Natural Resources and Conflict in Africa: Isolating Facts from Fiction

The Causality Question
Evolutionary biologists have long argued that competition over scarce natural resources is one of the key drivers of violent conflict within and across species. The competition in Africa appears to be over the revenue generated from scarce natural resources which often leads to violent conflict. Chilling examples of conflict in Angola, Democratic Republic of Congo, Rwanda, Sudan and other resource rich regions of Africa often tempts scholars into arguing that resources cause conflict. Figure 1 somewhat visualises this narrative whereby a rebel in Angola is seen to be guarding a diamond mine. Yet establishing causality has remained elusive largely due to the obvious limitations associated with cross-country studies and the lack of useful data for Africa.


Fig 1: A Rebel Guarding an Illegal Diamond Mine in Angola

The sketchy and weak evidence have not prevented scholars from speculating on the potential causes of conflict. Some have argued that armed conflict is often the result of greed rather than grievances. For example, access to an oil rig or a mine could provide lucrative financial opportunities to rebel leaders to build and sustain rebel organisations which would encourage armed conflict. This could override atypical social motives such as inequality, political repression, and ethno-religious division. Even though intuitively appealing, these theories and explanations do little in terms of resolving the issue one way or the other. For that we would require hard evidence which could be interpreted as causal. Our recent CSAE working paper embarks on a project to address precisely that.

The Study
In particular, the paper aims to systematically explore the causal effect of giant and major oil and mineral discoveries on internal armed conflict onset and incidence in Africa at the grid level corresponding to a spatial resolution of 0.5 × 0.5 degree covering the period 1946 to 2008. Using geocoded data on resource (oil and mineral) discoveries and conflict the paper is able to construct a quasi-natural experiment to establish causality. In other words, we are able to test whether giant resource discovery as an exogenous news shock has any bearing over conflict onset and incidence at the local level in Africa. While doing so, the paper uses a novel geocoded dataset of resource discovery at the grid level. In particular, the new dataset is able to distinguish between 11 different minerals[1] and oil discoveries.

The popular discourse both within the academy and the press is that competition over resource wealth in Africa is the root cause of armed conflicts. Yet this is not borne out in our grid level geocoded data. A quick snapshot of the continental map of mining and conflict locations in figure 2 reveal very little correlation between locations of resource discovery and armed conflict onset. In fact contrary to the conventional wisdom of resource riches triggering conflict in Africa, we find oil and mineral discoveries significantly reduce the likelihood of conflict onset up to ten years post resource discovery in a simple pooled cross-section set up with a sample of 47 African countries drawn from the Peace Research Institute Oslo Grid (PRIO-GRID) dataset over the period 1946 to 2008. The effect remains negative but statistically insignificant or weakly significant in most specifications when we control for grid-specific fixed factors and time varying common shocks. We observe little or no heterogeneity in the relationship across resource types (minerals or oil), size of discovery (giant or major), pre and post end of the cold war, and quality of national political institutions measured by Polity2 score. We also analyse the effect of resource discovery on conflict incidence using a panel of 47 countries covering the period 1989 to 2008. Even though the negative effect of resource discovery on conflict incidence remains in a pooled cross-section set up, the trajectory of the coefficient appears to be somewhat different once we control for grid and year fixed effects. The effect stays negative for the first four years post discovery and then it turns positive and peaks at eight years post discovery. This could be due to a decline in the incidence of ongoing conflict up to four years post discovery news shock. Beyond that point production starts in these newly discovered locations which trigger an increase in the conflict incidence. The heterogeneous effects of discovery and production appear to be consistent with the observation that the prospect of future production affects conflict onset and incidence less than the actual past production.


Figure 2: Oilfield and Mineral Discovery Locations and Armed Conflict Onset

The obvious question here is via what channel resource discovery affects conflict. Natural resource induced higher income at the local and national levels could improve state counter-insurgency capacity and reduce individual incentives to fight. Alternatively, natural resource induced higher state revenue could incentivize capture and trigger conflict. Evidence appears to be favoring the former theory rather than the latter. Our analysis finds that resource discovery (oil and minerals) improves luminosity at the grid level which in turn reduces conflict onset and incidence.

For the technically minded, our identification strategy relies on the exogenous variation in the discovery dates of giant and major mineral deposits and giant oil deposits. Our dataset codes a mineral deposit as giant if it has the capacity to generate at least USD 0.5 billion of annual revenue for 20 years or more accounting for fluctuations in commodity price. Similarly, major mineral deposits are those which could generate an annual revenue stream of at least USD 50 million but not as long life as a giant reserve. A giant oil or/and gas (including condensate) field is defined as a field that contains at least a total of 500 million barrels of ultimate recoverable oil or gas equivalent. Even though it is possible to identify the area where minerals or oil are likely to be found using geological data, it is not possible to accurately predict the timing of giant and major discoveries. Therefore, the discovery dates of giant and major reserves are exogenous. One might argue that politicians and government could manipulate the announcement of the precise timing of discovery. Our data is immune to such possibility as the discovery dates are independently verified and documented using multiple sources.

In spite of her colonial and post-colonial history as a supplier of raw materials, a vast majority of African natural wealth remains untapped. These resources are expected to be exploited over the coming two to three decades amid increasing global demand for raw materials. The expected steady depletion of natural resources and the favorable global commodity prices presents Africa with an opportunity to harness this wealth for improving state capacity and living standards. Our research suggests that both of these factors could significantly contribute towards eradication of civil conflict from the African continent.

Arezki, R., S. Bhattacharyya, and N. Mamo (2015). “Resource Discovery and Conflict in Africa: What Do the Data Show?” CSAE Working Paper No. 2015-14, University of Oxford.

[1] The minerals are gold, silver, platinum group elements (PGE), copper, nickel, zinc, lead, cobalt, molybdenum, tungsten, uranium oxide.

Posted in Conflict | Tagged , , , , , | Leave a comment

Ethics in Development Economics


On 3rd Nov we had a seminar presentation by Johannes Haushofer on a RCT in Kenya. In this experiment, some households  in villages is Western Kenya were given unconditional cash transfers of either USD 404 or USD 1525. The researchers measured psychological wellbeing and found, unsurprisingly, that the lucky ones were happier and that their unlucky neighbours were unhappy. The paper is aptly titled “Your Gain is my Pain”. During the seminar everybody appeared to be trying to understand the identification strategy and thinking of ways to criticise it.

I was just overwhelmed by a sense that this type of research should not be done at all. No matter how clever the identification strategy is. Am I the only one to think that is not ethical dishing out large sums of money in small communities and observing how jealous and unhappy this makes the unlucky members of these tight knit communities? In any case, what do we learn from a study like this? The results indicate that we should not single out households and give them large sums of money. So I hope this puts an end to unethical RCTs like this one.


The picture is taken from


Posted in Policies to Protect the Poor | Tagged , , , , , | 7 Comments

Universal access to HIV treatment: fiscal sustainability and incentives.

Before the advent of anti-retroviral therapies in the nineties, being infected by the HIV virus was equivalent to a death sentence. Now, thanks to remarkable advances in medical research, people who are HIV+ have the possibility to considerably increase their life expectancy and live relatively normal lives as long as they take anti-retroviral drugs properly. In sub-Saharan African countries, the yearly cost of generic antiretroviral drugs is less than USD150 for one person receiving treatment[1]. While this cost seems low compared to world GDP per capita (about USD 11,000), it may prove unaffordable in poor countries characterized by a generalized epidemics. From a dramatic medical issue, HIV is turning into a large and growing threat to fiscal stability in high-prevalence and poor countries. This blog post, written as part of ReThinkHIV, explores the implications for Africa[2].

Our argument starts from the proposition that if an individual can be saved from direct deadly threat with certainty at the cost of a few hundred dollars, there is a reasonable presumption that somewhere, globally, there resides a moral duty to rescue this individual. At a cost of less than USD150 per year for generic drugs and about the same amount for laboratory and health care costs, it would (only) require USD11 billion to provide free drugs to the 37 million individuals infected with HIV. To put these numbers in perspective, this represents 0.015% of world GDP; 1.6% of the military budget of the U.S.; 46% of U.S. federal funding for fighting HIV/AIDS domestically; and 3.4% of what OECD countries committed to give as international aid at the 2005 G-8 Summit in Gleneagles. This cost is further reduced by 23% if treatment is only provided below a CD4 count of 500 cells/mm3, and reduced by 57% if treatment is only provided below a 350 cells/mm3. Provided with these numbers, most people would agree that there exists somewhere a moral duty to provide free ART to HIV+ individuals who cannot afford to pay for their treatment.

HIV is a long-term infection causing the progressive failure of the immune system for which no definitive cure exists yet. Even if antiretroviral drugs are taken properly, the HIV virus will not fully disappear from the infected blood, and the viral load is expected to rebound if treatment is stopped. ART needs to be sustained in the long run, and is therefore more similar to the treatment of a long-term chronic disease than to a one-off treatment of a fatal disease. This difference has important economic implications in the long run. Offering life-long antiretroviral treatment to HIV+ people needing treatment generates a long-term financial liability which is analogous to the liability of sovereign debt.

Estimating the fiscal burden

Figure 1 compares the total cost of providing ART therapies to all those who need it with the external debt stock of countries. The blue bars represent the external debt stock of countries as a proportion of their GDP. The red bars are estimates of the discounted cost of full ART coverage in the long run for high-prevalence countries as a fraction of their current GDP. Each red bar represents the amount that a national government should save now in order to pay ART for their citizens in the long run. Figure 1 clearly shows that the moral duty of rescue constitutes a significant addition to the sovereign liabilities of high-prevalence countries. The fiscal impact seems particularly dramatic for countries like Malawi, Lesotho and Zimbabwe. A rule-of-thumb of the IMF proposes that the debt of developing countries should not breach 40% of their GDP. The analysis of Figure 1 shows that the moral duty to rescue imposes a financial liability which may be considerably higher than this threshold, thereby justifying the intervention of the international community.

As shown in Figure 2 for the case of Malawi, the yearly cost of providing ART to all those who need it is expected to rise in the short to medium term as (1) more people get infected; (2) more HIV+ individuals need treatment; and (3) more HIV+ individuals need costly second-line therapies. These estimates are a lower bound for what government will have to pay: the ageing HIV+ population on treatment faces an increased risk for many non-communicable diseases (NCDs) compared to an age-matched uninfected population, including cardiovascular disease, non-AIDS malignancies, liver and kidney disease, and osteoporosis.

Slide1Figure 1: Discounted total cost of full ART coverage (350 cells/mm3) and external debt stock in 8 countries

The crucial role of prevention

Once these future obligations are recognized, it is sensible for governments to minimize them. In this context, prevention can be understood as an investment to reduce quasi-liabilities resulting from the moral duty to rescue. In particular, interventions that prevent a person becoming HIV positive are worth paying as long as their cost is less than the cost of the obligation. Money should be devoted to prevention up to the point at which the marginal dollar spent on prevention reduces the cost of the moral duty to rescue, as measured by the discounted cost of treatment, by one dollar.

While this theoretical rule provides a simple way of determining the optimal level of investment in prevention, estimating the two sides of this equation in practice is likely to be an empirical challenge given the complexity of data required. Context-specific data on the epidemiology of HIV and on the marginal cost of treatment and prevention are much needed to refine this analysis and be able to orient policy.

Slide2Figure 2: Long-run evolution of the cost of ART provision as a % of GDP in Malawi, for different GDP growth scenarios.

How to share the burden?

Even with optimal investment in prevention, the financial liability is likely to be unaffordable for poor and high-prevalence countries. This is a clear case for the international community to take over part of the quasi-liability, as indeed it has done in the past.

Exactly how the burden should be shared between the country and the international community is a normative issue beyond the scope of economics. However, there are some parameters that can assist in reaching a defensible position.The Arusha Target for health investment in Africa can be seen as providing a basis for determining how much countries should pay domestically for health care. The burden of HIV in the total burden of diseases then provides the amount that countries could reasonably allocate to their HIV epidemics. The debt relief provided to poor countries during Jubilee 2000 provides another basis for determining whether the quasi-debt burden of HIV is too large to bear. Econometrics can also be used to unveil past preferences of donors and thereby determines how the financial burden should be shared.

Once a sharing rule has been determined between the international community and recipient countries, appropriate incentives should be put in place in order to foster adequate investment in prevention. Two important principles should guide how financing responsibilities for past infections, future infections and prevention are shared. The first principle states that the international community should finance the cost of past infections to minimize the value of its own contribution. Indeed, because the discout rate of recipient countries is expected to be higher that the discount rate of donors, financial support for HIV for past infections will be perceived as more valuable by recipients than financial support for prevention and treatment of future infections.

The second principles states that if one party accepts to finance a share σ of the liability created by the treatment of future infection, it should also support at least a share σ of prevention efforts. Only this rule can avoid moral hazard and provides adequate incentives to invest in prevention.


[2] This blog post is based on the paper: Collier, P., Sterck, O., & Manning, R. (2015). The Moral and Fiscal Implications of Anti-Retroviral Therapies for HIV in Africa. CSAE working paper No. WPS/2015-05.

This blog post is a summary of the CSAE working paper 2015-05, Paul Collier, Olivier Sterck, Richard Manning (2015). “The Moral and Fiscal Implications of Anti-Retroviral Therapies for HIV in Africa“.

This paper was written as part of RethinkHIV. RethinkHIV is a consortium of senior researchers, funded by the RUSH Foundation, who evaluate new evidence related to the costs, benefits, effects, fiscal implications and developmental impacts of HIV interventions in sub-Saharan Africa. Other members of the consortium have made valuable contributions and suggestions on the paper, and we specifically acknowledge Rifat Atun, Angela Chang, Tim Hallett, Judith Kabajulizi, Mthuli Ncube, Osondu Ogbuoji, Michelle Remme, Sachin Silva, Mariana Siapka, Mikaela Smit, Anna Vassall, Charlotte Watts and Alan Whiteside. We also thank David Miller and Markus Haacker for useful comments and discussions.


Posted in Policies to Protect the Poor | Tagged , , , , , | 1 Comment

The Burundi crisis: Local Grievances, Ethnicity, and the Economy.

President of Burundi Pierre Nkurunziza visits South Africa, 3-4 Nov 2014 (photo credit: GovernmentZA).

President of Burundi Pierre Nkurunziza visits South Africa, 3-4 Nov 2014 (photo credit: GovernmentZA).

When president Nkurunziza announced his intention to run for a third term, serious unrest exploded in the streets of Bujumbura. In the last three weeks, violent clashes with the police have triggered fears of a new civil war and destabilisation of the entire African Great Lakes region. So far, the international community has been unable to calm the situation and over 110,000 Burundians have already fled the country. The Oxford Central Africa Forum hosted a round-table on the 15th of May to examine the causes of the current situation and discuss the prospects for peace and democracy in Burundi and the region. It gathered six researchers who have conducted fieldwork in Burundi:

  • Professor Patricia Daley, School of Geography and the Environment [PD]
  • Dr Andrea Purdekova, African Studies Centre [AP]
  • Dr Olivier Sterck, Centre for the Study of African Economies [OS]
  • Rowan Popplewell, School of Geography and the Environment [RP]
  • Jean-Benoît Falisse, Department of International Development [JBF]
  • Benjamin Chemouni, Department of International Development, London School of Economics [BC]

This blog post presents some of their analysis and findings. This blog post is organized in three section. First, we explore the forces present on the ground. In the second part, we discuss the role of local grievances, ethnicity, and the economy in the conflict. In the last part, we discuss regional implications of the crisis.

The Burundi crisis: (local) forces in presence

The power balance between the forces present on the ground is obviously a key determinant in the current crisis. In this first section, the panellists sought to better understand some of the key forces in current crisis and address, among others, the following questions: Who are the protesters? What is the political opposition doing and how is it different from the civil society? What is the role of the army and the police?

So far most of the protests have been concentrated in Bujumbura, with a few incidents reported in the Mugamba region and in Burundi’s second city, Gitega. The protesters mostly come from the city but they have included both Hutu and Tutsi, a lot of young people and students and also, on some days, women and middle-aged people. Virtually all the neighbourhoods (communes) of Bujumbura have been affected, with particularly explosive situations in Musaga, Cibitoke, and Nyakabiga. The situation in the countryside is harder to analyse, especially since a prime source of information, the independent media, has been shut down by the government and their facilities were destroyed during the coup attempt. Historically, Pierre Nkurunziza and the CNDD-FDD have had a strong support in the countryside, and since the beginning of the protests the government has maintained that “99% of the country is peaceful”. Yet, the 110,000 Burundians who have fled the country are mostly not from Bujumbura, which seems to indicate that the situation is also very tense, and potentially violent, in the countryside. Efforts to map the current violence show a few incidents in the countryside as well. [JBF]

The political landscape is fragmented. The CNDD-FDD (National Council for the Defence of Democracy/Forced for the Defence of Democracy) historically had most of its votes from rural areas, as shows the map of the results of the 2010 municipal elections. The support for Pierre Nkurunziza and the CNDD-FDD was overwhelming in most of the country. There was also, during these elections, a strong and clear link between support for CNDD-FDD and the presence of demobilized ex-combatants from CNDD-FDD (correlation is 0.6). [OS]

(1) Demobilized ex-rebels returns

(2) Results of the 2010 municipal elections













Contrary to the CNDD-FDD, the support for opposition parties (FNL, FRODEBU, UPRONA and MSD) in 2010 was concentrated in Bujumbura Mairie and in the South West of the country. During the last elections, many opposition parties boycotted the presidential and parliamentary election, denouncing massive electoral fraud during the municipal elections and claiming that municipal elections were not free and fair and the next polls unlikely to be so. This strategy has proved disastrous for them as the CNDD-FDD consolidated its control over all state institutions. Pierre Nkurunziza has been travelling very frequently to rural areas, and seems may have preserved his advantage there, where he used to be very popular. Over the past ten years, the CNDD-FDD has created an impressive network of socio-economic associations, party buildings and “office hours” (permanences) in the countryside.

The opposition parties face several barriers that may prevent them to gain the support they need to become a more visible force of the Burundi society, especially in the countryside. Firstly, they lack funding -and the ruling party has secured the control of key resources. Secondly, they are the victims of alleged repression in the form of arbitrary detention and violence, as evidenced by recent UN and NGO reports. Thirdly, recent laws limit the possibilities of the opposition. For instance, a 2012 law determines fines and prison sentences against opposition leaders guilty of “falsehood or slander with the intention to disturb peace and security”. Another law also prohibits the formation of coalition outside periods of electoral campaigning, and yet another law states that “organizations should notice four days in advance to local authorities if they plan a demonstration of a party meeting. These can be cancelled or forbidden unilaterally if there is a risk to disturb public order”. In sum, it seems complicated to see a strong opposition emerge in such a fragmented and repressive landscape. [OS]

The CNDD-FDD used to be a complex political animal, with its different internal factions sharing and disputing power. The recent stance of senior party members against the third mandate has led to the exclusion of most dissident voices and the Nkurunziza camp establishing its dominance over the party apparatus. The practice of exclusion of dissident voices is not new for CNDD-FDD. After the 2005 election already, the fractious CNDD/FDD expelled party members who might have posed a challenge to the presidency. The party split following the arrest and imprisonment of party leader, Hassan Radjabu – a key strategist in the civil war – divided members, leading to key figures, such as Alice Nzomukunda and Pascaline Kampayano deserting the party and forming new offshoot parties. In 2007, Radjabu was sentenced to 13 years imprisonment for treason, but ‘escaped’ from prison in March 2015. How much people will differentiate the party from Mr. Nkurunziza remains a question. [JBF & PD]

The civil society has been at the forefront of protests, which emerged against a backdrop of rising tensions with the Burundian Government. They helped to bring people onto the streets, successfully tapping into widespread public anger over recent events, as well as long-term frustrations generated by sustained political and economic marginalisation and ongoing human rights violations. The Government responded by temporarily detaining one civil society leader and issuing arrest warrants for other activists. Although leaders of the civil society campaign against the third mandate publically condemned the coup, they called for protests to continue. However it is important to note that only a small proportion of civil society groups are involved in the protests; namely those groups that are based in Bujumbura who are focused on governance and human rights issues and regard themselves as watchdog responsible for holding the government to account. In a country where political opposition is weak and fragmented, these civil society groups are often regarded as the de-facto political opposition. Their involvement in recent protests, alongside members of opposition parties and their youth militias, will do little to challenge this perception among Government supporters. [RP]

A key factor affecting whether the country descends into more violence is the role in the current violence of the Police, the ruling party’s youth militia (imbonerakure) and the SNR (National Intelligence Service), as well as that of the more equally ethnically divided military. The Police were the instrument of repression before and after the 2010 elections and seemed to be playing the same role in the contemporary period. The SNR was also used repressively, and has also been used to root out opposition to the President among Hutus in Burundi and in the diaspora in neighbouring countries. The fact that Nkurunziza’s office was able to denounce the coup whilst he was still in Tanzania and stated that security forces were looking for the perpetrators, suggests that the SNR is far more resourced and effective than in the past. This is partly due to the training they would have received as part of Burundi’s contribution to international peace-keeping missions. [PD]

During the 2010 elections, the imbonerakure were already active in the intimidation of opposition candidates and their supporters. Since the ville morte days of the mid-1990s, Burundi’s youth have been mobilized by political parties to perpetrate political violence. Reports suggest that the imbonerakure are armed and the flight of over 110,000 Burundians to neighbouring countries is partly a consequence of the intimidation being carried out by the youth militia. Fifty per cent of Burundi’s population is under 17 years. Rates of youth unemployment are high at around 80 per cent. Consequently, unemployed and disaffected youth are drawn into complex patronage relationship with political officeholders and are easily mobilized into youth wings of political parties. [PD]

The military is well trained and better armed due to Burundi having the highest military expenditure of any country in the East African region and support from the US and Belgium; over 2.39% of Burundi’s 2012 Gross Domestic Product. In effect, the military is more professional and its level of ethnic integration (50/50 Hutu/Tutsi) means that it is likely to take a less divisive stance, unless non-sectarian senior officers are removed. Following the Arusha agreement and the demobilization, the army has been heavily transformed. Its composition is more or less balanced in terms of belonging to former rebel groups. Efforts have been made to transform the army into a professional corps which, according to the Arusha agreement, should remain neutral: “the members of the new force are prohibited from taking part in any political activities or demonstrations”. These transformations have been successful in improving the image of the army. In comparison, the police and the intelligence service are perceived as untrustworthy institutions controlled by the CNDD-FDD. [PD & OS]

Local Grievances, Ethnicity, and the Economy.

The question of the constitutionality of a third term is absolutely central in the present crisis. The protesters have repeated their view that the constitution does not allow for a third term and expressed their fear of a carving up of the 2000 Arusha Peace agreement. The issues around the third term and Arusha have been widely discussed in the media and other essays. The roundtable tried to take a slightly different viewpoint and explored the key issues of the present crisis beyond the question of the third term: local grievances and post-war reconciliation and integration, the economy (and especially agriculture and aid), and whether the “ethnic factor” is still relevant.

Even well before the election, there has been a palpable discontent with the government on a number of issues. There is no question that the Arusha peace agreement brought in a profound political change. The power sharing agreement changed the historical ethnic composition of government and ended the exclusion of Hutu from politics. Despite this landmark change in representation and the composition of government, local narratives are profoundly about continuity rather than change, taking issues with the very notion of ‘transition’. Fieldwork conducted in 2013 found that people even question the very notion of ‘peace’, repeatedly highlighting that Burundi has only ‘a little bit of peace’ or no peace at all. People’s notion of security expands beyond physical security, to material security and survival, and hence to economic concerns, prime amongst which is access to land. The story of continuity also ties to perceptions of impunity – no one in Burundi has thus far been prosecuted for political murder-, to lack of progress on questions of justice and memory, lack of trust in the newly established Truth and Reconciliation Commission (which interestingly only appeared on the scene months before the election, in December 2014), and perceptions of corruption, abuse of office, the sense of being ‘used’ by those in authority (people speak of being their “bridge” to other objectives). [AP]

The land issue is an explosive one, but again does not affect all parts of Burundi equally. Bururi province is particularly affected (and the soon to be province of Rumonge) as well as Makamba. These are the areas that produced the majority of refugees in 1972 and that in turn saw the largest inflow of returnees after the war. Most came back only to find their lands occupied. Instead of the promised recovery of their lands, many had to strike land-sharing agreements as an interim measure. The size of this problem is far from negligible (around half a million people returned after the war.) and it is also an open and unresolved issue. The CNTB (The National Land Commission), established in 2006, the very institution meant to resolve the land conflict, has recently fuelled it further. Under the controversial leadership of Serapion Bambonanire, many returnees were awarded the land in full, effectively pushing the “residents” out without compensation. This has resulted in stand-offs and fights on a number of hills, and in the deterioration in community relations at an already tense time (Since then the CNTB has been suspended, and Bambonanire sacked). [AP]

Burundi’s population overly relies on agriculture, which employs around 90% of the population. The Burundian performance in agricultural production since the end of the war is dismal. Food production barely keeps up with population growth. As a result, food prices are constantly increasing, feeding the general inflation that has regularly exceeded ten percent since the end of the war. The second consequence is that child malnutrition is the third highest in the world. Burundi has also the highest score on the Global Hunger Index. The roots of this problem are mainly political. Comparing to neighbouring Rwanda, whose food production is in a much better state, Burundi has more conducive ecological conditions for agriculture, such as vast plains. Yet, corruption and patronage undermine efforts to boost production. For instance, the sector of improved seeds and mineral fertilizers have been important sites of rent capture for individuals close to the ruling party. [BC]

The rest of the Burundian economy struggles to develop and few industries have been built since 2005. Burundi is lagging far behind its neighbour Rwanda, which has a similar size, population, and natural resources endowments (very few minerals, but tea and coffee). The apparently decent 4-5% GDP growth rate of Burundi in the past few years is unlikely to help rapidly alleviate poverty when once the very fast population growth rate is added to the equation. In 2013, international aid amounted to 20% of the GDP and 49% of budget (World Bank). Rampant corruption at all levels has been denounced by the protesters and is visible in ostentatious villas and cars in Bujumbura. Burundi ranked 159 out of 175 in the 2014 Transparency International Corruption Perception Index. A recent Gallup survey found that 38% of the Burundian population is “suffering”, and another 58% is “struggling”. Between 2009 and 2014, the proportion of people who find that the economy is getting worse has increased from 51% to 72%. [JBF]


The current crisis is worsening the economic situation and seriously endangers the ambitious –and often popular- social programmes launched by the president in the past ten years such as the exemption of health-care user fees from pregnant mothers and children below five, free schools, a reform of primary school, and a new health insurance scheme. Key donors (Belgium, EU, USA, the Netherlands, Germany) have already frozen aid programmes and threatened to pull out totally if the situation continues to deteriorate. [JBF]

Ethnicity failed to rear its head in the 2010 elections, with hardly any political opponents using the discourse of ethnicity during the campaign. The exception was Pasteur Habimana (from a small party – FNL-Iragi rya Gahutu) who, whilst campaigning in Rumonge Commune, Bururi Province, called on Hutus to ‘stop Tutsis from ever returning to power’. He was, apparently, condemned by the main Hutu-dominated parties. In popular discourse extremist Hutus hold the view that Tutsis still retain power, despite Hutu dominance in government. So far, the conflict is not framed in ethnic terms, for two main reasons:

Firstly, ethnic divisions do not necessarily overlap with political divisions, which are at the epicentre of the current crisis. The leader of the failed May 2015 coup, Major-General Godefroid Niyombare, is a Hutu and a former member of Nkurunziza’s rebel group and CNDD/FDD. He and other prominent party members were expelled by Nkurunziza from the party in February 2015 for criticizing his bid to run for a third term (Niyombare seems to have escaped capture but his Tutsi collaborators have been arrested). Yet, political in-fighting within CNDD/FDD and its contestation with FNL (the National Liberation Front) suggest that ethnicity is not the fault line that it was in the past. The bulk of political violence against other parties since the end of the war was towards the FNL, a Hutu-dominated party (and more recently against MSD). It is useful to note that historically (and more recently, by virtue of the constitution), Burundi’s political parties have always been multi-ethnic, drawing on regional alliances. This did not prevent most of them from operating in the interest of the ethnic group that dominated the party. For example, the longest ruling party in Burundi’s history, the Union for National Progress (UPRONA), was always seen as a Tutsi-dominated party with Hutu members. In the same vein, one could argue that the ruling party, CNDD-FDD and its main opposition FNL are predominantly Hutu parties with representations from other groups. Cross-ethnic alliances are more meaningful on the ground and, therefore, stronger among the poor urban and rural communities.

Secondly, the fact that the ruling CNDD-FDD is a Hutu-dominated party went a long way in fighting the ethnic prejudice of the eternal Tutsi exploiter. It has showed the Hutu majority of the population that Hutu can also exploit Hutu. Yet, it is not to say that ethnic conflict could not erupt if violence goes on. The Tutsi, by the simple fact of being a minority, may bear the brunt of violence as a group, following the Swahili proverb that “when elephants fight the grass gets hurt”. Tutsi could become an easy scapegoat – part of them are indeed opposed to the president and Tutsi are known to be very present in Bujumbura. What is worrying about the recent political violence in Bujumbura is that a very visible part of it has been occurring in areas of the city that used to be predominantly Tutsi, in the neighbourhoods of Ngagara, Cibitoke, Musaga, Nyakabiga and Mutakura – and that this might be used by unscrupulous politicians as evidence of an ethnicized opposition to Nkurunziza’s regime. [PD & BC]

Regional Implications

The important attention the Burundi crisis has received from the international media shows not only concerns with the safety and future of millions of Burundians but also fears of a regional contagion. Among the key questions are the regional support of president Nkurunziza, the likelihood of a foreign intervention, the lessons for the power-sharing model of peace-building Burundi incarnated, and possibilities for regional and national actors to solve the crisis.

Nkuruniziza would not have been able to return to Burundi after the May 13th attempted coup without some support of the East African states and the international community. For them, stability in Burundi is paramount for three reasons. Firstly, because of Burundi’s role in the AMISOM (African Union Mission in Somalia) – providing 5n542 troops to the peace-keeping force in Somalia. Civil war in Burundi would lead to the immediate de-legitimization of the peace-force. Since 2005, the deaths of Burundians, however small the number, have been tolerated because of the grander project – the war against Al Shabaab. Secondly, the East African political elite were seeking to strengthen economic links across the region with Bujumbura as an important hub for resource flows. And finally, heads of states in the region are also seeking third terms and Nkurunziza’ success will legitimize and enforce their bid among their electorates. Burundi is a first out of a string of key elections in the Great Lakes Region. 2016 is election year in the DRC and in 2017 eyes will be pinned on Rwanda. The Burundi elections will serve as litmus tests of democratisation (are incumbents willing to give up power?) and the depth of peace (can elections proceed without instability and violence?). Burundi’s fragile peace is at stake, and that of the region more broadly. [PD & AP]

With regard to Burundi’s neighbour Rwanda more specifically, there are two key questions that arise: What does the 2015 election in Burundi mean with regard to Kagame’s potential bid for a third term in 2017? Will Rwanda seek to intervene in Burundi if things escalate further? On the first point, it is interesting to observe Kagame’s rhetoric on Burundi, and the role of the framing of the crisis that he deploys. “If your own citizens tell you we don’t want you,” he suggested recently, “how do you say I am staying whether you want me or not?” Nonetheless Kagame repeatedly underlines that the crisis is not really about the third term, it is about “delivery” (referencing Nkurunziza’s delivery to his people). In this manner, he is effectively sidelining or subsuming constitutionality to performance. Importantly, Rwanda’s political space is more tightly controlled than that in Burundi and it is very unlikely that we will see weeks of protest action in the streets. Opposition will be less visible, and internal pressure thus less pronounced. [AP]

Rwanda is certainly watching the situation in Burundi closely, and not only from the outside. There is talk of Rwandan presence in the form of spies and most recently, Antoine Masozera, the manager of Econet, was given 48 hours to leave the country on accusation of espionage. Will Rwanda seek to intervene militarily? While this is not impending, the official discourse in Rwanda is keeping that possibility open. Rwanda has already expressed concerns that the FDLR might take the advantage of the current situation in Burundi and destabilize Rwanda, which could help legitimise a pre-emptive intervention. There has been talk also of the FDLR joining forces with the Imbonerakure; an allegation that the Burundi government has officially denied. Another issue that could open the door to Rwandan intervention is the perceived threat of ethnic violence. The Rwandans have expressed fears that the Imbonerakure are receiving training directly from the FDLR, are being ‘brainwashed’ with genocidal ideology, and that anti-Tutsi leaflets are being circulated in the North of the country. Rwanda’s Minister of Foreign Affairs Louise Mushikiwabo said quite openly that “while we respect Burundi’s sovereignty in addressing internal matters, Rwanda considers the safety of innocent populations as a regional and international responsibility.” [AP]

It is now incumbent on the African Union and the international community to persuade or put pressure on Nkurunziza not to take revenge against the demonstrators or coup plotters (both are from different constituencies), and to respect the Arusha agreement by stepping down, postponing the elections, and allowing democratic process to function. A peace-keeping force made up of West Africans would provide the space for mediation to take place between the various political actors in Burundi. It would be tragic if the region and the international community were to stand by and watch Burundians killed as they demonstrate for their democratic rights, the country disintegrates into civil war, and ethnicity used instrumentally by unscrupulous politicians. [PD]

Burundi is a unique case on which to judge what power-sharing mechanisms can and cannot achieve in what Arend Lijphart termed as ‘deeply divided’ societies. After the civil war, Burundi implemented a complex scheme of consociational power-sharing, carefully instituting ethnic proportionality at all levels of the government. But what have been the effects of power-sharing with regards to salience of ethnicity, democratization and stability? To begin with, and despite what some critics would have predicted, the saliency of ethnicity in politics has diminished. On the other hand, power-sharing has prevented neither a gradual democratic backsliding with every progressive election, the current one notwithstanding, nor insecurity and destabilization. As Stef Vandeginste has previously argued, power-sharing in Burundi can be considered at most a ‘fragile safety valve,’ to some extent limiting but ultimately not preventing the consolidation of power in the hands of the dominant party. Power-sharing thus cannot guarantee, in itself, sustainable democratisation or long-term stability. The latter issue is closely related to the former, as the current crisis clearly demonstrates. [AP]

Despite the successful ethnic re-fashioning of power in Burundi in the past 15 years, a deeper change is necessary, one which includes but goes beyond free and fair political competition. What is needed is a more credible break with past styles of power and ways of relating to the society at the local level, a government able to directly address past injustice and core grievances today. [AP]

Posted in Uncategorized | Tagged , , , , , , , | Comments Off on The Burundi crisis: Local Grievances, Ethnicity, and the Economy.

Poverty as a multi-dimensional and intertemporal phenomenon

Over the last few years, two major developments in the way poverty is conceptualised and measured stand out. These regard, respectively, poverty’s multi-dimensionality and its dynamic nature over time. Developing measurement techniques which appropriately capture these important aspects of poverty has been a very active area of research lately. The vast majority of contributions though have focused on capturing either the multi-dimensional aspect, or the temporal aspect, and have not attempted to deal with both simultaneously.

In a recent CSAE working paper co-authored with Jing You and Sangui Wang, we apply the intertemporal poverty indicator recently developed by Dutta et al. (2013) to two important dimensions – income and nutrition – and propose a generalised recursive selection model to investigate the determinants of households’ multi-dimensional and intertemporal wellbeing in two less developed provinces in northwest China, namely Gansu and Inner Mongolia.

China has experienced an economic miracle of a sustained 8-10% annual growth rate in GDP per capita, and this has enabled it to reduce its poverty rate from 84.3% in 1981 to 6.3% in 2011 (based on the US $1.25-a-day (PPP) line). In sum, 753 million people in China have escaped poverty in this period, and this comprises three quarters of the total number of people in the entire East Asia & Pacific region who were lifted out of poverty during this time. Nevertheless, in 2011, 85 million people in China remained poor. Two-thirds of these people live in western areas (Poverty Monitoring Report, 2011).

Our dataset was collected by the National Bureau of Statistics (NBS) local branches, and covers 1,500 households in 150 villages from 15 counties. Of these, 700 households from 7 counties are in Gansu and the remaining 800 households are from 8 counties in Inner Mongolia. Gansu is one of the poorest provinces in China; its real per capita GDP has consistently ranked in the bottom 3 out of 31 provinces for two decades (1990-2010). By contrast, Inner Mongolia was ranked in the middle of these provinces (16th-20th) throughout this period. The latter is characterised by rich natural resources, typically coal mines, animal husbandry and related processing industries such as meat, dairy and cashmere.

The trend in aggregate income poverty incidence in these two provinces, denoted by the red lines in Figure 1, is found to be broadly decreasing over the period, notwithstanding a 4 percentage point increase between 2000 and 2001 due to a severe drought in Inner Mongolia. Strikingly, however, the reduction in income poverty was not accompanied by an improvement in nutrition – a vital dimension of human well-being. Figure 1 indicates a clear trend of increasing under-nutrition under a threshold of 2,100 kcal per person per day. The nutritional poverty rate (gap) increased by 12% (14.4%) from 2000 to 2004, while household equivalent per capita net income grew by 31% concurrently. On average, 70% of households in the sample lived on less than 2,100 kcal per person per day.

Figure 1 Poverty rates in Gansu and Inner Mongolia


Note: The three (red) lines at the bottom are income poverty rates. The two (blue) lines at the top are nutritional poverty rates.

In this study we estimate Dutta et al.’s (2013) intertemporal poverty measure for income and nutrition, respectively. An individual’s poverty status over the entire five-year period is aggregated in a way that accounts for:

(1) Previous hardship. Specifically, poverty in a given period is treated as being exacerbated if it was also preceded by a poor spell. This reflects the fact that prolonged spells in poverty can be particularly detrimental, for a wide variety of reasons including physical, psychological and social factors.

(2) Previous affluence. Non-poor spells can mitigate the deprivation experienced in subsequent poor spells through a variety of non-income channels such as assets, health, social networks, human capital and so on.

In the nutritional dimension, we find that a major problem is a persistent shortage of nutrition, at a household level, in remote rural areas: 81% of those without sufficient nutrition in a given period continue to suffer in this way in the subsequent period. Overall, our intertemporal nutritional analysis paints a much more pessimistic picture than the widely reported (and, nonetheless, clearly extremely welcome) statistics on the greatly reduced numbers of people in income poverty.

The aids and subsidies in rural China, as in many other developing countries, are targeted to the poor, as defined by the national income poverty line. Those “not-so-poor” who live barely above this line are excluded from these projects, yet such people are still vulnerable to poverty and various risks and shocks, and this seems to be reflected in their poor nutritional status in our sample.

To the extent that intertemporal poverty is the focus of policy makers, our analysis suggests that household-focused interventions generally outperform village-level instruments. In the early stages of development, and when agriculture is a dominant element of individuals’ livelihood, improvement in agricultural production, both in terms of increased agricultural labour productivity and of larger farm size and investment in productive assets, still holds the key to reducing both intertemporal income and nutritional poverty. This is true whether focusing on the poorest of the poor, or when the richer poor are also included, by adopting higher poverty lines. The finding is also robust to different normative assumptions underlying intertemporal poverty aggregation over time. Furthermore, when there is evidence suggestive of intertemporal poverty-nutrition traps, higher labour productivity in agriculture in the long-term holds more potential for breaking the vicious circle, in both income and nutritional dimensions, than does local non-agricultural production or out-migration.

Our results suggest that the purported positive roles of local non-agricultural production and out-migration in reducing monetary poverty might be overblown, at least as an intertemporal phenomenon. They are certainly not magic bullets. Rather, whatever effect they have on both income and nutritional dimensions is likely to be highly context specific and dependent on other factors.

Dissemination of agricultural technologies and better living conditions also help reduce intertemporal poverty to some extent, but only in the short-term. Moreover, their impact is highly sensitive to which particular dimension of deprivation policymakers aim to tackle, and which specific technology or mechanisation households adopt. Such instruments need to be implemented with caution and an awareness of contextual specifics, as our analysis suggests that they can have unintended negative consequences, both for household income and nutrition.


Table 1 Intertemporal Poverty Profiles

Group Income Nutrition
Full sample 0.063 0.521
Gansu 0.067 0.318
Inner Mongolia 0.058 0.698
Han 0.121 0.759
Minorities 0.058 0.500
Labour allocation
Agriculture 0.069 0.513
Local non-agriculture 0.040 0.544
Circular migration 0.090 0.510



Posted in Uncategorized | Tagged , , , , , | 1 Comment

Affecting fertility behaviour in Zambia

Giving women control over their fertility is at the forefront of the agenda for improving the well-being of women in developing countries. This issue was explored in the Plenary Lecture at the 2015 CSAE conference, given by Nava Ashraf, an Assistant Professor at Harvard Business School.

There is growing evidence that households do not behave as one unit, in contrast to what earlier literature would suggest (e.g. Becker 1981). Instead, individuals in households appear to have different preferences, so that household-level decisions such as the sharing of resources and fertility involve some kind of weighting of these preferences.

Surveys have shown that women often want fewer children than men in developing countries (figure 1); this suggests preferences over the number of children differ. As long as women’s outside options are low, they have little power in implementing their own preferences within the household. In Zambia, for example, around 41% of births are unwanted at the time of conception (Zambia Demographic and Health Survey, 2007).

figure 1Figure 1: Gender gap in mean ideal number of children (Westoff, 2010)

Ashraf, Field and Lee (2014) implemented an experiment in Zambia that gave some women access to concealable contraception on their own, while other women were given contraception together with their husbands. They find that women given access with their husbands were 25% less likely to use the contraception and 27% more likely to give birth. This supports the idea that in these households, spouses do not agree over family planning and that women have a stronger preference for contraception than men. However, they also find that women given access to contraception on their own reported a lower level of happiness after the study, which the authors interpret as suggesting that hiding contraception also bears a psychosocial cost. Nevertheless, it is difficult to be conclusive about the mechanism behind this. One possibility is that because the study affected childbearing, this in turn may have affected subsequent happiness, either directly or because women in the couples-treatment had more children, which is known to be correlated with a lower probability of divorce.

Of course, the real question is why men prefer to have more children than women, and one hypothesis is that men are not as well informed as women about the health costs of pregnancy. For example, husbands whose wife had a previous difficult labour are better informed about the risks of pregnancy. Ashraf goes on to discuss work in progress that attempts to educate men on these risks. This is certainly an important topic and the results of this study will be of interest to policymakers.



Ashraf, Nava, Erica Field and Jean Lee. “Household Bargaining and Excess Fertility: An Experimental Study in Zambia.” (2014)

Becker, Gary. A Treatise on the Family. Cambridge: Harvard University Press (1981).

Westoff, Charles F.,”Desired Number of Children: 2000-2008,” DHS Comparative Reports, 2010, (25).

Posted in Uncategorized | Tagged , , , , , , | 1 Comment

Migration and labour mobility

The third plenary, and final session, of the CSAE Conference 2015 featured three presentations on different aspects of migration, and its relationship with the broader theme of economic development.

Michael Clemens (CGD): Skilled migration and development

Michael Clemens of the Centre for Global Development started with a presentation that questioned the standard narrative of skilled migration leading to adverse consequences for the origin country of the migrants. Much of the existing research assumes that skilled migration leads to a fall in the human capital of the home country, but recently collected data provides little evidence to support this claim (see figure).

figureFigure – The yellow line is the predicted relationship between human capital within the country (y axis) and the extent of skilled migration or human capital outside the country (x axis), based on the assumption that one skilled migrant from a country leads to a proportional decline in the number of skilled workers available within the country. The blue arrows show movements of individual countries between 1990 and 2000, which suggests that few countries have followed the predicted pattern.

Clemens suggested a number of fruitful areas for future economic research that extends the standard production function by identifying alternative channels through which migration affects development. His own research based on a natural experiment in Fiji indicates that anticipated participation in the global labour market could lead to an increase in the demand for tertiary education and a higher accumulation of skills at home. Other channels through which skilled migrants can promote economic development is the transfer of technology and capital back to their home countries, the facilitation of increased trade between home and destination countries, and, interestingly, the promotion of certain institutions, such as democracy and freedom of expression. Finally, one the most important questions of the development research agenda – what Clemens terms the ‘’biggest game’’ – is the identification of human capital externalities, which are yet to be empirically documented. If human capital externalities are limited, then the effect of skilled migration on both home and destination countries might be much lower than previously predicted.

Melanie Morten (Stanford University): Internal migration in sub-Saharan Africa

International cross-border migration may be the most visible form of migration, but in terms of numbers it is dominated by internal migration within the borders of a country. Melanie Morten of Stanford University presented a number of stylized facts on internal migration based on data from 18 censuses in sub-Saharan African countries. First, migration rates are high: as much as 30% of the population migrates from its region of birth, on average. Second, while households in the agricultural sector are less likely to migrate than not, a significant number – around 40% on average – do choose to migrate. Third, migrants are able to find employment, at least at the same rates as the non-migrant population. The only exception is South Africa where migrants are much more likely to be employed than non-migrants. Fourth, and most strikingly, in spite of high migration rates, large wage gaps continue to persist across regions.

Understanding why, in the presence of large wage gaps, people do not migrate from low-wage to high-wage areas is critical to determine a suitable policy response to migration. Researchers have not been able to explain the entire wage gap by differences in the skills and preferences of migrants. The explanation most favoured by Morten is that the wage gap reflects the high costs associated with migration, possibly due to the lack of adequate infrastructure. If we believe that there are gains to migration, and that high transport costs may be constraining migration, then the return to building better roads, for example, would be higher than previously estimated. Exploring these implications of migration on policy, says Morten, is an exciting area for future research.

Clement Imbert (University of Oxford): Seasonal migration in India

Clement Imbert of the University of Oxford made the final presentation on seasonal migration. Drawing on his own research in India, he drew a profile of the typical migrant who travels in search of employment in response to the low incomes of the agricultural lean season. Most migrants are male, poor, and small landowners from remote, rural areas, who migrate to well-connected, urban centres. They usually travel by rail and are able to cover large distances of 300-400 kilometres on average due to the low marginal costs of travel by train. They typically work in construction and small-scale manufacturing, and may be recruited at their destination or from their home villages.

Imbert examines the effect of the world’s largest workforce programme on seasonal migration. The National Rural Employment Guarantee Scheme (NREGS) provides employment to workers living in rural areas during the agricultural lean season. Imbert and Papp (2015) find that the scheme crowds out seasonal migration; in fact, workers are willing to stay at home for a relatively modest increase in household income (Rs 60 earned per day under NREGA against the Rs 120 they could earn if they migrated to cities). One reason for this wage gap, according to Imbert, is the perceived risk associated with finding work in urban areas; another could be the very poor and often dangerous living and working conditions encountered by migrants in the cities. The other effect of the NREGS is a rise in wages in urban centres in response to a fall in the supply of migrant workers, leading to higher labour costs for sectors that are heavily dependent on migrant labour. Government policies can, therefore, play an important role in shaping the extent and form of seasonal migration; space-based policies could reduce seasonal migration, while policies targeted at reducing travel costs could encourage migration.

Getting the words right

A lively discussion followed the presentations and perhaps the most thought-provoking insights related to the impact of language on research and policy. In particular, as Clemens emphasised, the use of normative and even pejorative language to describe migrants – such as the popular phrase, “the brain drain” – preempts a conclusion to the analysis of the effects of migration. Imbert, too, pointed out that the description of seasonal migration as “distress migration” might even lead government to formulate counterproductive policies. One hopes that researchers and policymakers follow the lead of our presenters; however, on a subject as politically motivated and emotionally charged as migration, this may well be easier said than done.

Posted in Uncategorized | Tagged , , , , | 1 Comment

The height production function from birth to maturity

Starting in the 1970s, anthropometric measures have increasingly been used in the social sciences as indicators of social well-being. Since then, adult height has been considered an indicator of the general health status in life, of the relative risk of survival, and labor productivity (Fogel, 1986). In particular, Case and Paxson (2008) explain the positive correlation between adult height and labor productivity by showing that height is positively associated with cognitive ability. They show that both cognition and height are driven by early childhood investments and therefore cognitive achievements are correlated with height, and wages are affected by cognitive skills. Poor health can explain both low height and low labor productivity. This is more evident in developing countries where living conditions are poor. However, adult height is just the final result of a process of growth that involves many different mechanisms and variables, and most of the common evidence looks at adult height (Steckel, 2009). It is therefore necessary to investigate the factors driving height, since understanding the determinants of height is important in order to understand health (Deaton, 2007). The determinants of height can be divided into non-genetic factors, genetic factors and the age when height is measured. The principal non- genetic factor is net nutrition which is the difference between food intake and the losses to activities and to diseases (Eveleth and Tanner, 1991).

To study the determinants of height from birth to maturity I use in my last CSAE working paper the Cebu Longitudinal Health and Nutrition Survey (CLHNS), which is a rich longitudinal survey of a cohort of Filipino children followed from conception, in 1983-84, to 2005. The distribution of height by age and sex is shown in Figure 1.

Figure 1: Height distribution by age and gender of the child. The dots correspond to the raw height measurements, while the lines correspond to spline interpolation of the mean height per gender and age of the child.

I build and estimate a height production function which depends on age, previous investments in the child, biological endowment and a shock. To study empirically the height production function, I assume (A) that the child’s biological endowment is determined at conception and is constant over time, and (B) that the height production function is linear in the inputs and in the unobserved endowment, and that the effects of the inputs depend on the child’s age. I can then specify a model where the current height is regressed on a function of age (linear and a quadratic age), current and past inputs (nutrition and diseases), a child’s fixed effect and a shock.

The nutrition inputs considered are the caloric intake and breast feeding. I distinguish between diseases experienced during infancy, pre-puberty and puberty because they are age-dependent and might have a different impact on a person’s growth, and moreover the data contain different measures of morbidity that differ between waves. In particular, I consider diarrhea episodes experienced during infancy, length of stay in the hospital during pre-puberty, and hospitalization during puberty (dummy if the child has been hospitalized).

In order to estimate this model, I consider a within-child fixed effects specification (FE) where I estimate the change in height between two consecutive measurements, remaining only with the contemporaneous inputs. However, contemporaneous inputs could respond to previous shocks causing endogeneity because correlated to unobserved parental preferences regarding their children’s nutrition and preventative care. I address endogeneity of both nutrition and diseases by using variation in village-level food prices, village and household characteristics, and climatic shocks as instrumental variables (IV) to estimate the production parameters via IVFE.

The IVFE results show that the diarrhea episodes experienced during infancy, and in particular in the second year of life, have the largest and negative effects on height. If experienced in the first year of life, an increase by one of the diarrhea episodes decreases the height of a boy by 1.376 cm, and the height of a girl by 2.113 cm. Experiencing an extra episode of diarrhea in the second year of life, reduces the height of a boy by 2.214 cm, and the height of a girl by 3.171 cm. Diseases experienced later in life and measured with time spent in the hospital or number of hospitalizations do not seem to affect height. Pre-puberty (age 8 and 11) is dominated by the effects of nutrition. The IVFE estimates show that if caloric intake increases by 100 kcal then height increases in 8 year old boys by 0.199 cm, in 11 year old boys by 0.246 cm, and in 11 year old girls by 0.200 cm. It might be that the pre-puberty years prepare the body for the final phase of growth, and that might explain the relevance of nutrition inputs at age 11 and also age 8 for boys.

Overall, these results show that growth in infancy and in the pre-puberty years turn out to be critical stages in the process of height formation. This is in line with the increasing literature on the long term effects of early life conditions (Almond and Currie, 2011) and with new studies that reveal the importance of later critical periods (van den Berg et al., 2014). Most importantly the paper shows that some critical periods are important because diseases experienced during infancy play a major role compared to nutritional intake. This might be important to design policy interventions that target individuals in these critical periods to improve their health and potentially their socio-economic outcomes later in life.


Elisabetta De Cao – Centre for Health Service Economics & Organisation, University of Oxford

New Radcliffe House (2nd floor), Walton Street, Oxford, United Kingdom (e-mail:;


Almond, G. and J. Currie, 2011. Human capital development before age five. In O. Ashenfelter, R. Layard, and D. Card, editors, Handbook of Labor Economics, volume 4B, chapter 15, pages 1315–1486. Amsterdam, The Netherlands: North- Holland.

Case, A. and C. Paxson, 2008. Stature and status: Height, ability, and labor market outcomes. Journal of Political Economy, 116(3):499–532.

Deaton, A., 2007. Height, health, and development. Proceedings of the National Academy of Sciences, 104(33):13232–13237.

Eveleth, P. and J. M. Tanner, 1991. World-Wide Variation in Human Growth. Cambridge. Cambridge University Press, second edition.

Fogel, R., 1986. Physical growth as a measure of the economic well-being of popula- tions: The eighteenth and nineteenth centuries. In F. Falkner and J. M. Tanner, editors, Human Growth: A Comprehensive Treatise, volume 3, pages 263–281. New York: Plenum Press.

Steckel, R. H., 2009. Heights and human welfare: Recent developments and new directions. Explorations in Economic History, 46(1):1–23.

van den Berg, G. J., P. Lundborg, P. Nystedt, and D. O. Rooth, 2014. Critical periods during childhood and adolescence. Journal of the European Economic Association, 12(6):1521–1557.



Posted in Uncategorized | Tagged , , , , | 1 Comment

Community-Based Development initiatives: Who in the village hears about them and who doesn’t? And how?

Community-based development programmes, known for placing greater control of resources and decision-making in local hands, have long been an important part of development policy. But these programmes are also behest with their own challenges and limitations. One of these is the spread of information: those who might stand to benefit the most from these programmes (such as the elderly and disabled) might be the least likely to hear about them. It is hard to benefit from something you’re unaware of. Community-based development initiatives often rely heavily on village meetings and media campaigns (e.g. radio, television and newspapers) to sensitise communities and to encourage participation. But vulnerable groups such as the elderly and disabled may not have (easy) access to such direct sources of information because of physical or socio-economic constraints.

Village meeting

We tend to learn information from our peers, so it is possible that interactions within a person’s social network might compensate for a lacklustre information campaign. One of my recent CSAE working papers sets off to investigate the extent to which neighbours may influence the awareness of elderly and disabled groups in rural Tanzania. My paper examines this in the context of Tanzania Social Action Fund II (TASAF II), one of Tanzania’s flagship community-based development projects, funded by the World Bank. Launched in May 2005 as an important component within the national framework of the Tanzanian National Strategy for Growth and Reduction of Poverty (popularly known as MKUKUTA), TASAF II followed its predecessor TASAF I (2000–2005) in the aim of empowering local communities to participate in development initiatives. Among the communities explicitly targeted by TASAF II are vulnerable groups comprising orphans, disabled, elderly, widows/widowers and those infected by HIV/AIDS. These vulnerable groups could collectively apply for funding to start up income generating group activities.

Most of the activities implemented by TASAF during the first two years of implementation (2005–2007) were aimed at creating programme awareness, by means of national and regional radio programmes, newspapers, newsletters, television shows, technical launch workshops, village meetings and the distribution of brochures, posters, t-shirts and caps. And yet, as has recently been shown by Baird et al. (2013), in 2008 less than half of TASAF’s eligible non-beneficiary households reported to have ever heard about the program, and that was after the funds had already been allocated.

Baird et al. (2013) identify a household’s direct access to information (such as education and village meeting attendance rate) and their immediate access to local politicians and other decision makers (e.g. being blood related to the village elite) as the dominant determinants of being informed about TASAF. I contribute to this evidence base by considering whether social interactions with neighbours played a role in the information reaching more widely and, if so, whether different households benefitted differently from information in their neighbourhood.

To answer this empirical question I used household data from a listing exercise organised by the World Bank in April–November 2008 collected in the context of an impact evaluation of TASAF II. This unique dataset captures the details of 30,339 households, including all elderly and disabled people (19,916 households), in 100 villages from 5 regions in rural Tanzania. The dataset includes, among other things, global positioning system (GPS) coordinates of each of these households in the dataset and information about their knowledge of TASAF.


Using a two-stage least squares estimation strategy derived from Bramoulle et al. (2009) and De Giorgi et al. (2010), I find that, on average, for each additional informed neighbour in a set of 10 nearest neighbours, the respondent of a household with disabled or elderly members was approximately 8 percentage points more likely to be informed about TASAF. To put this into context, compare this estimated peer effect with the impact of attending village meetings (i.e. the sensitisation medium most heavily relied on by TASAF): having two informed neighbours among one’s ten nearest neighbours had, on average, a similar impact on the probability of being informed as had attending a village meeting (15% and 14%, respectively).

My paper also looks at which types of elderly and disabled households were most and least receptive to information in the neighbourhood, and which households (vulnerable and non-vulnerable) were most effective in transmitting information to others. The results suggest that households with members holding political positions were relatively more effective in the transmission of information to others. Moreover, the economically worse-off households benefitted most from the presence of well-informed neighbours. This is not surprising, as better off households are likely to have better access to primary sources of information, such as radio and newsletters and, therefore, are less therefore likely to rely on peers for information.

Importantly, the following groups turned out to be least sensitive to information in the neighbourhood: the oldest of the elderly members, female-headed households living in a male-dominated neighbourhood, those who do not have blood relatives living in the village, those who are uneducated and/or unemployed and/or those who do not belong to the main religion of the village.

Although we should be careful in generalising these results to other demographic groups and other geographic contexts, this new piece of evidence points out that there are people who are not only restricted in their access to direct sources of information (e.g. village meetings, radios, television, etc.) but who also lack indirect access to information through social networks. If the aim is to maximise inclusion in  community-based development programmes, especially those targeted at the most vulnerable people, it may be desirable to target some demographic groups separately with complementary – specifically designed – sensitisation strategies which take these constraints into account (e.g. through targeted home visits).

Bet Caeyers – Institute for Fiscal Studies (

Posted in Policies to Protect the Poor | Tagged , , , , , | Comments Off on Community-Based Development initiatives: Who in the village hears about them and who doesn’t? And how?

Gender equity in schools in Muslim countries: it can be done

Muslim countries worldwide have problems with gender equality. They dominate the bottom ten countries in the Global Gender Gap Report and none of the ten most successful countries offering equal opportunities for men and women is Muslim.

Girls lag behind boys in school attendance, making up 54% of the out-of-school child population in the Arab states, a figure that has not changed since 2000. Of the ten countries that fare the worst for child school attendance rates, seven are Muslim. These are Nigeria, Pakistan, Ethiopia, Cote d’Ivoire, Burkina Faso, Niger and Yemen: countries that are often considered hotspots for acts of violence against women and school girls.

The near-fatal attack on Malala Yousafzai by a Taliban gunman in Pakistan two years ago reminds us of the challenging circumstances in which girls attend school in many Muslim countries. Her subsequent fight for education for girls and children worldwide won her this year’s Nobel Peace Prize.

Earlier this year, as many as 276 school girls were abducted for similar reasons in the Borno state of Nigeria. Their captors from the extremist Boko Haram group consider secular education a grave threat to Islam. Frequent attacks on schools in Nigeria have forced many parents to withdraw girls from education. In some states schools have even closed down for fear of insurgent attacks. More recently a Boko Haram-style armed group warned schools in Pakistan against co-education. It’s little wonder Nigeria and Pakistan together account for a quarter of the world’s out-of-school children.

Many Muslim nations fare poorly for gender equity, but there are steps they can take and examples they can follow to make sure girls are properly educated. Flickr/Nevil Zaveri, CC BY-SA

Why this inequity in Muslim nations?

In many Muslim countries women are subjected to patriarchal norms and varying degrees of restriction on economic participation. This reduces the value of girls’ education in society.

Some scholars blame culture and religion for this problem. Others say the economic structure of some Muslim countries is not conducive to women’s development. They argue patriarchal norms persist because oil‐rich economies limit the role of women in the paid workforce and restrict women’s participation in politics.

Not all Muslim nations suffer gender inequity in education

Within the Middle East, a region widely considered to lack progress in girls’ education, Turkey is very close to eliminating the gender gap in schooling. In Quwait, UAE, Bahrain and Libya, more girls are in secondary school than boys and there is gender parity in primary enrolment.

Much clearer success stories are emerging outside the Arab world. In Indonesia, the world’s most populous Muslim country, equal numbers of girls and boys are in school. In neighbouring Malaysia, boys even lag behind girls at almost all levels of education.

Muslim nations in South-East Asia have a better record of educating girls. Flickr/Andy Maluche, CC BY

In the Borno state of Nigeria, half of ten-year-old girls remain out of school, a situation Malaysia overcame nearly three decades ago. A similar pattern is visible in Bangladesh where girls outnumber boys in primary as well as secondary school. Neither religious orthodoxy nor income poverty could keep girls away from schools in this Muslim majority nation.

What explains these contrasts within the Muslim world?

The answer lies partly in different policy choices. Even though public spending on education hasn’t increased significantly, the Bangladesh government prioritised women’s development through gender-responsive budgeting and partnered with non-state providers to make up for gaps in public investment.

The government introduced nation-wide scholarship schemes for school-aged girls. NGOs were also encouraged to run development programs targeting girls and women in rural communities. The coordinated push for greater educational opportunities for girls was matched by large-scale deployment of rural women as providers of social services as well as micro-entrepreneurs.

This helped to quickly shift the cultural norm in girls’ social participation. The absence of such a “big push” in Pakistan probably explains why it has fallen behind Bangladesh in terms of girls’ education. The Global Gender Gap index ranks Bangladesh (75) much higher than other Muslim majority countries such as Saudi Arabia (127), Iran (130) and Pakistan (135).

Neither poverty nor patriarchy stands in the way of women’s development in some Muslim countries. Flickr/Guy Nesher, CC BY

Malaysia’s approach to girls’ education was more conventional. Inspired by the success of high-performing East Asian economies such as South Korea, Malaysia prioritised education for all. Long before experiencing double-digit growth rates, Malaysian education spending was much higher than that of other developing countries. According to the World Bank’s World Development Indicators, by 1970 gross female primary school enrolment in Malaysia was nearly three times higher than that of Nigeria.

In Malaysia, democracy, development and equitable growth in early years reduced the demand for extreme views. This may explain why female schooling has never been undermined in Malaysia by religious extremism or violence caused by “Islamic” radical groups like Boko Haram.

What is happening in Nigeria is a reflection of long-term deprivation of basic health and education services. Despite plentiful natural resources, human development was never a priority. The country’s oil rent was not used productively to build schools and boost literacy rates.

The absence of a functional democracy, poor governance and widespread corruption may have fuelled demand for a puritanical Islamic regime and allowed extremists groups like Boko Haram to thrive. Boko Haram has emerged in the poorest part of Nigeria where nearly three-quarters of the population live in abject poverty. Nigeria remains one of the few developing countries where more than 20% of primary school-aged children will remain out of school by 2015.

The success stories of Indonesia and Malaysia highlight the importance of sustained public investment. And Bangladesh proves that if gender equity is prioritised, rapid progress is possible despite limited economic capacity. These three countries show that neither poverty nor patriarchy can stand in the way of women’s development in Muslim societies.

The lesson for the rest of the Muslim world is that schooling gender gaps can be narrowed, as long as there is a political consensus on achieving education for all. A lack of consensus can encourage religious extremism and opposition to inclusive development. This will further limit the prospect of women’s empowerment through education.

* Niaz Asadullah is Professor of Development Economics and Deputy Director of the Centre for Poverty and Development Studies (CPDS) at the University of Malaya. He is a Research Affiliate of the CSAE.

The article was first published on The Conversation.


Posted in Uncategorized | Tagged , , , , , , , | Comments Off on Gender equity in schools in Muslim countries: it can be done