The Economic Legacy of Civil War

Abstract

This article positions itself among the very rare microeconomic analyses on the consequences of civil war. Up to now, most analyses on this topic are based on household surveys. The originality of the present study is that it investigates for the first time the likely predominant route by which civil conflict affects the economy, specifically through firms. The context of the study is Sierra Leone, a country that was ravaged by violent conflict from 1991 to 2002. The approach is to use geographical variations in the intensity of conflict to estimate the impact of violence on firms, on which we have data from the World Bank 2007 Employers’ Survey. The proposed theory is that during conflict, violence affects production through a form of technical regress and demand through a reduction in income. The persistent post-conflict effects are less obvious. We assume that war forces a prolonged contraction in output skills, which slows the pace of recovery. We termed this phenomenon “forgetting by not doing”. The results confirm our theory: the size of firms in 2006 is negatively affected by the intensity of the war in the area it operates. The analysis of training needs clearly corroborates the long-lasting lack of skills experienced as a result of the war in areas where the conflict was more intense. Yet, the analysis cannot identify robust recovery patterns.

Keywords

post-conflict civil war reconstruction firms

Civil war is a difficult context for applied economic research. To date, the economic consequences of civil war and its aftermath have been investigated predominantly at the macroeconomic level, using national accounts data to estimate its impact on gross domestic product (GDP; Barro 1991; Collier 1999; Collier and Hoeffler 2002, 2004, 2008). Such analyses have revealed the titanic consequences of civil war. Using macro-level data, Collier and Hoeffler (2008) estimate the total cost of a war at US$65 billion. War-torn countries experience what the authors qualify a “development in reverse”. At the national level, the loss of economic growth induced by civil war amounts to 2.2 percent for each year of conflict. In addition, the costs induced by the war keep on accumulating once peace is achieved with an estimated fourteen years typical for post-conflict countries to reach their counterfactual GDP. Meanwhile, the relatively rare—but growing—microeconomic analyses have been based on household survey data, focusing on vulnerability to violence (Deininger 2003) and the effects of being a victim of violence upon subsequent political participation (Bellows and Miguel 2009). Yet, the predominant route by which civil war affects the economy is likely to be through firms. To our knowledge, the present study is the first to use firm-level data to investigate the effects of civil war and post-conflict recovery. The added value of such an analysis lies in the microeconomic evidence. We, however, fully acknowledge that the limitations of the data set, as often the case in developing countries affected by long wars, calls for further investigation to allow for robust conclusions.

The context of the research is Sierra Leone. Between 1991 and 2002, Sierra Leone was ravaged by an extremely violent civil war. The war led to over 50,000 deaths, the displacement of 500,000 civilians, and the wholesale destruction of the national economy. By the end of the war, Sierra Leone had sunk to the bottom of the United Nations Development Programme’s (UNDP 2009) human development index (HDI). In 2002, peace was achieved by the intervention of British troops: the insurgent Revolutionary United Front rapidly collapsed. Since then security has been maintained without significant challenge, aided by a continued British military presence. Yet, the pace of recovery has been modest: in 2008, the country remained at the bottom of the HDI.

While the conflict in Sierra Leone was intense and prolonged, it was concentrated in particular locations. Household data analyzed by Bellows and Miguel (2009) provide a reasonable measure of this spatial variation. Our approach is to use this variation to estimate the impact of violence on the growth of firms, on which we have data from a firm survey. Clearly, even firms in areas not directly affected by violence were nevertheless operating in a country at war and this is likely to have affected their performance. During war, government policies will deteriorate, trade with conflict-affected areas will be reduced, and firms may fear that violence will spread to their own areas. Hence, our approach is likely to underestimate the full effects of civil war and so should be thought of as a lower bound.

How might exposure to the violence of civil war affect firms? In the second section, we propose a simple theory. The effects during war are straightforward: violence affects production through a form of technical regress and demand through a reduction in income. However, the persistent post-conflict effects are less obvious: we propose a phenomenon that we term “forgetting by not doing.” In essence, war forces a prolonged contraction in output, skills atrophy through neglect. This slows the pace of recovery. In the third section, we describe the context of the war in Sierra Leone, and in the fourth section, our data. The results are presented in the fifth section.

The Effect of Civil War on Firm Performance

The most striking visual images of violence are those of physical destruction. These effects are potentially persistent even after the conflict has ended. However, in low-income countries they may be relatively minor: there is little capital at risk and after the conflict replacements can swiftly be imported. Miguel and Roland (2011) study the effects of the bombing of Vietnam, probably the most destructive civil war in any developing country and clearly far more destructive than that in Sierra Leone. They find that the bombing left no discernable long-term economic effects. Cerra and Saxena (2008) observe that output partially rebounds after a civil war, in contrast to financial crises. Using panel data, they find that half the loss is recuperated after four years while the other half of cumulative loss remains after a decade. They explain this specific rebound by the fact that infrastructure can be repaired within a short time. However, they do not formulate any hypothesis about the causes of the persistent loss. Rather than physical destruction, the most important effects of civil war on firms are likely to be through the disruption of production through the flight of employees, unreliability of transport, and fear of looting. Faced with unreliable transport, firms would normally have carried larger inventories, but the fear of looting would warrant the opposite response. Such costs of disruption can be characterized as technical regress in the private sector of the economy, and so raise the unit cost of its output.

Firms are also affected by a decline in demand. This is driven by the decline in incomes, and more particularly cash incomes, as people move liquid assets abroad for safety and shift into subsistence activities. The demand for the output of the private sector is thus reduced through a combination of the higher unit cost and the reduced cash income of the wider economy.

These effects in themselves appear to be readily reversible, so that post-conflict recovery should be rapid. With secure peace productivity, demand would be restored. Why might the cost of conflict be more persistent? We propose that in low-income countries such as Sierra Leone, the key channel for persistence is not physical, but human, capital and that the key route by which human capital is lost is through the atrophy of skills.

Skills are maintained through use. The theory of “learning by doing” (Arrow 1962) proposed that “doing” was the main mechanism for learning but more evidently it is essential to the retention of task-specific knowledge. The technical regress induced by civil war reduces the maintenance of skills through two effects. Directly, technical regress amounts to the reversion to production practices, which in normal times would be inefficient. Faced with no electricity, a lack of inputs, stolen equipment, or limited scope for repairing machines, workers may need to switch to manual operations. In effect, they learn reversionary techniques that peace will make redundant and, if the shock persists over a long period,¹ forget techniques that peace will again make feasible and superior. Indirectly, technical regress reduces the maintenance of skills through its aggregate effect on income and hence on demand. Indirectly, the contraction of output in response to the decline in demand leads to a contraction in the labor force, though less than proportionately due to the productivity decline. The decline in employment then deskills that part of the labor force that loses employment in the activity, in a process analogous to the deskilling of workers who remain unemployed for long periods. Pissarides (1992) uses a stylized model to emphasize that the loss of skills associated with unemployment persistence can have long-lasting consequences and may explain the observed slow adjustment of the labor market after a temporary shock on employment. This relates to our theory that the most persistent consequence of war is the destruction of human capital. A parallel can be drawn with the literature on the effects of economic recession. The “effect of the business cycle” (Dynarski and Sheffrin 1990; Baker 1992) translates into an increase in length of unemployment during economic recession. Indeed, the outflow rate for long-term unemployment collapses during recession as employers have larger pools of labor to choose from. In developed countries, wages would adjust and unions would play a major role. In Europe, Machin and Manning (1999) conclude that higher rates of long-term unemployment put an upward pressure on wages, which tend to be higher in periods of high unemployment. In post-conflict situations, however, unions are nonexistent and most workers are employed in the informal sector, which is likely to lead to a longer recovery phase.

The atrophy of skills through reversionary technology and reduced employment is only of consequence once peace is restored. Having adjusted to the conditions of conflict, with the onset of peace firms find themselves in a favorable disequilibrium. Demand increases and there is now scope to abandon reversionary technologies. In response to the disequilibrium, firms expand, but in doing so they encounter a shortage of skilled labor. Firms in conflict-affected areas must therefore make do with lower-quality workers and hence have lower productivity.

This simple theory of the effect of conflict and its legacy on firms has testable implications. The most evident is that during conflict those firms in areas most affected by conflict will contract relative to those in less-affected areas. Similarly, post-conflict, we would expect this pattern to be reversed. A less evident testable implication is that, in the post-conflict phase, there would be an apparent paradox: the fastest-growing firms would be less productive. Finally, in the post-conflict phase, those firms in conflict-affected areas would manifest the most severe shortage of skilled labor.

Taking a Cobb–Douglas production function where $Y_{t}$ the production level at date $t$ depends on $K_{t}$ the stock of capital at time, $L_{t}$ the stock of labor, and $A_{t}$ the stock of knowledge. Arrow (1962) defines the learning by doing effect as

A_{t} = B K_{t}^{Ø},

with $> 0$ ; $Ø > 0$ and $K_{t} = s Y_{t}$ where $s$ is the saving rate. This, however, does not enable us to capture the impact on the stock of knowledge $A$ of a substantial decrease in production from the previous periods, which leads to the resurrection of reversionary technologies, and so a loss of skill in cutting-edge technologies. One way to model such an effect would be to introduce the concept of previous peak output, and allow $A$ to decline in response to shortfalls of current output from previous peak output. The period since the peak level is long enough for the shock to be persistent. Thus, to capture this “forgetting by not doing” effect we rewrite equation (1) as

A_{t} = B K_{t}^{Ø} + C δ_{Δ Y_{t} < 0},

where $δ_{Δ Y_{t} < 0} = 1$ if $Δ Y_{t} < 0$ and 0 otherwise and $C < 0$ . In other words, when $Δ Y_{t} > 0$ , we observe the standard “learning by doing” effect, but when $Δ Y_{t} < 0$ , which is characteristic of conflict environment, the “learning by doing” effect is replaced by a “forgetting by not doing” effect. In the case of a civil war, $C = f (i, T, Δ Y_{t})$ . The magnitude of the effect depends on $i$ the intensity of conflict, $T$ the duration of the war, and $Δ Y_{t}$ the magnitude of the fall in production. As production shrinks during the war when activity reverts to subsistence, the level of knowledge of the economy, $A$ , decreases from its previous peak level.

This article is an attempt to test the veracity of our “forgetting by not doing” theory using country-level data on Sierra Leone. Without a doubt, the main contribution to the existing literature therefore stands at the micro level. The idea is to investigate whether firms in zones highly affected by the war behave differently five years after the end of the conflict. In particular, do those firms face higher shortages in human capital and as a result a higher demand for skills? This would confirm that firms in conflict-affected areas have to make do with lower-quality workers and consequently, that the persistent effect of war on firms is through human capital. In order for us to fully validate our theory, we need to make sure that our analysis does not capture a completely different story. Indeed, a higher demand for human capital could serve as a substitute for physical investment. Equipment might be unavailable or too expensive to import, and as a consequence, employers would invest into human capital to compensate the lack of machines. If this constitutes reality, there is a danger in interpreting investment in human capital as validating the “forgetting by not doing”, while in fact the effects of war persist through the lack of physical capital.

To avoid misinterpretation, we discuss the physical capital–labor complementarity versus substitutability in a country like Sierra Leone. However, as is always the case in analyses focusing on post-conflict environments, we face problems with the quality and availability of data. One of the main limitations is that we have very little information on firms prior to the shock. In particular, we have no information on the capital–labor intensity by sectors prior to the war; neither can we estimate a production function with precision, as we do not have access to data on wages or costs of capital. Teal (2000), while investigating the fall in real wages in the 1990s in Ghana, finds that the elasticity of substitution between unskilled labor and capital is higher than between skilled labor and capital. Duffy, Papageorgiou, and Perez-Sebastian (2004) use a cross-section panel of seventy-three developed and developing countries over twenty-five years to investigate the labor–skill complementarity. Using a lower than usual threshold for skilled capital—they find some evidence of capital–skills complementarity. This confirms the mild evidence found by Fallon and Layard (1975) using a cross-country data for the year 1963. Nevertheless, most of the existing country-level literature on the subject is based on developed countries data, from which we cannot draw any robust conclusion for Sierra Leone.

It seems, however, that labor–skill complementarity might be more likely, even more so for skilled labor. Empirically, if physical investment translates into a lower need for staff training, it is likely that physical and human capital are substitutes, if, on the contrary, higher investments in capital translate into a higher need to train staff, then complementarity can be assumed. If physical and human capital are complementary, which the empirics seem to confirm, our analysis constitutes a lower bound. Indeed, in areas where the conflict was more intense, violence would have destroyed more physical capital, which as a result would have a negative effect on the need for training.

The Shadow of the Civil War

Like most of sub-Saharan Africa, Sierra Leone is noticeable for its very young population. Most of its youth, especially those from rural areas grew up during the decade-long “Dirty War” (Gberie 2005). As a result, both their human capital and transition to adulthood have been dramatically affected adding the difficulties of being an ex-combatant or a victim, the burden of forced migration and psychological long-lasting traumas to widespread poverty. Many children at the time of the conflict were prevented from going to school.² Consequently, the literacy rate in Sierra Leone is particularly low.³ Thirty-five percent of the fifteen- to twenty-four-year-old and 63 percent of the twenty-five- to thirty-five-year-old never attended school (World Bank 2007). Both age groups constitute the core of today’s workforce as life expectancy barely reaches forty-one. Promoting employment opportunities for youth has thus been identified as a core challenge in the Sierra Leone Poverty Reduction Paper and is widely recognized as a necessity for the political stability as a result of the predominant role played by youth during the civil war.⁴

GDP per capita dramatically fell during the 1980s, eventually leading to war and ten years of gradual destruction of the whole economy. Between 1980 and 2000, the GDP per capita was halved from a mere US$300 to US$150 (in constant US$2000) with negative growth rate skyrocketing at more than –15 percent in the 1990s. Since 2000, the economy has been gradually recovering. Yet, despite a successful period of recovery from the conflict characterized by a GDP per capita growth rate close to 4 percent since 2003, the government of Sierra Leone still faces colossal challenges to promote sustainable development. In 2006, 70 percent of the population still lived under the national poverty line, with an average GDP per capita below US$250. In 2008, the West African state still lay at the bottom line of the UNDP’s HDI. Recovery from two decades of war and negative growth is yet to be achieved and Sierra Leone suffers from an overall lack of economic opportunities. Meanwhile, a striking return to agriculture was observed, highlighting a clear switch toward subsistence activities in the early 1980s when the Sierra Leonean economy started to collapse and war started. At the end of the conflict, the size of the service sector had shrunk to a quarter of what it was in 1980, from almost 50 percent of GDP (in value added) to a mere 15 percent in 2000, while the agricultural sector represented more than 60 percent of GDP. Since peace, the relative sizes of sectors are gradually returning to their prewar levels. By 2007, the value added of the agricultural sector as a percentage of GDP followed a decreasing path to just over 40 percent and the service sector size reached 30 percent, double of what it was in 2000 (but still half its relative size in 1980).⁵

On the business environment side, the situation is comparable. Investment is still shy as underlined by the rather small number of large firms operating in the country. Moreover, existing businesses face massive constraints on growth and expansion such as access to finance (due to a completely underdeveloped formal credit system), insufficient access to electricity,⁶ and an apparent lack of resources (World Bank 2007). The 2007 World Bank study also reports that availability of labor does not constitute a constraint in itself for firms, but the availability of skilled workers remains a major problem for the largest ones. The rating of the Doing Business Initiative, which provides a national quantitative measure of different business regulations, confirms that Sierra Leone lies among the countries where it is the hardest to do business. In 2006, when the survey was undertaken, the report emphasized the difficulty of hiring and firing workers as well as of getting credit. In the 2009 Doing Business report (World Bank 2009), which covers the period from April 2007 to June 2008, Sierra Leone ranks 156th of the 181 countries. If it seemed relatively easy to start a business in 2007/2008 relative to the previous year (53rd rank/+41), Sierra Leonean employers face among the worst conditions in dealing with construction permits, but also in employing workers, registering property, getting credit, paying taxes, trading across borders, enforcing contracts, and in closing a business. In more detail, the study reports that it takes 283 days to build a warehouse. Firing costs amount to 189 weeks of salary as opposed to a regional average of 68.3 weeks. On the credit side, the scope, access, and quality of credit information available through public registries or private bureaus is close to nil. The costs of enforcing contracts as a percentage of claims are three times higher than in the region. Overall, this difficult operating environment explains in part the slow growth of the private sector.

In the context of this article, it is necessary to highlight some of the particular features of the war. The war erupted in 1991 at the border with neighboring Liberia and later on spread out over the entire country, reaching Freetown on January 6, 1999. The war techniques included, among others, the systematic targeting of civilians as a terror tactic, rapes, amputation of limbs, abduction of child soldiers, and the intensive looting of diamond resources located mainly in the areas bordering Liberia. The decade of fighting was rhythmed by advancements of the rebel forces toward Freetown followed by defeats against the government army, and later by Economic Community of West African States (ECOWAS) peacekeepers pushing rebels back toward their stronghold in the east. The fighting was not dictated by the destruction of particular economic sectors, apart from mining. In other words, the variation in conflict intensity was not systematically related to the structure of the economy. On the contrary, the primary incentive was to control Freetown, and as rebel forces advanced toward their objective whatever was in the way—villages, villagers, towns—was destroyed. The war tactics were characterized by extensive looting of anything with any value, during Operation Pay Yourself, and by widespread killings of civilians as an effective terror tactic such as Operation Burn House. The deadliest illustration of outrages committed against the civilian population was in the course of Operation No Living Thing, whose title is self-explanatory.

Data

The Sierra Leone Employers’ Survey (SLES) undertaken by the World Bank in 2006 constitutes the core of our database.⁷ The survey covers five districts of the eleven in total: the western urban area, which includes the capital city Freetown; the western rural area surrounding Freetown; Bo district in the south; Bombali district in the north; and Kailahun district, the Revolutionary United Front’s stronghold, in the east at the border with Liberia.⁸ The data coverage well reflects the differences observed in Sierra Leone in terms of economic development, as well as how the country has been affected by civil war in the 1990s.

For our study, which focuses on firms, we relied on data in urban areas that gather information on 419 formal and 248 informal businesses.⁹ All 30 large- and 136 medium-sized firms were surveyed along with 502 small businesses.¹⁰

We use the number of firms per chiefdom¹¹ to control for the state of the economy before the war. Information is provided by the Directory of Business and Industry for the Western Area and Multiunit firms (Sierra Leone Central Statistics Office 1970) and the Directory of Business and Industry for the Northern, Southern and Eastern Provinces (Sierra Leone Central Statistics Office 1968). These are, to our knowledge, the only existing documents that give relevant economic information disaggregated at the chiefdom level before the war. A brief description of the activities of firms allows for their classification by sector matching the SLES classification. Nevertheless, we need to acknowledge that the number of firms obtained in this manner raises some problems. First, in the Freetown area, we cannot ascertain in which of the eight chiefdoms the firm was located. This consequently constrains us to treat the entire western urban area as one single location, instead of eight distinct chiefdoms.¹² Second, we only have a picture of the economy in the early 1970s leaving a twenty-year gap until the beginning of the war in 1991. However, the civil war broke out after decades of mismanagement and corruption since independence from Britain in 1967. The decades following independence appear to premise the coming war. On this basis, the data from 1970 present a good snapshot of the economy before the process of collapse that culminated in the war.

It is important to acknowledge the shortcomings of the SLES data as only war survivor firms are surveyed. Results might consequently be biased—if firm exit was not random—yet, prewar data are insufficient to enable analysis of firm exit patterns and correction of a potential selection bias. Table 1 in the annex reports sectorial variations from 1970 to 2006 in the surveyed chiefdoms, but the information contained in the censuses is too scarce to establish any pattern of firm destruction. In particular, it is impossible to investigate whether firms that disappeared during the conflict were significantly different, that is, in size or productivity level. The impossibility to address this potential selection bias in our results certainly constitutes a recognized limit to the econometric analysis. Yet, in 2006, the bias might be partially offset by business creation/recreation in the immediate postwar years in the recovering sectors.

Table 1.

Descriptive Statistics.

	N	M	SD	Min	Max
Firm level—main variables
Conflict intensity	847	0.4	0.07	0.17	0.8
Freetown	847	0.62	0.49	0	1
Age of the firm	678	9.08	10.71	0	86
Years of operation during conflict	830	2.19	3.7	0	10
No. of employees	691	13.18	59.15	0	1350
Income (approx. using median)	406	9,315,271	9,939,713	500,000	27,500,000
Willingness to invest in training	678	0.2	0.4	0	1
Willingness to invest in training, skilled staff	678	0.15	0.36	0	1
Financial constraint	689	0.39	0.49	0	1
Acquire assets	675	0.31	0.46	0	1
Primary education, g	847	−0.12	0.37	−1	0.33
Secondary education, g	836	0.74	0.53	−1	1.17
Labor turnover	691	0.032	0.11	0	1
% of females	684	0.14	0.21	0	0.9
Sector variations—1970–2006 in % of firms
	SIC	Census (1970)	SLES (2006)
Fishing	900	0.38	1.93
Forestry	800	0	0.59
Poultry	2,015	0.08	0.45
Livestock	200	0	0.45
Mining and quarrying	1,400	0.46	0.45
Manufacturing	3,990	1.86	8.32
Electricity, gas, and water supply	4,900	0.53	0.59
Construction	3,531	1.33	2.97
Wholesale and retail trade	5,040 to 5,990	81.56	47.7
Pharmacy	5,912	0.53	4.31
Repairs	7,600	2.24	4.75
Hotels and restaurants	7,000	1.37	4.9
Transport	4,700	0.46	2.08
Communication	4,800	0.08	1.78
Financial intermediation	6,000/6,100	0.61	1.34
Real estate	6,500	0.08	0.59
Public administration	2	0.11	0.3
Education	8,200	0.08	2.53
Health and social work	8,000/8,300	0	2.53
Agrobusiness	100	1.63	0.59
Other	73	6.64	10.85
Total		100	100

Note: SLES = Sierra Leone Employers’ Survey; SIC = Standard Industrial Classification.

The intensity of war, our variable of interest for this analysis, is calculated following Bellows and Miguel (2009). We compile the average answer by chiefdom to four questions asked to households:

Were any members of your household killed during the war?

Were any members of your household injured/maimed during the war?

Did anyone from this community/neighborhood die as the result of the conflict?

Were any members of this community/neighborhood injured/maimed as a result of the conflict?¹³

We are then left with four variables by chiefdom ranging from 0 to 1. We define the intensity of conflict as the mean of these four variables. It is important to mention that almost 80 percent of households surveyed moved to their community before 2002. We therefore expect that our intensity variable represents reality and is not based on population movements. The variations in intensity observed in the data are reported in figure 1.¹⁴

Figure 1.

Conflict intensity by district and chiefdom.

A problem of endogeneity may arise from the fact that higher intensity areas might have had intrinsic characteristics making them more likely to be affected by hostilities. This could lead to biased estimators in our regressions. To avoid such a possibility, we instrument the intensity of conflict using the distance to Monrovia from the epicenter of the chiefdom in kilometers. The assumption is that the closer to Liberian capital the chiefdom is, the more intense the conflict.¹⁵ On the contrary, distance from Monrovia should not impact the dependent variables. Chiefdoms and firms closer to Monrovia should not be systematically different from chiefdoms and firms located further from Monrovia. Chiefdoms close to Monrovia could be systematically different if roads from Sierra Leone to Liberia were important trade routes. In fact, roads between the two countries are in poor condition and require the use of high clearance vehicles. It takes approximately eight hours to travel from Kenema, the largest town close to the border on Sierra Leone’s side, to the border gate at Bo-Waterside. The last fifty kilometers are gravel roads. Heavy precipitation around the border further increases travelling difficulties in the area, especially during the rainy season (May to December). In addition, the main border bridge to cross the Mano River was closed from 1990 due to war in both countries. It was only reopened in June 2007, after the data were collected. Distance to Monrovia, as a tradable route, is therefore unlikely to have played a role in the recovery of the surveyed firms. First-stage regressions confirm that the distance to Monrovia is a strong predictor of the intensity of conflict. It is highly statistically significant at a 1 percent threshold in all regressions. High R ² values are reported for each of the first-stage regressions. However, we cannot statistically conclude that the instrument validates the exclusion restrictions as we instrument with one exogenous variable only, therefore cannot perform a standard Sargan test. Yet, it is worth underlying that the inclusion of district effects should reduce the likelihood of violation of the exclusion restrictions.

Results

We investigate the effect of war on the size of firms, income, and the willingness of the entrepreneurs to pay to train their employees. We introduce binary variables for the Western Area, Bo, Makeni, and Kailahun districts to reduce the impact of unobservable variables. Standard errors are clustered at the chiefdom level. As a consequence, our estimates are extremely conservative. All regressions are weighted according to the weight of each stratum. In second-stage regressions, the intensity of conflict is instrumented using the distance to Monrovia in hundreds of kilometers. The instrumented variable is then interacted when appropriate. The coefficients of the instrument in the first-stage regressions are reported, as well as the $R^{2}$ values.

The Effect of War on the Size of Firms

We first look into the possibility that the intensity of conflict had an impact on the size of existing firms¹⁶ while estimating:

{S i z e}_{k, j} = α + β_{1} X_{i} + β_{2} X_{k} + β_{3} X_{i} + β_{4} I_{i} + I_{t} * H_{k} + ϵ_{i, k},

where $X_{i}$ are characteristics of chiefdom $i$ and $X_{k}$ of firm $k$ . $H_{k}$ is a set of variables capturing the history of firm $k$ relative to the conflict and value of 1 for the sector $j$ . $I_{t}$ is the conflict intensity level of chiefdom $i$ .

The results are reported in Table 2, columns (1) and (2). The coefficient of the number of years of operation during conflict is positive and significant, underlying that the older the firm the bigger it is. We find a significant effect of the intensity of conflict on the size of the firm when interacted with the number of years of operation during conflict. As expected, the sign of the coefficient is negative. The higher the intensity of conflict, and consequently the disturbances of activities, the smaller the firm is given the date of creation of the firm. The absolute value is about twice that of the direct effect of the number of years during the conflict that the firm operated. Thus, firms operating in chiefdoms that had more intense fighting are in 2006 significantly smaller than other firms, ceteris paribus. A possible explanation could be that they faced a shortage of labor during the conflict as workers joined the armed forces or were displaced. This leads to a drop in terms of inputs available in a labor-intensive economy as well as a fall in demand. Take an example of three similar firms A, B, and C, assuming all were created before the war, and are located in chiefdom with similar characteristics. All survived through eleven years of war. However, firm A—our benchmark firm—was located in a (fictive) chiefdom where no was conflict reported, while firm B was operating in a low-intensity conflict zone (say 0.2 of 1) and firm C was located in a high-intensity conflict area (0.6 of 1). Using the results of column (2) where the intensity of conflict is instrumented, firm C, everything else being equal, will have twenty employees fewer than firm B, which is located in the low-conflict zone, and twenty-nine fewer employee than firm A, which operates in a chiefdom with no conflict at all. Setting the initial size of the three firms at the average of the sample (in 2006), thirteen employees, firm A have grown to thirty-seven employees by 2006, firm B to twenty-seven employees, while Firm C have contracted to eight employees.¹⁷

Table 2.

Size and Income of the Firms.

	Size		Income
	OLS	IV	Ologit	IV
	(1)	(2)	(3)	(4)
Ln no. of firms in 1970	–0.388 (–1.594)	–0.317 (–1.191)	–0.066 (–0.368)	–0.022 (–0.176)
Ln no. of firms in 2006	–0.533 (–1.446)	–0.472 (–1.248)	0.459* (1.684)	0.693*** (3.348)
Ln population in 1985	1.838** (2.381)	1.688** (2.304)	–0.433 (–1.017)	–0.820** (–2.436)
Intensity	–0.930 (–0.130)	–0.942 (–0.087)	–2.213 (–0.877)	–8.785*** (–3.918)
Age of the firm	0.174 (0.906)	0.174 (0.902)
No. of years during conflict	2.082** (2.145)	2.211** (2.154)
No. of years during conflict × intensity	–4.157** (–2.521)	–4.481** (–2.457)
No. of years since conflict	–1.425 (–1.450)	–1.645 (–1.479)
No. of years since conflict × intensity	3.561 (1.412)	4.126 (1.435)
No. of employees			0.021* (1.719)	0.021* (1.722)
Cons	–11.210 (–1.559)	–10.570 (–1.193)
R ²	.02	.02
Log likelihood			–6,207.38	–6,184.13
χ²			15.850	34.894
Number of observations	610	610	406	406
First-stage regression
Instrument: Distance to Monrovia		–0.153 (0.024)***		–0.150 (0.025)***
R ²		.91		.94

Note: OLS = ordinary least squares. T-stats in parentheses. District effects not reported; standard errors clustered by chiefdom.

***p < .01. **p < .05. *p < .1.

We also investigate possible recovery patterns by interacting the number of years since conflict with the intensity of conflict. The sign of the coefficient suggests a recovery phenomenon in more affected areas. Nevertheless, the effect is not significant when the standard errors are clustered at the chiefdom level. The data thus suggest that firms affected by conflict will not catch up, in terms of size proxied by the number of employees, once peace is restored. This implies that those firms will always lag behind, at least in term of size, without a targeted intervention. Note that the coefficient of the numbers of years during conflict interacted with conflict intensity is significantly different from the coefficient of the numbers of years since conflict interacted with conflict intensity. In other words, firms in high-intensity areas have a very different experience during and after conflict.

The Effect of War on Firm Income

Studying the impact of conflict on the income $Y_{}$ of the firms requires the use of an ordered probit: employers were asked to answer questions on income using ranges. The income variable thus varies from 1 to 7, 7 being the highest range.¹⁸ Hence, we estimate the following equation:

P R (Y_{i, k} = c {) = P r (k}_{c - 1} {< β}_{1} X_{i, c} {+ β}_{2} X_{k, c} {+ β}_{3} I_{i, c} + μ_{i, k, c} \geq k_{c}),

where $Y_{i, k} ϵ [1; 7]$ . $β s$ are estimated together with the cut points, $k_{1}, k_{2}, k_{3}, \dots, k_{6}$ .

We performed a Brant test to verify that the proportional odds assumption is not violated. The result suggests that the assumption is violated for the number of employees and the population of the chiefdom in 1985 but that the proportional regression assumption stands for the intensity of conflict variables. Moreover, the estimated coefficients are similar when we use an ordered logistic model where the proportional odds assumption is relaxed for the variables concerned. Therefore, in Table 2, only the first set of regressions are presented in columns (3) and (4). The coefficients of the intensity of conflict are negative suggesting that the more violent the conflict, the lower the level of income of the firm ceteris paribus. The coefficient is only significant when the intensity is instrumented. According to the existing literature on firm destruction in developing countries, small firms and large unproductive firms face the highest odds to disappear (Söderbom, Teal, and Wanbugu 2005; Frazer 2005; Söderbom and Teal 2000; Hardling, Soderbom, and Teal 2006). Potentially, this pattern of exit could have been accentuated in highly intense conflict zones with a higher destruction. As a result of the potential selection bias, as the data might miss firms that have exited, the coefficient of the intensity of conflict variable could be upward biased. Consequently, the negative effect on the income of firms might have been significantly larger if smaller firms and unproductive large firms had survived.

Similar to the study of the size of firms, we looked at the impact of the war history of the firm while introducing the number of years of operation during and after conflict as well as interactions with the intensity of conflict. None of these variables prove to have a significant effect on income. They are consequently dropped to allow for a larger sample size. Although conflict intensity does impact income negatively, the history of war does not seem to matter. We cannot find robust evidence of a recovery pattern.

The Effect of War on Skills

The intensity of conflict has a negative effect on both the size and the income of the firm. The findings imply that the conflict affected the availability and quality of capital. Hence, it gives us a first sense, which confirms the theory developed earlier that conflict affects firms by destroying human capital. The design of the SLES employers survey enables us to investigate this hypothesis more deeply through data on staff training. Specifically, we investigate whether the intensity of conflict affects the reported willingness of employers to pay for training of their staff.¹⁹ The underlying assumption is that the more employers feel a need to train their employees, the more severe their shortage of qualified labor must be. We therefore estimate the following equation using a probit model:

{T r a i n i n g}_{i, k} = α + β_{1} X_{i} + β_{2} X_{k} + β_{3} I_{i} + ϵ_{i, k},

where ${T r a i n i n g}_{i, k}$ is a dummy variable on whether the employer of the firm $k$ operating in chiefdom $i$ wants to invest more in the training of employees. $X_{i}$ and $X_{k}$ are vectors of variables relative to the chiefdom $i$ and to the firm $k$ . We include the previous year labor turnover as well as the percentage of female workers in the firm’s characteristics, assuming that those could negatively impact the willingness to pay for training. $I_{i}$ refers to the intensity of the war in chiefdom $i$ .

We acknowledge that the willingness to pay for training variable might not be the ideal proxy for human capital and could in fact capture unobserved characteristics such as wealth effects, credit constraints, or the costs of physical capital. Nevertheless, we control for whether financing is identified by the employer as one of the main constraints for growth and expansion of the business as well as whether physical assets were acquired in the two years preceding the interview. Before introducing those as controls in the regression, we investigated whether there is a significant relationship between the financing constraint, physical capital investment, and the intensity of conflict. The results are reported in Table 3. We observe a positive impact of the intensity of conflict on the probability that firms face major financing constraints (columns 1 and 2). This is not surprising, as more destruction will require in return higher investment for firms to recover. Yet, the results are not significant. But recall that conflict has a negative effect on revenue, which itself significantly reduces financial constraint. By contrast, there seem to be a negative and significant relationship between the intensity of conflict in the chiefdom and the proxy for capital investment (columns 3 and 4). The coefficient of the interactive term between financial constraint and the intensity of conflict is positive and significant at 10 percent after instrumentation. Firms operating in zones where the conflict was more intense and report financing as a major constraint to growth appear to have invested more in physical capital over the past two years. This potentially explains the financial constraint. This positive effect offsets the direct negative impact of the intensity of conflict in the chiefdom. However, we do reject the hypothesis that the direct and combined effects of the intensity are null (column 4). In fact, physical capital levels might have started to caught up faster in areas of greater destruction as the data were collected five years following the signing of the peace agreement, but there is still a persistent negative effect of the intensity of war.

Table 3.

Financial Constraint and Investment in Physical Capital.

	Financial Constraint		Physical Capital Investment
	Probit	IV	Probit	IV
	(1)	(2)	(3)	(4)
Ln no. of firms in 1970	0.146*** (0.050)	0.135*** (0.049)	0.031 (0.035)	0.019 (0.027)
Ln no. of firms in 2006	–0.115* (0.065)	–0.093 (0.064)	0.095** (0.038)	0.076** (0.032)
Ln population in 1985	–0.038 (0.090)	–0.062 (0.086)	–0.251*** (0.074)	–0.208** (0.081)
No. of employees	0.001 (0.001)	0.001 (0.001)	0.001 (0.001)	0.001 (0.001)
Firm revenue	–0.059** (0.029)	–0.060** (0.029)	0.065*** (0.016)	0.066*** (0.016)
Intensity	1.143 (0.731)	0.615 (0.751)	–1.837*** (0.669)	–1.138* (0.683)
Financial constraint=1			–0.236 (0.190)	–0.267 (0.166)
Financial constraint=1 x intensity			0.623 (0.491)	0.702* (0.424)
Log-likelihood	–249.34	–250.11	–236.11	–237.79
χ²	31.450	61.666	82.405	100.476
Number of observations	398	398	398	398
Adjusted R ²	.054	.051	.099	.093
First-stage regression
Instrument: Distance to Monrovia		–0.151		–0.151
Instrument: Distance to Monrovia		0.025***		0.025***
R ²		.943		.943

Note: Standard errors in parentheses. Marginal effects, district effects not reported, standard errors clustered by chiefdom.

***p < .01. **p < .05. *p < .1.

The results of the analysis of training partners are reported in Table 4. Columns (1) and (2) refer to the training of all staff while the dependent variable in columns (3) and (4) is restricted to skilled and management staff.

Table 4.

Willingness to Pay for Training.

	Training		Training (skilled only)
	Probit	IV	Probit	IV
	(1)	(2)	(3)	(4)
Ln no. of firms in 1970	–0.099*** (0.032)	–0.109*** (0.039)	–0.012 (0.031)	–0.010 (0.025)
Ln no. of firms in 2006	–0.078** (0.033)	–0.075** (0.035)	–0.080 (0.060)	–0.136*** (0.051)
Ln population in 1985	0.446*** (0.080)	0.390*** (0.082)	0.147*** (0.049)	0.219*** (0.055)
Primary schooling, g	–0.324*** (0.097)	0.145 (0.272)
Secondary schooling, g			0.093 (0.081)	0.167** (0.083)
No. of employees	0.001 (0.000)	0.000 (0.000)	0.000 (0.000)	0.000 (0.000)
Firm revenue	0.023** (0.011)	0.020* (0.011)	0.018** (0.008)	0.017** (0.008)
% of females	–0.103 (0.111)	–0.128 (0.111)	–0.096 (0.097)	–0.100 (0.095)
Labor turnover in 2005	0.167 (0.189)	0.194 (0.186)	–0.038 (0.057)	–0.027 (0.056)
Intensity	0.208 (0.491)	2.739** (1.267)	0.825 (0.596)	1.779*** (0.545)
Acquired new assets = 1	0.068 (0.064)	0.084 (0.065)	0.053 (0.039)	0.053 (0.038)
Financial constraint = 1	0.785*** (0.156)	0.755*** (0.156)	0.570** (0.264)	0.506** (0.291)
Financial constraint = 1 × intensity	–1.460*** (0.457)	–1.368*** (0.394)	–0.917* (0.491)	–0.808 (0.508)
Log-likelihood	–172.71	–172.35	–132.60	–132.01
χ²	335.049	317.473	1,236.836	354.227
Adjusted R ²	0.137	0.139	0.129	0.133
Number of observations	385	385	380	380
First-stage regression
Instrument: Distance to Monrovia		–0.125 (0.038)***		–0.141 (0.013)***
R ²		.940		.974

Note: Standard errors in parenthesis. Marginal effects, district effects not reported, standard errors clustered by chiefdom.

***p < .01. **p < .05. *p < .1.

Similar to previous regressions, we control for the number of firms by chiefdom in 1970 and 2006 as well as the population in 1985 (Sierra Leone Central Statistics Office 1992). We also control for the variation in the levels of education, as the decline in schooling might be an underlying cause of a higher need for training. We construct the variable using the NPS 2005 section containing information on the education level of the respondent.²⁰ In order to capture the variation between education levels before and after the war we construct two groups, one composed of respondents who were born before 1980, which means that they were in age of having completed primary school at the beginning of the war (i.e., they were eleven years old), and those born after 1980, whose schooling was potentially disturbed by the conflict. We use the same method for secondary education, comparing those born before or after 1970. The results suggest that firms located in chiefdoms where the primary education rate decreased—potentially where conflict was intense—are more likely to be willing to invest in training (column 1). However, this observation is not robust to the instrumentation of conflict intensity. On the contrary, firms are more likely to be willing to train their skilled staff in chiefdoms, where secondary schooling rates have increased relative to prewar levels.

At the firm level, we control for the size (proxied by the number of employees) as well as the income of the firm. There is no significant size effect and while the income effect is significant and positive it remains relatively small. In addition, we introduce variables that capture the gender composition of the firm and the reported labor turn over in the preceding year (2005) as both can influence the willingness of an entrepreneur to train staff. Indeed, an entrepreneur might be less likely to train women and be less willing to invest in staff when high turnover was experienced in the past. The signs of the coefficients confirm the intuition though the coefficients are not significant. We also control for potential investments in physical capital (proxied by a binary variable whether the firm acquired new assets in the past two years). The coefficient of the variable is positive but never significant. If capital and labor were substitutes, we would expect firms that invested in physical capital to be significantly less willing to train staff. On the contrary, the results suggest that investing in new assets does not impact significantly on the need for skills. If anything, the coefficient of the variable capturing recent investment in physical capital is positive. We can therefore confidently argue that human and physical capital are complements rather than substitutes. As previously discussed, the analysis therefore constitutes a lower bound. Financial constraint, on the other hand, seems to increase the need to train staff, which might imply that when employers are constrained to invest, they train their staff. Yet, the coefficient of the interacted term between intensity of conflict and financial constraint is negative and significant. In high-conflict zones, the positive effect on training is (partially) offset.

We find a positive, large, and significant effect of the intensity variable (column 2). Thus, conflict appears to cause a scarcity of skills in those chiefdoms where the war was more intense, as indicated by the greater need to train staff. The reported marginal effects are strikingly high. This is consistent with our theory of “forgetting by not doing”: in areas where the conflict was more intense, employers show a significant higher need for human capital than did those in regions less affected by conflict, when controlling for both financial constraint and physical capital investment.

We also investigate the willingness to pay for training of skilled staff as a robustness test. As discussed earlier, skilled labor elasticity to physical capital is lower than unskilled labor. The results presented in columns (3) and (4) display similar patterns, confirming direct effect of the intensity of conflict on the needs to train staff, hence the “forgetting by not doing” effect.²¹

We also attempted to investigate potential variations between sectors. We introduced in the above regressions sector dummies interacted with the intensity of conflict variable. With standard errors clustered at the chiefdom level, we observe little variations between sectors, and therefore, we cannot pretend to identify strong trends. The lack of significant results may be due to the small sample size and panel data could have potentially led to other results. The results are not reported.

Conclusion

Using a firm survey, we investigated the economic legacy of the 1991–2002 war in Sierra Leone. The theory proposed is that conflict results in a significant loss in human capital stock as a result of a “forgetting by not doing” phenomenon, broadly analogous to learning by doing. Civil war induces the reversion to more subsistence activities and less-sophisticated techniques, so that skills atrophy in a manner similar to the deskilling observed after long periods of unemployment. During the post-conflict phase, the growth of firms is therefore slower as human capital has become scarce and takes time to rebuild.

Our results support this hypothesis. Using geographical variations in the intensity of conflict, we find that the more the firms were exposed to conflict, the smaller they were in 2006. Nevertheless, investigating the recovery patterns, we observe no significant catch-up phenomenon in the areas badly hit by war. If this observation is not the result of conservative estimates, it implies that firms in areas affected by war will always lag behind without any targeted intervention. The analysis of training patterns confirms the lack of skills experienced as a result of war. Entrepreneurs’ willingness to pay for the training of their staff is significantly higher in those areas of the country most affected by the conflict, indicating a more acute shortage of skilled labor, controlling for both investment in physical capital and financial constraint. As a result, post-conflict governments should prioritize training in sectors that produce nontradable goods, especially the capital-intensive nontradable goods.

The “forgetting by not doing” phenomenon has important implications for the recovery of war-torn countries. Indeed, firms have to make do with lower-quality workers. This negatively impacts their productivity and the acquisition of new techniques of production. As a result, the growth process is likely to be slowed down. In fact, post-conflict environments experience “supranormal growth” (Collier and Hoeffler 2002), underlying a catching up of the economy, possibly driven by physical reconstruction. The rate of catching up is, however, diminished by the destruction of human capital, which takes long to recover its counterfactual, if ever. Firms are also less competitive, and exports will suffer accordingly. Our results suggest that firms operating in conflict zones five years after the end of the civil war still lag behind. If the lack of skills in post-conflict economies is tackled early, it is likely that recovery would be much faster and more sustainable.

The lack of skills can have much longer term consequences. Indeed, civil war not only deskills the existing labor force, but conflict also creates important disturbances in schooling. The education system collapses (often completely) and children miss vital years of schooling. As mentioned earlier, in Sierra Leone, children were forcibly enrolled in the guerilla forces, some as young as seven years old. Once peace is settled, there is a clear necessity to reconstruct the education system and encourage rehabilitation through training, but there are little competencies in the economy to achieve this. If not addressed, generations will continue to be lost. Education and training are a necessity, but an important lack of competencies needs to be overcome. One solution is to encourage the diaspora to return and pass on the knowledge acquired abroad. Psychological support and social work will also be needed to allow those youngsters who grew up in violent circumstances to benefit fully from training programs in order to integrate into society and participate efficiently in the labor force.

Footnotes

Acknowledgments

The authors thank Daniel Ali, Jean-Claude Berthélémy, Fabricio Corricelli, Klaus Deininger, Jed Friedman, Damien De Walque, Eliana La Ferrara, and Patrick Guillaumont for useful comments as well as two anonymous referees for guidance. The authors are also grateful to the participants of the 2008 CSAE conference and the doctoral days in CERDI for their comments. The second author is thankful to Jean-Marc Chataigner, Bernard Mokam, Bernd Eckhardt as well as the entire UNDP Sierra Leone country team for their support, and to Pia Peters for sharing the data. All mistakes are ours.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Marguerite Duponchel acknowledges funding from the Université Paris 1—Panthéon Sorbonne.

Notes

References

Arrow

1962. “The Economic Implications of Learning by Doing.” Review of Economic Studies 29 (3): 155–73.

Baker

1992. “Unemployment Duration: Compositional Effects and Cyclical Variability.” American Economic Review 82 (1): 313–21.

Barro

1991. “Economic Growth in a Cross Section of Countries.” Quarterly Journal of Economics 106 (2): 407–43.

Bellows

Miguel

. 2009. “War and Local Collective Action in Sierra Leone.” Journal of Public Economics 93:1144–57.

Cerra

Saxena

S. C.

. 2008. “Growth Dynamics: The Myth of Economic Recovery.” American Economic Review 98 (1): 439–57.

Collier

1999. “On the Economic Consequences of Civil War.” Oxford Economic Papers 51 (1, Symposium on Trade, Technology, and Growth): 168–83.

Collier

Hoeffler

. 2002. “Aid, Policy and Peace.” Defence and Peace Economics 13:435–50.

Collier

Paul

Hoeffler

2004. “Conflicts.” In Global Crises, Global Solutions, edited by Lomborg

. Cambridge: Cambridge University Press.

Collier

Hoeffler

. 2008. “Aid, Policy and Growth in Post-Conflict Societies.” European Economic Review 48:1125–45.

10.

Deininger

2003. “Causes and Consequences of Civil Strife: Micro-Level Evidence from Uganda.” Oxford Economic Papers 55 (4): 579–606.

11.

Duffy

Papageorgiou

Perez-Sebastian

. 2004. “Capital-Skill Complementarity? Evidence from a Panel of Countries.” Review of Economics and Statistics 86 (1): 327–44.

12.

Dynarski

Sheffrin

S. M.

. 1990. “The Behavior of Unemployment Durations over the Cycle.” The Review of Economics and Statistics 72 (2): 350–56.

13.

Fallon

P. R.

Layard

P. R. G.

. 1975. “Capital-Skill Complementarity, Income Distribution, and Output Accounting.” Journal of Political Economy 83 (2): 279–302.

14.

Frazer

2005. “Which Firms Die? A Look at Manufacturing Firm Exit in Ghana.” Economic Development and Cultural Change 53 (3): 585–617.

15.

Gberie

2005. A Dirty War in West Africa: The RUF and the Destruction of Sierra Leone. London: Hurst.

16.

Hardling

Soderbom

Teal

. 2006. “The Determinants of Survival among African Manufacturing Firms.” Economic Development and Cultural Change 54 (3): 533–55.

17.

IRCBP (Institutional Reform and Capacity Building Project) of Sierra Leone. 2005. National Public Services Database. Freetown: Government of Sierra Leone.

18.

Machin

Manning

. 1999. “Chapter 47 The Causes and Consequences of Longterm Unemployment in Europe.” In Handbook of Labor Economics, edited by Ashenfelter

Orley C.

Card

David

, 3085–139. Princeton, NJ: Elsevier.

19.

Miguel

Roland

. 2011. “The Long Run Impact of Bombing Vietnam.” Journal of Development Economics 96 (1): 1–15.

20.

Pissarides

C. A.

1992. “Loss of Skill During Unemployment and the Persistence of Employment Shocks.” Quarterly Journal of Economics 107 (4): 1371–91.

21.

Sierra Leone Central Statistics Office. 1968. Directory of Business and Industry for Northern, Southern and Eastern Provinces. Freetown, Sierra Leone: Central Statistics Office.

22.

Sierra Leone Central Statistics Office. 1970. Directory of Business and Industry for Western Area and Multiunit Firms. Revised ed. Freetown, Sierra Leone: Central Statistics Office.

23.

Sierra Leone Central Statistics Office. 1992. 1985 Population and Housing Census of Sierra Leone. Freetown, Sierra Leone: Central Statistics Office.

24.

Smillie

Gberie

Hazleton

. 2000. The Heart of the Matter: Sierra Leone, Diamonds and Human Security. Waterloo, Canada: Ploughshares Monitor.

25.

Söderbom

Teal

. 2004. “Size and Efficiency in African Manufacturing Firms: Evidence from Firm-Level Panel Data.” Journal of Development Economics 73 (1): 369–94.

26.

Söderbom

Teal

Wanbugu

. 2005. “Unobserved Heterogeneity and the Relation between Earnings and Firm Size: Evidence from Two Developing Countries.” Economics Letters 87 (2): 153.

27.

Teal

2000. “Real Wages and the Demand for Skilled and Unskilled Male Labour in Ghana’s Manufacturing Sector: 1991-1995.” Journal of Development Economics 61 (2): 447.

28.

UNDP (United Nations Development Programme). 2009. Human Development Indicators. New York: United Nations Development Programme.

29.

World Bank. 2007. “Improving the Opportunities for Sustainable Youth Employment in Sierra Leone.” Technical Report, Washington DC.

30.

World Bank. 2008. World Development Indicators. Washington, DC: World Bank.

31.

World Bank. 2009. Doing Business Sierra Leone. Washington, DC: World Bank.