Long-term health trends
The determinants of long-term health improvements
In theory, two main channels can be identified to improve health and living conditions of a given population. First, public provision can improve the health environment, which consists both of infectious agents (prevalence of malaria for instance), and of the protection brought by health facilities and hygiene. The collective health environment is generally improved through the implementation of a health system, a set of resources, medical care (preventive and therapeu-tic), hygiene and sanitation. Second, health outcomes can be improved through an individual channel, i.e. nutrition and the private demand for health.
This dichotomy between public provision and private determinants of health can be found in Fogel’s and Deaton’s work. Both recognize that the two channels can play a role, but they do not identify the same factor as the starting point of health development. According to Fogel (1994), the escape from hunger and premature death was mainly an escape from a nutritional trap where we could not work to produce food because we were too weak, and we were too weak because we could not work to produce food to make us strong. Nutrition was a key element of the “historical great escape”, and led to an improvement in health standards. This process has been observed over the last three centuries in Europe and in the United States, with both increasing heights and a secular decline in mortality rates. This thesis, putting forward nutrition as the main channel for health improvements, clearly insists on the importance of income. Yet, private determinants are not the only channel linking living conditions and health. Even though Deaton (2007) acknowledges that private income affects health outcomes through nutrition, he surely insists on the importance of the immediate health environment. According to him, “public health measures, particularly the provision of clean water and better sanitation, were the fundamental forces for mortality reduction during the century from 1850 to 1950”.
Depending on which channel is the most important between supply-side policies and pri-vate determinants of health, policy recommendations should not be the same. Of course, health supply and private demand are complementary. Whatever the level of demand, this demand needs a minimum provision of health to be met. Yet, if private determinants are the main force behind health improvements, the most efficient policy would focus on individual preferences and economic growth. On the contrary, if infectious diseases are the main driver of morbidity and mortality, public health policies should be targeted first.
The evolution of health standards in Africa
Describing what happened in Africa during the 20th century would contribute to shed light on what needs to be further done to improve health policies in the continent. Getting new insights on this question is particularly important in the context of recent health progress made in Africa. We know that Africa has the lowest level of health investments and the worst health conditions in the world today. What happened over the last decades, which could explain such a fact?
A first explanation is that the continent had extremely low health standards to start with, compared to the rest of the world. As a consequence, despite huge progress in health outcomes since colonial independence, it did not catch up with other regions of the world. The most salient example is the case of maternal mortality (Figure 0.1). Sub-Saharan Africa made huge progress from 1990. But it started from such high levels of risk that it was unable to catch up with other regions. The lifetime risk of maternal death was above six percent in 1990 in sub-Saharan Africa, followed by South Asia, with less than a 2.5% risk. The risk in sub-Saharan Africa is still 2.6% today, compared to 0.5% in South Asia. The same conclusion can be made about DPT (diphtheria, pertussis, and tetanus), or measles immunization rates (upper graphs in Figure 0.2). South Asia and sub-Saharan Africa experienced the same trends as other regions over the last 30 years. However, both regions started from much lower levels, and are still lagging behind. Similarly, in sub-Saharan Africa, the proportion of population with access to improved sanitation, or improved water source, did increase as much as in other regions from 1990 (lower graphs in Figure 0.2). Yet, today, only 65% of the sub-Saharan population has access to an improved water source (30% has access to improved sanitation facilities), compared to around 90% in South Asia (40% for sanitation), its closest region.
A second explanation for Africa’s low levels of health is that the continent experienced lower health progress than other regions of the world, for some health dimensions. These lower health progress could be related to the fact that welfare levels have been deteriorating in Africa. Chen and Ravallion (2004) estimate that the world poverty rate nearly halved between 1981 and 2001, declining from 40% to 21%. Meanwhile, the level of poverty in sub-Saharan Africa increased from 41.6% to 46.4% during the same period. While the under-five mortality rate decreased from 23% to 2.4% in Middle-East and North Africa from 1960 to 2013, the sub-Saharan mortality rate only decreased from 26% to 9.2% (upper graph in Figure 0.3). Similarly, sub-Saharan Africa and South Asia started from very close life spans in 1960: 40 years old in sub-Saharan Africa, and 42 years old in South Asia (lower graph in Figure 0.3). However, from the 1980’s, the rate of increase slowed sharply in Africa compared to other regions of the world. Today, life expectancy has reached 67 years old in South Asia, an increase of 57%, compared to only 56 years old in Africa, an increase of 40%.
A third explanation is that Africa is still lagging behind in terms of health inputs. African countries represent twelve percent of the world population, but account for only one percent of worldwide health expenses and two percent of medical staff. There were no significant trends in health expenditures and health inputs over the last decades (Figure 0.4). Sub-Saharan Africa and Middle-East and North Africa approximately have the same level of public health investments as a percentage of GDP (upper-left graph in Figure 0.4), well above South Asia. However, when looking at the total health expenditure per capita, the picture changes, due to the relative disadvantage of sub-Saharan Africa in terms of GDP. Sub-Saharan Africa has the lowest level of total health expenditure per capita, together with South Asia (upper-right graph in Figure 0.4). The region also has the lowest number of physicians per capita in the world, and is second regarding the number of hospital beds per capita, after South Asia (lower graphs in Figure 0.4). Furthermore, the number of physicians per capita did not increase since 1990, while it did in all other regions.
These descriptive statistics give a view on health standards and their progress from the independence of African countries. But what happened before that? How did health inputs and outcomes evolve during colonial time? What can be said about the evolution of health standards along the 20th century?
The study of the long-term evolution of health standards in African countries is very chal-lenging because of a lack of reliable census data, even for the contemporaneous period. Very few births and deaths are officially registered in the continent. In particular, there are almost no mortality data for the first half of the 20th century. Until 1920-1930, only the European mor-tality is registered. From then on, the African mortality is sometimes reported in colonial data, but only when it happens within a medical institution, or when it is recorded by a member of the medical staff. Similarly, child mortality and birth rates are only available for a sub-sample of the population, which is not representative of the whole population.
Given this lack of data, standard measures of health – such as life spans or mortality – are not available to study the evolution of health standards in Africa; alternative measures need to be used. A few economic papers use anthropometric data to look at health standards in colonial Africa. For instance, Austin, Baten, and Moradi (2007) and Moradi (2009) on Ghana and Kenya and Cogneau and Rouanet (2011) on Cote d’Ivoire and Ghana, study long-term height trends. On former French West Africa, Huillery (2009) relates contemporary child z-score to early colonial investments in health, measured by medical staff. These works focus on one specific health outcome, height. However, as stated by Bleakley (2010a), “both theory and evidence suggest that we should stop thinking of health as a univariate object”. Ideally, the study of the long-term evolution of health conditions should look at various dimensions of health: health policies on the one hand, and various health outcomes, on the other hand.
The starting point of this dissertation is a research article, co-written with Denis Cogneau before I started this PhD. In Cogneau and Rouanet (2011), we show with survey data that the pace of increase in height stature experienced by successive cohorts born in Cote d’Ivoire and Ghana during the late colonial period (1925-1960) is almost as high as the pace observed in France and Great Britain during the period 1875 to 1975, even when correcting for the bias arising from old-age shrinking. By contrast, the early post-colonial period (1960-1985) is char-acterized by stagnation or even reversion in Cote d’Ivoire and Ghana. Trends are shown in Figure 0.5. This article argues that the selection effects linked for instance to measuring the height of women rather than men, mothers rather than women, and, most importantly, the in-teractions between height and mortality, cannot account for these figures. This paper provides evidence that a significant share of the increase in height stature may be related to the early stages of urbanization and cocoa production.
Building on this work, two research questions emerged. First, what is the magnitude of the selective mortality bias on adu lt heights? To what extent does the validity of anthropometric studies,158 and of some heal t h interventions’ impact evaluations, can be threatened by such a Height bias? Second, what c a n be said about health in Africa before 1960? How do health progress and health policies re la te to the colonial presence? More generally, this dissertation aims at opening the black156 box of what happened to h ea lth in Africa during and following the colonial period.
Book 1.indb 60 2011/12/13 12:10:41
There is an urgent need Yearto ofhavebirth a clear view on what happened in this continent regarding 1920 1940 1960 1980 LSMS DHS health inputs, heal th outcomes and f ertility patterns, during the 20th century. This thesis tries Figureto 2:partlyHeightof fillwomeninin thisCôted’Ivoire,gapLSMSbyandinvestigatingDHSurveys several aspects of the evolution of health conditions in Africa, making three contributions to the literature.
The identification of a causal impact of health interventions on various outcomes (produc-tivity, income, health outcomes) is challenging. The first two chapters of this dissertation study 60two specific challenges in the African context. The first chapter addresses the issue of the endo-geneity of health policies, asking whether the provision of health policies is specific as regards to other colonial policies. The colonial strategy for health2011/12/1312:10:41policies’PM provision is compared with the strategy for other colonial policies, with a linear model and a two-tier model at the colonial district level. The second chapter tackles the question of selective mortality in the context of the “double African Paradox” in West Africa. It develops and estimates a new model of height differential between survivors and deceased, in order to assess the magnitude that could be reached by the selective mortality bias. The third chapter deals with another specificity of African health: fertility and gender preferences. It develops a new indicator of gender prefer-ences based on birth spacing, which is applied to Africa. This chapter also studies the role of structural patterns and of socioeconomic factors in shaping gender preferences.
Table of contents :
1 The Provision of Colonial Policies in Former FrenchWest Africa: Are Health Policies Specific?
1.2 Literature and historical context
1.3.1 Colonial data
1.3.2 Colonial statistics: main challenges
1.4 Colonial policies: first descriptive evidence
1.4.1 Colonial health policies: what was done?
1.4.2 Increasing trends in colonial investments
1.4.3 Health provision across space
1.4.4 Colonial policies and urbanization
1.4.5 Colonial policies and Europeans
1.5 Health policies and other colonial policies: context and empirical strategy
1.5.1 The decision-making chain of colonial investments
1.5.2 The provision of colonial policies: what are the main drivers?
1.5.3 The empirical strategy
1.6 Results: what are the main determinants of health provision ?
1.6.1 Linear results
1.6.2 Intensive and extensive margins of colonial investments
1.6.3 Robustness checks
2 The Double African Paradox: Height and Selective Mortality in West Africa
2.2 Existing literature and contribution
2.3 Data and descriptive results
2.3.2 Descriptive statistics on the paradox
2.4 The height-mortality relationship
2.4.1 The mechanisms at stake
2.4.2 What are the potential biases?
2.4.3 The empirical strategy
2.4.4 A positive correlation?
2.5 A model of height differential between survivors and deceased
2.5.1 The adult height equation
2.5.2 The height differential between survivors and deceased
2.5.3 Identification assumptions of the model
2.6 Estimation of the model and implication for the “double African paradox”
2.6.1 Estimation of the height differential between survivors and deceased
2.6.2 The level paradox
2.6.3 The trend paradox
2.6.4 Adult heights, child heights: why do conclusions differ?
2.7 Robustness of the results and main implications
2.7.2 Main implications
3 Gender Preferences in Africa: A Comparative Analysis of Fertility Choices with Pauline Rossi
3.2 Gender preferences in Africa
3.2.1 Theoretical motives for gender preferences
3.2.2 Empirical evidence so far
3.3.2 Descriptive statistics
3.4 Empirical Strategy
3.4.1 A duration model of birth intervals
3.4.2 Relating durations to the proportion of sons
3.4.3 Identification assumptions
3.5.1 Comparative descriptive analysis
3.5.2 Key drivers of gender preferences
3.5.3 Mechanisms: individual choices or social norms?
3.6 Robustness Tests
3.6.1 Testing the child mortality bias
3.6.2 Testing the sample selection of mothers
3.6.3 Investigating heterogenous effects across birth ranks