Growth-reducing structural change

Dani Rodrik finds an interesting fact about economic development: Workers in Latin America and Africa have moved to lower labor productivity sectors in the period 1990-2005.

The basic facts about world poverty

In 1981, 52 percent of the world population were poor. By 2005, the number has been reduced to 25 percent. This reduction is mostly due to the achievement in China.

That's the summary of findings by two World Bank economists Shaohua Chen and Martin Ravallion. By "poor", they mean their daily consumption is below 1.25 US dollars, which is the amount of money you need to spend to satisfy the basic energy requirements in the poorest 15 countries.

A theory of the origin of the state

I always wonder what caused the emergence of the state. The answer to this question has an important implication to poverty reduction in developing countries because one of the biggest causes of underdevelopment is state failure (the best example of this is Somalia).

Google search takes me to an interesting theory proposed by an anthropologist named Robert L. Carneiro in his 1970 article. In a nutshell, the state emerges in agricultural land surrounded by mountains, seas, and deserts. In such an area, when all the cultivable land is cultivated, villages start fighting each other to conquer more pieces of land to sustain the population. When the conquerors rule the conquered, there is a need to create the administrative structure. This is the origin of the state.

But this theory was proposed 40 years ago. There must be a refutation to this theory, or some refinement to this theory by now. If I learn those, I will write a sequel to this post.

What is democracy: a case of Egypt

A long time ago, I listed up the requirements for a country to be a democracy often mentioned in social science studies. Here's the list:

A1: Suffrage is universal - (nearly) all adults can vote.
A2: The executive is directly elected or indirectly elected by legislature.
A3: Elective legislature exists.
A4: Opposition parties are allowed to exist.
A5: No gerrymandering (ie. manipulating electoral rules and districts in favour of the ruling party) is undertaken.
A6: Press freedom is ensured.

B1: Opposition candidates can freely run for office.
B2: Opposition candidates can freely conduct their electoral campaign.
B3: Voters can freely cast their ballots.
B4: Votes are neutrally counted.

C1: The executive is checked and balanced by legislature and judiciary.
C2: Non-elective veto players do not intervene politics.

The Economist magazine recently reported how the Egyptian ruling party has held power for a long time.

First, those who would oppose the ruling party find it difficult to register as voters.

Few people register because the legal period for doing so is short and comes many months before elections. Besides, registration involves a visit to a police station, which many Egyptians prefer to avoid.

This can be seen as violation of condition B3 above.

Second, condition B1 is violated:

The parties allowed to run for the People’s Assembly, Egypt’s parliament, are selected by a committee controlled by the ever-ruling National Democratic Party (NDP), which is headed by Mr Mubarak. Independents can stand, which is how the Muslim Brothers, banned as an organisation, field their candidates. But they risk arrest on some pretext, and harassment even without one. (...) Since 2005 there have been elections for the presidency, too, replacing the previous embarrassingly unanimous referendums. The next one of those is scheduled for September 2011, albeit under restrictive rules that, in essence, allow the NDP to choose not only its own candidate but his opponents as well.

Condition B2 is violated, too, if we regard the post-election period as the campaign period for the next election:

In the previous presidential poll, in 2005, nine selected opponents were allowed to run against Mr Mubarak, yet he still grabbed 89% of the vote by the official count. His closest challenger, Ayman Nour, a youthful lawyer, was locked up in prison on flimsy charges of forgery soon afterwards and released only last year.

Condition B3 is also more obviously violated:

In the parliamentary election of November 2005 voting was spread over three rounds because a ruling by the constitutional court required that judges should monitor it closely. The Brotherhood, wary of showing too much strength and so provoking a backlash, fielded candidates for only a third of the 454 seats. In the first round they trounced most rivals. In the second and especially the third round police simply locked many polling stations and sent everyone home.

The last half of the article talks about Mohamed ElBaradei, a potential presidential candidate. The strategy for the government to sabotage his campaign is to punish his supporters:

Kuwait, a country friendly to Egypt’s regime, summarily expelled a group of Egyptian workers who had innocently created a local support group for Mr ElBaradei. In advance of his scheduled visit to the rural province of Fayoum in May, companies that rent equipment for public events received warnings not to work with him or risk having their equipment seized. Private television stations have been “advised” to pay less attention to the upstart, well aware that their owners’ useful government links may be at stake.

These episodes violate condition B2 in a broad sense. But we could treat them as an additional condition for democracy: opposition supporters are not economically threatened by the government.

ABCDE 2010 in Stockholm (Day 1)

The annual World Bank's conference on development economics (abbreviated as ABCDE) is currently held in Stockholm. I attended a couple of sessions and below is what I learned. Some are backed by solid evidence while others are speculation.

From Abhijit Banerjee's speech:

The failure of microfinance to have a large impact on firm creation and firm growth (Banerjee et al. 2009; Karlan and Zinman 2009; a summary by FT) may be due to its targeting of poor people and small-scale firms. It is medium to large scaled firms that are missing in less developed countries (LDCs) if the firm size distribution is compared to the one from developed countries (Hsieh and Klenow 2009), and that type of firms is probably the driver of economic growth. Governments in LDCs and development assistance agencies, however, have ignored these medium and large scaled firms.

The lack of evidence for the positive impact of education on economic growth may be due to the falling return to primary schooling relative to secondary schooling while education policies have been targeting primary school attendance in the past.

Evidence from conditional cash transfers shows just a small amount of cash incentives dramatically improve human capital investment. We do not have a good conceptual framework to explain this phenomenon.

The issue is how to get policy-makers to stop and think when we do not know at all how to change institutions as the fundamental cause of economic development.

From Geoffrey Heal's speech:

A solar panel power generator today costs 1.8 dollars per KW, cheaper than a diesel power generator. Like mobile phones which have revolutionized telecommunication in LDCs, solar panel may revolutionize power generation in LDCs in the near future.

From Keijiro Otsuka's speech:

Nigerians used to eat rice only on festivities like Christmas, wedding parties, and funerals. Now they eat rice every Sunday.

Back in the 1960s, rice yields were the same in Africa and in Asia. Now yields in Asia double while those in Africa stay the same.

The green revolution in Asia was initiated purely due to technological reasons, not driven by policy or institutions. Then the market and policy responds to promote the use of high-yielding varieties by selling fertilizer or by installing irrigation. In Africa, this didn't happen because of the price control imposed by the government.

Rice yield can be higher in some parts of Africa (Senegal, Tanzania, and Uganda) than in Asia. But rice farmers in Africa do not know even simple ways to increase yields such as creating bunds between fields and leveling fields (which are the norm in Asia).

From Sabine Bruntrup-Seidemann's speech:

In Benin in the 1990s, many civil servants were fired due to the structural adjustment programme. University students were no longer guaranteed government jobs upon graduation. These skilled, but unemployed people formed NGOs because it was when the international development assistance community started putting money to NGOs rather than the governments of LDCs.

From Tomoya Matsumoto's speech:

In a randomized experiment in Uganda (a joint work by Matsumoto and Takashi Yamano from GRIPS), giving fertilizer for free to maize farmers (and one-off 2 hour training of how to use fertilizer) increases the amount of fertilizer purchase in the following growing season. There was a similar intervention in Uganda several years ago. At that time, farmers didn't respond. What has changed in Uganda since then is the scarcity of land due to the rapid population growth and the opportunity to sell maize produce for good prices (merchants now come to rural areas to purchase produce). Maize farmers in Uganda, therefore, do care about how to maximize the amount of yields now.

Developing Rural Areas

I listened to a podcast of Esther Duflo's lecture on rural development at LSE on 24 September, 2009. It is an excellent lecture in the sense that she synthesizes the latest findings in development economics research from the viewpoint of rural development. This kind of effort has long been missing. The past decade saw an emergence of high-quality research on economic development. But there are so many pieces of such research that it is difficult to grasp a big picture of what's known and what's not in a systematic way.

Anyone interested in economic development should listen to this lecture, available as a video streaming (scroll down within the video player window to 24 September 2009) and as an MP3 file (search for Duflo and click "mp3").

Long-lasting insecticide-treated bed nets in Madagascar

There are always things to learn which you would not imagine unless you visit the field.

Long-lasting insecticide-treated bed nets (LLINs) are the main tool for combating against malaria promoted by the international aid community. These nets have insecticide ingrained into the fabric so that mosquitoes with malaria parasites will die as soon as they touch the net. People talk about its durability in terms of how long the effect of insecticide lasts.

Today the head of the malaria control department at the Ministry of Health in Madagascar told us the following:

In 2006, pregnant women and mothers with children under the age of five in the east coast area received LLINs for free at health centers. Since then, the malaria infection rate had been stable around 10 to 20 percent. Last year, it suddenly went up to nearly 50 percent. We investigated what happened. It turns out that those LLINs distributed in 2006 have holes so that mosquitoes can come inside the bed net. Why? People in the east coast tend to have the kitchen inside the house. They burn charcoal to cook. The cooking smoke makes bed nets dirty. So people wash LLINs quite often. After two years, LLINs are worn out.

It's not about the durability of insecticide but that of nets themselves that matter to Malagasy people.

Reactions from a Control Group in Ghana

Randomized control trials (RCTs for short) have been the fad in development economics research during the past five years or so. One usual concern for this research methodology is that it's ethically incorrect. Usually, the trials evaluate the effect of offering poor people in poor countries with something that is supposed to be good, such as medicine, money, savings accounts, etc. And this offer is given to people randomly selected. So there are people who won't get any offer like this. Such people may be offended.

In Ghana, at least, this concern appears to be minimal. See this post from the blog of IPA, a non-profit organization founded by Dean Karlan, a leading development economist who made a fame from his RCT research in development economics. (He's actually coming to my workplace next month.)

Impact of insecticide-treated bed nets on infant mortality

I've done a bit of surveys on the impact of insecticide-treated bed nets (ITNs) on infant mortality. For the uninitiated, ITNs are known to be effective for preventing child deaths due to malaria infection. ITNs do not just protect you from mosquito bites that can transmit malaria parasite into your body. They also kill mosquitos that touch the ITNs as the bed net fabric contains insecticide. As malaria is transmitted only through mosquitos, ITNs reduce the risk of malaria infection in these two ways.

But how much do they reduce infant mortality in a village in Africa?

To identify randomized control trials of high quality on this topic, I look at the Cochrane Review (Lengeler (2004)), which identifies the list of such trials with no bias in the estimates. There are five studies looking at the impact on child mortality, and Lengeler (2004) concludes that ITNs reduce child mortality by 5.5 deaths per 1,000 children.

However, I'm interested in infant mortality (the death rate among those aged under 12 months old), not in child mortality (the death rate for children under the age of 5 years). Also, it is not clear to me whether these estimates refer to the intention-to-treat effect (the impact of distributing ITNs) or the effect of the actual use of ITNs. So I look at the original studies.

It turns out that one study looks at insecticide-treated curtains. Another looks at the impact of treating bed nets with insecticide (because the study is conducted in The Gambia, where the use of bed nets is common). The third study does not report infant mortality. So only two studies actually evaluate the impact on infant mortality. And the intervention they evaluate is actually the distribution of ITNs to a community, not the actual use of ITNs by villagers (though both studies find that the usage rate of distributed ITNs is rather high).

Binka et al. (1996) report that, in Kassena-Nankana district in Ghana, before distributing ITNs, the mortality among children 6 to 11 months old is 49.7 per 1,000 for treated villages while it is 55.1 per 1,000 for control villages. After the intervention, 73.2 for the treated and 90.3 for the control. The difference-in-difference estimate yields a reduction by 11.7 per 1,000. In addition, the authors report that the mortality for children aged 1 to 5 months is comparable between the treated and the control before the intervention (they don't report exact numbers) and that it becomes 77.4 for the treated and 100.7 for the control, implying that the impact is a reduction of 23.3 deaths per 1,000. Combining these figures, the impact on infant mortality is 35 deaths per 1,000, much higher than the effect for all the children under the age of 5 years concluded by Lengeler (2004).

The other study, Phillips-Howard et al. (2003), reports a reduction of infant mortality by 31 per 1,000 due to the distribution of ITNs in Asembo and Gem areas in western Kenya, which is surprisingly similar to what Binka et al. (1996) find. This similarity may be explained by the fact that malaria is endemic (ie. there is risk of infection throughout the year) in both study sites.

By the way, Hawley et al. (2003), another piece of paper by the same research team as Phillips-Howard et al. (2003), reports the reduction of child mortality by similar magnitude in villages without ITNs but near those villages with ITNs distributed. This finding suggests that the main mechanism through which ITNs reduce child mortality is not the protection of children (and pregnant mothers, as malaria infection during pregnancy can cause the malfunctioning of placenta and thus reduce the birth weight of newborns, which increases the risk of infant death) by bed nets but the eradication of mosquitos by insecticide.

Bloch, Genicot, and Ray (2008, JET)

The paper's abstract is available at http://dx.doi.org/10.1016/j.jet.2008.01.008.

Consider a village in a developing country where households engage in bilateral insurance schemes with directly connected households (relatives or friends, for example). If a household deviates from the insurance scheme (ie. fails to transfer money to directly connected households that have negative income shocks), the victim household will tell such behavior to connected households who in turn terminate the insurance scheme with the deviating household as a punishment.

The question is what aspects of the household network characterizes a stable household network, that is, a network in which no household has incentive to deviate from the bilateral insurance schemes.

Denote by q the number of links that the victim's information is passed through. For example, q equals two if the victim tells its directly connected households which in turn tell their own directly connected households, but the information does not spread beyond that.

Then the stable network is characterized by the maximum length of the smallest cycle connecting any three households in the network being q+2.

Why? Suppose household i fails to transfer money to household j. Household j tells its suffering to its connected households with the path length of less than or equal to q. These informed households cut the link to household i. Since the length of the smallest cycle of any three households in the network is at most q+2, the information from household j reaches all the households with (direct or indirect) links both to i and j. As a result, household j will be cut off from the network that includes household i, by deviating from the insurance scheme. A smaller network yields a lower payoff (intuitively, pooling income risk among more households provides better insurance against negative income shocks). Therefore, household i has no incentive to deviate, which means that the household network is stable.

Day of an Assistant Professor (57)

Work on the climate project with my colleagues.

Attend a seminar given by this week's visitor, Pascaline Dupas. Based on a randomized field experiment in Kenya, she presents the first convincing evidence that allowing the poor to save in a bank account increases their welfare.

Go for dinner with today's seminar speaker and my colleagues.

Day of an Assistant Professor (54)

1. Have a chat to a PhD student on her research.

2. Have a chat with another PhD student on her research.

3. Review some PhD applications.

4. Read Guttman (2008):

Assortative matching of borrowers under joint liability contracts may not occur due to dynamic incentives, because forming a group with safe borrowers decreases the probability of being denied credit after group defaulting MORE for a risky borrower.

How development economics research is actually done

Belatedly, I found Robert Jensen's guest posts to Steven Levitt's Freakonomics Blog. It's an unusual account of how development economics research is actually done.

Robert Jensen's research is characterized by the testing in developing countries of basic economic theory that is yet to be empirically verified in a convincing fashion. As a result, it contributes both to economics in general and to development economics in particular. He's one of my favorite economists.

An Economist's take on HIV in Africa

A well-done lecture on what we really know about HIV in Africa, delivered by Emily Oster:


Summary: (1) An emphasis on education may be misleading because preventive behavior appears to be linked with life expectancy. (2) The HIV prevalence rates in African countries, reported by US Census Bureau, are likely to exaggerate the actual prevalence. (3) Poverty do not seem to be the cause of the spread of HIV infection.

Day of an Assistant Professor (32)

Read Maccini and Yang's paper entitled "Under the Weather: Health, Schooling, and Economic Consequences of Early-Life Rainfall" (forthcoming in American Economic Review) for the climate change project. Indonesian women are taller and more educated if it rains more than average during the first year of life. No such relationship between early childhood rainfall and men's height and education.

Day of an Assistant Professor (29)

CORRECTED on 22 Feb.

1. Attend a lunch-time seminar at Handels.

2. Continue working on the revision of the African democracy paper.

3. While waiting for regressions to be run, read Deaton (2009) again to summarize his points on randomized control trials (RCTs) in development economics:

(1) RCTs gives us an unbiased estimate of the average treatment effect due to the fact that the expectation is a linear operator, but it is not informative about other features of the distribution of the treatment effect such as median. (Pages 25-7)

(2) RCTs should be designed to learn not whether a particular project works but why it works, because "projects can rarely be replicated while the mechanism underlying success or failure will often be replicated and transportable" (pages 30-1). See Section 5 for more on this.

(3) Calculation of the standard error of the average treatment effect is not straightforward if the treatment effect is heterogenous, because the variance in the error term will then be different between the treated and the control group. (pages 31-3)

(4) Controlling for covariates to improve the precision of the treatment effect estimate can be hazardous with the presence of heterogeneous treatment effects. (Pages 34-6)

(5) If compliance to the treatment is incomplete, reduced-form evidence (ie. intention-to-treat effect) is more informative than the instrumental variables estimation because the latter yields the local average treatment effect which may not be of interest. (pages 14-20, 36-8)

Political Economy of Reform

There is a small theoretical literature on the political economy of reform. Here's an attempt to relate each paper to each other.

The seminal paper by Fernandez and Rodrik (1991) considers two kinds of reform. First, welfare-improving reform. Second, majority-benefiting reform. Both types of reform can be stalled due to individual-level uncertainty (whether or not each citizen benefits from reform is unknown to themselves).

Jain and Mukand (2003) show that welfare-improving reform can be implemented even with individual-level uncertainty if voters can vote on whether to compensate the losers from reform after reform is enacted. However, majority-benefiting reform still cannot.

Dewatripont and Roland (1995) instead analyze the case where citizens do not know whether each reform improves welfare.

All the above papers analyze the political feasibility of reforms in terms of conflicts of interest among citizens. Majumdar and Mukand (2004) instead look at it in the political agency framework (conflict of interest between government and citizens), showing why policy-makers often resist scrapping their failing policies and adopt high-risk policy experiments.

Who must pay bribes and how much?

Development economists interested in bribery should read Svensson (2003) before designing their research. I've seen a couple of papers on corruption which completely ignore the contribution made by this piece of research.

He takes great care of designing the firm survey to solicit reliable information on bribe-paying behavior, which should be followed by anyone who wants to collect bribe data by surveys.

His empirical findings include:

(1) Firms are more likely to pay bribes if they are required to deal with the public sector officials (using public infrastructure, engaging in international trade, and paying taxes). In other words, formal sector firms are more likely to pay bribes than informal sector ones. A theoretical reason is that such firms lose a lot by refusing to pay bribes.

(2) More profitable firms are as likely to pay bribes as less profitable firms. But more profitable firms pay a larger amount of bribes than less profitable ones, conditional on bribe payments. This is an interesting fact. Barely profitable firms still need to pay bribes, if a very small amount, so that they can keep operating.

Any theory on bribery contradicting to these findings is hard to appreciate (unless counter-evidence is provided). If your empirical finding is different from these findings, you need to explain why.

Impact of Successful Assassinations on Political Regime and Wars

My memory has been weakening recently. As an attempt to remedy this, I'll start writing down on this blog what I learn from reading papers, books, news media, etc.

Empirical findings from Jones and Olken (2008) "Hit or Miss? The Effect of Assassinations on Institutions and War":

When an assassination is attempted against the national leader...

1. If the dictator is killed, democratization is likely to follow.

2. It does not matter for subsequent political regime change if a leader in democracy is killed.

3. If the leader facing a small-scale war (less than 1000 battle-related death per year) is killed, the war is likely to be intensified.

4. It does not matter for the onset of a war whether the leader is killed.

5. If the leader facing a large-scale war (1000 deaths or more per year) is killed, the war is likely to be terminated.

Private property rights enforcement

A great example of private property rights enforcement when the state is weak. HT: Dani Rodrik.