most of the other regions are much less clearly coalition that defeated it, and the evident failure of defined both as to nature and as to interests; a its economic program jointly implied a revulsion process of sharpening this definition should logi- from these three attributes. cally precede the process of coalition-building. In The unifying theme in the history of govern- the meantime, it is extraordinarily difficult either ment in Ethiopia, running from Tewodros through to start negotiations, or to agree on a concrete form Menelik and Selassie to Mengistu, was the attempt for the ensuing federal structure. to create an ever more powerful centralized state, There has been a great deal of activity on while imposing the progressive subordination of devolution, with proclamations defining (at least regional movements and nationalities. In the later broadly) the physical boundaries and rights and years, this saw the government fighting a succes- responsibilities of the regional governments, and sion of unwinnable wars, in ever more desperate announcements of budgetaryallocations. However, attempts to enforce unitary control, in the face of the real shape of the future structure, including endemic economic failure and in the face of in- how truly federal it proves to be, remains pro- creasing alienation of the population,even in areas foundly obscure (Bevan, 1993~).On the resolution which were not in revolt. Since defeat was of this issue, much depends. It appears that there achieved by forces from the periphery, and not by can be no going back to the old centralized pattern, an insiders' coup, it never seemed likely that this but what has replaced it is still evolving. It is centralizing thrust would be maintained after the unclear whether this evolution will settle in a fall of the Dergue; a more serious danger is that the configuration which permits a healthy federated whole Ethiopian state will unravel after the fashion Ethiopia to flourish; in one which is plagued with of the USSR. instability, political tension and unproductive It is clear that national unity could not be ethnic rivalry; or whether the country will end up tearing itself apart with all the horrors of ethnic effectiveness in difficult circumstances. cleansing culminating in further secessions. The direct efficiency costs of retaining a large It is also difficult to discern how complete a public sector presence in productive activities need transition will be achieved from Ethiopia's authori- not, therefore, be high, but two aspects of it may tarian tradition of government to a more respon- prove more problematic. The first relates to the sive, democratic one. The original timetable within general perceptions of the business community, which the Transitional Government undertook to who remain deeply skeptical of the government's hold national elections has already been extended. willingness to create an appropriately enabling While the scale and urgency of the agenda for environment for their activities, and whose skepti- economic reform and the need for prior work on a cism is reinforced by its intention to remain a new constitution make this understandable, the major player on the allegedly level playing field. precedents of deferred elections elsewhere in the The other danger is that current arrangements may continent are not encouraging. Questions have also encourage the regionalisation of public enterprise been raised about the conduct of the regional activity. Regional governments are likely to be elections, which have already been held. While short of revenue sources, and it will be tempting to continuation of an Ethiopian state clearly requires utilize their control of industrial land to engage in that a viable form of power sharing be developed, extensive joint or sole enterprise activities using there is no guarantee that this will be embodied in land both as a bargaining counter and as a potential democratic processes. These issues should begin to barrier to competitive entry. Even if the successor be clarified in the very near future. regime in central government adopts a stance more Since the nature of the successor regime is committed to private enterprise, it might not be unclear, so is the extent of future disengagement able to ensure that this was pursued nationally. from socialist interventions. The retreat from Whatever the balance betw'een public and wholesale market intervention seems to be well private participation may be, the two sectors established, and there does not appear to be any jointly face a daunting challenge in mobilizing a constituency for systematic reversion. However, very large and substantially unemployed urban while the liberalizing thrust of economic reforms labor force (Mengistae, 1992). A survey conducted may be secure, attitudes to the ownership and in 1990 implied that half of the: younger urban control of productive assets are much more ambig- adult population was then unemployed, and the uous. The unwillingness of government to contem- situation will have worsened follbwing demobili- plate outright sales of land has already been dis- zation, since many ex-soldiers remained in the cussed. There is also a great deal of policy ambiva- towns, particularly Addis Ababa. This level of lence on the relative roles of public and private unemployment, apart from being a potent source of sectors in production, with no unambiguous com- poverty and its associated misefy, is a potential mitment to privatization. Indeed, the TGE has source of political unrest. To date, however there reserved a number of industrial activities to be the has been relatively little sign of increased crime or exclusive preserve of the state, on the ground that banditry following the end of the war and the they are "essential for the development of the demobilization (Collier, 1993). economy" (Eshetu Chole, 1992). On the other hand, despite the large extent of the state-owned 5. Conclusion enterprise sector in Ethiopia, it has probably been less problematic than elsewhere on the continent. Amongst the many unknowns that make any While the previous structure of incentives, price present assessment of Ethiopia's prospects so controls and protection meant that they were speculative the most important are the ways in sometimes foreign exchange absorbing rather than which different agents, both public and private, foreign exchange generating, they were usually respond to the opportunities and challenges facing profitable, unlike the experience of many African them, and to each others' actions. Circumstances countries (World Bank, 1985, 1989). Indeed, now permit major beneficial shifts in behavior, but Ethiopia has been remarkable in the region in its these require both that ingrained habits be broken capacity to maintain a high level of managerial and that individuals have some confidence in the stability and reliability of the new economic and rathertransferable between differentactivities. The political environment.Unfortunately governments, problem, as indicated earlier, may be more a whether old ones proclaiming they have reformed, matter of residual mistrust of government by the or new ones claiming they are different, have a business community, coupled with some genuine seriouscredibility\problemin announcing that the remaining "objective" difficulties in thebwayof environment will in future be more supportive to private productive activity. Experience of struc- private agents. The only reliable way of establish- tural adjustment in Africa providesa warning that ing a reputation is to earn it in execution, by private investment may take a very long time to building an observable track record. If private recover, and it is difficult to know where to place agents reasonablydecibe to defer irreversible and Ethiopia in this context. On the one hand, the scale costlydecisions, @d wait and see, the payoff to the and depth of the intrusion by thestate was extraor- new government strategy will certainly be post- dinarily great, so that suspicion and distrust might poned, and\may be aborted altogether. be more entrenched than elsewhere. On the other, While there may be no short cut in this pro- the fall of the previous regime has been particu- cess, it is certainly possible to prolong it, by post- larly decisive and irrevocable, and there does exist poningcrucial choices, leaving major policy issues a substantial diaspora of successful emigre Ethio- ambiguous, and generating mistrust and uncer- pian businessmen. In consequence, the familiar tainty. The Ethiopian authorities are in somewhat gap between inexperienced and illiquid local of a cleft stick; it has been mandatory that they entrepreneursand a reluctantinternational business orchestrate a major break with the past, and the community probably does not hold in the case of choices involved, such as those over the new Ethiopia. constitutionand the regions, require a great deal of Recovery in the rural economy is also subject time. Indeed, in some respects it would be more to uncertainties. Apart from that associated with appropriate to criticize them for moving too the climate itself, the most significantis the ques- quickly rather than the converse. However, there tion of tenure. One complicatingfactor is that the areareas in which it would have been possible and resettlement and villagisation programs have left desirable to have moved more quickly, and which many farmers in locations they are desperate to have still not been resolved. One is the vexed issue leave. Thereare other displaced persons in need of of finding a mechanism for ensuring adequate relocation, notably many demobilized soldiers. tenure of land; another is the need to demarcate Some relocation has taken place spontaneously, very clearly the conditions under which private but the process is very far from complete. There is enterprise may operate and the safeguards on also the ruinous state of much of the transport which it may rely, not least from incursions by the system. In these circumstances, the rate at which state. investmentand innovation will occur is also highly The speed with which the economy is able to uncertain. emerge from its current state depends on releasing There are two main conclusions from this energies which have been dormant or directed into discussion, neither surprising. First, whilethe main privately but not socially profitable uses. One problems and opportunities facing Ethiopia are difficultyis that, just as honing the skills required relatively easy to characterize, the transitional to win a guerilla war may be a poor preparation for process is at an early stage, and it is profoundly government by coalition, so the skills required to unclear how it will evolve. Second, the sheer scale flourish in a set of black markets may not lay the of the damage, and the scope of the changes that foundation for successful productiveentrepreneur- are needed means that theadjustment horizon must ship. In this case, private agents with liquidity may be very long indeed. Fortunately, the end of the not be best placed fruitfully to invest it, while the war does not appear to have unleashed an unsus- banking system has no experience of the type of tainable revolution in expectations, so that Ethio- intermediation required to resolve the difficulty. pia may be spared the intolerable political pres- However, experienceof the subsequent success of sures under which some adjusting governments those who made fortunes in wartime elsewhere have to operate. suggests that entrepreneurial skills may in fact be Chapter 9 Economic Aspects of the Ugandan Transition to Peace Paul Collier and Sanjay Pradhan 1. Introduction: The Economic Meaning of 'War' This paper applies aspects of the analytic papers Asian repatriation, Northern Uganda has not yet developed in the study to the case of the Ugandan been entirely integrated into the present civil civil war and the subsequent recovery. Section 2 society. considers the macroeconomy and some sectoral Once a governmentdeclares'economicwar' on disaggregation. Sections 3 and 4 consider private a part of its population, it abandons the central and government behavior respectively.Section 5 economic functionsof government: the provision considers the third important actor, namely the of impartial arbitration, protection and contract donors. First, however, I discuss the economic enforcement. While some of these functions can in meaning of the Ugandan war. principle be provided privately, in Uganda as in Whereas international wars can invariably be most societies, the state had reserved to itself the precisely dated, civil wars are less precise. Here I monopolyof supply and so the sudden cessation of will date the start of the war to 1972, the time of state provision could not be met from alternative the declaration by the Amingovernment of 'Eco- sources. Hence, a defining economic feature of the nomic War' on its Asian community and the Ugandan civil war was that public officials, mili- resulting expulsion of that community. Civil war tary personneland privateagents were to a consid- subsequently engulfed far larger groups than the erable extent able to behave outside due process of Asian community:during the Amin era (1972-79) law. This extended beyond the expropriations up to 500,000 Ugandans died as a result of the suffered by Asian entrepreneurs. Many of the regime. Exiles in Tanzania mounted two invasion Asian businesses were acquired in a rather con- attempts, the latter in 1979 being successful. fused and ad hoc manner by soldiers, ministers and Whereas the 1979 campaign was short and deci- other members of the political elite. There was sive, the first part of the 1980s was one of extreme generallydefacto control but not the acquisitionof military instability as successive governments good title, giving rise to insecurity of tenure and gradually lost control of territory to insurgent non-marketability of the enterprises. In these forces. By 1985 some 7% of the population was circumstances there was a tendency for those who displaced or refugees. In January 1986 the Na- currently controlled the enterprise to strip it of its tional Resistance Army gained control of Kampala assets. and secured more complete territorial control the The economic consequences of civil war in following year, this marking the start of the return Uganda are not, therefore, primarily related to to peace. Just as the Asian communitywas the first military expenditureand material destruction, they to be embroiled in the war, it has been among the are about the removal of legitimate authority. It is last to be incorporated in the peace. The return of this which makes the economic consequences of confiscated assets is still in process, and to date civil war quite different from those of a conven- only a small proportion of the community and tional international war: civil war removes legiti- those assets which it expatriated have returned mate authority. Consequently, the restoration of despite strenuous efforts on the part of the NRM 'economic peace' is not the automatic corollary of government. In addition to the slow process of militaryvictory, it is the reconstruction of systems economy is still far short of 1971 per capita of legitimacy. production. In the first two years of peace there The longer a society stays in a state of civil was rapid growth of over 8% p.a., but this was war the more do conventions of legitimateconduct followed by four years in which growth was decay. In this sense, the period 1972-86 was a modest. Only in 1992/93 did growth recwer. It fairly continuous descent from civil society, while may be that peacedelivers some readily rctliizable the post-1986 period is betterconceptualizednot as economic gains, but that sustained rapid recovery post-war material reconstruction but as the gradual depends upon the restoration of prl-,*qtesector re- emergence of the institutionsand conventions confidence wnich in Uganda is only r w happen- of civil society. 1971 marks the last full year of ing. peace, 1986 marks the nadir of economic civil war, During the war the decline in per capita GDE and 1986-93 marks the period of a graduai and exceeded that in GDP. This was for both exoge- partial return to economic peace. nous and endogenous reasons. Exogenously, the terms of trade deteriorated though with major 2. Aggregate Consequences of War and fluctuations. Endogenously, the disproportionate Peace: Production,Expenditure and Their decline in the export sector (discussed below) Composition reduced the real income of the economy, being equivalent in terms of income loss to the introduc- I begin by describing the salient changes in the tion of severe trade restrictions. A second endoge- levei and composition of production and expendi- nous income loss was the decline in aid. Neither ture during and after the war. I then introduce a the Amin regime nor its successors were very disaggregation which attempts to rank activities by attractive to donors and had little appeal to com- their vulnerability to war. mercial banks. Hence, relative to need, Uganda Between 1971 and 1986 GDP declined by was considerably underborrowed by 1986. In- 13%. Since fertility far exceeded mortality, this is deed, during the first half of the 1980s the balance an amazing scale of contraction. For example, a of official merchandise trade was approximately normal slow-growing African economy would zero. Further, as argued below in our analysis of have expanded by around 3.5% p.a. during this private response, there is likely to have been a period. Had the Ugandan economy grown at this private capital outflow. There were thus three rate it would have been double its actual 1986 endogenous reasons why expenditure relative to size. To a small extent the decline in output GDP was lower than in the counterfactual. Post- relative to this counterfactual is due to endogenous 1986 there was a more rapid reversion in aggre- losses of the laborforce through emigration. gate expenditure than in aggregate production. Although only a small proportion of the labor First, the shift of resources back into the tradable force left Uganda, this was disproportionately the sector produced a trade-liberalization effect, most productive. The expelled Asian community enhancing real income for a given level of pro- was highly skilled, and the Arnin regime systemat- duction. ically targeted the educated. Hence, not only did Secondly, there was a resumption of aid flows these groups have the best earning opportunities on a large and conceivably unsustainable scale. abroad, they had pressing reasons to leave. How- This seems to be a common pattern in war-peace ever, a decline on this scale must predominantly transitions, for exarr,ple in the first three post-war be explained by a reduction in the productivity of years Zimbabwe was able to run a payments a given stock of labor. This was partly a matter of deficit approaching ten percent of GDP. Thirdly, reduced productivity in all sectors, and partly there was some repatriation of private capital. In reallocation between sectors to those with initially 1992 private capital inflows totalled $200m. lower returns. With the return to peace post-1986, Offsetting these endogenous gains there was an some of this output loss has been recovered. exogenous loss due to the collapse in the world Between 1986 and 1992 GDP has grown by 37%. coffee price. The resulting collapse in export Althoughabove the growth rate for the region, the earnings is temporary in that the coffee price will probably recover somewhat from its historic low, and the Ugandan economy will diversify its characteristics: transactions-intensity and asset- exports away from the extraordinary dependence intensity. Transactions depend, to a varying upon coffee which was a by-product of the war, degree upon the institutionsof civil society, but at however, during the post- 1986 period the endog- a minimum expose the transactors to a degree of enous gain was approximately offset by the exoge- visibility and potential predation. Assets are nous loss. vulnerable because their continued possession The severe decline in the size of the economy depends upon defensible property rights which are was accentuated by a change in its shape, re- eroded by civil war. Because activities vary in sources moving to less productive activities. Table these two intensities, civil war changes their 9.1 depicts the change in the composition of GDP relative cost and so shifts resources between them. during the period. The Table is organized around The second mechanism by which the composition the concept of transactions costs pioneered by of activity is changed is that some other sectors North (1992). There are two broad mechanisms supply either transactions services or assets to the by which civil war changes the composition of an directly productive sectors and these face a col- economy. First, directly productive activities vary lapse in demand. in their vulnerability to civil war according to two Table 9.1. The Composition of GDP by War-Vulnerability (% share of GDP) at 1987 constant prices Manufacturing Marketed agric Subsistence 30.5 39.2 39.2 35.8 Transaction-providing Transport Commerce Asset-vulnerable Livestock (inc sub) Construction Note: The National Accounts provide data at 1987 prices for 1981-92. The 1971 data has been put onto an approximation of 1987 prices as follows. Both 1971 and 1981 data are available at constant 1966 prices. The estimated share of sector i in GDP in 1971 at 1987 prices is then its share at 1966 prices multiplied by (share of i in 1981 at 1987 priceslshare of i in 1981 at 1966 prices). Table 9.2. Index of Components of GDP by War-Vulnerability (index, 1972=100) value-added at 1966 prices, 1971-86 at 1991 prices, 1986-92 Manufacturing 100 37 73 Export crops 100 (75 88 Marketed food 100 { 97 Subsistence 100 100 127 Transport 100 85 113 Commerce 100 58 86 Livestock (inc sub) Construction 100 45 96 Memo GDP 100 87 119 Government 100 166 198 Four sectors might be distinguished as 'di- channels can be highly informal and food is rectly productive', but having very different sufficiently standardized that transactions can be degrees of vulnerability to war: manufacturing, anonymous with little cost, since there is little role export crops, marketed food production, and the for reputation. Although more vulnerable than subsistence economy. Of these four sectors it is subsistence in that a transaction is involved, evident that the subsistence economy is the least marketed food production involves less asset- transaction- intensive. However, it is still exposed holding than subsistence since output can be sold to war. First, it is still to an extent dependent immediately after harvest. Arguably more vulner- upon transactions, since it is only 'subsistence' in able than marketed food is marketed export crops. the sense that its output is not sold, whereas it The decision to produce these crops is the deci- depends upon purchased inputs. Secondly, it sion to sell or stor2 them, since they cannot be depends upon assets some of which are acutely consumed. Hence, if there is a risk of increased vulnerable to war, namely stores of seed grain. A predation, rhe safe course of action is to grow hungry army will attempt to asset-strip subsistence food crops, the marketing decision being post agriculture. These two negative effects may ponable. Further, export crops by their nature explain why, although the subsistence sector have a higher value-to- weight ratio than food gained very considerably relative to other sectors, crops and so predation (for purposes other than its total output was no higher in 1986 than in hunger) is more attractive, and they must be 1971. The next activity up the scale of vulnerabil- transported right through the country rather than ity is likely to be marketed food. The marketing just to the nearest purchaser of food. The most vulnerable of the directly productive sector. but its output is irreversible investment activities is manufacturing. First, it is transac- which cannot be removed from the country. The tion-intensive. The ratio of the value of transac- willingness to make this sort of commitment tions per shilling of value-added is much higher evaporates during the uncertainties of war and so than in agriculture and the number and variety of the construction sector suffers a demand collapse. transactions is also higher. Further, the output is This behavior has persisted through to the present. less standardized and so reputation is more impor- The supply response to the change in transac- tant, requiring repeated transactions with the same tions costs and the resulting changes in demand agent and a higher profile. In addition to transac- for transactions services and asset-related activi- tion- intensity, manufacturing is asset-intensive ties in turn changed relative prices. The sector- and its scale economies imply that assets are specific GDP deflators are shown in Table 9.3, geographically concentrated. however, for much of the period they are unfortu- This pattern of vulnerability is consistent with nately highly unreliable. For what they are worth, the changes in the shares of the four activities prices are shown relative to the deflator used for during and after the war. The sector which bene- the subsistence sector, and with the relative price fited most from the war was subsistence agricul- structure in 1971 set to unity. Thus, the numeraire ture, which increased from 30% to 40%of GDP. is subsistence output. For example, the figure of The sectors which contracted most severely were 152 for commerce in 1986 implies that the price market agriculture and manufacturing. Official of a unit of value-added in commerce rose relative exports, which were an important subset of mar- to a unit of subsistence by 52% between 1971 and ket agriculture, declined in overall volume by 1986. The changes in transaction-using prices around 60%. Indeed, only coffee survived as an relative to subsistence are modest. The deflator on official export activity, all other official exports subsistence is, however, derived from that for declining by close to 100%. Manufacturing more marketed agriculture and so would be unlikely to than halved as a share of GDP. Post-war, the deviate substantially from it. The manufacturing structure of output has tended to revert but only deflator may, however, indicate that despite the partially. Subsistence activities have grown least war the trade regime was relatively liberal. That (25%), though remain at 36% of GDP, so that is, the collapse of the customs seIivice implied de only around half of the retreat in subsistence has facto trade liberalization,so that domestic manu- been reversed. Manufacturing output has more facturers were restricted in their capacity to raise than doubled, but is still one third smaller relative prices. Manufactures entered the economy to GDP than in 1972. through informal trading channels which were not The next group of activities identified in substantially inconvenienced by the war. On this Tables 9.1 and 9.2 supply transactions services: account, the rise in unit costs in manufacturing transport and commerce. In aggregate these due to its transactions intensity could not be services indeed reached their nadir as a share of passed on to consumers and so resulted in a yet GDP in 1986 and have thereafter partially recov- more severe quantity contraction. Turning to the ered. The argument is that since these sectors are transaction-providingsectors, the apparent fall in supplying transactionsservices they are peculiarly the relative cost of transport during the war period vulnerable. They are also intensive in lootable is evidently spurious, presumably failing to take assets. into account quantity rationing. The sharp rise and The final pair of activities identified in the subsequent fall in the unit price of commerce is Tables are asset-related. Livestock is an activity consistent with what might have been expected. which is peculiarly intensive in lootable assets and The sharp rise in construction costs despite the so particularly vulnerable: it is harder to conceal collapse in demand for construction services livestock than to conceal grain. The construction demonstratesthat there was a severe cost shock to sector is not only subject to a transactions- inten- the sector. sity similar to that of the manufacturing Table 9.3. Relative Price Changes by War-Vulnerability (relative to subsistence, 1971 relative prices = 100) Transaction-using Manufacturing Marketed agric 100 Subsistence 100 Transport 100 Commerce 100 Construction 100 Less is known about the change in the compo- increase in government expenditure was skewed sition of expenditure than that of production towards consumption. Whereas private consump- because there are not yet any expenditure National tion rose by 20%, public consumption rose by Accounts. However, the war gave rise to two nearly 50%. major shifts in expenditure composition. First, To summarize, war caused a collapse in expenditure shifted away from investment. The aggregate output and an even more severe col- share of investment in GDP fell from 12% in lapse in expenditure. It shifted production out of 1971 to 2.9% by 1978 (the peak of the coffee transaction-intensive, transaction providing and boom) and thereafter if anything declined further. asset- vulnerable sectors, and expenditure out of Secondly, expenditure was privatized: the share of investment and the public sector. During nearly government expenditure in GDP fell from 20% in seven years of peace the economy has only par- 1971 to 10% by 1986. Between them, these two tially reverted to its pre-war characteristics. responses cushioned the other expenditures of Expenditure levels and composition have reverted private agents against the 40% fall in per capita much more swiftly than production. Both the slow production. Both of these changes in composition recovery of production and the swift recovery of were reversed post-1986. Investment increased by expenditure warrant further analysis. The incom- about 70% during 1986-91, whereas consumption plete recovery of production might reflect an rose by only around 27%. Within investment, incomplete fall in transactions costs rather than there was a modest switch towards non-tradable just slow adjustment to a complete reversion in capital goods as opposed to imported capital. transactions costs. Below we therefore investigate Most of the investment recovery was private: further what is involved in their reduction. The private investment more than doubled whereas rapid expenditure reversion has been achieved by public investment rose by around 40%. By 1991 the public sector raising its consumption whereas private investment was slightly higher than public the private sector has raised its investment. The investment. The leading role played by the private recovery in the share of public expenditure re- sector in investment recovery was despite a strong flects both the increase in the tax base and the recovery in public expenditure. By 1991, govern- effects of aid. The recovery in the tax base is ment expenditure had fully recovered its pre-war brought about because taxation falls dispro share of GDP, being back to 21%. However, the portionately upon the tradable sector of the econ- omy and least upon the subsistence and smuggling the economy is subject to external aid shocks. parts of the economy, so that the shift in the Hence, private agents attach some probability to structure of the private economy endogenously a high-inflation scenario. Interviews with the raises the share of public revenue. However, in business community in March 1993 established Uganda this effect has been fairly modest pre- that the consensus expectation%f the exchange cisely because the private recovery has been so rate for the end of 1993 was 1500 shillings per incomplete. Whereas in 1972 revenue was 15%of dollar, as against 1200 at the time of the inter- GDP and at its trough in 1986 was only 4.5%, it views. The differential between domestic and has only recovered to 8%. The major reason for foreign interest rates prevailing at the time were the recovery in government expenditure has consistent with those expectationsof depreciation. therefore been due to the channelling of aid However, in the event, the government's tight through the budget rather than directly to private macroeconomic policy was sustained so that by agents. Since the sustainable recovery of public late-November 1993 the actual exchange rate had expenditure depends upon the recovery of taxable appreciated to 1150. The result was that real activities, the primary focus of attention should interest rates for borrowing denominated in therefore be upon the impediments to that recov- domestic currency were around 30%, whereas the ery. Why has the private sector not shifted re- real interest rate for those who were financing sources back into the tradable and marketable investment in foreign currency Was around zero. components of the economy more swiftly? I turn This created considerable tewion, since the to private responses to the return of peace. entrepreneurs with foreign assets were dis- proportionately Asian, whereas newly entering 3. Private Responses to Peace African businessmen needed to borrow locally and were thus at a heavy disadvantage. This configu- Asset Responses ration of incredible disinflation leading to high real interest rates for a part of the business com- In War, Peace and Private Po~olios(Collier and munity is not unique to war-peace transition, but Gunning, 1993) we suggested that during civil it may be a common feature given the conjunction wars entrepreneurs build up large holdings of of dollarisation of financial assets and the likeli- financial assets and that in Uganda these assets hood of bouts of high inflation during the transi- took the f ~ r mof dollars. Government estimates of tion period as the budget is subject to shocks and coffee smuggling imply that during the coffee the demand for money function shifts in an unpre- boom of 1976-79 unofficial exports were $520m. dictable manner. For example, no matter what a While some of this will have been counterbal- post-war Angolan government were to do, it anced by unofficial imports, this gives some idea would be unable initially to generate a credible of the possible scale of foreign asset holdings. I expectation of sustained single figure inflation. begin with an implication of this high liquidity in The domestic financial sector is doubly dam- foreign assets for domestic financial markets and aged by civil war and its aftermath. First, using then discuss the switch back into domestic real the classification of the previous section it is assets. Because of the large accumulated holdings transactions-providing, and so contracts sharply of foreign exchange, the capital account is de during the war. Additionally, it is intensive in a facto open even prior to de jure liberalization. government-providedservice, namely a numeraire Hence, once the domestic financial market is and this is undermined both during and after the liberalized, the domestic interest rate becomes set war. In those sophisticated economies in which by the conventionalopen capital account condition the governmenthas a history of high and variable that it should equal the interest rate on dollars plus inflation, such as much of Latih America, the the expected depreciation of the shilling against government nevertheless provides a numeraire the dollar. In turn, the expected exchange rate through the publication of a reliable CPI, permit- reflected the credibility of the government's ting indexation of contracts. In Uganda the gov- macroeconomic strategy. A transition government ernment is not trusted to produce a price index. In has had little opportunity to acquire reputation and fact the government has invested in an accurate and rapid consumer price index, but in interviews while peace enabled fairly small-scale land trans- it was evident that its use for purposes of index- actions to take place. transactions in other assets ation of contracts was not practicable given pri- remained very difficult. In effect, social conven- vate skepticism. Indeed, an article in Uganda tions related to land transactions were sufficiently Confidential had claimed (inaccurately) that the strong that the restoration of basic social order CPI was manipulated by the government. Hence, was sufficient to permit them without recourse to the domestic financial sector lacks a numeraire the more complex legal framework needed for which can be trusted to maintain its value. The other asset transactions. A second marked change government therefore faces the dilemma that it was the reduction in urban-to-rural migration. either grants the private sector the inflation which This is again consistent with the reversal of the it expects, achieving low real interest rates but retreat to the rural economy induced by the war. high inflation, or maintains low inflation. yielding Private investment tended to be concentrated high real interest rates until private expectations in three types of capital, housing, transport equip- adjust. The latter option continually wrong-foots ment and machinery for manufacturing. The the private sector. For example, during 1992-93 housing boom is picked up in the National Ac- the government disinflated from an annualized counts, where the construction sector expands inflation rate of 230% to 0%. Although this was substantially more rapidly than GDP, and in the unexpected,so that real interest rates became very Bigsten and Kayizzi-Mukerwa survey, which high; the disruptioncost was remarkably low. The found a high incidence of house construction. The economy grew at 7.2% during the year. The concentration on vehicles and machinery is found reason for the low cost of disinflation was that from the breakdown of investment in data sup- there were few long term contracts denominated plied by the Uganda Investment Authority. As in domestic currency and in particular few credit discussed in Collier and Gunning (1993, 1993a), transactions. Thus,the legacy of the civil war was if private investors are worried about a reversion an inability of the domestic financial sector to to 'war' this may be the expected pattern. The function and this in turn made disinflation an advantage of transport equipment is that it can be unusually cheap option. An implication is that the removed from the country if conditions deterio- early post-transition years, during which the rate, and since it does not last long, the investor domestic financial sector is inevitably truncated, does not have to take a favorable view of the are a good time for the government to invest in a long-termfuture. Manufacturing machinery is also once-and-for-alldisinflation. However, the price readily removable as long as it is chosen with this of this will be that the credit market will be in mind. Housing has neither of these features: it bifurcated into a predominant low real interest depreciates only slowly and it cannot be removed rate foreign currency denominated market, and a from the country. However, it is relatively im- minor but politically sensitive high real interest mune to war since the flow of its services is not rate domestic currency market. transactions-intensive. The housing boom has 1now turn to the switch from foreign financial taken two forms, dwellings for owner-occupation assets to domestic real assets. As noted, private and dwellings for rental to expatriates ('dollar investment recovered rapidly, but even by 1991 it houses'). The former are part of the subsistence was only around 7% of private income. This economy. In effect, were agents to view peace as suggests that upto that date there had been little if only temporary, they would use the o p p o m t y of any net repatriation into real assets. The Kampala the temporary reduction in transactions costs, to household survey of 1990 (Bigsten and Kayis undertake the transaction-intensive business of si-Mugerwa (1992)) asked some long recall investment, but locate the investment in the questions on asset transactions and migration, and subsistence part of the economy. The dollar so providessome basis for comparing the immedi- houses are superficially precisely the sort of ate post-war situation with that four years after investment which might appear most difficult to recovery. One marked change was the increase in attract: irreversible investment in the export sector land transactions. They argue that this reflected of the economy. However, once the investment the pent-up need for asset transactions and that, has been made, the activity is again transac- tions-extensive. The services provided by the non-tradable capital; instead of the non-tradable house are not very dependent upon a production sector being the targeted recipient, it is the trad- process and so are not transactions intensive. able sector. Hence, the appropriate policy is When they are sold to expatriates, the latter can under-valuationof the exchange-rate. - pay rent direct fromone foreign bank account into 2 another, so that the main transaction in this export Risk-bearing and Market Integration activity is relatively immune from government control. Finally, in the worst case scenario in The retreat into a subsistence economy reduces which expatriates leave, the house can be occu- average income sharply, and also 1,eaves some pied by the owner. Although not internationally private agents more exposed to risk: the represen- mobile, the capital can thus be switched from the tative urban household becomes more diversified export sector into the subsistence sector. Collier whereas the representative rural household be- and Gunning argue that the social return on comes less diversified. The urban response is irreversible investments in transactions-intensive because of the collapse of incomes in the formal sectors is higher than the private return, because sector. As Bigsten and Kayizzishow, by 1990 the the latter must allow for the risks of 'war'. Non- typical urban household with a base in the formal tradable investments are irreversible (whereas sector nevertheless drew a substantial part of its imported capital can often be exported if neces- incomefrom informalactivities, including agricul- sary). The private sector will therefore under ture. By contrast, the representative rural house- invest in non-tradablecapital in the tradablesector hold loses remittances from urban households (the latter being transactions intensive). The (there is evidence of a sharp decline in remit- evidence for this is that of the investments re- tances), loses the chance to work in the rural labor ported to the Uganda Investment Authority (all market as the latter contracts disproportionately, large scale investment in the economy), currently withdraws from the non-coffee export crops only around 3% is going into agriculture, which almost completely, and reduces its earnings from is the sector most intensive in non-tradable capi- coffee. Finally, as discussed belaw, whatever tal. Tea estates, for example, could be rehabili- domestic financial assets it is holding evaporate tated at a far faster rate. Hence, there is a case for due to inflation. Although there are large foreign public inducement of private investment of non- asset holdings in the economy, these are clearly tradable capital in the tradable sector. They argue not held by the poorest households. Hence, that during the 'war' private investment collapsed exposure to risk increases in the rural economy by more than private savings so that there was a just as the mechanisms for coping with risk are substantial acquisition of foreignfinancial assets. reduced. These features are altered only gradually Hence, private investment is not severely finan- by the restoration of peace. cially constrained so much as being deterred by To some extent, risk reduction in the rural high war-related transactions costs, and perceived economy is an externality provided by other risks of a reversionto 'war'. The complete private agents. For example, opportunities in the labor investment problem then is to encourage repatria- market reduce risk even for those who choose not tion into the purchase of non-tradable capital to enter the market and so enable more lrislq but goods for the export sector. The policy instrument higher- yielding production decisions to be taken. which best achieves the desired subsidy is the Hence, there is a social premium upon the recon- exchange rate. In the 1970s an over- valued struction of markets because of the externality of exchange rate implicitly subsidized the use of risk-reduction. There is therefore a case for public domestic resources for the purchase of imported subsidy of transactions. Although it might seem capital and made the non-tradable sector relatively that such a subsidy would be difficult, in fact the attractive. Each of these features of the 1970s government has a readily usable instrument at its implicit subsidy is now needed in reverse. Instead disposal. One private transactions cost is the of domestic resources being used to finance the inflation tax on money. Agents need to hold investment, it is repatriated assets; instead of money for transactions purposes and so loss of imported capital being the target of subsidy it is value of money implied by inflation is a cost which is born not by economoic activity in gen- higher. If the 1971 structure is viewed as normal, eral but by transactions. If the recovery of the and so in some sense the target to be restored, the transactions part of the economy is socially too most striking feature is the failure to contract the slow because the risk-reducing effect is an ex quantity of government post-1986. ternality, then there is a case for subsidizing During the war the government had not been transactions. It is very difficult for the government in a good position to borrow either abroad or directly to subsidize transactions, however, indi- domestically, and so it had resorted to the infla- rectly it is taxing them through inflation. Since the tion tax. Adam (1992) estimates that the long term government is short of revenue (see below) there revenue-maximizing inflation tax rate in Kenya would be a case for setting the inflation tax at the has been around 10% and Adam et al. (1993) revenue-maximizing level which may be around show that this sort of range is probably broadly 10% (see Adam (1992) and Adam et al. (1993)). true also in Ghana and Tanzania. Were it to apply However, the case for subsidizing transactions approximately to Uganda, it would suggest that tends to offset this case for an implicit tax. Logi- the actual rate of inflation had grossly exceeded cally, it is possible to subsidize transactions by the revenue- maximizing rate: the government had means of a falling price level. However, a falling snatched a short term gain in exceeding it, but this price level gives rise to side effects to the extent gradually reduced the real demand for money and that there are nominal rigidities such as wage so reduced the sustainable yield. By 1986 the rates. It might therefore be better to aim for price government had few tax handles. The economy stability: the government foregoes implicit taxa- had shifted towards subsistence and transactions- tion of transactionsin the interest of compensating extensive activities. The civil service had decayed for the externalities which transactions generate to the extent that revenue collection was arbitrary for risk- reduction. and ineffective. Because of the exodus of the professional classes, enterprises were operated on 4. Government Responses an informal basis with poor book-keeping, and tax officials were not proficient in applying normal Fiscal Responses accountancy rules. The conjunction of these features made the taxation of enterprisesa process The war had undermined government revenue and of coercion countered by bluff. Although thereare thereby gradually reduced expenditure. However, always elements of this in a tax system, the ex- whereas the share of government expenditure in treme form reached in Uganda implied that until GDP was radically reduced between 1971 and the restorationof a well- functioningcivil society, 1486, as Table 9.2 shows at constant prices extra tax revenue from enterprises would to an government was much the most rapidly expanding extent be at the price of a higher incidence of sector. In terms of quantities, the government coercion and arbitrariness: in other words, extra sector doubled relative to GDP. Hence, the fall in revenue from this source would be at the price of government expenditure relative to GDP was regression in the move to 'peace' in its wider because of a radical fall in the unit price of gov- meaning. ernment. During the war the country acquired a The government had relied heavily on coffee very large and very low price governmentsector. taxation. The fall in the world coffee price in- Post-war the share of government expenditure in duced the governmznt to repeal this tax, but it GDP has reverted to its 1971 level, but the quan- attempted to collect import duties more vigor- tity of government has expanded less rapidly than ously. This switch from export taxes to import GDP. Hence, the increase in expenditure has taxes in analytically immaterial. By the Lerner reflected an increase in the unit price of govern- equivalence theorem the two have commoneffects ment. Comparing 1992 with 1971 the government with two exceptions. First, that portion of imports was much larger relative to GDP in quantity terms which is not financed by exports now pays the tax though as a share of expenditure it was similar. whereas under export taxation it would not. The unit cost of government was much lower and However, most of these imports are financed by the quantity of government was correspondingly government sales of foreign aid. Since private agents pay the market-clearing price for imports. (which is there to finance recovery in both the if the government taxes them then private agents public and private sectors) should be passed on will simply offer correspondingly less for the aid indirectly to the private sector by reduced tax dollars. The government is therefore paying its effort relative to the no-aid counterfactual, and own import duty by selling the foreign exchange even conceivably in absolute terms. That is, even more cheaply than it otherwise would. Secondly, though the economy was generating very little that component of exports which is not coffee, government revenue, increasing that would not be was untaxed during the export tax on coffee a high priority. Donor conditionality of revenue phase, but is implicitly taxed when trade taxes are recovery, by gearing up aid with tax revenue, levied on imports. In 1986 non-coffee exports forces a rapid recovery in the public sector which were negligible and so this was not a significant may be at variance with private sector needs. consideration. Further, since it is very much in During the war the Ugandan government was the economy's interest that exports re-diversify predatory and this is how government is perceived out of coffee, it is not clear that the government by the population. During the early stages of should want to tax them in the short term. Thus, peace this predation needs to be downplayed: if the case for dropping taxation of coffee was a what is most scarce during 'war' is not the mate- good one, and it probably was, then there is little rial services of government, but coherent and to be said for replacing it with taxes on imports. restrained authority. Services can be expanded So far I have suggested that the government's massively by the infusion of aid and any change major revenue options of extra taxation of enter- through revenue is peripheral. In Uganda, public prises and extra taxation of imports should not be expenditure has more than doubled as a share of exercised in the short term. Further, the risk GDP, increasing by 11 percentage points of GDP, externality argument set out above, implies that Revenue has also nearly doubled, but its increase the government should forgo the revenue- maxi- of 3.6 percentage points is evidently a minor part mizing inflationtax. This implies that the govern- of financing. The question is thus whether, given ment should not place much emphasis upon that public expenditure has increased by around revenue recovery in the early phase of transition. 150%, the marginal 18% has been worth incur- There is a final argument for low taxation ring the 100%increase in predation which it cost? during the transition phase. As discussed below, Might it not be more appropriate for there to be a aid is endogenously very high during the early temporary decrease in predation. If the objective stages of peace. Since aid is channelled to the is to re-establish the private activity conditioned government, this enables public expenditure to on peacetime authority patterns, the critical path recover more swiftly than private expenditure. As might be first to remove the wartime authority we have seen, public expenditure doubled as a patterns rather than attempt to get public expendi- share of GDP in the first five years of peace, fully ture upto its peacetime share regardless of the regaining its previous level. It is arguable that this authority patterns which that effort must entail if pace of recovery is too fast relative to the private it is to succeed. sector. The most slowly adjusting component of the economy is a wide variety of private activities, Social Service Provision and Private Priorities including private consumption. Some private recovery is indeed dependent upon public recov- There is one area in which the government can ery, and so to this extent it is justifiable for the usefully increase revenue and this is through the public sector to recover at a more rapid rate. sale of its social services, especially in rural areas. However, public recovery financed by higher The government is a supplier of health and educa- taxation is directly at the cost of private recovery tion services. As discussed above, these have and so the indirect effects need to be substantial declined severely during the period of war, how- for the net effect to be beneficial. As suggested ever, the government is still a fairly large supplier above, some of the indirect effects, such as the in these markets. Bigsten and Kayizzi (1992) in a arbitrary coercion by the tax authorities, as nega- survey of the use of health facilities, found that in tive. There is therefore a case for part of the aid rural areas 27%of treatments were supplied by government services, and in Kampala 13%, the and was deeply integrated into the market. Road rest being either mission or private. The govern- improvements were consequently very important. ment is a much larger supplier of education. As The high valuation of road improvements in rural reported above, the main demand on the part of Uganda 1986-90 must therefore indiczre not the rural population is for increased public provi- merely that peasants recognized that this was the sion of health services. The suggestion here is that main focusof the public expenditure commitment, in the short run these services should be run at a but that the expenditure was actually found useful. profit so that their expansion enhances the budget. Thus, there must have been a desire to remegrate The sale of social services has three potential into the market. advantages. First, it directly improves the budget. The Bigsten and Kayizzi-Mugerra survey also Secondly, it permits service provision to be investigated what people now most wanted the expanded when there is evidently a demand. government to provide. Further expenditure on Thirdly, by raising the need of the rural roads was fairly low on the list of priorities. The population for cash, it encourages re-integration most desired improvement was for health facilities into the market economy. Since both curative (40%) and the next most wanted improvement services and education services are used dis (27%)was the streamlining of the functions of the proportionately by the better off, the distributional local administration (the resistance councils). implications of raising revenue in this way are These responses can again be compared to the probably now worse than the alternatives, whereas priorities given in rural Kenya in 1982 and rural the incentive effects are probably more favorable. Tanzania in 1983. As in Uganda, what people Evidence on private prioritization of services most wanted in the future was different from what is provided by a survey of rural households they had most appreciated in the past. In both conducted in 1990 (Bigsten and Kayizzi-Mugerwa Kenyan provinces the most desired expenditure (1992)). Respondents were asked to name the was water supply, and in Tanzania it was health single biggest improvement in their area brought services. Thus, in Kenya although the reduction in about by the government since 1985. For 34% the transactions costs implied by road expenditurewas biggest improvement was peace and for 48% it most appreciated, the new demand was in effect was the road network. These responses can be for more labor time, since piped water replaced a compared with those to identical questions asked very labor-intensiveactivity. The Kenyan prioritis in surveys conducted in Kenya in 1982 and Tanza- ationscan be interpreted as the move from market nia in 1983 (Bevan et al. (1989)). In Kenya (in integration to its consequence, an increased value Central and Nyanza provinces) the most valued of labor time. The Tanzanian valuations, a switch public expenditure over the preceding seven years from primary schooling to health, can be inter- was also roads, chosen by 43% and 55% of preted as the response of a peasantry which saw respondents respectively. Hence, the Kenyan less value in market integration and consequently average was virtually the same as the Ugandan. In had a lower opportunity cost of labor. The provi- Tanzania, by contrast, road expenditure was sion of free primary education in villages had chosen by less than ten percent of respondents, largely sated the demand for this social service, the most valued service being primary schooling. leaving health care as the next social service Rural Tanzania in 1983 was in one important demand. The Ugandan prioritizationis distinctive, respect analogous to rural Uganda in 1986, but it is closer to the Tanzanian than to the Ken- namely, there had been a retreat from the market yan. It is distinctive in that the high riority (though for somewhat different reasons). The low placed on the reform of government local stitu- valuation placed upon road expenditure 1975-83 tions has no parallel. It is akin to Tanzania in that in rural Tanzania was not that the government had social services rather than output-enhancing not made such expenditures, but that they were services are demanded. However, within social irrelevant to a peasant society which during this service provision the ranking is somewhat differ- period was in retreat from the market. By con- ent. In Tanzania although health care was now the trast, the Kenyan peasantry in 1982 had over the top priority, this was because of the satisfaction of previous seven years experienced the coffee boom the previous priority of primary schooling. In Uganda. primary schooling featured neither as a agricultural technology transfer in various parts of past priority nor as the future one. Hence, the Africa (Oehmke and Crawford (1993)) finds that prioritization responses are atypical for the region Uganda during the period 1986-91 is a very rare in that the dcsign of local government is seen as exception to a general pattern of quite high re- an important issue, and greater weight is placed turns. In Uganda the return was negative (Laker- upon distress-related social services than either Ojok (1992)). The negative return was attributed education or production-enhancingservices. to the effects of the war, in part the costs involved It might seem surprising that the second most in reconstruction of buildings and staffing, and in popular request should be something to do with part the lack of an adequate distribution system local administration reform. However, this takes for seed and an adequate market for output. In us back to the wider notion of how civil war has such a context investment in innovation, whether affected the economy. Our argument has been that public (i.e. technology transfer programs) or the predominant route is not through the physical private (i.e. education) does not pay. Bigsten and damage or risk of violence inherent in war, but Kayizzi-Mugerwa also investigate the entry rather in the more generalized breakdown in the decision into non-farm enterprises using logit institutions of civil society. While the former type analysis and find similarly that in the small scale of cost very largely ceased (except for a high business sector education plays no role in the incidence of crime discussed below), the rebuild- entry decision. This again is likely to be the case ing of civil society after a civil war is a slow only while the range of business activities is process. Four years after the end of it, the second highly conservative. Essentially, until the econ- most important perceived need in rural Uganda omy has surpassed its previous frontier, there is was that the power of public officials should be no need for innovation and so the private returns contained and clearly demarcated. That this is to education are likely to be low. This showed up ranked above everything except health care in the Bigsten and Kayizi-Mugerwa analysis of suggests that it is regarded as really important. rural income distribution. The high income group The high relative demand for health facilities was landed rather than educated. partly reflects their chronic deterioration during The evidence from prioritization of past and the war. The World Bank (1993) estimates that by future public expenditurestherefore points to the 1985 government real expenditure on health Ugandan peasantry re-enteringthe market (and so services had declined by 917% from its 1972 level. valuing roads unlike their counterparts in Tanza- Additionally, the legacy of warfare and vagrancy nia in 1983) but not yet to the extent that the was an unusually high level of sickness and valuation of labor time has risen much (so water debility. As discussed in 'Demobilization and supply is not demanded as it is in Kenya, but Insecurity', this has implicationsfor the design of rather social services as in Tanzania). Within demobilization. social services, both private priorities and objec- However, the high ranking of health is also tive evidence from earnings functions and the partly due to the relatively low demand for public return on agricultural technology transfer point to expenditure on education. This is consistent with the low return to education. The take-up of school evidence on the private returns to education in places is correspondinglylow in rural Uganda and rural areas. Bigsten and Kayizzi-Mugerwa con- we now consider whether this is a socially optimal struct agricultural production functions and find response, given the circumstances which have that formal schooling does not contribute to produced a low rate of return, or whether there is agricultural productivity in Uganda. Previous a case for public intervention. studies have found different results for Kenya, but The decline in private investment in education the contrast is consistent with the Schultz hypothe- need not be socially sub-optimal. While the sis that while peasants are confined to traditional private economy has retreatedinto subsistence and activities they are efficient, so that education is non- tradable activities, the social return to educa- only useful in modernizing environments. War tion has fallen and so the case for public invest- I produces a 'traditionalising' environment. A ment in education is correspondingly reduced. recent comparative study of the rate of return to However, the low current returns to education are likely to be temporary. Once the economy returns purchase some of the social expenditure which to the production frontier the returns to education they and their constituentsdesire, without interfer- will then rise. Since the gestation period on ing with the budget. At present, only a small share primary education is extremeIy long, the returns of expenditure upon rural health and education now are a poor guide to the returns on new invest- services goes through the budget, since mostly it ments in education. There is a case for social is private. Rather than donors attempting to intervention in that the present signals generating increase government expenditureson these items, private decisions are a temporary reflection of the which would first require that the ministries effects of war and so are predictably lower than function much better than at present (which in the returns which will apply to children educated turn requires very large increases in salaries), the now. Private decisions reflect current returns, donors might simply by-pass the government for social decisions can legitimately anticipate what at least part of this component of their aid, en- the returns will be once the economy has recov- couraging NGOs or private initiatives. The lower ered to the old frontier. priority of the government for social expenditures Further, there is evidence that private invest- might be quite appropriate given the priorities of ment decisions in East African agriculture are reconstruction, and yet incompatible with donor influenced quite strongly by social learning (Bur- constituencies. ger et al. (1993)): households copy each other A related donor priority has been accountabil- rather than just calculating the returns based on ity for program aid. Much aid takes the form of their own information. The popularity of an program aid because of the very limited capacity investment is therefore self-reinforcing. In the of post-war government to implement projects. context of the spread of a new technology social Hence, an aid boom must be program-unspecific. learning can be socially beneficial, accelerating Yet the high inherited level of military expendi- adoption. However, in the context of the decline ture produces an acute concern that scrutiny is in the return to an existing asset, it can lead to necessary to avoid the money being wasted on excess disinvestment.There is therefore a role for armaments. In Uganda the donor demands for the state in offsetting 'social herding'. scrutiny delayed the reform program. The object of scrutiny was the import program, and the 5. Donor Responses method chosen was the inspection of import documentation. Because of the collapse of the Peace brings an aid boom. Uganda and Ethiopia, normal institutionsof government during the war, being in addition extremely poor and relatively notably the customs service, the only reasonably high profile in the media, are receiving substantial reliable way in which these documents could be inflows of private charity through NGOs. Both the generated was by linkingthem to the disbursement donors and the NGOs have their own priorities. of foreign exchange. Thus, the apparatus of It is quite possible that the government of a foreign exchange control was needed for donor post-war state has a greater interest in fiscal scrutiny and the move to an inter-bank market was probity and output-enhancing public expenditure consequently resisted by the government until than some of the donors. Both the NGOs and November 1993. In the Ugandan context it was some of the bilaterals have a constituency moti- reasonable of donors to demand scrutiny of the vated by immediate perceived social need rather composition of impcrts. However, the imposition than longer-term considerations of economic of documentation requirements failed to take into management. For example, in Uganda Danida account the extent of the decay in government during 1993 and the EC during 1992 were Iobby- institutions. The donors had a ready alternative in ing for increased public expenditure at the price of that they could have acquired more accurate more rapid inflation. Most of the donors have evidence on imports from Direction of Trade emphasized social expenditures as a priority and statistics rather than from import documentation. have attempted to attach conditionality to the This is indeed how the Ugandan Statistics Depart- compositionof expenditure with this in mind. An ment (bureau of statistics) itself gathers data on alternative approach may be for the donors to Ugandan imports, having decided that customs data is unusable. The grounds which made it tracted. It was therefore a weak lobby and a lobby sensible for donor contributions to be atypically much more concerned with the ending of preda- large during the post-war phase, and for much of tion than the removal of trade restrictions which this to be project-unspecific, also make it appro- were in any case notional. With the gradual priate for the mechanisms of monitoring to be reduction in predation, the import-substitute modified. manufacturing sector has been able to double between 1987 and 1993 despite trade liberaliza- 6. Conclusion: Commitment and Adjust- tion. ment in a Frightened Society Secondly, people have become used to flux and so have avoided the irreversible specific Dornbusch has recently suggested that the adjust- commitments which are the source of the re ment costs of policy reform are higher in societies distributiveeffects of policy reform. The Ugandan which are polarized. The two key adjustments disinflation is a remarkable instance. In most which he has in mind are disinflation and trade societies such a rapid disinflation would have liberalization. The underlying argument is that produced enormous inter-private transfers. In polarized societies are particularly sensitive to Uganda the credit market was so limited that such redistributionand are organized so as to block it. transfers were small. Uganda does not accord well with the Dorbusch Thirdly, in Olsen's hypothesis, defeat in hypothesis. Whatever is meant by a 'polarized' internationalwar breaks up the domesitic coalitions society, it would be hard to maintain a definition which normally block policy change. By exten- on which in the aftermath of a civil war polariza- sion, if civil war is resolved by defeat of the tion is other than extreme: onto the disputes which government as in Uganda and Ethiopia, the vested provoked the war are added the legacy of interests which must be challenged are predomi- atrocities committed during the war. Yet in nantly those of the enemy, and the fact that the Uganda during 1989-91 there was a comprehen- enemy has just been defeated means that the new sive trade liberalization, and during 1992 there regime has the power to impose change. was a spectacular disinflation from 230% to -17%. Hence, the period of the transition to peace is The ease of reform in a post-civil war context is a particularly suitable time for radical policy perhaps explicable in terms of three features. reform despite the high degree of polarization. First, few people benefit from a wartime The policy inheritance makes reform necessary, regime. Government expenditure is so heavily and the uncertainty surrounding prospective skewed towards the military that, other than the reform will reinforce the reluctance to make political problem of demobilization, there is irreversible decisions until reforms have been unlikely to be a major group which suffers from implemented. Far from increasing uncertainty, recomposition of public expenditure. speedy reform will reduce it. This is important not In Uganda non-military public expenditure only because the reforms are desirable in them- was such a small fraction of GDP that no group selves, but because it is the reluctance of the needed to be squeezed. The same is broadly true private sector to make irreversible decisions of trade policy. As suggested by Table 9.3 above, which is perhaps the major impediment to rapid in Uganda the borders were sufficiently porous recovery. After a civil war private agents are during the war that the underlying trade regime frightened. They are frightened of each other, and was free trade other than for a certain level of most of all they are frightened of the government export confiscation. In particular, it was not in its various forms. The transition to peace is not possible for industries to gain protection from primarily the reconstruction of darnaged infra- imports by quantitative restrictions. Secondly, as structure, it is the transition from fear and the discussed above, the transactions-intensive activi- defensive responses which have become in ties such as import- substitute manufacturing were grained. particularly vulnerable to civil war and so con References Adam. C.S. 1992. "On the Dynamic Azam, J-P. 1992. Grain Market Decontrol in Specification of Money Demand." Ethiopia (l990), Background paper to Journal of AJi.ican Economies, Vol. 1. D.I.A.L. report on How to PassJi.om a War to a Peace Economy: the Ethiopian Adam, C. 1993. Time series econometricsfor Case. Afiica, AERC and CSAE (mimeo). Azam. J-P. 1993. Democracy and development: Adam, C.J., B. Ndulu and N. Sowa. "Financial a theoretical framework. Public Choice, Liberalisation, Exchange Rate forthcoming. Unification and Seigniorage Revenue in Kenya, Ghana and Tanzania." mimeo Azam, J-P. 1994. La faisabilite politique de (Oxford, CSAE, 1993). I'ajustement en CBte d'Ivoire (I 981- 1990). in Azam, J.-P. and Morrisson, Adam, C.J., "On the Dynamic Specification of Ch.: Lafaisabilite' politique de Money Demand in Kenya", Journal of I'ajustement structure1(Cbte d'lvoire, AJi.ican Economies, Vol. I, No. 2 Maroc), Paris: Etudes db Centre de (August 1992),pp. 233-70. DCveloppement,OECD, forthcoming. Africa Watch. 1991. Evil Days. Thir~yYears of Azam, J-P. 1995. "How to Pay for the Peace." War andFamine in Ethiopia, New York: Public Choice. Human Rights Watch. Azam, J-P., P. Collier and A. Cravinho. "Crop Alemayehu Lirenso. 1987. Grain Marketing and Sales and Portfolio Choice under Pricing in Ethiopia. AStudy of the Shortages: an Application to Angola." Impact of Grain Quota andFixed Grain forthcoming in the Journal of Prices on Grain Producers, Research Development Studies. Report no. 28, Addis Ababa: Institute for Development Research. Azam, J-P., "Liberalisation du commerce de grains en Ethiopie" (The Ethiopian Grain Alesina, A. and R. Perotti. 1993. "Income Market during the Transition to Peace), Distribution, Political Instability and Revue d'Economie du Developpement, Investment", mimeo, Harvard Vol. 2, No. 4 (December 1993), pp. 95- University, Department of Government. 116. Alesina, A., S. Ozler, N. Roubini and P. Swagel. Banerjee. A., J. Dolado, J.W. Galbraith, and D.F. 1992."Political Instability and Economic Hendry. 1993. Co-integration, Error- Growth." Working Paper, NBER. Correction, and the Econometric Analysis of Non-Stationary Data, Alexander, C. and J. Wyeth. 1994. Advanced Text in Econometrics. Oxford: "Cointegration and Market Integration: Oxford University Press. An Application to the Indonesian Rice Market." Journal of Development Banerjee, A., J. Dolado, D.F. Hendry and G.W. Studies. Smith. 1986. "Exploring Equilibrium Relationships in Econometrics through Static Models: Some Monte Carlo Evidence." Oxford Bulletin of Economics andStatistics, Vo1.48, pp.253-77. Bevan, D.L., P. Collier and J.W. Gunning. 1989. Burger K., P. Collier and J.W. Gunning. 1993 Peasants and Governments: an "Social Learning: an Application to EconomicAnalysis. Clarendon Press, Kenyan Agriculture." mimeo, CSAE, Oxford. Oxford. Bevan, D.L. 1994. "Fiscal Aspects of the Bryceson, D. 1993. Liberalizing Tanzania's Food Transition from War to Peace: with Trade, London: James Currey. Illustrations from Uganda and Ethiopia." Centre for the Study of African Chole, E. 1992. "Privatization and Economies Working Paper Deregulation in Ethiopian Industry: no.WPS194.7, Oxford. Problems, Prospects and Impact on the Economy", DIAL, Addis Ababa. Bevan, D.L. 1993. "Economic Aspects of the Ethiopian Transition to Peace." mimeo, Oxford, CSAE. Clapharn, C. 1988. "Transformation and Contimiry in Revolutionary Ethiopia", Bevan, D.L. 1993a."Fiscal Aspects of the CUP. Transition from War to Peace." mimeo, Oxford, CSAE. Collier, P. and J.W. Gunning, 1992. "Aid and Exchange Rate Adjustment in African Bevan, D. 1992."FiscalAspects of the Ethiopian Trade Liberalisations." Economic Transition",CSAE and DIAL, Oxford. Journal, vol. 102, 1992, pp. 925-939. Bevan, D.L., P. Collier and J.W. Gunning. 1990. Collier, P. 1993. "Demobilisation and Controlled Open Economies, Oxford, Insecuriry."CSAE, Oxford. Clarendon Press. Collier, P. and J-W Gunning. 1993. "Some Bevan, D. 1993a."me Ethiopian Budget: EconomicConsequences of the Peace." Macroeconomic Perspectives and CSAE Oxford and VU Amsterdam. Implications of Devolution",CSAE and DIAL, Oxford . Collier, P. and J.W. Gunning. 1993. "Trade and Development: Protection, Shocks and Bevan, D. 1993c. "Regionalisation", Liberalization', in: L.A. Winters and D. (Background Paper to the Ethiopia Greenaway (eds.), Surveys in Public ExpenditureReview), CSAE, International Trade, London: Oxford. Macmillan. Bigsten, A. and S. Kayizzi-Mugerwa.1992. Collier, P. 1994. "Demobilisation and Insecurity: "Adaptionand Distress in the African a Study in the Economics of the Economy: a Study of Retrenchmentand Transition from War to Peace." Journal Reform in Uganda." mimeo. (Unit for of International Development, 1994(3). Development Economics, Departmentof Economics, University of Gothenbur. Collier, P. and J.W. Gunning. 1995. "War, Peace and Private Portfolios." World Bigsten, A. and S. Kayizzi-Mugerwa.1992. Development. "Adaptionand Distress in the African Economy: A Study of Retrenchmentand Collier, P. and J.W. Gunning. 1994a."Policy Reform in Uganda."Unit for Uncertainty, Repatriation and Development Economics, Departmentof Investment." mimeo, Centre for the Economics,University of Gothenberg. Study of African Economies, Oxford. Collier, P. and J.W. Gunning. 1993. "Policy (September 1993), pp. 1180-89. Uncertainty, Repatriation and Invest- ment", mimeo. CSAE, Oxford and Free Ensminger, J. 1992. Makinga Market, University, Amsterdam. Cambridge University Press. Collier, P. and J.W. Gunning. July 1992. Engle, R.F. and W.J. Granger. 1987. "Macroeconomic Aspects of the "Cointegration and Error Correction: Economic Reforms in Ethiopia." mimeo, Representation, Estimation and Testing." CSAE, Oxford and Free University, Economeaica, Vo1.55, No.2, pp.251-76. Amsterdam. Engle, R.F., D.F. Hendry and J.F. Richard. 1983. Collier, P. and S. Pradhan. 1994. "Economic "Exogeneity."Economeaica, Vo1.51, Aspects of the Ugandan Transition to N0.2, pp.277-304. Peace." mimeo, Oxford, CSAE. Engel, C. and K.M. Kletzer. 1991. "Trade Policy Collier, P. and S. Pradhan. (forthcoming) under EndogenousCredibility', Journal "Economic Aspects of the Ugandan of Development Economics." vol. 36, Transition to Peace" in H.B. Hansen and pp. 213-228. M. Twaddle (eds.) Developing Uganda, James Currey. Franzel, S., F. Colburn and Getahun Degu. 1989. "Grain marketing regulations: Impact on Dercon, S. 1993. "The Consequences of peasant production in Ethiopia." Food Liberalisation and Peacefor Food Policy, November, pp. 347-358. Markets in Ethiopia", CSAE Oxford and Catholic University of Louvain. Goodwin, B. and T. Schroeder. 1991. "Cointegration Tests and Spatial Price Dercon, S. 1994. "The Consequencesof Linkages in Regional Cattle Markets." Liberalisation and Peace for Food AmericanJournal of Agricultural Markets in Ethiopia." mimeo, Centre for Economics, Vo1.73, no. 2, pp.452-64. the Study of African Economies,Oxford. Government of Uganda. 1993. Key Economic Devereux, S. 1987. "Entitlements, availability Indicators, 13th Issue. Statistics and famine: a revisionist view of Wollo, Department, Entebbe. April 1993. 1972-74."FoodPolicy (August). Granger, C.W.J. and A. Escribano. 1987. "The Dixit, A. "Exit and Entry Decisions under Long-Run Relationship Between Prices Uncertainty."Journal of Political from an Efftcient Market: The Case of Economy, Vo1.97, No. 3 (June 1989), pp. Gold and Silver." University of 620-38. Californi,San Diego. Minheo. Doornik, J. and D.F. Hendry. 1993. The Hausman,J.A. 1978. "Specification Tests in Econometrics of PC-GIVE 7.0. Oxford: Econometrics."Econometvica, Vo1.46, Institute of Economics and Statistics. no.6, pp.1251-71. Dowd, K. and D. Greenaway. "Currency Holmberg, J. 1977. Grain Marketing andLand Competition, Network Externalitiesand Reform in Ethiopia, Research Report Switching Costs: Towards an Alternative no.41, Uppsala: The Scandinavian View of Optimum Currency Areas." Institute of African Studies. EconomicJournal, Vol. 103, No. 420 Keen, D. 1994. The Benefits of Famine: a Mosley, P. and J. Harrigan. 1991. "Evaluating Political Economy of Famine and Relief the Impact of World Bank Structural in Southwestern Sudan. 1983-89, Adjustment Lending." Journal of Princeton University Press. Development Studies, vol. 27, pp. 63-94. Kello, A.B. 1992. "Institutional Mechanisms of MSEC. 1993a. The Uganda Veterans Assistance Access to Resources in Smallholder Board: Needs Assessment Survey, Agriculture", DIAL, Addis Ababa. Management Systems and Economic Consultants Ltd., Kampala. Legesse Dadi, Asfaw Negassa and S. Franzel. 1992. "Marketing maize and tef in MSEC. 1993b.The Lkanda Veterans Assistance western Ethiopia: implications for Board Veterans Assistance Programme: policies following market liberalization." a Final Monitoring and Evaluation Food Policy, June, pp.201-213. Report, Management Systems and Economic Consultants Ltd., Kampala. Lemma Merid. 1986. "The implicationsof State Procurementand Marketing of Nakamura, A. and M. Nakamura. 1981. "On the AgriculturalProducts to Market Prices: Relationships among several Ethiopia'sCase." Ethiopian Journal of Specification Error Tests presented by Development Research, vol 8, no.2, Durbin, Wu and Hausman." pp.15-48. Econometrics, vo1.49, No.6, pp.1583-88. Laker-OjokR. 1992. "The Rate of Return to Oehmke J.F. and E.W. Crawford. 1993. "The AgriculturalResearch in Uganda: the Impact of Agricultural Technology in Case of Oil Seeds and Maize." mimeo, Sub-Saharan Agriculture." mimeo, Department of Agricultural Economics, Department of Agricultural Economics, Michegan State University, East Michegan State University, East Lansing. Lansing. Little, P.D. and I.B. Lundin de CoIoane. Palaskas, T.B. and B. Harriss. 1993. "Testing "Trading to Eat: Petty Trade and Income Market Integration: New Approaches Strategies in the Peri-Urban Areasof with Case Malerial from the West Maputo, Mozambique." Development Bengal Food Economy." Journal of Anthropology Network, Vol.11, No. 2 Development Studies, September. (Fall 1993), pp. 1-8. Little, P.D. "Traders, Brokers and Market 'Crisis' Pesaran, H.M. and R. Smith, 1990, 'A Unified in Southern Somalia."Afiica, Vol. 62, Approach to Estimation and No. 1 (March 1992), pp. 94-123. Orthogonality Tests in Single Equation Econometric Models', Journal of Malkiel, B. 1987. "'Efficient Market Econometr ics. VoI.44, pp.41-66. Hypothesis," in: J. Eakwell, M. Milgate and P. Newman, The New Palgrave RavalIion, M. 1986. "Testing Market Dictionary of Economics,London: Integration." AmericanJournal of MacMillan Press. Agricultural Economics, pp.102-9. Mengistae, T. 1992. "Incomes. Employment and Ravallion, M. 1987. Markets and Famines. Industry: The Challenges of the Crisis in Oxford: Clarendon Press. the Urban Rural Nexus."Dial, Addis Ababa. Rodrik. D. 1989. "Credibility of Trade Reform: a Takayama, T. and G. Judge. 1971. Spatial and Policy Maker's Guide." The World Temporal Price and Allocation Models. Economy, vol. 12, pp. 1-16. Amsterdam: North Holland Publishing Company. Rodrik, D. 1991. "Policy Uncertainty and Private Investment in Developing Countries." Urbain. J.-P., 1992. "On Weak Exogeneity in Journal of Development Economics, vol. Error Correction Models." Oxford 36, pp. 229-242. Bulletin of Economics and Statistics, V01.54, No.2, pp.187-207. Sen, A. 1991. "War and Famine." Centre for Development Economics, Discussion Wijnbergen, S. van. 1985. "Trade Reform, Paper, LSE. Aggregate Investment, and Capital Flight: on Credibility and the Value of Sims C.A., J.H. Stock and M.W. Watson. 1990. Information." Economics Letters, vol. "Inference in Linear Time-Series with 19, pp. 369-372. Some Unit Roots." Econometrica. V01.58, pp.1 13-44. Wijnbergen, S. van. 1992. "Trade Reform, Policy Uncertainty, and the Current Account: a Soares, 1992. F.B. "Agricultural Supply Non-Expected Utility Approach." Response. " DIAL, Paris. American Economic Review, vol. 75, pp. 626-633. Stock, J.H. 1987. " ~ s ~ m ~ t ' oProperties of t i c Least-Squares Estimators of Co- World Bank. 1985. "Ethiopia: Industrial Sector integrating Vectors', Econometrica, Review." Washington. v01.55, pp.1035-56. World Bank. 1989. "Ethiopia:Strategies and Taylor. M.P. and I. Tonks. 1989. "The Policies for Improved Performance. " Internationalisation of Stock Markets dustrial Sector Review, Washington. and the Abolition of UK Exchange Control." Review of Econonlics and World Bank. 1990. "Ethiopia's Economy in the Statistics, Vo1.71, pp:332-36. 1980s and Framework for Accelerated Growth." Washington. Policy Research Working Paper Series Contact Author Date for paper WPS1374 The Evolution of Trade Treaties and Sarath Rajapatirana October 1994 J. Troncoso Trade Creation: Lessons for Latin America 37826 WPS1372 AdministrativeCharges in Pensions Salvador Valdes-Prieto October 1994 E. Khine in Chile, Malaysia,Zambia, and the 37471 United States WPS1373 Firm Behavior and the Labor Market Simon Commander October 1994 B. Keller in the Hungarian Transition Janos Kollo 35195 Cecilia Ugaz WPS1374 Infrastructure Finance: Issues, Anand Chandavarkar November1994 M. Gellei Institutions, and Policies 31393 WPS1375 Policy Lessons from a Simple ShantayananDevarajan November 1994 C. Jones Open-Economy Model Delfin S. Go 37754 Jeffrey D. Lewis Sherman Robinson Pekka Sinko WPS1376 How Trade, Aid, and Remittances Maurice Schiff November 1994 J. Ngaine Affect international Migration 37947 WPS1377 Macroeconomic Adjustment to Vittorio Corbo November 1994 R. Vo Capital Inflows: Latin American Style Leonardo Hernandez 33722 versus East Asian Style WPS4378 Mexico after the Debt Crisis: Daniel Oks November 1994 S. Harbi Is Growth Sustainable? Sweder van Wijnbergen 37143 WPS1379 Financing lnfrastrlicture in Developing Barry Eichengreen November 1994 WDR Countries: Lessons from the Railway 31393 Age WPS1380 Transfers and the Transition from Kathie Krumm November1994 K. Krumm Socialism: Key Tradeoffs Branko Milanovic 34263 Michael Walton WPS1381 Welfare Economics, Political Ravi Kanbur November 1994 M. Youssef Economy and Policy Reform in Ghana 34614 WPS1382 Saving, Investment and Growth in Klaus Schmidt-Hebbel November1994 E. Khine DevelopingCountries: An Overview Luis Serven 37474 Andres Solimano WPS1383 Rural Demand for Drought Insurance Madhur Gautam November 1994 C. Spoones Peter Hazell 30464 Harold Alderman Policy Research Working Paper Series Contact Title Author Date for paper WPS1384 Fiscal Decentralization and David Sewell November1994 G. Langton IntergovernmentalFinances in the Christine I. Wallich 38392 Republic of Albania WPS1385 Fiscal Federalism Dimensions of Tax Robin Boadway November 1994 C. Jones Reform in Developing Countries Sandra Roberts 37754 Anwar Shah WPS1386 EU Bananarama Ill Brent Borrell December1994 G. llogon 33732 WPS1387 Fiscal Decentralizationand the Size Saber Ehdaie December1994 C. Jones of Government: An Extension with 37699 Evidence from Cross-CountryData WPS1388 Does Voice Matter? For Public Samuel Paul December1994 B. Moore Accountability, Yes 35261 WPS1389 Ruble Overhang and Ruble Shortage: Patrick Conway December 1994 L. Suki Were They the SameThing? 33974 WPS1390 RegionalIntegration and the Baltics: Piritta Sorsa December1994 J. Ngaine Which Way? 37947 WPS1391 Where in the World Is Population Jeff Kling December 1994 Growth Bad? Lant Pritchett WPS1392 Some Economic Consequences of Jean-Paul Azam December1994 C. Jones the Transition from Civil War to Peace David Bevan 37699 Paul Collier Stefan Dercon Jan Gunning Sanjay Pradhan