Public Disclosure Authorized

28136
Public Disclosure Authorized

The Challenge of Ugandan Reconstruction, 1986-98

Paul Collier
World Bank
Public Disclosure Authorized

This draft November 2, 1999

Public Disclosure Authorized

The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They
do not necessarily represent the views of the World Bank, its Executive Directors, or the countries they
represent.
1. Introduction

Ugandan economic performance is now among the most successful in the world. Rapid
growth is reducing poverty, prices are stable, and investor confidence has increased more
than anywhere else in Africa. This is the main example of successful African economic
liberalisation and so it is important to understand the process. However, Uganda’s current
success can only be understood in terms of its past. Over the period 1971-86 both the
Ugandan economy and Ugandan society collapsed. By 1986, when the National
Resistance Movement captured Kampala and formed a government, Uganda had suffered
the predations of Idi Amin and three further transient presidents, including civil war,
mass emigration of the skilled, and mass murder. Current success is thus a recovery from
conflict. Indeed, Uganda is the main example of successful African post-conflict
recovery.

The preceding collapse made the attainment of rapid growth easier in some respects, and
more difficult in others. It was easier in that there were resources to draw upon, for
example, there was the scope to induce repatriation of human and financial capital. It was
more difficult in that social and institutional collapse left a persistent inheritance, such as
low trust and high opportunism. This makes Ugandan success more complex, but also
potentially even more illuminating than if it were a straightforward story of economic
liberalisation. Most of Africa’s currently problematic economies will need to replicate
Uganda’s transition to rapid growth from an inheritance of social decay. Our study thus
aims to understand the twin processes of liberalisation and reconstruction, and their
interrelation. The need for social reconstruction changed the liberalisation program both
by constraining it, and by introducing new priorities. The need for liberalisation similarly
changed the reconstruction program.

This chapter analyses post-1986 performance from the perspective of the inheritance of
socio-economic collapse. What is distinctive about an economy newly emerged from
internal conflict? How successful was the Ugandan government in meeting the twin
needs of enhanced security and poverty reduction? First, we attempt to quantify the
inheritance of social disorder which faced the NRM in 1986. How prone was the society
to further conflict and how had the level and structure of economic activity been altered?
We then turn to the implications for government security policy. As we show, the
government had little choice but to attach some priority to making Uganda a safer
society. We ask how this might have been achieved, and to what extent NRM policy
succeeded in reducing the risk of conflict. Finally, we turn to the implications for
economic policy. How should the inheritance have changed priorities, and to what extent
was this recognised in practice?

2. The Inheritance of Disorder

In 1972 President Idi Amin declared `economic war’ against the large and commercially
dominant Asian community in Uganda. This marked the beginning of an economic
collapse and of escalating political and social disorder. By 1979 when Amin was
overthrown, up to 500,000 Ugandans had died as a result of the regime and there had

2
been two insurgency attempts by exiles. The overthrow of Amin produced a period of
extreme instability and mass murder. By 1986, which marks the return of political order,
some 7% of the population were displaced. Since 1971, per capita income had declined
by around 40% and the composition of economic activity had changed radically.

The causes of political breakdown in Uganda were various. It is important to consider

them so as to distinguish between those factors which make Uganda intrinsically prone to
conflict, and so must be addressed in the post-conflict phase, and those which were
idiosyncratic to the episode itself. In particular, we need to distinguish between the extent
to which Uganda was simply unfortunate to be led by particular personalities, notably Idi
Amin, and to what extent personalities were merely the instruments of deeper social
forces. Naturally, there is no way of providing a definitive answer to such a question, but
recent analyses of the causes of domestic conflict can serve as a guide. Globally, four
factors have been shown to have powerful effects upon the probability of conflict (Collier
and Hoeffler, 1999). Countries with little education and high poverty, high dependence
upon natural resources, partial democracy, and ethnic and religious homogeneity are most
at risk. In Table 1 we apply the probit regression which Collier and Hoeffler find best
explains the risk of civil war globally, to estimate the risks to Uganda in 1965 and 1985.

Table 1: The Risk of Civil War in Uganda in the Ensuing Five Years; 1965 and 1985

Factors determining risk coefficient contribution to risk

Factor value in 1965 value in 1985 1965 1985
Education 1.19 1.54 -0.237 -0.282 -0.365
Primary exports 0.16 0.16 11.692 1.871 1.871
Primary exports2 0.0256 0.0256 -20.843 -0.534 -0.534
Fractionalization 0.376 0.337 -1.66 -0.624 -0.560
Democracy 7 0 0.226 1.582 0
2
Democracy 49 0 -0.025 -1.225 0
Population(ln) 15.9 16.46 0.326 5.183 5.366
Constant -7.129 -7.129
TOTAL [ln(p/(1-p)] -1.158 -1.351
p (= risk of war) 0.239 0.206

Notes: Education is the mean years of education per head of population; primary exports are the share in
GDP; fractionalization is the square of the product of ethno-linguistic fractionalization and religious
fractionalization. On a scale of 0-100, measuring the probability that two randomly drawn people will be
from different groups, ethnic fractionalization in Uganda is estimated for the entire period at 90, and
religious fractionalization at 68.13 in 1965 and 64.33 in 1985. Democracy measures the extent of
democratic rights according to the Polity III scale, with 7 being full rights and 0 being no rights. The
coefficients on these variables are taken from Collier and Hoeffler. The contribution to risk measures the
product of the value of the variable and its coefficient. The sum of these contributions shows the value of
[ln(p/(1-p)], where p is the probability of civil war in the ensuing five years.

In 1965, prior to Amin’s coup, Uganda is predicted to have had quite a high risk of
conflict: a nearly one-in-four chance during the ensuing five years. The factors causing

3
this were the combination of high dependence upon natural resources, and a low level of
education. Natural resources are readily lootable, and so both increase the incentive for
rebellion and make it more readily financable. A low endowment of education implies
low returns to labor, and hence a low opportunity cost of participating in rebellion.
Offsetting these disadvantages, Uganda’s major advantage was its very high level of
social fractionalization. Ugandan politics reflected its social divides of which two perhaps
stood out: that between the Buganda and the non-Buganda, and that between the
predominantly Moslem and poorer north and the predominantly Christian and more
commercialised south. However, these divisions were mitigated by many further ethnic,
and religious distinctions; indeed it was among the most highly fractionalized societies in
the world. In popular perception this would be a factor increasing the risk of civil war.
Differing social groups usually dislike each other, and so highly fractionalized societies
are typically characterised by a noisy discourse of inter-group grievance. However,
empirically, fractionalization is a powerful source not of risk but of safety. The reason for
this is that rebellion in the pursuit of justice needs to overcome a collective action
problem which fractionalization makes more difficult. A successful rebellion requires a
degree of mass cooperation which highly fractionalized societies find difficult.

Uganda’s other advantage as of 1965 was that it had a fully democratic constitution. The
level of political rights was, however, in rapid decline as President Obote used the army
to crush the political opposition of the Buganda, the largest ethno-linguistic group,
accounting for around 30% of the population. The relationship between political rights
and the risk of conflict is not straightforward. As rights decline from full democracy, the
risk of war initially increases. However, further political decline into severe repression
reduces the risk of war. Between 1965 and 1975 Uganda made a rapid transition from full
democracy to complete repression.

With a one-in-four risk of civil war in each five year period, it was more likely than not
that Uganda would experience a war within the first thirteen years of independence. To
the extent that it is possible to form a judgement, it therefore seems reasonable to
conclude that Uganda was intrinsically prone to conflict. This is not, of course, to say that
it was inevitable. The society was evidently unfortunate to face the repression of political
rights under President Obote, and to have someone with the character of Idi Amin as head
of the army.

Social conflict and political breakdown have various effects (see Collier, 1999). First,
conflict produces the physical destruction of assets. Both government and rebels become
predatory. For example, many households lost livestock to soldiers. Secondly,
transactions become disrupted as the mechanisms for contract enforcement start to break
down. Faced with opportunism by the political leadership, the legal system and the civil
service become opportunistic. Businesses respond to this by retreating into a hard core of
known relationships, not daring to risk new clients. For example, the major banks in
Uganda were extremely reluctant to take on new business customers. Transactions
thereby became both more costly and less competitive, fostering the emergence of
informal cartels. Thirdly, government expenditure is diverted from its economically
productive functions to the objectives of predation and defence. A graphic example of

4
this in Uganda was the contraction of the police force and the rise of the army. Fourth,
because conflict causes a phase of economic decline, there are strong incentives for
households, firms and government to dissave: since incomes are perceived as temporarily
low, there will be an attempt to maintain consumption. Fifth, households will shift their
assets, including human capital, out of the country, to environments where they are both
safer and more productive. The process of shifting assets may itself be very costly. For
example, consider the problem of shifting the assets constituted by a building, a coffee
tree, or an installed machine out of Uganda. In one sense, these are all immovable.
However, to remain as productive assets they all need expenditures upon maintenance.
Hence, the owner can gradually liquidate the asset by cutting maintenance expenditures.
If the money saved on maintenance is used to purchase illegal foreign exchange and this
is held in a foreign bank, the immovable asset is gradually transformed into flight capital.
By 1986 over half of Uganda’s private wealth was held abroad.1 These five effects of
conflict affect both the level and composition of economic activity.

Table 2: The level and composition of economic activity in Uganda, 1971, 1986,
1993/94 and 1996/97

1971 1986 1993/94 1996/97

GDP per capita (Uganda shillings ‘000) 127.7 149.9

War-vulnerable activities:
(percentage share of GDP at 1991 constant prices)
activities intensive in assets and in 8.8 4.4 6.0 8.6
transactions (manufacturing)

transaction-providing activities 21.2 16.1 17.2 17.7

(transport and commerce)

asset-providing activities 12.5 3.5 5.5 7.6

(construction)

total 42.5 24.0 28.7 33.9

War-invulnerable activities
subsistence activities excluding
livestock and construction 20.5 36.0 26.9 21.0

Unassigned Activities 37.0 40.0 44.4 45.1

Source: Collier (1999), updated from Statistical Abstract, 1997

1
Collier, Hoeffler and Pattillo (1999) provide estimates of how private wealth was held for 56 countries as
of 1990 based on cumulating investment and capital flight since 1970. The figure reported here uses their
data but measures the stock as of 1986.

5
The level of economic activity is reduced through two routes. The stock of physical and
human capital is diminished by destruction, dissaving, and flight. Additionally,
productivity is reduced by disruption and expenditure diversion. As shown in Table 2, per
capita GDP declined by around 40% between 1971 and 1986. This severe decline was
larger than is globally typical for periods of domestic conflict. Collier (1999) estimates
that on average such conflicts reduce growth by 2.2% per annum relative to
counterfactual growth. The Ugandan economy declined in absolute terms by 1% per
annum, and given rapid population growth, counterfactual growth could scarcely have
been less than 2-3%. Hence, either the Ugandan economy was atypically vulnerable to
social disorder, or the disorder was severe even by the normal standards of civil war.
Since, as we show below, subsistence agriculture is relatively invulnerable to disorder
and Uganda had a large subsistence sector even prior to the outbreak of conflict, its
economy should not have been abnormally vulnerable to disorder. Hence, the more likely
interpretation is that Uganda suffered an unusually severe social collapse. The decline in
aggregate expenditure was even more severe than the decline in output. Aid inflows
ceased and the government was manifestly unable to borrow commercially. As foreign
flows to the government ceased, an illegal outflow of private capital started.

The composition of GDP is altered by social disorder because activities are affected
differentially. As transactions costs rise and the capital stock declines, the activities
which are intensive in transactions and in capital, such as manufacturing, decline relative
to those which have opposite characteristics, such as arable subsistence agriculture.
Similarly, the sectors which supply transactions services, such as transport and finance,
and which supply capital, such as construction, decline because the demand for their
output falls. Table 2 classifies the components of GDP according to these principles into
three groups, those likely to be differentially vulnerable to social disorder, those which
are likely to be differentially immune, and those for which, a priori, there is no particular
reason to expect a differential effect. As Table 2 shows that based on this classification
there was a remarkable change in the composition of activity between 1971 and 1986.
Subsistence activities (excluding livestock and construction which were asset-dependent)
increased from 20% of the economy to 36%. The share of manufacturing and the other
vulnerable sectors virtually halved.

To summarize, the economy inherited by the NRM in 1986 was both much poorer than
that of 1971, and it had retreated into subsistence activities. Each of these had
implications for post-conflict recovery.

3. The Restoration of Peace

The policy objectives of a post-conflict government such as that of Uganda in 1986 are
likely to differ from those of a government conventionally embarking upon economic
reform in that greater priority will be given to the security of the state. Sometimes such a
reaction is sensible and sometimes it reflects needless paranoia. It is sensible if the fact of
the civil war is information that the country has underlying characteristics which make it
particularly prone to conflict, or if the civil war has changed characteristics in such a way

6
as to have increased risk. It is paranoid if the society is intrinsically safe but has simply
been extremely unlucky: a householder whose house has been hit by lightening should
not spend large resources defending it from a future strike. Above, we have suggested
that even as of 1965 Uganda was intrinsically prone to conflict. Between 1965 and 1985
the economy and the society changed, partly due to the experience of social disorder, and
this may have changed the risk of conflict through several routes.

Firstly, there had been a radical decline in per capita income of around 40%. Collier and
Hoeffler find that internationally the best proxy measure for the opportunity cost of
rebellion is the endowment of human capital, and despite the exodus of skilled labor, this
did not decline. As shown in Table 1, when human capital is measured as the average
number of years of education of the population, the expansion in the schooling system
more than offset the skill exodus. However, it is evident that the returns to schooling
must have declined radically over the period, so that the increase in the mean years of
education is a misleading indicator of how the opportunity cost of rebellion changed.
Given the decline in income, the opportunity cost almost certainly fell, increasing the risk
of rebellion. This had severely increased poverty and so increased the risk of conflict.
The young men who form the recruits for rebel movements had less to lose by joining up
in the late 1980s than their equivalents of the 1970s.

That poverty is a cause of conflict has two distinct implications for policy. First, to the
extent that the government can raise the incomes of those most likely to rebel it enhances
security. Secondly, other things being equal, the poor are more likely to rebel than are
higher income groups.

Poverty reduction reduces the risk of conflict by raising the opportunity cost of
participation in rebellion for the young men who might join rebel forces. Such an
increase in opportunity cost is partly achieved as a by-product of economic growth. On
average, historically, income growth has been distributionally approximately neutral
within societies, so that the opportunity cost of conflict tends to rise across the society.
However, it is possible to use policy to target income growth towards particular groups
and a government concerned to enhance its security should target income growth towards
those people most likely to rebel. The most straightforward way in which income growth
can be targeted is spatially. Infrastructure always have a spatial dimension. Expenditure
of public services such as education can be differentiated spatially. There is evidence that
education reduces the risk of conflict more effectively than generalised income growth,
perhaps because it is better targeted to raise the opportunity cost of rebellion for the age
range and the income groups most likely to find rebellion attractive. Further, the structure
of economic activity often differs spatially: for example, cotton is grown in northern
Uganda and coffee in the south. The ability to target income growth spatially is important
because often rebellions have a spatial dimension: a particular locality rebels against the
government. Hence, if it can be determined how the risks of rebellion differ spatially,
then spatially targeted income growth can be used to enhance security.

Four considerations might guide the estimation of spatially-specific risks of rebellion.

First, since poverty is a cause of conflict, the poorer districts are more likely to rebel.

7
Secondly, given that Uganda is bordered by several politically problematic states, the
border regions might be more prone to rebellion: rebels could use neighbouring states
both for safe havens and for logistical support. Third, those regions with the largest stock
of former soldiers are potentially the easiest recruiting grounds for rebel organisations.
Finally, those regions least represented in the government are least able to achieve their
objectives by political means. On all these criteria as of 1986 the most risk-prone region
was the North. It was much the poorest region with the least education. It bordered on
Sudan and Zaire. It had been the principal recruitment ground for Amin’s army, a force
which had subsequently been largely demobilised. It was somewhat underrepresented in
the NRM: in the 1996 presidential elections, which were the first nationally comparable
measure of electoral support, the North was the only region in which the opposition
secured a majority. There was therefore a strong case for targeting poverty reduction
efforts in the north.

We now turn to the second underlying cause of conflict. Globally, the larger are export of
primary commodities as a share of GDP the greater is the risk of conflict. This is partly
because during the conflict rebels can extort revenue from the export trade which helps to
finance them, and partly because the prospect of the eventual control of the tax revenue
acts as a lure. Both in 1965 and in 1985 Uganda was highly dependent upon primary
exports. Although coffee had been highly taxed during the period of disorder, output was
relatively inelastic with respect to this taxation, and so its share of economic activity had
remained fairly constant. In terms of Table 2, primary exports were neither war
vulnerable nor war invulnerable. However, in absolute terms the coffee sector had
declined along with the overall economy, and its revival was central to the government’s
development strategy, since it would augment the supply of foreign exchange. This gave
rise to a potential trade-off. The revival of the coffee sector was central to economic
recovery and to poverty reduction, and so through this route tended to enhance security,
but by increasing the importance of primary commodity exports in GDP it would tend to
reduce security. The government could hardly neglect the recovery of coffee on these
grounds, but the negative effects of primary commodity exports implied that the
diversification of the economy into other activities would be desirable on security
grounds.

The third underlying influence on the risk of conflict is the level of political rights. By
1985 Uganda was classified on the Polity III scale as being completely lacking in
democratic practices. Especially in highly fractionalized societies, a lack of democracy is
a major impediment to economic growth. Collier (1999a) finds that in a society as
fractionalized as Uganda, the change from repression to full democracy will typically
increase the growth rate by nearly 3%. However, repression is globally fairly effective in
reducing the risk of civil war. While repression and full democracy generate similar risks
of conflict, Uganda started from repression and so would probably only reach full
democracy by a transition process in which democratic rights accumulated gradually.
This unambiguously increases the risks of conflict: partially democratic societies are
considerably more prone to civil war than are repressed societies. Hence, Uganda again
faced a short-term trade-off between economic growth and security from conflict. Of
course, in the long term the faster growth would raise the opportunity cost of conflict, and

8
once full democracy was reached the society would in any case not be significantly more
conflict-prone than when repressed. However, this long term outcome could only be
reached by a transition phase which would increase the risks of conflict.

Finally, the underlying high fractionalisation of Ugandan society was potentially a major
asset. However, a sense of political identity is not a given, but depends upon history and
government behaviour. The experience of conflict may have tended to polarise Uganda’s
initially highly fractionalised society into North versus South and Buganda versus non-
Buganda. Hence, to restore security the government could attempt to rebuild diversity,
encouraging the historic identities not only to re-emerge, but to generate a multiplicity of
political parties and allegiances.

In addition to the changed risks associated with these four factors, there are various other
reasons why the experience of conflict might have made Ugandan society more or less
prone to further conflict. On the negative side, the large armies of both the former
government and the NRM would be expensive to maintain, but if demobilised, their
members might either take to violent crime, or be ready recruits for new rebel
movements. Additionally, violence had become normal as a means of conducting
political disputes and there was a legacy of atrocities which legitimated revenge. On the
positive side, Ugandans had come face to face with the horror of societal breakdown. Just
as Germany’s experience of hyperinflation convinced a whole society that price stability
was essential, so Uganda’s experience of conflict might have convinced its citizens of the
need for peace. Hence, in principle, post-conflict societies can be either more or less
prone to renewed conflict as a result of their experience. Collier and Hoeffler (1999)
provide a composite estimate of these other effects. They add to the variables of Table 1 a
dummy variable which is activated if the society has had a previous conflict (since 1945).
They find that it is insignificant. On average, post-conflict societies are neither more nor
less prone to renewed conflict than are societies which have the same underlying
characteristics but which have been at peace for a long time.2 Hence, in aggregate, all
these other possible influences on the risk of conflict appear to cancel each other out.

On the analysis of Table 1, the society was slightly less risky in 1985 than it had been in
1965. However, this is probably spurious because, as discussed above, the increase in
human capital probably mis-states the change in the opportunity cost of rebellion.
However, even on the reduced estimate of Table 1, the society was still highly prone to
conflict, with a risk of around one-in-five that there would be a conflict during a five-year
period. Thus, unless it reduced the causes of conflict, the NRM government faced a
chance no better than one-in-two that it would reach the millennium without civil war.
Further, its economic and political priorities of reviving the primary export sector and
democratising would, in the short term, worsen these odds. Evidently, the government
needed to attach considerable priority to improving the prospects of internal peace
through whatever means it had available.

2
A further implication is that the variables included in Table 1 are a complete account of the factors which
are both persistent and which are significant in causing civil war. If there were omitted variables of this
type then those societies with multiple conflicts would tend to have high values of them and would appear
collectively as a significant dummy variable.

9
How did the government achieve this core objective of reducing the risk of conflict, and
how successful was it? Since 1986 there have been many minor domestic military
challenges to the government, especially in the North. However, none has quite escalated
to the point where it meets the definition of a civil war. Given the initial odds, the
Ugandan government has thus either been fortunate, or it has succeeded in reducing the
underlying risks which the society faces. We now assess whether its policies have indeed
tended to reduce the risks of conflict.

As poverty is reduced, the opportunity cost of participating in conflict rises and so the
risk of conflict declines. As Appleton and Guillaume show in Chapter 5, national
economic growth has very substantially reduced poverty. However, the reduction has
been least in the North, where from a security point of view the need was greatest. The
North has indeed harboured the most substantial rebellions and so the failure to achieve a
more substantial reduction of poverty in that region is, from the perspective of security, a
serious policy weakness. Further, the government initially attached only a low priority to
education. It was not until 1997 that resources were shifted into rural schooling.

With respect to primary commodity exports as a share of GDP, policy has been
successful in generating growth across many sectors. Thus, although a key objective was
the restoration of coffee exports, which was achieved by radical reduction in export
taxation and the liberalisation of marketing, other sectors have broadly kept pace with
export growth. By the end of the 1990s, primary exports were less than 10% of GDP.

In terms of political rights the government has made remarkable progress in view of its
inheritance. While there is still an embargo on formal campaigns by political parties
during elections, so that members of parliament campaign as individuals, in other
respects Uganda has achieved a highly open society. The press is free both de jure and de
facto, and several major figures have returned to Uganda from exile. There has been a
genuinely contested presidential election. This democratization will in the short run have
increased the risks of conflict, in comparison with the previous repression. The
government has also encouraged the restoration of Uganda’s diverse social identities. The
kingships have been restored, and there is extraordinary diversity of the press. While it is
too early to determine how this will translate into allegiance to political parties, it is
possible that the incipient polarisation during the period of disorder may have been
reversed, restoring the main strength of Ugandan society, which is its diversity.

The government has also been largely successful in its demobilisation strategy. Major
demobilisation was delayed until 1993. Once they were demobilised, soldiers were taken
to their home villages and provided with transitional financial and material assistance.
Soldiers appear to have reintegrated. Over the first year of demobilisation there was no
tendency for crime to increase in those districts in which many soldiers were
demobilised. However, the demobilisation could have been even more successful: overall
some 12% of demobilised soldiers reported in advance of their demobilisation that they
did not have access to land. These soldiers were around one hundred times more likely
than the average Ugandan to commit crimes one demobilised (Collier, 1994). Hence, in

10
retrospect, the ideal policy would have either retained the landless in the army or made
greater provision for them. While there is no direct evidence as to whether demobilised
soldiers have joined rebel groups in significant numbers, the fact that they did not
differentially resort to crime suggests that they did not differentially resort to rebellion,
since from the perspective of the potential recruit these choices are somewhat similar.

Overall, the policies of poverty reduction, sectorally diversified growth, democratisation,

social pluralism and demobilisation which the government has implemented, would be
predicted to reduce the risk of conflict. Since internal peace has been broadly maintained,
it therefore seems reasonable to judge conflict reduction policy as having been largely
successful. Table 3 attempts to quantify the impact of policies on risk, although such an
estimate must of course be treated with extreme caution. For what it is worth, the
estimate implies that the risks of conflict fell by around a quarter between 1986 and 1998.
As compared with the initial inheritance in 1965, Ugandan society was very substantially
safer on the basis of these estimates, the risk of conflict having fallen by around 40%.

Table 3: An Estimate of How the Risks of Conflict have been Reduced, 1985-98

Contribution to
Factors determining risk Value in 1985 Value in 1999 risk in 1999

Education 1.54 c4.0 -0.95

Primary Exports 0.16 c0.09 1.05
Primary Exports2 0.0246 c0.0081 -0.17
Fractionalization 0.337 c0.337 -0.56
Democracy 0 c7 1.58
Democracy2 0 c49 -1.22
Population(ln) 15.9 16.86 5.50
Constant -7.129
Total [ln(p/(1-p)] -1.90
P (= risk of war) 0.15

The policies which enhanced security sometimes involved a trade-off with the objective
of economic growth. Two policy options which were much more important for security
than for overall growth were the targeting of growth on the north, and the expansion in
education. The north was the poorest region because it was the least accessible and the
least fertile, and so probably offered the lowest returns on public investment. Investment
in education would have a long gestation period before returns began to be realised.
Hence, in economic terms there were probably higher priorities: as we show below,
rehabilitating the road network offered a very high rate of return. However, education
would probably have reduced the incentive to join rebel forces, since they drew their
recruits disproportionately from illiterate young men. For example, a 1993 census of the
Ugandan army, which by then combined both the previous official army and the
originally guerilla forces of the NRM, found that its recruits had far less education than
the Ugandan average. As shown above, it is typically the best single measure of the

11
opportunity cost of rebellion. Each extra year of schooling per capita, reduces the risk of
conflict by around 20%.

In the event, rebellion has continued to be a costly problem for the economy and so it
may have been desirable to accept a slower rate of growth in the first decade of the NRM
government in order to reduce the scale of the problem in the second decade. Since 1986
the north has been the region most prone to rebellion, which was predictable. It has
grown more slowly than other regions, whereas on security grounds it would have been
desirable for it to have grown more rapidly. Similarly, if in the NRM’s first decade more
children been educated, by its second decade there would have been fewer illiterate
youths without prospects for rebel movements to recruit into their ranks. From the
perspective of security, UPE should have been implemented a decade earlier.

4. Growth policies in the context of post-conflict

We now turn from the objective of security to that of economic growth. As of 1986,
Uganda had in place a range of policies and institutions which were likely to be
impediments to growth, so that in one respect the government needed to embark upon a
conventional liberalisation strategy. However, the fact that Uganda was a post-conflict
economy created both opportunities and constraints which would not otherwise have
existed. In this section we consider how, given the objective of economic growth,
Ugandan economic policy needed to be distinctive because of its post-conflict
inheritance. We return to the five effects of conflict discussed above.

First consider the policy implications of the destruction and dissaving of assets and the
diversion of public expenditure from economic to military purposes. These effects
created an evident need for infrastructure reconstruction. This is the aspect of post-
conflict policy with which donors are most familiar. The government needs to make
various infrastructure investment expenditures at a time when its own resources are very
limited. In Uganda, the key infrastructure deterioration was probably in the road system
which had not been maintained. This was particularly important because the key
structural change in a post-conflict economy is the shift back out of subsistence and into
the market, and this process needs roads. Collier and Pradhan (1998) show that the rate of
return upon Ugandan transport projects during the first years of peace was an astonishing
40%. Hence, these projects needed to be prioritised over other expenditures with lower
payoffs.

The spending priorities of the Ugandan government broadly reflected these needs.
Military expenditure was reduced as a share of total expenditure, the police force was
rehabilitated, and in 1994 there was a substantial demobilisation of the army. However,
military expenditure remained quite high as the government built a better-equipped force,
partly reflecting the continued threat to internal security. The most favored sector was
road construction. A survey by Bigsten and Kayizzi-Mugerwa (1994) found that the
public service improvement which rural households most appreciated was the road
network. However, not all of the restoration of infrastructure was so successful.
Electricity, which had received no investment post-1971, gradually became a severe

12
problem as supply failed to keep up with demand. Since Uganda is landlocked and so
cannot cheaply rent mobile power stations, while the gestation period for increased
supply is long, it was all the more important to plan electricity supply carefully. By 1998
Ugandan firms reported electricity as their single most important constraint. Around a
third of manufacturing investment was going into private generators, which could only
produce electricity at high-cost. Hence, the failure to rectify the neglect of electricity
investment must count as one of the major policy errors of the period.

Now consider the policy implications of the disruption of transactions through the rise of
opportunism and the decline of the police. Such a decline poses a potentially major
problem of transition back to a society of trust. Tirole (1994) shows how both high
opportunism and low opportunism societies can be locally stable equilibria. In the
Ugandan context, the shortening of horizons and breakdown of contract enforcement
mechanisms during the period of disorder seems likely to have shifted the society from a
low-opportunism to a high-opportunism equilibrium. The restoration of peace may itself
be insufficient to shift the society back to the low-opportunism equilibrium even though,
had peace been maintained throughout, the society would have remained in the low-
opportunism equilibrium. If this is indeed the case, then what are the implications for
policy?

The high level of opportunism becomes a constraint upon policy and the reduction in
opportunism becomes a priority. As a constraint it implies that many aspects of
government will function badly because the professional ethics which normally police
conduct will have eroded. This was markedly the case in Uganda. The erosion of ethics
affected the legal, accountancy, medical, education and civil service professions. Its
consequences were far-reaching. If the legal profession no longer enforces standards of
conduct then the last-resort means which firms normally use for contract enforcement
become unreliable. One important result is that banks would find it difficult to enforce
assets as collateral. A survey of banks in 1996 (Kasekende and Atingi-Ego, 1997) found
that their single greatest need was for a fast-track legal settlement procedure to enable
foreclosure on collateral. In the absence of this, banks will lack profitable lending
opportunities and so either restrict credit or encounter a high default rate. If the
accountancy profession no longer enforces standards of conduct then accounts cannot be
trusted. This impairs both banks and the tax authorities. Banks cannot lend on the basis of
balance sheets, and tax authorities cannot levy taxes on the basis of audited profits. In the
medical profession standards of prescription had declined. A survey of the leading
hospital in the country found a mis-prescription rate of around 50%. Further, many
medical staff pilfered drug supplies and sold them privately. As a result, although imports
of drugs were approximately adequate for the population, there was continual severe
scarcity at the level of the public clinics. As a result, there was considerable hoarding of
drugs at the household level. In consequence, households were typically taking home-
stored drugs, many beyond their expiry date, on the basis of self-prescription.
Presumably, this reduced the efficiency of public drug expenditures in reducing illness. In
the teaching profession there was a high incidence of non-attendance. Headmasters added
ghost teachers to payrolls and ghost students to school enrolments. The decline in civil
service conduct impaired public sector performance both in service delivery and revenue

13
collection. For example, Ablo and Reinikka (1998) show that only 37% of the money
released by the Ministry of Finance for non-salary primary school expenditures actually
reached the schools. Reinikka and Svensson (1998) show that private sector firms paying
taxes believed that on average much of the money which they paid never reached the
government.

The decline in professional standards had two key implications for policy. First, since
government activities are by their nature atypically dependent upon self-policed standards
of conduct in the professions, the efficiency of government expenditure would be
unusually low, and the cost of tax collection unusually high. Both of these implied that
the optimal size of government would be smaller than in an economy without a post-
conflict inheritance. Secondly, some private activities, notably banking, were also highly
vulnerable to opportunism.

The government recognised the constraints on its own efficiency and accommodated
them appropriately. Both recurrent public expenditures and revenue were kept low
relative to other developing economies with less severe problems of professional conduct.
It also attempted to relax the constraints by five strategies. In tax collection the
government established a new service, the Uganda Revenue Authority, in which staff
were paid well above civil service pay scales and held more directly to account. In
customs collection it privatised many functions, using SGS. To improve public service
delivery it reduced the size of the civil service and used the expenditure savings to raise
salaries. It also decentralised to the distinct level, with the intention of subjecting services
to the scrutiny of local electorates and politicians. Finally, it prioritised the courts,
bringing in some foreign judges. However, even by 1998 these strategies had had only
limited success. Tax revenues remained low, while the Revenue Authority was the
subject of considerable complaint in terms of corruption. The productivity of public
service delivery remained very low. Hence, the attempt to rebuild professionalism had
only limited success. The strategies applied to revenue collection were not extended to
service delivery. For example, the health service was not transformed from civil service
pay and conditions into a URA equivalent, nor were some functions privatised as with the
customs service. The government recognised these constraints and accommodated them
appropriately. Both recurrent public expenditures and revenue were kept low relative to
other developing economies with less severe problems of professional conduct.

With respect to the private sector, the activity most likely to be damaged by opportunism
was banking. The Ugandan banking sector inherited by the NRM government had four
private banks, all of which had survived by being conservative, and a large government-
owned bank, which had all the problems of the rest of the public sector. The number of
private banks was too few to support competition, and so it was desirable to liberalise the
sector and attract new entrants. However, this priority was in danger of conflicting with
the constraint implied by the high degree of opportunism in society. The latter implied
that the policy priority in the sector was for the central bank to combine the design of
incentives and the supervision of banks sufficiently to avoid dishonest banking practices.

14
In the event, policy probably placed too much weight on the need for liberalisation, and
too little recognition of the constraints implied by the high degree of opportunism. By
1998, several of the new banks had got into severe difficulties, and the privatisation of
the government-owned bank had proved a fiasco.

Finally, consider the policy implications of the massive shift of assets abroad which had
taken place during the period of disorder. This created both a major problem and a major
opportunity for the government. Recall that by 1986 some 60% of private wealth was
held abroad. As a result, the private capital stock per member of the laborforce had been
declining, so that it was among the lowest in the world. Uganda was thus chronically
short of private capital. In principle, it had three means of recapitalising the economy:
through domestic private savings, through attracting in foreign capital, and through
inducing the repatriation of its own wealth. Of these, the first was neither feasible nor
desirable. There was, of course, some scope for raising the private savings rate, but since
incomes had fallen to very low levels, it was important to increase consumption as
rapidly as possible. The second faced the impediment that as a result of its history of
social disturbance Uganda was not seen as being suitable for foreign investment. A good
measure of this is the Institutional Investor risk rating, which is in effect a poll of
international business opinion. As of 1986 Uganda had the worst rating of the twenty-five
African countries which were then rated. Further, there is evidence that the ratings are
quite persistent (al Haque et al, 1999), so that even a rapid improvement in the underlying
investment environment would not lead to a rapid reappraisal of Uganda on the part of
foreign investors. Repatriation was therefore the most realistic option for recapitalising
the economy. If Uganda’s own private wealth could be recovered, the private capital
stock could more than doubled.

Globally, three features influence capital flight: risk, exchange rate over-valuation and
the amount of wealth per capita (Collier, Hoeffler and Pattillo, 1999). Table 4 applies this
global relationship to estimate the share of Ugandan private wealth held abroad in 1985
and 1998, the approach being analogous to that used in Table 1 to measure the change in
the risk of war.

15
Table 4: The Predicted Change in Flight Capital as a Proportion of Ugandan
Private Wealth, 1985-1998

Variable Value of the variable in Uganda Coefficient effect on flight:

1985 1998 change (change in % of
private wealth)

Risk rating 5.1 20.3 +15.2 -0.497 -7.6

Exchange rate misalignment 198 35 -163 -0.000878 -33.0
Wealth ($) 383 395 +12 0.000889 +0.0

Total predicted -40.6

(Actual, 1986-97) -10.0
Notes: The risk rating is that of Institutional Investor, averaged for the year. The exchange rate
misalignment measure is the Dollar distortion index which measures exchange rate misalignment due to a
variety of factors. The value used in the regression is the square of this index. The stock of private wealth
per worker for 1985 is taken from the data generated by Collier, Hoeffler and Pattillo (1999). It is estimated
for 1998 using their methodology, whereby the ratio of wealth/GDP is assumed constant, so that wealth in
1985 is simply scaled up by the growth in GDP. The coefficients on these variables are taken from Collier,
Hoeffler and Pattillo, Table 2A. The contribution to capital flight is the product of the variable and its
coefficient in each year.

The high level of capital flight as of 1986 is no mystery. The two main drivers of capital
flight, poor risk ratings and exchange rate misalignment, were both extreme problems.
The risk ratings, as measured by the Institutional Investor survey of private investment
opinion, were among the lowest in the world at only 5.1 (out of 100). Similarly, the
Dollar distortion index, at 198, showed the Ugandan shilling to be among the most
misaligned currencies in the world. The loss of capital gradually reduced the private
capital stock per worker: by 1986 capital per member of the laborforce was around 10%
below its level of 1971. Uganda thus faced a severe problem of undercapitalization.

The progress of policy reform divides into two clear phases: 1986-92 and 1992 to the
present. In the first phase neither of the major determinants of capital flight were much
improved. The Institutional Investor risk ratings were essentially unchanged by 1992,
standing at only 5.2. There was greater progress on reducing exchange rate misalignment,
but by 1992 the Dollar distortion index was still at the very high level of 87. Hence, our
analysis would predict that, because of the lack of reform of these two variables, there
would be little change in the experience of capital flight. This indeed was the case. In
most of the intervening years capital flight continued to be a severe problem, and as a
result, the private capital stock per member of the laborforce continued to decline. By
1992 it had fallen to only 70% of its 1971 level.

During the second phase, that is post-1992, progress has been remarkable. The
government rebuilt investor confidence by a wide range of measures. It returned
confiscated property so that property rights could both be clarified and restored. This was
a politically difficult step since it involved handing back assets to an emigrant ethnic
minority, sometimes requiring the expulsion of those currently living in properties. The

16
second step was to restore investor confidence through simplifying procedures through an
Investment Authority. The third step was to attain macroeconomic stability and
structural liberalization. Inflation was reduced to single-digit levels and held there,
foreign exchange reserves were accumulated, the exchange rate was gradually made fully
convertible, and trade policy was liberalized. The fourth step was to signal strong
political commitment to the reform process. For example, during the year of the
presidential election around a third of the civil service was dismissed. This signalling was
reinforced by the response of the International Financial Institutions: Uganda was the first
county to gain HIPC status, and thus qualify for debt relief. Cumulatively, these measures
had dramatic effects on confidence. The Institutional Investor risk ratings have improved
sharply, from 5.2 to 20.3, overtaking countries such as Côte d’Ivoire. Similarly, the
liberalization of the exchange rate led to a massive improvement in the Dollar distortion
index, which fell from 87 to 35 (by 1997). Our analysis would predict that this would
lead to a major reversal of capital flight. Indeed, using the coefficients of the Collier-
Hoeffler-Pattillo model, these reforms predict a massive capital repatriation, resulting in
around 40% of private wealth returning to Uganda. The C-H-P model is a cross-section
equilibrium model, and so gives no indication of how long private wealth holders would
take to adjust to this new equilibrium. It might, for example, take a decade for portfolios
fully to adjust to the new policies. However, the Ugandan data on capital flight show an
immediate response. There are four different measures of capital flight available (World
Bank, Dooley, Morgan and Cline). Taking the average of these four measures, there has
been a clear and dramatic reversal, from capital flight to capital repatriation (Table 5).

Table 5: Recent Capital Flight (-) and Repatriation (+)

1991 -$17.3m
1992 -$15.0m
1993 +$16.8m
1994 +$160.3m
1995 +$59.0m
1996 +$108.9m
1997 +$311.3m

Source: World Bank, Development Research Group.

The scale of the repatriation as of 1997, although very substantial, was still well short of
that predicted by the model. Specifically, by 1986 60% of Ugandan private wealth was
outside the country. By 1991 this had worsened to a peak of 67%. By 1997 (the most
recent data) repatriation had reduced the figure to 50%. Thus, 17% of Ugandan private
wealth had been repatriated from abroad between the start of the 1992 reforms and 1997.
The model predicts that around 40% of wealth would be repatriated. This suggests that
portfolios had yet fully to adjust to the new equilibrium, so that repatriation could be
expected to increase further in subsequent years.

To summarize, by 1997 the Ugandan government had turned the tide on capital flight. In
only five years 17% of private wealth had returned to the economy. Yet even in 1997,

17
some 50% of private wealth remained abroad. There was thus enormous potential for
continued repatriation: the private capital stock could be doubled.

The above effects created both constraints and opportunities for a post-conflict
government which would not be present to the same degree in economies embarking on
liberalisation from an inheritance of peace. On balance, there was some reason to expect
that the opportunities outweighed the constraints. Collier (1999) finds that growth during
the first five years of peace after a civil war depends upon the duration of the previous
conflict. After long conflicts growth rebounds, presumably in part because the exodus of
capital has already been completed, so that peace induces repatriation. After a fifteen year
conflict growth rebound by 6% per year over the underlying performance of the
economy. Prior to 1986, Uganda experienced disruption over a period of fifteen years.
Although this disruption was intermittent rather than continuous, recall that the
cumulative effects on output were even more severe than would typically have been
experienced by a continuous civil war of fifteen years. In the period 1986-98 the
economy has considerably out-performed the average, but not by more than might have
been expected for an economy in the early years of recovery from a long internal conflict.
Even by 1998 the economy had not returned to the per capita GDP level reached in 1971.

This is not to belittle the achievement of the government of Uganda. The recovery from a
long conflict is in no sense automatic, and requires careful policy management. Rather, it
suggests that the government has taken the opportunities provided by a post-conflict
situation. However, it cautions against treating Uganda as simply a conventional case of
liberalization: the same policy sequence in a society that had not emerged from a long
conflict might yield a slower pay-off in terms of growth.

5. Conclusion

Post-1986, the Ugandan government has been faced with two policy challenges: reducing
the risks of conflict in a conflict-prone society, and reducing poverty in a very poor
economy. The risk of conflict comes partly from the economic and political inheritance at
Independence, and partly from the economic and political consequences of prolonged
conflict between 1971 and 1986. The economic inheritance implied a high risk of conflict
because of natural resource dependence and low levels of education. The government has
done much to diversify the economy, and to increase educational attainment. The
economic consequences of the prolonged conflict included massive capital flight and the
breakdown of many institutions such as the professions. The government has achieved
considerable success in reversing capital flight through a range of measures which have
rebuilt investor confidence. It has embarked upon the task of rebuilding institutions, but
this is necessarily a long process. The political inheritance at Independence was a paper
democracy, but democratic institutions did not survive when tested. The consequences of
prolonged conflict were a legacy of suspicion and bitterness. The government has made
considerable progress in reconciliation, by encouraging the return of former political
leaders and traditional rulers. It has also built more robust democratic institutions, most
notably a free and active press, and has decentralised decision making.

18
Overall, by 1999 the society was considerably safer from internal large scale conflict than
it had been both at Independence and at the start of the Museveni government. During the
1990s the economy had staged a remarkable recovery from the collapse which had
occurred during conflict, having one of the highest growth rates in the world. However,
even with this impressive performance, it had yet to exceed the level inherited at
Independence.

19
References

Ablo, E. and R. Reinikka (1998) ‘Do Budgets Really Matter? Evidence from Public
Spending on Education and Health in Uganda,” World Bank Policy Research
Working Paper No. 1926.
Collier, P. (1994) ‘Demobilisation and Insecurity: a Study in the Economics of the
Transition from War to Peace’, Journal of International Development, Vol. 6, pp.
343-51.
Collier, P. and A. Hoeffler (1999) ‘Justice-Seeking and Loot-Seeking in Civil War’,
mimeo, , Development Research Group, World Bank.
Collier, P. (1999) ‘The Economic Consequences of Civil War’, Oxford Economic Papers,
Vol. 51, pp. 168-83.
Collier, P. (1999a) ‘Ethnicity and Economic Performance’, mimeo, Development
Research Group, World Bank.
Collier, P. A. Hoeffler, and C. Pattillo (1999) ‘Flight Capital as a Portfolio Choice’,
Development Research Working Paper 2066, World Bank.
Collier, P. and S. Pradhan (1998) ‘Economic Aspects of the Transition from Civil War’,
in H.B. Hansen and M. Twaddle (eds.) Developing Uganda, James Currey, Oxford.
Bigsten, A., and S. Kayizzi-Mugerwa (1994) ‘Adaption and Distress in the African
Economy: a Study of Retrenchment and Reform in Uganda’, Unit for Development
Economics, University of Gothenberg.
Ul Haque, N.(1999) ‘Rating Africa: The Economic and Political Content of Risk
Indicators’, in P. Collier and C. Pattillo (eds.) Investment and Risk in Africa,
Macmillan, London.
Kasekende, L., and Atingi-Ego, M. (1997), ‘Financial Markets and Monetary Policy in
Sub-Saharan Africa: The Experience of Uganda.’ Final Draft Report AERC
Collaborative Research Project, Nairobi, Kenya.
Reinikka, R. and Svensson, J. (1998) ‘Investment Response to Structural Reforms and
Remaining Constraints: Firm Survey Evidence from Uganda’, mimeo, Africa Region,
World Bank.
Tirole, J. (1992) ‘Persistence of Corruption’, Institute of Policy Reform, Working Paper
No. IPR55:2-32.

20