Jump to ContentJump to Main Navigation
The Political Economy of Hunger: Volume 3: Endemic Hunger$

Jean Drèze and Amartya Sen

Print publication date: 1991

Print ISBN-13: 9780198286370

Published to Oxford Scholarship Online: January 2008

DOI: 10.1093/acprof:oso/9780198286370.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (oxford.universitypressscholarship.com). (c) Copyright Oxford University Press, 2020. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use. date: 31 October 2020

5 Undernutrition in Sub‐Saharan Africa

5 Undernutrition in Sub‐Saharan Africa

A Critical Assessment of the Evidence

Chapter:
(p.155) 5 Undernutrition in Sub‐Saharan Africa
Source:
The Political Economy of Hunger: Volume 3: Endemic Hunger
Author(s):

Peter Svedberg

Publisher:
Oxford University Press
DOI:10.1093/acprof:oso/9780198286370.003.0006

Abstract and Keywords

From 5% to 45% of the population of sub-Saharan Africa appear to be undernourished, depending on the indicator and sources consulted. This enormous discrepancy calls for a diagnosis of the extent of undernourishment in the region. This chapter argues that both FAO and IBRD have based their estimates on the FAO calorie availability data, which are downward biased thus leading to an inflated figure for undernourishment. This exaggeration is an upshot of biased methods, non-representative data, and imprecise and ambiguous conception of undernutrition. The chapter makes extensive use of anthropometric evidence to establish these substantive conclusions. It suggests that even when sample studies are representative and unbiased, supplementary socio-economic data are required to understand the source of undernutrition.

Keywords:   FAO/IBRD method, anthropometric methods, estimation bias, calorie availability, undernutrition

5.1. Introduction

Most of the sub‐Saharan African countries have experienced what cannot be labelled anything but an economic and political crisis during the 1970s and 1980s. The annual growth of GDP per capita for the region as a whole went down from 2 per cent in the 1950s and the 1960s to 0.8 per cent during the 1970s; in the 1980s, it was −2.5 per cent. The share of sub‐Saharan Africa (SSA) in world exports has declined dramatically and real export earnings of many of the countries in the region have dropped significantly since the early 1970s (Svedberg 1991a). In recent years, almost every government in the SSA countries has been faced with the inescapable fact that drastic economic policy changes have to be undertaken during the 1990s. More than a dozen countries with previously highly overvalued exchange rates have already devalued substantially and also initiated, or are in the process of initiating, reform in a large number of areas (see Svedberg 1991a and IBRD 1989).

Not only have the overall economic performances of most SSA countries been miserable over the past two decades. The conventional wisdom in the international organizations is that the world's food problems are now concentrated in Africa. According to the FAO, food production per capita has declined over the 1970s and 1980s and the per capita ‘availability’ of food for human consumption is now only 80 per cent of the FAO/WHO recommended intakes. The FAO further claims, on the basis of estimates of how the food is distributed, that between one‐quarter and one‐fifth of the population in the region does not have enough food to be able to work or pursue any form of physical activity. Estimates by the World Bank purport that almost half the population in the region is at least moderately undernourished.

The food consumption problems will most certainly continue to be one of the main policy concerns in the SSA during the 1990s. Food will be an issue both in its own right and as a constraint on policy reform in other areas. That food will be a major question of policy concern is self‐evident even if the problem is of a less alarming order of magnitude than suggested by the international organizations. Food is after all the most basic of all human needs and the notion that eradicating hunger and undernutrition is a top priority aim for policy is shared almost universally. The nutritional problems will also enter as serious constraints on the structural adjustment programmes that will dominate the African scene during the 1990s. New policies mean shifts in (p.156) relative prices and redistribution of real incomes. This will affect the food entitlements of different groups. It is, then, of the utmost importance that the nutritional situation is known. If only 5–10 per cent of the population in a country is at risk, there will be many more degrees of freedom in the pursuit of new general economic policies than if it is 30–60 percent. Moreover, one needs to know who the people at risk are; otherwise it will be impossible to reach them with targeted policies or to see to it that they are compensated if they suffer from side‐effects of general policies.

The objectives of this chapter are threefold. The first is to bring together as much as possible of the available evidence on nutritional standards in the SSA (section 5.2). The second objective is to compare the different pictures of the nutritional situation that emerge when different indicators are consulted (section 5.3). The third aim is to try to delineate how much of these differences is explained, on the one hand, by errors and biases in the different estimation methods (sections 5.4–5.7) and, on the other hand, by the fact that the different indicators measure different things, i.e. rest on very different notions of what ‘undernutrition’ is all about (section 5.8).

5.2. Indicators of food standards and undernutrition in SSA

Food consumed by the population in various countries has conventionally been estimated in two different ways. One has been to estimate food availabilities from the supply side. This is the approach that has long been favoured by the FAO. The other method is to estimate the actual intake of sample populations and extrapolate to more aggregate levels. However, the food ‘available’ to, or actually consumed by, households, even if accurately estimated, does not say much about the nutritional status of the population as a whole or the prevalence of undernutrition. In order to be able to say something on these issues, the distribution of the food within the population and the (im)balance between the energy intakes and ‘desired’ expenditures of individuals must be known.

Several different methods have been used to estimate the incidence of undernutrition in different populations. The most aggregate estimates are derived by the FAO and the World Bank. On the basis of the FAO estimates of calories available in various countries and assumptions of how the calories are distributed across households, the FAO and IBRD estimate the share of the population that does not meet the ‘desired’ calorie expenditures, i.e. that is undernourished. A number of studies have also been conducted at the level of villages, where people's actual intakes have been estimated and related to their assumed energy requirements. A third set of studies has used anthropometric and biochemical and related methods to assess the nutritional status of samples of individuals. A fourth set of assessments focuses on the (presumed) consequences of undernutrition, such as child mortality, without considering the (p.157) ‘underlying’ food intake, energy expenditure or the anthropometric status of the population. One can also learn about people's food and nutrition standards by examining their economic expenditures in general and food expenditures in particular. (The latter type of investigation will not be discussed in this paper, but is dealt with in Svedberg 1987, 1991b).

In the following subsections, we shall describe in more detail how the various estimates of food standards and undernutrition mentioned above are arrived at, and discuss the results that have been obtained for the SSA countries.

(a) The FAO calorie availability estimates

Method

The FAO describes its estimation method as follows:

The total quantity of foodstuffs produced in a country added to the total quantity imported and adjusted for any change in stocks that may have occurred since the beginning of the reference period gives the supply available during that period. On the utilization side, a distinction is made between the quantities exported, fed to livestock, used for seed, put to industrial and other non‐food uses, or lost during storage and transportation, and food supplies available for human consumption at the retail level, i.e. in the form food leaves the retail shop or otherwise enters the household. The per‐caput supply of each food item available for human consumption is then obtained by dividing the food supplies available for human consumption by the related data on the population actually taking part of it. Data on per caput food supplies are expressed in terms of quantity and also, by applying appropriate food consumption factors, in terms of nutrient elements (calories, protein, etc.). (FAO 1980)

Much of the food production data that underlie the FAO calorie availability estimates have been supplied by national governments in ‘the form of replies to annual FAO questionnaires’. Where no official or semi‐official figures are available from the countries themselves, the FAO makes its own estimates (FAO, Production Yearbooks, introduction).

Estimates

The FAO data suggest that in 1961–3, the per capita calorie availability for the region as a whole was 2,014 (Table 5.1). By the mid‐1980s, the number had dropped to 1,876, but the year‐to‐year variations have been too marked for a statistically significant trend to be discernible. In an international comparison, the SSA region comes out the worst. The Near East has seen the per capita availability of calories increase drastically since the early 1960s; in the Far East and Latin America, there have been improvements. Among individual countries, India has experienced a small increase, while for Bangladesh, the FAO estimates suggest a deterioration.

Table 5.1 Estimated per capita calorie availability and requirement, by selected countries and regions

Number of calories

Requirementsa

1961–3

1970–2

1980–2

1983–5

Sub‐Saharan

Africab

2,014

1,896

1,982

1,876

2,340

India

2,038

2,054

2,075

2,161

2,200

Bangladesh

1,938

1,953

1,879

1,859

2,315

Developing

market economies

2,069

2,187

2,338

2,363

Latin America

2,381

2,518

2,692

2,700

2,380

Near East

2,225

2,415

2,879

2,947

2,200

Far East

1,962

2,080

2,186

2,239

2,230

Africa

2,055

2,103

2,200

2,129

2,340

aRequirements according to FAO/WHO (1973) standards.

bThe FAO (1987) tapes provide no separate data for sub‐Saharan Africa. The above estimates have been derived by correcting the average for Africa as a whole with the weighted averages of the North African countries (Algeria, Egypt, Libya, Morocco, and Tunisia).

Source: Derived from FAO (1987).

Since 1970, the estimated per capita calorie availability has improved (statistically significantly) in a little more than a third (18) of the 44 sub‐Saharan African countries according to the FAO. In some of these, the improvement is substantial, e.g. in Congo, Gabon, Lesotho, Mauritius, Niger, and Tanzania. There are, on the other hand, a dozen countries where there has (p.158) been a statistically significant deterioration; in a few cases (Chad, Ghana, and Mozambique) by more than 1 per cent annually over the 17 years. In 14 countries the year‐to‐year fluctuations have been very pronounced and there is no statistically significant trend. The FAO data thus suggest a rather varied experience across Africa since the early 1970s (see Svedberg 1991b: Table 5.2).

Table 5.2 Prevalence of undernourished people in developing countries, by major regions, as estimated by the IBRD and the FAO

Region

No. of people (m.)

% of population

1970

1980

Change

1970

1980

Change

IBRD

Sub‐Saharan Africa

High estimate

115

150

+35

43

44

+1

Low estimate

60

90

+30

21

25

+4

East Asia and Pacific

High estimate

93

40

−53

41

14

−27

Low estimate

47

20

−27

21

7

−14

South Asia

High estimate

341

470

+129

47

50

+3

Low estimate

136

200

+64

19

21

+2

Middle East and North Africa

High estimate

53

20

−33

35

10

−25

Low estimate

31

10

−21

18

4

−14

Latin America and Caribbean

High estimate

59

50

−9

20

13

−7

Low estimate

25

20

−5

10

6

−4

Developing countries

High estimate

664

730

+64

40

34

−6

Low estimate

298

340

+42

18

16

−2

FAO

Africaa

High estimate

81

99

+18

29

26

−3

Low estimate

57

70

+13

20

19

−1

Far East

High estimate

303

313

+10

31

25

−6

Low estimate

208

210

+2

21

17

−4

Near East

High estimate

34

25

−9

22

12

−10

Low estimate

23

16

−7

15

8

−7

Latin America

High estimate

53

56

+3

19

16

−3

Low estimate

36

38

+2

13

11

−2

Developing countries

High estimate

472

494

+22

28

23

−5

Low estimate

325

335

+10

19

15

−4

aAfrica, excluding Egypt, Libya, Sudan, and SAU.

Sources: IBRD (1986: Tables 2–3 and 2–4); FAO (1985a: Table 3–4).

The most notable development over the 1980s is that the per capita supply of calories has continued to fall in the countries where there has been a secular deterioration since the early 1970s: Chad, Ghana, Guinea, and Sierra Leone. According to the FAO, there is no single country in which there has been a notable increase over the first half of the 1980s. The FAO further suggests that the per capita availability of calories in 1983–85 was lower in 30 of the 44 countries in SSA than in India. In 7 African countries, the situation was worse than in Bangladesh, the FAO claims.

(b) The dietary calorie intake estimates

Methods

Three main methods have been used to estimate food consumption at the level of individuals and households in Africa. The most common (and inexpensive) procedure is to collect qualitative information through interviews (recalls). That is, one simply asks people how much of different foods they have consumed over a specific time period, e.g. the past 24 hours, or the last week. The second method is to measure purchases and/or changes in food stocks and convert these into ‘consumption’ flows. The third (and most expensive) method is to weigh (the equivalent of) the food actually observed to (p.159) have been consumed by the individual or the household. The nutritional content of the estimated food consumed, whatever the method used, is usually derived from more or less standardized conversion tables.

Estimates

The estimated daily per capita calorie intake in 85 sample studies from African villages lies in the 1,800 to 2,200 range in most cases. There are thus notable intervillage differences and the average fails to meet the FAO/WHO recommended per capita calorie requirement norms, which are about 2,340 calories for the African countries (Schofield 1979: Table 5.5; Hulse and Pearson 1981: Table 5). The estimated (unweighted) average per capita calorie consumption for all the 85 studies together (two studies are covered in both surveys) is 1,950.

Table 5.5 Estimated calorie savings possible for adult males

FAO standard reference male

Nutritionally constrained reference male

Height (cm)

172

160

Weight (kg)

65

51

BMI

22

20

BMR/kg/day

26

26

Total BMR

1,690

1,326

BMR/hour

70

55

Hours spent in

Sleep

8; 1.0 BMR

8; 1.0 BMR

Work activity

7; 2.7 BMR

7; 2.7 BMR

Social activity

2; 3.0 BMR

0; 0.0 BMR

Residual activity

7; 1.4 BMR

9; 1.4 BMR

Total calories for external activity

1,303

847

Total calorie expenditure

2,976

2,173

% of FAO reference male

100

73

Source: FAO/WHO/UNU (1985) (FAO ‘standard reference male’).

(c) The prevalence of undernutrition: FAO and IBRD

Method

The FAO (1985) and the IBRD (1986) use an indirect method to estimate the incidence of undernutrition.1 The procedure comprises three steps. The first is to estimate the food energy ‘available for human consumption’ at the country level. The second is to estimate ‘desired’ energy expenditures. The third is to model how the distribution of the available calories relates to the distribution of calorie ‘requirements’. Thus they do not aim to measure people's actual intakes directly (the dietary approach) or the outcome of energy deficiency (clinical and anthropometric approaches).

Model

The main features of the FAO/IBRD model can be described by a simple graph (Fig. 5.1). Let us assume that we have a population of a given number of households (H) for which all differences in size, composition (age, sex, etc.), and other factors that affect their calorie ‘requirements’ have been normalized. All households thus have the same per capita requirements, R in Fig. 5.1. It is further assumed that there is a given ‘pool’ of calories available to this population. On a per capita basis, the availability of calories to the total population is given by A. The households are different in one important aspect, however: they have different incomes and the available calories are allocated in proportion to incomes. The households are ranked in ascending order of their per capita calorie intake (PCCI) along the horizontal axis in Fig. 5.1. The household with the lowest income and per capita calorie intake (measured along the vertical axis), is located at the left‐hand end of the horizontal axis (at 0). Households with higher per capita calorie intakes are placed successively to the right. The household with the highest income and intake is located at the extreme right of the graph (at H). The DD' line gives the absolute distribution of the calories across the H households; in terms of area, OAA'H is equal to ODD'H.

5 Undernutrition in Sub‐Saharan AfricaA Critical Assessment of the Evidence

Fig. 5.1. Simple version of the model used by the IBRD and the FAO to estimate the prevalence of undernutrition in developing countries

(p.160)

Out of the H households, U 1 have an estimated per capita intake that fails to meet the per capita calorie requirement norm R. These households are thus undernourished according to the FAO/IBRD estimation procedure. In the particular case depicted in Fig. 5.1, the per capita availability of calories exceeds the per capita calorie requirement (A > R), signifying that had the calories been allocated evenly over the population, no one would be undernourished. It is further notable that the main interest of the FAO and IBRD is in the share of the population that falls below R (as defined; see below), not in the amount by which the most destitute are short of calories. It is also important to remember that the ‘undernourished’ households are not identifiable with the FAO/IBRD method; they comprise an anonymous ‘poor’ section of the population.

Data

Both organizations work with the same aggregate database, i.e. the FAO per capita availability of calorie estimates described above. The ‘requirement’ norms are different, however. The FAO uses two cut‐off points to define undernutrition. Under the assumption that people have given energy requirements for internal body functions, i.e. a constant ‘basal metabolic rate’, BMR, per kilo of body weight, they set the calorie requirement norm at 1.4 times the BMR. This norm allows an individual to exert the minimum of external physical activity needed for the most basic personal undertakings (dressing, washing, etc.) and to maintain cardiovascular and muscular fitness, but no physical activity (such as work) above that. The other cut‐off point used by the FAO to estimate the prevalence of undernutrition is 1.2 BMR. This norm is derived on the same assumptions as the higher one, with one exception: the human body is assumed to have a built‐in mechanism that ensures that the energy in food is more efficiently metabolized when food is scarce.

(p.161) The IBRD also uses two alternative calorie requirement norms; not on the basis of different assumptions of the energy requirements human beings have, but rather in order to distinguish between those who are ‘moderately’ and ‘severely’ undernourished. The two norms applied by the IBRD are 90 and 80 per cent of the recommended dietary allowances (RDAs) of calories supplied by the FAO/WHO (1973). These cut‐off points are higher than those used by the FAO.

The distribution of the available calories across households is assumed by both organizations to be determined by income. The IBRD is explicit in showing what income distribution data and what calorie‐income elasticities it has used to map the distribution of calorie consumption from the distribution of incomes. The FAO has undertaken its own distribution and elasticity estimations, but has not published the details. One further difference is that the IBRD works with discrete income groups, such as the 30 per cent poorest, while the FAO uses a continuous income scale.

Estimates

The IBRD and FAO estimates for the years 1970 and 1980 are reproduced in Table 5.2. The highest estimate of the incidence of ‘moderate and severe’ undernutrition in sub‐Saharan Africa for the year 1980, 44 per cent of the population, is suggested by the Bank, applying its higher requirement norm. The smallest estimate, 19 per cent, is from the FAO, derived on the assumption that people can ‘adapt’ to low calorie intakes.

The two organizations did not find a significant change in the relative incidence of undernutrition in the SSA region as a whole between 1970 and 1980, the two points of time examined. In absolute terms, however, the organizations report an increase in the number of undernourished people in the 13–35 million range over this period. Between 1980 and 1986, the FAO estimates of per capita food availability in the SSA show a decline. This implies that had the relative incidence of undernutrition in Africa been assessed by the mid‐1980s, using the FAO and IBRD methods, one would have registered an increase (other things being equal). However, preliminary data for more recent years suggest positive growth in per capita domestic food production in some countries, which may have led to increased per capita food consumption as well. It is thus too early to say whether the FAO and IBRD are likely to find that the prevalence of undernutrition has increased or decreased over the 1980s in the SSA.

According to the Bank's estimates, the incidence (in percentage terms) of severe undernutrition in the SSA is higher than in South Asia (although the share of the population suffering from moderate undernutrition is higher in the latter region). It should further be noted that the Bank provides background tables of estimates for individual countries, which suggest that in 1980 the situation in about a dozen SSA countries was worse than in Bangladesh (Svedberg 1987: Appendix Table 4). (p.162)

(p.163) (d) The prevalence of undernutrition: anthropometric evidence

Method

The basic presumption behind the anthropometric methods is that one need not estimate people's actual calorie intakes and expenditures to assess their nutritional status. Any imbalance between intake and ‘desired’ expenditure will show up in reduced body weight and retarded growth in stature (for children and adolescents). If weight and/or stature fall below some anthropometric norm, people are defined as undernourished.

The anthropometric studies that have been conducted in Africa have almost exclusively focused on children; very few studies of adults exist. On the other hand, there are a fair amount of height and weight observations for adults, made by anthropologists and human physiologists. Eveleth and Tanner (1976) brought together and made comparable such observations of average height and weight of grown‐ups in almost three dozen populations in the SSA. Unfortunately, these data are reported without estimates of variance, so that it is not possible to derive the percentage of the different populations that falls below some anthropometric norm that would define them as undernourished. One can only compare the average height and weight in these populations to the norms. Eveleth and Tanner did not provide such norms, however, which means that we have to construct them ourselves.

The height norm for adults that will be used here (as a first approximation) is established on the presumption that all ethnic groups have roughly the same genetic potential for adult stature. This is what most human biologists claim today. The growth potential is further assumed to have been achieved in the populations that have the de facto highest stature. These are found in the Scandinavian countries, where adult males and females reach 180 and 167 cm on the average (in the mid‐1980s). For the time being, it is also assumed that inadequate nutrition is the only constraint on achieving the full genetic potential for growth in the SSA populations. The norm is further based on the assumption that the achievement of the full genetic potential has a value in its own right, irrespective of whether modest stunting is linked to dysfunctions or not.

The weight norm is established on the notion that the individual can vary his or her weight within a range without increased health risks and that there is no ‘genetic potential’ weight. The latest FAO/WHO/UNU (1985: 183) expert committee has endorsed a study suggesting that a weight for height corresponding to a ‘body mass index’ (BMI) ranging from 20.1 to 25.0 for adult males and from 18.7 to 23.8 for adult females is consistent with health and unimpaired mental and physical capabilities. The weight norms used in the following have been derived from the lower ends of these ranges, with a slight upward revision for females because of the high incidence of pregnancies in the SSA. The weight norms used here are thus derived from a body mass index of 20.1 for men and 19.0 for women.

For small children (below 6), there is a very large number of anthropometric (p.164) studies. These usually report the proportion of children that fall below the norm for one or more of the three most commonly used anthropometric indicators: weight for height (wasting), height for age (stunting), and weight for age (combined wasting and stunting). The most commonly used weight for height norms are 80 per cent of, or two standard deviations from, the median in the NCHS or Harvard reference populations. The conventional height for age norms are 90 per cent of, or two standard deviations below, the median decile of reference populations.

Estimates: adults

The numbers shown in the first column of Table 5.3 give the average height of the various adult SSA populations as a percentage of the height norm (180 and 167 cm for men and women, respectively). The table reveals that there are enormous differences in actual heights across ethnic groups in the SSA. The 36 populations covered include the Dinkas (southern Sudan), the tallest people measured in the world. Also included are the Bunia pygmies (Zaire), the shortest people measured in the world. The average adult Dinka male, at 181.6 cm, is 25 per cent taller than the average Bunia male (at 145 cm). Contrasted to the Scandinavian norm, the African populations examined have an average height ranging from 82 (Bunia pygmies) to 101 per cent (Dinkas).

Table 5.3 Average height and weight of adults in selected African populations (% of height and weight norm)

Country

People/place/group

Ethnic origina

Year

Heightb

Weight for heightc

Botswana

Bushmen

(B/P)

1962

89

96

Bushmen

(B/P)

1970

88d

Chad

Sara/rural

(N)

1969

97

112

Sara/urban

(N)

1972

98

Ethiopia

Debarech

(N/H)

1969

93

104

Adi‐Arkai

(N/H)

1969

93

100

Gambia

1952

94

104

Kenya

Students

(B)

1961

91d

Samburu

(N)

1969

97d

88d

Malawi

Lilongwe

(B)

1972

92d

Bantu

(B)

1970

92

105

Mozambique

Recruits

(B)

1968

94d

105d

Namibia

Dama

(B)

1969

93d

Nigeria

Akufo/Yoruba

(S)

1970

93

107

Ibadan/well‐off

96

126

Ibadan/slum

94

108

Lagos

1970

94

118

Rwanda

Tutsi

(N)

1965

98

99

Hutu

(B)

1965

93

108

Sudan

Dinka

(N)

1963

101d

88d

Shilluk

(N)

1963

99d

91d

Nilo Hamites

(N/H)

1961

99d

114e

Tanzania

Hadra

(B)

1972

90

108

Bantu

(B)

94

Tanganyika

(B)

1961

91d

Kasanioja

(N)

1961

99d

Uganda

Students

(B)

1961

92d

Baganda/rural

(B)

1969

92

116

Zaire

Fulero

(B)

1965

88d

94d

Tutsi

(N)

1965

96d

92d

Congolese

(B)

1970

94

104

Twa pygmies

(B/P)

1972

89d

100d

Mbaiki pygmies

(B/P)

1967

84d

100d

Bunia pygmies

(B/P)

1962

82

98

Kasai

(B)

1964

93d

107d

Katanga

(B)

1964

91d

105d

aB/P = Bushmen/Pygmies, B = Bantu, N = Nilotic, N/H = Nilo Hamites, S = Sudanese.

bThe height norm is based on estimations of adults in the Scandinavian countries (180 cm for males, 167 cm for females).

cBased on a body mass index of 20.1 for males, and 19.0 for females.

dObservations of males only.

eObservations of females only.

Source: Eveleth and Tanner (1976: Appendix tables 44, 45, 77, and 78 (height and weight data)); FAO/WHO/UNU (1985: Appendix table 2.c. (weight norm)).

The average person in 19 of the 28 populations for which there are weight data in Table 5.3 (second column) is above the weight for height norm. As with height, there are considerable differences in weight for height across the samples. The Dinkas and the Samburies (in Sudan and Kenya, respectively) are the thinnest, at 88 per cent of the norm weight. The well‐off from Ibadan in Nigeria are the heaviest.

Estimates: Children

The results of anthropometric examinations of 23 random samples of children in 17 sub‐Saharan countries over the 1973–84 period are summarized in Table 5.4. The first criterion, height for age, gives an indication of the incidence of chronic undernutrition. In the seven countries for which these data are available, an estimated 16 to 28 per cent of all children (up to 5 or 6 years old) are stunted by US standards (10 per cent or two standard deviations below these norms). There are notable differences across the countries. Unfortunately, few surveys provide data on the prevalence of severe stunting (i.e. below 80 per cent of the reference median), but the two that do find it low.

Table 5.4 Percentage share of undernourished children according to selected anthropometric indicators, Africa, 1973–1984

Country (age or height group)

Year (season)a

Category

Size of sample

Height for age below 90% of norm

Weight for height below 80% of norm

Weight for age below 80% of norm

Reference norm usedb

USAID surveys

Cameroon (3–59 months)

1978 (Oct.–Apr.)

Rural

3,942

22

1

23

NAS

Urban

1,733

15

1

12

Total

5,675

22

1

21

NRGc

505

4

4

Lesotho (0–59 months)

1977 (—)

Rural

1,421

24

4

25

NAS

Urban

285

17

3

17

Total

1,706

23

4

22

NRG

293

11

5

6

Liberia (0–59 months)

1975–6 (AH)

Agr.

2,502

20

2

26

NAS

Non‐ag.

977

14

1

20

Total

3,479

18

2

24

NRG

285

9

3

13

Sierra Leone (0–71 months)

1978 (—)

Rural

2,937

26

3

32

NAS

Urban

1,943

14

3

24

Total

4,880

24

3

30

NRG

361

2

1

5

Swaziland (3–59 months)

1983–4 (BH)

Rural

3,475

17

0

12

NCHS

Urban

658

13

0

10

Total

4,133

16

0

12

Togo (6–71 months)

1977 (—)

Rural

20

2

16

NS

Urban

11

1

9

Total

19

2

15

NRG

Other surveys

Benin (0–59 months)

1976 (BH)

Nationwide

127

6

H

Botswana (0–59 months)

1978–81 (MH)

Nationwide

c.50,000

27

NS

Burkina (1) (0–9 years)

1973 (BH)

Sedentary

132

38

H

Migratory

43

49

Total

175

41

Burkina (2) (65–115 cm)

1974 (BH)

Nationwide

875

9

SM

Burkina (3) (0–71 months)

1978 (BH)

Nationwide

320

14

H

Chad (65–115 cm)

1974 (BH)

Nationwide

779

22

SM

Gambia (6–35 months)

1981–2 (DS)

Urban

6

NCHS

Kenya (12–47 months)

1977

Rural

c. 3,000

24

NS

1979

27

NS

1982

28

NS

Malawi (0–59 months)

1981 (DS)

Rural

32

NS

Mali (1) (65–115 cm)

1974 (BH)

Nationwide

625

11

SM

Mali (2) (0–59 months)

1976 (BH)

Migratory

208

9

H

Mali (3) (0–71 months)

1978 (BH)

Rural

122

15

H

Mali (4) (0–71 months)

1979 (MH)

Rural

249

6

H

Maurit. (1) (70–120 cm)

1973 (BH)

Sedentary

781

8

SM

Migratory

410

17

Total

1,191

14

Maurit. (2) (65–115 cm)

1974 (BH)

Nationwide

875

10

SM

Niger (65–115 cm)

1974 (BH)

Nationwide

774

11

SM

Senegal (0–71 months)

1979 (AH)

Rural

347

9

H

aThe following abbreviations for season have been used: BH (before harvest); AH (after harvest); MH (mid‐harvest year); DS (average for different seasons).

bThe standards applied are: NAS: National Academy of Sciences; NCHS: National Centre for Health Statistics; SM: Stuart Meredith; H: Harvard; NS: Not stated.

cNational Reference Group.

Sources: USAID (1978a: Tables 21–4) (Cameroon); USAID (1976: Table 50) (Liberia); USAID (1977a: Tables 36, 38, and 40) (Lesotho); USAID (1986: Tables 4.82–4.83) (Sierra Leone, Togo, and Swaziland); Kloth et al. (1976: Table 1) (Chad, Mali (1), Mauritania (2), Burkina Faso (2), and Niger); Greene (1974: Table on p. 1094) (Mauritania (1)); Benefice et al. (1981: Tables 5, 12, and 13) (Mali (2)–(4), Burkina Faso (3), Benin, and Senegal); IDRC (1981: 22) (Burkina Faso (1)); Tomkins et al. (1986: 536) (Gambia); Maribe (1984; Figs. 1 and 2) (Botswana); Chiligo and Msukwe (1984: 25) (Malawi); CNSP (1984: Table 1) (Kenya).

When it comes to acute undernutrition, as measured by weight for height, the picture looks unambiguously more favourable. As Table 5.4 shows, the prevalence of mild to modest (between 60 and 80 per cent of weight standard) acute undernutrition among children is only a few per cent in most countries. Not surprisingly, the incidence was significantly higher in the Sahel countries during the famine years 1973 and 1974; in these years also severe acute undernutrition (below 60 per cent of standard references) was found in a small (p.165) (p.166) (p.167) (p.168) percentage of children. The observations from Malawi and Botswana show relatively high prevalence of combined chronic and acute undernutrition as indicated by weight for age.

5.3. The conflicting evidence

In the preceding section, two different sets of estimates of the quantities of food that are consumed in the SSA countries on the average have been presented. First, there were the FAO estimates of the calories ‘available’ for human consumption on a per capita basis. Second, there were the estimates of the actual calorie intake in 85 different sample populations. If correctly measured, one would expect a reasonable congruence between the two sets of estimates.

In the previous section, two different sets of estimates on the prevalence of undernutrition in the SSA countries were also presented. The first was the estimates arrived at by the FAO and the IBRD at the very aggregate (country) level. The second was a large set of anthropometric studies of sample populations. In this section, the results obtained using the various types of indicators of food standards and the extent of undernutrition will be compared.

(a) Calorie availability vs. intake estimates

For the period 1960 to 1979, the FAO (1987) estimates the average per capita daily availability of calories in sub‐Saharan Africa as a whole at 1,964. The average per capita intake in the 85 sample studies consulted from the same period was 1,950. There is thus hardly any discrepancy between the FAO estimate, based on supply side, aggregate data, and the average of the sample estimates derived from demand side, disaggregated consumption survey data. The FAO further claims that if there is a bias in its estimates of the calories available for human consumption in Africa (and elsewhere), it is towards overestimation. This is because some of the available food is ‘wasted’. If this is correct, one would expect the FAO ‘availability’ estimates to be higher than those obtained through direct observation of food consumption in the respective countries.

The little difference there actually is between the FAO and the sample estimates is positive, but small, which may be interpreted to suggest that the ‘waste’ of the food available at the household level is minuscule. At first sight, one might be inclined to take this almost unbelievably close agreement between estimates obtained in completely different ways as proof of the robustness of the estimation methods. As will be evident in sections 5.4 and 5.5, however, such an interpretation is premature.

(b) Average height and weight vs. dietary intake

The average adult person in most of the samples of Bantu peoples was found to be 7–12 cm shorter than in the Scandinavian populations. This is an indication that as a child, the average Bantu person was deprived of food (although illness (p.169) can be an alternative explanation). However, the average adult in the SSA populations examined has a weight for height above the lower end of the range that is considered safe for health and physiological capabilities in the Western societies. On average, children in the SSA are also shorter (for their age) than Western children are and, more importantly, than children from well‐to‐do socio‐economic strata in the SSA countries themselves. They have, however, a weight for height that is relatively close to the Western children and well above (on average) what is considered safe for health.

If the current nutritional status is judged by weight for height, it thus seems that the average person in the SSA countries is at least somewhat above the level that is conventionally thought to imply undernutrition. This observation is not readily compatible with the FAO estimate, the dietary sample estimates, that the food ‘available for human consumption’, or actually consumed, corresponds only to about 80 per cent of the food needed to meet average calorie requirements (the RDAs). That is, if the average person has a weight for height above the safe level, one would expect that, by and large, food consumed is above what is needed to avoid undernutrition.

(c) Prevalence of undernutrition: FAO/IBRD vs. anthropometrics

Although the FAO and the IBRD use the same basic approach to estimate the prevalence of undernutrition in the SSA region, their models are different in two important respects. They use different cut‐off points to delineate the undernourished and they map the distribution of the calorie intakes from the distribution of income across households in different ways. These are also the easily observable explanations for the fact that the two organizations arrive at different estimates of the prevalence of undernutrition (see Beaton 1983). The main difficulty is to reconcile the estimated prevalence of undernutrition suggested by the FAO/IBRD methods with the anthropometric estimates.

The IBRD (1986) study claims that 44 per cent of the population in the SSA as a whole was at least moderately undernourished in the early 1980s, and that the situation then was no better than in the 1970s (Table 5.1). Dozens of sample studies from these years suggest that only 5 to 10 per cent of children were ‘moderately’ wasted, indicating acute undernutrition at the time of measurement. Only in a few cases (during years of famine) was the incidence of wasting among children greater.

(d) Can the conflicting evidence be reconciled?

It has been shown that anything from 5 to 44 per cent of the population in sub‐Saharan Africa is undernourished depending on what indicator and source are consulted. This diversity of results is, of course, highly unsatisfactory and, in a policy perspective, a more precise understanding of the underlying reasons is warranted. In the following sections, two hypotheses about the reason are to be tested. The first is that some or all of the indicators are derived on the basis (p.170) of models/methods and/or from data that are erroneous. It will be investigated whether the highest estimates of the prevalence of undernutrition, derived by the FAO and the IBRD, are built on a biased model and inaccurate data. A parallel investigation of the low estimates, based on anthropometric evidence, will also be conducted. The hypothesis is that after the various biases have been corrected for, the different indicators will show a less diverse picture. The other main hypothesis about why the different indicators, even after correction for ‘technical’ estimation biases, show different results is that they are derived from different notions of what constitutes undernutrition. To some extent this is simply due to the fact that different indicators aim to capture undernutrition that is more or less ‘severe’ in a single dimension, but this is not the whole story.

5.4. The FAO calorie availability estimates: errors and biases

In assessing the various pieces of evidence on food standards and undernutrition in the SSA, the FAO calorie availability estimates are central. First, these estimates indicate in themselves that the nutritional situation in the region is very serious. A per capita calorie availability of 1,856 (in the mid‐1980s) is only 80 per cent of what the FAO itself, and also the WHO and UNU, deem is required to feed everyone appropriately even if the food were allocated according to needs. Second, the importance of the FAO calorie availability data is enhanced by the fact that they underlie both the FAO and the IBRD estimates of the prevalence of undernutrition in Africa (and elsewhere).

(a) Estimation difficulties

Anyone trying to estimate the food produced and the food available for human consumption in the African countries will face formidable problems and costs. This is for a variety of reasons. Most African countries are large, sparsely populated,2 and span a multitude of climatological and cropping systems. A high proportion of the population derives its livelihood from agricultural activities and much of the food is produced for subsistence. Many peasants pursue mixed farming, i.e. both crop production and livestock holding. The number of minor crops is usually very large (Eicher and Baker 1982).

The problems with estimating harvested area for cereal crops, the dominant food, are much more severe in Africa than in most other parts of the world. First, few of the countries have land records, and the ones that exist only cover the most densely populated regions where cash crops dominate. In other parts of these countries, land is usually owned, not privately, but communally, and production is mostly for subsistence. Furthermore, since slash‐and‐burn (p.171) shifting cultivation is still an important mode of production in many parts of Africa, area harvested is difficult to define even with the best measurement technology. The harvested area also tends to vary significantly from year to year in response to rainfall and other natural vagaries.

The problems involved in estimating harvested area for roots and tubers, the most important food crops in many SSA countries, are especially great. This is because cassava, the major product, is grown in patches and not always harvested annually. Cassava can remain in the ground for up to three years without much loss in nutritional value (although the digestibility of the product deteriorates). The bulk of cassava production is for subsistence and the reason for ‘storing’ it in the ground is usually to even out yearly fluctuations in household access to food so as to improve food security.

Yields are also more difficult to estimate in the non‐commercial agricultural sector in Africa than in most other places. The climatological and soil conditions tend to vary sharply from region to region and the variety of cropping systems is enormous. In some areas, there is mixed cropping, in others multiple or continuous cropping, and also crop rotation (Eicher and Baker 1982). Under such circumstances, very refined and costly measurement methods are needed if reliable yield estimates are to be obtained. The number of sample cuts at each point of time has to be very large and the sampling has to be repeated several times over the year.

With practically no base data on acreages and yields derived with modern, scientific methods, it is, of course, impossible to produce reliable direct estimates of staple food production. Consequently, almost everyone who has taken a closer look at the FAO (and other) estimates of food production in the region comes to the obvious conclusion: they cannot be trusted. Lipton (1986: 3–4) goes as far as saying that, even for the main staples in the four largest countries (Nigeria, Zaire, Ethiopia, and Sudan), ‘we have no idea of the levels or trends in output or consumption … over the past 5–20 years’ (see also the other papers contributed to the two recent conferences on food output statistics in Africa: FAO 1985b and EEC 1986). According to Lipton's assessment, the available output estimates of main staple crops by small farmers are subject to unknown errors of at least plus/minus 20–40 per cent. When it comes to ‘minor’ crops, which taken together are important in many parts of Africa, nobody has offered even a guess as to the order of magnitude of the estimation errors that beset the FAO estimates.

In the absence of reliable production data, the FAO derives indirect estimates of the ‘availability’ of main vegetable crops, based mainly on the food that is marketed in each country. At least until recently, this meant a government marketing parastatal (see Ahrin et al. 1985). On top of that, a rough allowance is usually made by the national agency or the FAO itself for ‘subsistence’ production. The problem with this indirect method of estimating food production is that only major cereal crops are usually sold outside the village or district; and of these, often only a small part (20–40 per cent in some samples) (p.172) is marketed through official channels (Eicher and Baker 1982: 48) and (although imperfectly) measured. In the parts of rural Africa where minor cereal crops and various roots and tubers play an important role in the diet, the measurement problems are especially acute. To estimate accurately the food production that is not marketed (the greater part of total food production) is impossible in Africa as things stand.

(b) Estimation biases

Considering the large margins of error that beset not only the food production data, but almost all the stages in the long chain of estimates from food production to calorie consumption (see Svedberg 1991b: ch. 5), one cannot but conclude that the unreliability of the existing (FAO) calorie availability estimates is great indeed. The FAO itself admits that some of its estimates are incomplete and/or not totally reliable.

When it comes to the possibility of biases, however, the FAO (1984) stresses that ‘it is important to note that the quantities of food available relate to the quantities reaching the consumer but not necessarily the amount of food actually consumed, which may be lower than the quantity shown, depending on the extent of losses of edible food and nutrients in the household, e.g. during storage, preparation and cooking … plate‐waste or quantities of food fed to domestic animals and pets, or thrown away’ (italics added). That is, if there is a tendency to biases in the availability estimates, the FAO claims that it is towards overestimation.

The possibility of household wastage is the only source of bias that is explicitly discussed by the FAO. There is, however, a whole range of other possible biases in their estimates, most of which seem to be towards underestimation rather than overestimation. Given that so little alternative and reliable data can be found, it is difficult to arrive at a firm assessment of the extent of underestimation, but there are some indications. There is also a downward bias (admitted by the FAO if one reads the small print) arising from incomplete coverage of food sources (see below). Let us start, however, by discussing a few a priori reasons why there probably is a net downward bias even in the items actually covered.

Implications

If one accepts Lipton's (1986) assessment that the margin of error for main staples produced by smallholders in the SSA is plus/minus 20–40 per cent, is an overestimate of, say, 30 per cent in the production of ‘main vegetable food’ as likely as an underestimate by the same amount? If the production of main staples (including cassava) is overestimated by 30 per cent, the per capita ‘availability’ of (all) calories has to be adjusted downwards (ceteris paribus) by 15–20 per cent. We are then down to per capita calorie availability figures of 1,500 in the ‘typical’ SSA country. Such a figure implies that between half and three‐quarters of the population has an intake below 1.2 BMR (equal to a per capita intake of 1,500 calories). (This follows from the FAO/IBRD (p.173) model that is referred to in section 5.6e below.) The implication would be that almost half the African population has an intake which probably no human biologist or nutritionist would say is enough for biological survival, much less economic survival in the African context. We can thus be fairly sure that if there is a bias, it is not towards overestimating production of staple crops by 20–40 per cent. This, however, does not prove that there is a bias in the other direction, i.e. towards underestimation, but there are other indications to that effect.

Incentives

One reason to suspect that the estimates of staple food production in Africa are downward biased is that there are incentives for underreporting at all levels. The smallholder farm sector will certainly gain in a number of ways from giving the official national authorities the information—through whatever channel—that its production is lower than it actually is. First, in many African countries, until recently, trade in main staples has been more or less strictly monopolized by a government trade board, and the prices paid to the farmers have often been below those in the parallel, unofficial market (see Bates 1981; Eicher and Baker 1982; Ahrin et al. 1985). The incentives to sell on the unofficial, non‐registered, and sometimes illegal markets have thus been strong. Second, to the extent that farmers and pastoralists pay taxes, the taxable incomes are related to what they produce and a way to reduce the tax burden is to underreport production, whether of crops or livestock. Third, in order to qualify for government (input) subsidies of various kinds, it may be advantageous to give downward biased information on productivity.

The national government agencies that supply the FAO with base data also have an incentive to keep food production figures down in order to attract more food aid (and aid in general). The FAO itself has no incentive to overestimate food supplies; the organization's existence is largely based on the notion of severe food problems in the underdeveloped countries. One does not have to go as far as saying that there is a systematic and explicit falsification of data, either at the national or at the FAO level. One need only think that underreporting on behalf of the bureaucracies ‘reflects nothing more than the persistence of honourable men attempting to dramatize their case through exaggeration’ (Poleman 1977: 387). Considering the unreliability of the base data, there is always scope for choosing ‘low’ numbers within the confidence intervals without violating conventional practice.

Biased methods

There are reasons to think that the present estimates of acreages in the SSA are based on methods that generate downward biases in the estimated production of ‘major vegetable food’, the most important food category according to the FAO. In the absence of complete land records, the responsible national government agencies tend to register only the food crop land that is the easiest to identify (e.g. by ocular observation). Small fields in remote and non‐accessible areas tend to be incompletely covered. The (p.174) experience from India in the 1950s and 1960s shows that when complete land records replaced the earlier estimation methods, significant revisions of the cultivated acreage followed (Zarkovich 1962). The experience from India further suggests, however, that the introduction of more sophisticated yield‐sampling methods did little to improve the reliability of the production estimates.

Incomplete coverage

A further source of downward bias is that the ‘calorie availability’ estimates are derived from data on food supplies that are incomplete—in several respects. The data on ‘major’ food crops from national sources usually cover only part of what is actually produced; mainly the part that is marketed through official channels. Most livestock censuses do not include small domestic animals like pigs, sheep, and goats. The FAO admits that its estimates of some of these items are based on ‘incomplete’ coverage, which means underestimation. Chicken is underestimated according to the FAO itself. The coverage of various minor roots and tubers is incomplete and so are the data on domesticated fruits, berries, nuts, and honey, as explicitly admitted in the small print of the FAO reports. Food like game meat (especially from small animals), invertebrates, and undomesticated fruit, nuts, berries, roots, green leaves, and other wild flora are not included at all, or very incompletely so.

(c) The size of the bias

The above assessment of the estimates of calories ‘available for human consumption’ in Africa, as produced by the FAO, has not permitted us to say exactly by how much the estimates are downward biased. We have mainly been concerned with the domestic food production statistics which are used to estimate, with the help of various conversion techniques, the amount of food available for human consumption. This does not mean that all other links in the long chain of assumptions that the FAO is forced to make to estimate food calories are free from ambiguity (see Svedberg 1991b: ch. 5).

The quantitatively most solid evidence of underestimation stems from the incomplete coverage of ‘minor food items’. Adding up the gaps relating to the minor food items for which we have found it possible to say at least something in quantitative terms implies an underestimation of the total supply of calories by about 10 per cent on the average. When it comes to main staple food, the unreliability of the base data is large indeed, but little quantitative evidence on biases exists. There is one such indication, however. The incomplete agricultural land (acreage) enumeration in Africa probably means underestimation of staple food production. Considering the large share of the total supply of calories accounted for by vegetable staple food, an underestimation of acreage and, thus, production, by as little as 10 per cent would imply an additional 3–4 per cent underestimation of total food (calorie) consumption.

(p.175) 5.5. The dietary evidence: errors and biases

In the preceding section, several reasons to expect the FAO calorie availability estimates to be downwards biased were discussed. If this is the case, it must be that the perfectly matching sample estimates of actual calorie intake in the SSA are also downward biased. In the following we shall investigate this issue. The estimation of the habitual food intake in a population entails two general problems. The first is to find a representative sample of households at a representative time. The second is to get a complete and accurate coverage of the food actually consumed.

(a) Adverse selection

It is difficult and expensive to obtain a correctly stratified sample in countries where there are large inter‐ and intra‐regional differences in factors affecting household food consumption. It is thus not surprising that the sample populations in most of the dietary studies for the SSA are not representative. This is for several reasons.

First, in most cases, the intention was simply not to obtain a random sample. In a majority of the studies, the focus is on a specific rural population group that was identified as having nutritional problems before the examination; this being the very reason why the investigation was carried out in the first place. The sample population is unambiguously representative for the national population only in four out of the 51 studies covered in Dillon and Lajoie (1981). Schofield (1979: 11) does not discuss the representativeness of the samples in the African studies in her survey in any detail, but she notes that, ‘in general…investigations are restricted to small, unrepresentative samples’. All this means that the intake estimates derived for these unrepresentative groups must be lower than the national average for the respective country. How much lower cannot be ascertained, but if a population group is at a nutritional disadvantage, one would presume that for this to be detected in the first instance, the ‘disadvantage’ must correspond to more than a few per cent.

Second, almost all the dietary observations are from rural areas; urban and peri‐urban areas are seldom studied. The available anthropometric evidence suggests that food standards in urban areas are significantly higher than in rural ones throughout Africa (Svedberg 1987: Table 4). The sample dietary estimates from rural populations thus probably understate national averages, as between 15 and 50 per cent of the population in the various SSA countries dwell in urban and peri‐urban areas.

Third, there is the problem of obtaining estimates that are representative in the time dimension in countries with large intra‐year (seasonal) and inter‐year variations in food consumption. In 18 of the 51 studies from the Sahel countries surveyed in Dillon and Lajoie (1981), information is given on the time of the year when the investigation was carried out. In 16 of these 18 cases, the study was conducted in the pre‐harvest, lean, dry season (‘soudure’); in the (p.176) remaining two, shortly before that period. There is plenty of evidence showing large intra‐year variations in per capita calorie consumption in countries with marked seasonality in agriculture (the majority of the countries in the SSA). The difference between the pre‐ and post‐harvest months amounts to several hundred calories according to observations from West Africa (see Schofield 1979: 53–4; Hulse and Pearson 1981: Table 6, 7; Chambers et al. 1981: 45–50; Rosetta 1986; Tomkins et al. 1986; von Braun 1988). Moreover, it seems that most studies have been carried out in a below average year. It is notable, indeed, that three of the four studies based on otherwise representative samples listed in Dillon and Lajoie (1981) were conducted in the Sahel during the famine in 1972–4, clearly not representative years.

(b) Biased estimation methods

The most reliable dietary estimation method is to survey the individuals continuously (over several days at repeated intervals over the year) and weigh the equivalent of all the different food items they eat. The most reliable estimates of the nutritional content of the (estimated) food intake are obtained through mechanical and chemical decomposition of the (equivalent) food. These methods are very costly, however, and in most instances interviews and standard conversion tables are used.

Less than half of the 68 dietary surveys from the SSA covered by Schofield (1979: Table 4.1) are based on the food ‘weighing method’. The majority of these surveys rely on stocktaking, recalls, or qualitative assessment. The main disadvantage with the stocktaking method is that it usually covers only main meals consumed at home; snacks and away‐from‐home meals are automatically excluded. The recall method entails two main problems. First, when interviewed, people tend to forget minor items consumed and/or snacks in between meals. The second problem is that when people are asked about their food consumption habits, they are inclined to provide the information they think the investigators would like them to give, or what they think would benefit themselves. This may mean that poor people tend to ‘talk a good diet’, i.e. to exaggerate their food intake, being ashamed of their deprivation. There are thus two conflicting biases in the recall method.

Cross‐checking of results obtained by recalls with those obtained through weighing of the food consumed by sample populations in India suggests a net downward bias in the former, however, ranging from 10 to 40 per cent. Only in some recent studies has the underestimation been less.3 It is notable that the 85 dietary studies from Africa referred to above are all of the not‐so‐recent type (pre‐1979).

There seem to be three reinforcing reasons why many of the sample studies of per capita calorie consumption in Africa show figures that are not representative of the long‐term situation for the population of the respective countries. (p.177) In many cases, the sample comprises (1) a rural group that (2) was known a priori to have nutritional problems (3) during the lean season, or in a particularly poor year. Moreover, most of the sample studies have been conducted with (4) methods that have been shown to produce notable underestimates of food intake. This is a further indication that the FAO per capita calorie ‘availability’ estimates for sub‐Saharan Africa are too low in general, since these estimates are in very close agreement with the ones derived from sample studies of intake.

5.6. The FAO/IBRD estimates of undernutrition: errors and biases

The FAO and IBRD estimates of the prevalence of undernutrition rest on three presumptions that can all induce errors and biases in the results. The first is that the ‘food available for human consumption’ in the region is that claimed by the FAO. As we saw in preceding sections, there are several good reasons to expect these estimates to be downward biased. An underestimation of the food available will unambiguously imply (ceteris paribus) that the prevalence of undernutrition in the region is overestimated. However, there is also reason to think that there are biases in the other two exogenous parameters in the IBRD and FAO estimation models: the calorie requirement norm and the function that distributes the ‘available’ calories across households.

(a) Biases in the calorie requirement norms?

Through the years, the FAO/WHO has constantly been criticized for providing recommended daily allowances (RDAs) of calories that are too high. And over the years, the FAO has repeatedly adjusted its RDAs downwards. Several nutritionists seem to be of the opinion that, if at all useful for identifying undernutrition in a population, the latest RDAs (FAO/WHO 1973)4 are still biased on the high side (e.g. Mayer 1976; Poleman 1977). However, the notion that they are ‘too high’ can have at least two very different meanings.

One meaning is that the FAO/WHO estimates of the energy expenditure for specific internal or external activities are too high. The other is that the organizations have derived their estimates on the basis of too high a level of overall external physical activity. The first problem is basically one of deriving appropriate ‘technical’ coefficients for various types of activities in the human body. The other problem is partly normative and relates to the question of how to conceptualize and define undernutrition. In what follows, we shall be concerned with the first set of ‘technical’ problems. The normative discussion of the overall level of external physical activity that the requirement norms should allow for is postponed to section 5.8.

(p.178) Let us first discuss the lowest norm used by the two organizations, i.e. 1.2 times the BMR. This norm is derived by the FAO on the presumption that an individual should (1) cover his or her energy expenditure for internal body processes and (2) the minimum of external activity needed to maintain basic health, (3) after his or her body's metabolic rate has ‘adjusted’ to a permanently low intake (by becoming more efficient).

Is it possible to claim that the 1.2 BMR norm, which corresponds to only 65 per cent of the RDAs suggested by the FAO/WHO, is too high in any ‘technical’ sense? The 1.2 BMR norm does not allow for any physical activity beyond the minimal movements involved in sitting up for short moments, to dress and wash, etc. It is presented as a ‘baseline biological survival norm’. In one respect, however, this norm may be too high. The FAO/WHO estimate energy expenditure for BMR on the basis of a linear model with a positive intercept and only one independent variable: body weight. Many nutritionists have argued that a quadratic relationship between BMR and body weight has a better theoretical underpinning and fits the data better. With such a model, the estimated requirements for BMR per kilo of body weight will be some 10 per cent lower (with the same data set) for the people with relatively small bodies (Payne 1987).

There is thus one possible reason why the 1.2 BMR norm is ‘too high’. It is ‘too low’, however, for other reasons. First of all, it is based on the assumption that the human body has the ability to adjust its energy needs for basal metabolism and external activity to low intake within a substantial range without any ‘costs’. Although an increasing number of nutritionists seem to accept the notion of ‘intra‐individual adaptation’, it is still a controversial issue that has yet to be corroborated by empirical evidence.5 Second, this requirement norm does not allow for any physical work activity at all; it is a biological, ‘short‐term‐survival’ requirement norm. In a world where people have to expend energy in work in order to be entitled to food, the economic survival requirement must be set higher. Even if the ‘true’ requirement for BMR is some 10 per cent lower than purported by the FAO, the calories thus ‘freed’ do not allow for much work activity. On balance, it seems that the FAO lower norm is downward biased in a context where almost every (adult) person is engaged in physical agricultural work with few supplementary factors of production. The higher FAO norm, at 1.4 BMR, differs from the lower one only in so far as it does not allow for intra individual adaptation to low intake.

The calorie requirement norms used by the IBRD in its estimation of the prevalence of undernutrition in the SSA rest on the same assumptions about energy expenditure for BMR as the FAO ones do. However, the IBRD calorie requirement norms differ from the FAO norms in two important respects. The first is that both the IBRD norms are built on the assumption that there is no mechanism in the human body that ensures that the basal metabolism becomes (p.179) more efficient in periods of nutritional stress. The second is that the two IBRD norms allow for various amounts of physical external work activity.

The allowance for external physical activity in the form of work raises the question of how much work should be allowed for; this issue is discussed in section 5.8 below, dealing with the conceptualization and definition of ‘under‐nutrition’. However, the inclusion of an allowance for work activity, at whatever level, also raises the question of possible biases in the estimated conversions from a specific activity to the energy expenditure (requirement) involved. On that issue there is very little empirical evidence from actual field studies. There are scattered observations, however, suggesting that there has been a tendency for the FAO/WHO to classify many common agricultural tasks as ‘too heavy’ and, thus, too energy consuming (see Lawrence et al. 1985: 759).

(b) Biases in the calorie distribution estimates?

The third exogenous parameter in the FAO and IBRD estimates of the prevalence of undernutrition in the SSA region is the assumption of how the distribution of the calories within each country is determined. A fundamental fact is that neither the FAO nor the IBRD has any direct empirical knowledge about this. Their mapping of the distribution of the ‘available’ calories across households from the perceived distribution of incomes is based on a theoretical model that may not be of much relevance (see section 5.6d). But even within the confines of this particular model, very strong quantitative assumptions are made on the basis of weak and conflicting stylized facts.

A priori, it is quite clear that even the slightest difference in the assumption about how the available calories are distributed can have huge effects on the estimated proportion of the population that falls below the ‘requirement norm’. As an illustration, suppose that we accept the IBRD low cut‐off point (at 0.80 RDA = 1.5 BMR), corresponding to a per capita calorie requirement in the SSA as a whole of 1,856. Assume also that the FAO estimate of a per capita availability in the SSA in 1983–6 of 1,876 is correct. It is then theoretically possible that no one in the SSA was undernourished in these years. This would be the case if calorie intake were distributed according to individual expenditure requirement.

With a slight alteration in the assumed distribution, however, 90 per cent of the population in the SSA may turn out to be ‘undernourished’. That would be the case if 10 per cent (say the ‘urban rich’) had a per capita calorie intake of 3,000, and the remaining calories were distributed equally among the rest of the population. The whole of the latter group would then be ‘undernourished’ (with a per capita intake of 1,750 calories, or 9 per cent below the norm). These are just two out of many conceivable examples, but they underscore the basic point that the FAO/IBRD method is very sensitive to the assumed distribution of calorie intake, especially when per capita ‘availability’ is close to per capita ‘requirement’.

(p.180) (c) The FAO/IBRD calorie distribution model

The assumptions made by the FAO and IBRD regarding the distribution of the ‘available’ calories comply with the conventional wisdom of the mid‐1970s. The additional knowledge we have today is not very extensive, but taken together, it clearly suggests a more ‘even’ distribution of calorie intake in general than assumed by the two organizations. This is so whether or not the basic FAO/IBRD method of estimating the calorie distribution is accepted or not. Within the confines of the FAO/IBRD model, however, three sets of assumptions can be questioned.

The first concerns the assumption about how incomes are distributed in Third World countries. A re‐examination of the ‘old’ studies of income distribution in Africa (on which the IBRD/FAO estimates of the prevalence of undernutrition rest) has shown that these tend to underestimate the income of the poorest groups and, thus, to overestimate the maldistribution of income (van Ginneken and Park 1984). With the techniques used by the FAO and IBRD to estimate the prevalence of undernutrition, starting off with a higher share of income going to the poorest means (ceteris paribus) that calories also become more evenly distributed.

Second, the IBRD and FAO estimates of the prevalence of undernutrition are built on the assumption that the distribution of the available calories across households is a one‐to‐one transformation of the distribution of income. The general finding today is that income variations explain a very low share of the variation in calorie intake (as measured by the adjusted R 2). This may be for various reasons, including faulty econometric estimation techniques and white noise in the data (see Bouis and Haddad 1988). It is interesting to note, however, that in a recent study, based on the state‐of‐the‐art technique to estimate the conventional ‘calorie‐as‐a‐ consumption‐good’ model, the share of the variation in calorie intake that is explained by income is not even statistically significant (Behrman and Deolalikar 1987).

Third, still within the confines of the FAO/IBRD model, there is the question of the size of the income elasticities of calorie intake. The IBRD assumes that the calorie‐income elasticity for all African countries is 0.15 at the level of fulfilled requirements. For the lowest income groups, the elasticity is set at 0.55. (The FAO uses a slightly different method to map calories from incomes, but, in effect, the two versions are very similar; see Beaton 1983.) Regarding the first elasticity, we have to ask why it should be at all positive for people with a calorie intake at or above their requirements. If a person's calorie requirements (expenditure) are met, why would he or she indulge in more calories? It has to be recalled that a calorie intake above expenditure (requirement) does not disappear into thin air. By the first law of thermodynamics, or the Atwater formula in the nutrition context, calories consumed over and above those expended in physical activity will accumulate as fat.

The literature aimed at the estimation of calorie–income elasticities in the (p.181) poor(est) population segments, where intake supposedly are below requirements, has reached very conflicting results. The elasticity estimates range from −0.30 to +1.18 (see Bhargava 1988; Bouis and Haddad 1988). It seems, however, that the IBRD choice of parametric value for the calorie–income elasticity in the low income range, 0.55, is considerably higher than the estimates thought to be appropriate today, which are in the 0.10 to 0.15 range (Bouis and Haddad 1988).

(d) An alternative model of calorie distribution

So far we have discussed the conventional calorie ‘consumption’ model that underlies the FAO/IBRD estimates of the calorie intake distribution. We have also shown that the predictions of this model have not been easy to reconcile with empirical evidence. The conventional explanations of the weak results in the empirical literature are poor estimation techniques and weak data. An alternative interpretation is that the underlying theoretical model is misspecified. In that model, income is the main determinant of the demand for calories, which is considered an ordinary consumption good that enters the individual's utility function. In an alternative theoretical model (Svedberg 1988), it is assumed that calories are intermediate goods that enter the individual's production function.

In the ‘production’ model, the level at which a person's calorie intake and expenditure balance is not directly causally related to his or her income. In fact, both calorie intake–expenditure and income, as well as body weight, are endogenous variables. The exogenous variables that determine the individual's calorie intake–expenditure are a set of biological (e.g. height and metabolic efficiency) and economic characteristics (endowment of factors of production, technology, prices, etc.). These exogenous ‘characteristics’ tend to show large variations across individuals of the same sex and age.

The prediction of the production model is thus not that calorie intake is monotonically and positively associated with income. This model would rather lead us to expect that the highest‐income earners in a Third World population have a relatively lower calorie intake than the middle‐income earners because they expend less energy in physical work. The richer members of the population are usually not engaged in heavy manual work activities, and they tend to buy a higher share of the labour‐demanding services (such as water and firewood transportation) than the average person. Still, the individuals in developing countries that are permanently undernourished (by anthropometric or clinical standards) belong to the poorest income groups almost without exception. On the other hand, far from all individuals in the lowest‐income decile have an anthropometric status or show medical signs that indicate undernutrition.

What the production model predicts is that for the population in any one country, and in the middle‐ and high‐income ranges in particular, other factors than income determine the level at which calorie intake and expenditure (p.182) balance. If we accept that calorie intake does increase monotonically with incomes, the crucial question is to what extent the individual‐specific ‘desired’ expenditure is met by individual‐specific intakes. The FAO assumes that, within each income class, there is a perfect correlation between individual requirement and intake. Inserting this assumption in our type of model, where there are no ‘income classes’, while retaining the other assumptions, would lead to the conclusion that there is no undernutrition at all in the SSA by the standards suggested by the FAO.6

(e) Margins of error and biases in the FAO and IBRD estimates

In an important article, Beaton (1983) tried to disentangle the reasons why the FAO and the IBRD, both using the same basic estimation approach, arrive at such different estimates of the prevalence of undernutrition in various parts of the world. Beaton saw two main possible explanations for the divergencies: one was that different ‘requirement norms’ were used; the second that different assumptions were made about how the available calories are distributed. By combining the different assumptions made either by the FAO or the IBRD on these two accounts into new constellations, Beaton demonstrated that anything between 17 and 80 per cent of the population in a country like Tanzania can be estimated to be undernourished (Beaton 1983: Table II). If, on top of that, one allows the per capita availability of calories to vary by 10 per cent above and below the FAO figure, the range of estimated prevalence of undernutrition becomes even wider (8 to 90 per cent of the population; see Svedberg 1991b: ch. 6).

One major argument advanced above is that the FAO and IBRD estimation models are not only extremely sensitive to small variations in the size of the exogenous parameters; there also seem to be biases in most of the exogenous parameters. The FAO has used per capita requirement norms that are too low, while those used by the IBRD are too high. However, both organizations have based their estimates on the FAO per capita calorie availability data, which are biased on the low side. Moreover, both organizations have assumed that the ‘available’ calories are more unevenly distributed than predicted by modern theory and empirical investigations. On the whole, both organizations have arrived at figures on incidence of undernutrition that are upward inflated even when ‘undernutrition’ is defined in the very broad sense of the IBRD and the very narrow sense of the FAO (cf. section 5.8 below).

(p.183) 5.7. The anthropometric evidence: errors and biases?

In the previous section we have shown there are ‘technical’ measurement errors in the FAO and IBRD aggregate estimates of the prevalence of undernutrition in the SSA which, on balance, tend to overstate the seriousness of the situation. This claim is consistent with the earlier observation that, when anthropometric measures are consulted, the prevalence and severity of undernutrition in the region is considerably lower than purported by the FAO and IBRD. In this section we shall investigate the possibility that the anthropometric indicators themselves are biased in one direction or the other. There are two potential sources for ‘technical’ biases in the anthropometric estimates reported in section 5.2 above. One is that the sample populations are not representative of the SSA population as a whole. The second is that the anthropometric norms that were used to estimate relative height and weight status are not appropriate in the SSA context.

(a) Representative samples?

There is the possibility that the samples shown in Tables 5.3 and 5.4 are not representative of the African populations. First, the individuals examined in a given sample are not necessarily representative of the particular ethnic group to which they belong. Second, the 36 samples of ethnic groups may not be representative of the population in the SSA as a whole.

Unrepresentative individuals

That the individuals examined are unrepresentative for the particular population group to which they belong is quite likely in some of the studies of adults. In two of them, from Kenya and Uganda, the samples comprise university students. Since students tend to come from relatively well‐off families, their height and weight are probably above those of the average person in their ethnic group. In the study from Mozambique, the sample consists of army recruits. The average height of military personnel is sometimes above that of the base population since recruits with a height below a specific norm are rejected. The recruits’ weight for height may also be above average because army personnel is sometimes a priority group in terms of food allocation in Third World countries. In the remaining 33 samples, there is no indication that the individuals are unrepresentative. In the USAID surveys listed in Table 5.4, it is explicitly claimed that the children were selected with random methods; there is little that can be done to check whether random samples were achieved in practice. In some other surveys, there was probably an ‘adverse selection’ of children, e.g. when the sample comprises children brought to a health clinic.

Unrepresentative samples

According to a recent inventory, there are more than 1,000 ethnic groups in sub‐Saharan Africa (Oliver and Crowder 1983). (p.184) The question is thus whether the relatively small number of samples in Tables 5.3 and 5.4 are representative for all ethnic groups. Five of the 36 adult samples comprise Bushmen and Pygmy populations. Considering that these make up less than 1 per cent of the total population in SSA, they are clearly overrepresented. Whether the 15 Bantu populations and the 8 samples from Nilotic groups are representative is impossible to say, but there is no reason to believe that the picture is severely distorted. The variance across the Bantu samples is quite small and the same applies to the Nilotic ones. Most of the studies of children have a national coverage that ought to ensure representativeness.

(a) Biased height norms?

The use of Scandinavians as the reference population for height of adults in sub‐Saharan Africa poses two main problems. One is concerned with the genetic potential for growth in stature; the other with the existence of other, non‐nutritional, constraints on growth.

Differences in genetic potential

The height norm used above was based on the assumption that the genetic potential for growth of Africans is the same as for Caucasians. This is what most contemporary human biologists and geneticists claim (see Eveleth and Tanner 1976 and Roberts 1985). The standard method used to test this notion is to study children from families for which adequate nutrition and health care is not a problem and compare them with ‘sound and healthy’ children in Caucasian populations. More than a dozen examinations of preschool children from well‐to‐do families of Bantu stock have been conducted within the past two decades. With few exceptions, these children (0–6 years) are shown to have the same average height (and weight) as those given in the National Centre for Health Statistics (NCHS) height and growth charts (the most commonly used ones).

Since well‐to‐do Bantu children and adolescents have approximately the same height for age as their Caucasian counterparts, one would expect them to achieve the same final height. However, one cannot be certain because there are variations in the age at which growth terminates. The standard method for checking the genetic potential for final adult height in nutritionally constrained populations is to sample individuals who have lived for a generation or two in places (often abroad) where there are no such constraints. With this method, it has been shown that Blacks living in the US and the UK have the same average stature as the Caucasian population of the respective country (Eveleth and Tanner 1976).

All African ethnic groups are sometimes lumped together into one single ‘race’, which would be to say that they all have the same potential for growth in stature. At face value, this would imply that the enormous difference in de facto average height between individuals from Nilotic and Bantu ethnic groups, respectively, is phenotypically explained. That is, the differences have nothing (p.185) to do with genes; small (or great) stature is solely the outcome of external factors, such as nutrition and the health environment.7

No tests of the relative influence of genes and environment on the average height of Nilotic populations have been made, it seems, and only one from a Nilo‐Hamitic population. This sample comprises privileged children in Ethiopia, whose height was found to be very similar to Caucasian children (Eksmyr 1970). Therefore we do not know whether the potential for growth in Nilotic peoples is the same as for Bantus and Caucasians. (The ancestors of most of the US Blacks came from West Africa.)

The fact that the actual height of the Nilotic peoples significantly exceeds that of the Bantu peoples, and is at par with the tallest people in Europe, is intriguing. One theoretically conceivable explanation is that they do not face environmental constraints on achieving their full potential for growth. This hypothesis is not altogether convincing, however, considering the fact that the Nilotic peoples often share the environment with significantly shorter Bantu tribes (e.g. the Tutsies and the Hutus in Rwanda). A plausible hypothesis is thus that Nilotic peoples have a larger potential for growth in stature than Bantu peoples and Caucasians (while it is smaller for Bushmen and pygmies).

Nutrition vs. other constraints

The second problem with the Scandinavian height norm is the underlying assumption that an elimination of undernutrition would ceteris paribus bring the African populations up to these height standards. This would probably not happen, because there are other external constraints on the African populations' ability to fulfil their potential for growth in stature. Infectious diseases are the most obvious example, but there is a whole range of social and economic factors that correlate with actual height in relatively well‐off populations in Europe and America (Floud 1987). The elimination of undernutrition, without doing away with the other constraints would, presumably, raise the average height of the Africans, but not to the Scandinavian standards. (And the elimination of all other constraints than undernutrition would also raise average height but, again, not to the Scandinavian norms.)

In this perspective, the average height of Europeans at a time (1950–70) when undernutrition must have been almost non‐existent in Europe, but when large sections of the populations were still not achieving their full potential for growth because of other constraints (Floud 1987), is perhaps a more appropriate norm than the Scandinavian one. The average height of males and females from the various SSA populations reported earlier (male and female average) (p.186) have been reproduced together with the average heights at that time of the adult populations in most of the European countries in Fig. 5.2.

5 Undernutrition in Sub‐Saharan AfricaA Critical Assessment of the Evidence

Fig. 5.2. Mean height of adults in Europe and Africa (cm)

When it comes to males (for whom there are more observations than for females), the picture is quite clear. What we see is that the average height of the average man in the Nilotic populations (see Table 5.3 for identification) is the same as for males in the European countries with the tallest people (Scandinavia, the Netherlands, and the UK). In all the seven Nilotic populations, the average male is taller, by 3–4 cm, than the average male in the average European country. However, the average male in the various Bantu populations, at 167 cm (ranging from 164 to 169 cm), is 5–6 cm shorter (3–4 per cent) than the average European male. The picture is roughly the same for females.

(c) Biased weight norms?

The lower end of the ‘safe range’ of weight for height that was used as the norm in the above estimates has been derived from observations in developed countries. There are two main reasons to question this approach. One is that there may be genetically determined differences, not only in stature, but also in body build across ethnic lines. The other is that the ‘safe’ weight range may differ across environments.

Genetic differences

The very tall Nilotic populations are relatively thin and have a body mass index below the norm in half the samples. It would not be wise, however, to conclude that the Nilotic peoples have, on the average, a body weight for height that implies inflated health risk. It may simply be that these people are not only exceptionally tall, but also have a more slender (linear) body build for genetic reasons. Since there is a whole line of genetically determined differences in body composition and shape across ethnic groups (see Eveleth and Tanner 1976), there is no reason why some of these should not show up in the BMI. In examining the ‘Nilotic Physique’, Roberts and Bainbridge (1963) find at least 20 different body characteristics (from ‘thin, fragile‐boned face’ to ‘weak muscling of thighs’) of the Nilotic peoples that make them very light for their height.8

Environmental differences

For lack of independent knowledge of the safe range in particular African environments, we have relied upon the FAO/WHO/UNU expert committee's general weight for height recommendations. Considering the many different and severe health hazards in the SSA it may also be that a higher weight is warranted in environments with incomplete (credit) markets and high intra‐ and interseasonal variability in many people's access to food. Under such circumstances, a relatively high body weight (p.187) (p.188) provides an energy ‘buffer’ in times of prolonged illness and/or economic crisis. These issues are not discussed by the FAO/WHO/UNU expert committee, and there seems to be little evidence to be found elsewhere.

5.8. Different measures: different definitions of undernutrition

Some of the discrepancies in the extent of undernutrition suggested by the various indicators can be explained, as we have seen in the previous section, by the fact that there are biases in many of the estimation methods and in the data. However, even in the strictly hypothetical situation where we could obtain measurements that are unbiased in the ‘technical’ sense, the various indicators would still not provide the same picture of nutritional standards in the SSA region. This is for the simple reason that the different indicators are built on different assumptions about what constitutes undernutrition. As mentioned earlier, undernutrition is not a well‐defined and unambiguous concept that is easily quantified. In the following, we shall nevertheless attempt to make explicit the different definitions of ‘undernutrition’ that (often implicitly) underlie the various measurements and indicators. One such explanation is that the various indicators are aimed at capturing undernutrition of different ‘severity’, but this is only part of the story.

(a) The FAO/WHO and IBRD dietary norms

The recommended daily allowances of calories (RDAs) produced by the FAO/WHO (1973), which in scaled‐down versions underlie the IBRD (1986) estimates of the prevalence of undernutrition in the world, comprise two main components. The first is the energy requirement for internal body functions, the BMR, and other ‘baseline’ activities. The other component is the calorie requirement for external physical activity, such as work and social engagements. The estimation of energy requirements for BMR and other ‘baseline’ activities is largely a positive problem that can be resolved with scientific methods, at least in principle. In practice, several fundamental empirical and theoretical issues are still highly disputed, as we have seen in previous sections.

When it comes to energy requirements for work, it is a bit unclear on what principle the FAO/WHO norms are based. In their description of what constitutes ‘calorie requirements’, the organizations say that the norm should allow for the work that is ‘economically necessary’. They refrain from any attempt to explain: necessary for what? A priori, there are many conceivable ‘what's’. In this very context, however, the most straightforward ‘what’ is: to avoid undernutrition. Then the obvious next question is: undernutrition, in what sense? If the answer to the latter question is: undernutrition in the sense that the individual's health and functional capabilities are at risk because of his nutritional status (body weight and activity level; cf. below), there are good reasons to think that the RDAs arrived at by the FAO/WHO are too high.

First, the FAO/WHO reference individuals, both adults and children, have (p.189) a body mass (index) above what these organizations themselves have found to be compatible with unimpaired health and physiological capabilities (FAO/WHO/UNU 1985: 183). This has the implication that the RDAs are derived for a reference person with a weight above what is needed for health and related purposes and, thus, an inflated energy requirement for BMR and all external activities, which are derived as multiples of the BMR. (It may also be that the FAO/WHO/UNU have overestimated the BMR per kilo of body weight, as discussed in a previous section.)

The second and more important ‘exaggeration’ is in the requirements for work activity. The FAO/WHO assumption is that the reference adult man in the developing countries has to work 2,555 hours per year in moderately heavy physical activities in order to avoid undernutrition in the family. For a ‘reference’ work load to make sense, it has to be equal to the work load that the average man in the population must pursue to avoid undernutrition. The question is then how this reference work load corresponds to the assumed distribution of income and calories in the FAO and IBRD models. The answer is that both organizations assume that there is no correlation between work load (and, thus, calorie expenditures in physical activity) and income. The lack of correlation means (by implication) that the man with the average work load also has the average income.

If one is to trust the income distribution estimates from countries like Kenya, Zambia, and the Ivory Coast reported by the World Bank (IBRD 1987: Appendix table 26), the poorest 20 per cent of households earn less than one‐fifth of average income. These households would thus only be able to command less than one‐fifth of the calories that the average income household can afford to buy, a household in which the man works 2,555 hours per year and still barely avoids undernutrition.

The poorest households can, of course, adjust their energy expenditure (requirements) downwards to some extent by reducing their body weight and height and lower their non‐income‐earning external physical activity levels. As an illustration of the scope for such adjustments, consider the standard adult reference male in the developing countries. He is 172 cm tall and weighs 65 kilos (corresponding to a BMI of 22). He works for seven hours per day in ‘moderately heavy’ physical activities, spends two hours in physically demanding social activities, seven hours in ‘residual’ (very low‐energy‐consuming) activities and is at sleep for eight hours (see FAO/WHO/UNU 1985: ch. 6). Altogether, this means that he expends close to 3,000 calories per day (see Table 5.5). Another male adult, who was nutritionally constrained already as a child, at 160 cm and with a body weight of 51 kilos (corresponding to the lower end of the range of BMI that the FAO/WHO/UNU finds acceptable and safe for males) can work the same hours in equally demanding physical activities at three‐quarters of the calorie expenditure. This is because his energy demand for BMR is lower as a consequence of his smaller body size and because he has eliminated all social activities that are energy intensive.

(p.190)

However, a reduction of his energy requirements by 27 per cent is not sufficient if his income is 20 per cent of that of his ‘average’ neighbour, whose earnings are barely enough to avoid undernutrition. The inescapable conclusion is that the IBRD estimation technique implies that one‐fifth of the households in the typical African country will be undernourished on a very severe scale. The incidence of anthropometric and clinical signs of undernutrition in these households would be very high. As we have seen, this is not at all what the evidence at hand shows.

The conclusion is that the RDAs are not appropriate norms, even in the scaled‐down versions applied by the IBRD, for the identification of undernutrition in any medical or clinical sense. The RDAs may adequately state what is compatible with good nourishment, health, and a productive and rich social life, but they include a significant safety margin and overstate what is needed in a more fundamental sense. The RDAs are ‘poverty lines’ rather than calorie requirement norms.

(b) The 1.2 BMR and 1.4 BMR norms

While the RDAs, which, in the African context, are equivalent to 1.87 BMR, probably grossly overstate the calorie intake that the average person (or per capita equivalent) needs in order to avoid undernutrition in a medical sense, it is equally obvious that this person needs more than 1.2 or 1.4 times the BMR. (p.191) These norms are derived from what nutritionists conceive to be the energy requirements for biological baseline activities only; with no allowance at all for work (or any other form of external physical activity). In a world where food entitlements are dependent on the household's own work activities, which in the African region are basically non‐mechanized agricultural work, such norms make little sense. The FAO (1985a: 24) has stated en passant that it has set up the 1.2 and 1.4 BMR norms without any allowances for work because such allowances are impossible to derive accurately and without invoking normative judgements, the very argument used above in more explicit terms against the IBRD norms. However, this, as we have stressed at length, is not a valid reason for ignoring energy needs for physical work altogether. The logical conclusion is that to use standardized calorie norms in order to distinguish the undernourished from the well nourished in a population where there are de facto very large variations in calorie expenditure for work is a meaningless exercise.

(c) The anthropometric norms

The standard anthropometric norms cannot be used to single out the particular individuals that suffer from undernutrition in a clinical sense, and they cannot be used to provide exact numbers of ‘undernourished’ people in a population. They can be used, however, to derive a relatively reliable upper limit for the number of those who are at risk. With the anthropometric measures and norms that are used today, the estimated incidence of people at nutritional risk is, in fact, considerably higher than one would find on the basis of clinical observations of actual ‘undernutrition’. This is for three reasons.

First, anthropometric measures alone cannot be used to say whether the primary cause of low weight and stunting is lack of food at the household level or whether illness is the basic cause (secondary undernutrition). People who are found to have an anthropometric status below some norm will thus include both those who are primarily undernourished and those who are basically ill.

Second, the health and functional impairment that has been documented in the literature (such as child mortality, reproductive capacity in women, and work ability in adults) is usually only detectable at an anthropometric status below the standard height and weight norms. The latter are derived on the basis of such criteria as 90 per cent of height or weight of reference populations in the Western countries. These norms thus include at least some ‘safety margin’.

Third, the anthropometric norms in use are defined on probabilistic grounds. That is, the norm is set above, or at, the level where a low weight or height affects the probability of falling ill or being functionally impaired in one way or the other. This does not mean that all people with a below‐the‐norm weight will actually become sick; most will not, but all in the risk group will be defined as undernourished. As long as one accepts the paradigm that undernutrition is a state of health and physiological impairment, caused by energy (p.192) inadequacy, one can thus use the anthropometric measures to estimate the upper limit of the number of people at risk of being undernourished in the clinical sense. However, to live with a constant excess risk of illness is clearly welfare reducing, which is an argument for defining undernutrition in a probabilistic rather than a deterministic (clinical) sense.

5.9. Summary and conclusions

The nutritional situation in sub‐Saharan Africa looks very different depending on which indicator is consulted. The highest incidence of undernutrition is suggested by the aggregate estimates derived by the FAO and the IBRD. The first important conclusion of the above analysis is that the prevalence of undernutrition in the region is not of the enormous order claimed by the international organizations; not even if undernutrition is defined in the broadest possible sense (the IBRD). The estimates are upward biased because they are built on ‘calorie availability’ data that understate true food supplies and also because of assumptions of an unduly uneven distribution of calories across households. Moreover, the implied consequences of undernutrition of the magnitude and severity (the very low norms) claimed by the FAO are so far‐reaching that they are impossible to reconcile with demographic and anthropometric observations.

The high incidence of moderate stunting among small children, in the 10–30 per cent range, is the only (technically unbiased) indication of a high prevalence of undernutrition in the SSA. Lack of food at the household level may be an important explanation of this phenomenon, but, a priori, there are several other feasible explanations, such as disease, inappropriate weaning food, and intrahousehold discrimination. The fact that only a small proportion of the same children have a subnormal weight for height suggests that food intake is normally adequate. The extra calories needed for growth in stature constitute only a few per cent of a child's total energy requirements after the first year. Moreover, even if primary undernutrition is the reason for moderate stunting, the mortality and dysfunctions that have been observed in relation to short stature usually show up only at what is termed ‘severe’ stunting (below 80 per cent of reference height for age),9 which affects a few per cent of children in the samples studied.

On the basis of the studies available today, however, it is not possible to say exactly what share of the population in the various SSA countries suffers from undernutrition in various meanings that are relevant in different contexts and for different policy objectives. The available studies are too few and, above all, in most instances not conducted with appropriate methods (see also Svedberg 1991b). The aggregate estimates of the FAO and the IBRD have little meaning. (p.193) The hundreds of dietary and anthropometric sample studies that have been carried out in the SSA, even when representative and unbiased, seldom include supplementary socio‐economic data that are required if we are to understand the reason for the undernutrition that exists, irrespective of its severity and prevalence.

Without proper knowledge of which population groups are undernourished there is little hope that these people can be assisted in the near future. Even more important, without such knowledge there are large risks that already vulnerable groups will see their situation deteriorate further as unforeseen consequences of the structural adjustments that almost all the SSA countries will be engaged in during the 1990s.

(p.194) References

Bibliography references:

Ahrin, K. (1985), Marketing Boards in Tropical Africa (London: KPI Limited).

Bates, R. H. (1981), Markets and States in Tropical Africa: The Political Basis of Agricultural Policies (Berkeley, Calif.: University of California Press).

Beaton, G. H. (1983), ‘Energy in Human Nutrition: Perspectives and Problems’, Nutrition Reviews, 41.

Behrman, J. R., and Deolalikar, A. (1987), ‘Will Developing Country Nutrition Improve with Income? A Case Study of Rural South India’, Journal of Political Economy, 95.

Benefice, E., et al. (1981), ‘Surveys of Nutritional Status in Semi‐arid Tropical Areas (Sahel 1976–79): Methods and Results’, in IDRC (1981).

Bhargava, A. (1988), ‘Estimating Short and Long Term Elasticities of Food and Nutrients for Rural South India’, mimeo (University of Pennsylvania).

Bouis, H. E., and Haddad, L. J. (1988), ‘Comparing Calorie‐Income Elasticities Using Calories Derived from Reported Food Purchases and a Twenty‐Four Hour Recall of Food Intakes: An Application Using Philippine Data’, mimeo (Washington, DC: IFPRI).

Chambers, R., et al. (1981), Seasonal Dimensions to Rural Poverty (Exeter: Frances Pinter).

Chiligo, M. O., and Msukwe, L. M. (1984), ‘Nutritional Surveillance in Malawi’, background paper for workshop on Social and Nutritional Surveillance in Eastern and Southern Africa, Cornell Nutritional Surveillance Program.

Cornell Nutritional Surveillance Program (CNSP) (1984), ‘Trends in Nutritional Status of Children in Rural Kenya: Review and Policy Implications’, Working Paper 30 (Cornell University).

Dasgupta, P., and Ray, D. (1990), ‘Adapting to Undernourishment: The Biological Evidence and its Implications’, in the first volume of this book.

Dillon, J. C., and Lajoie, N. (1981), ‘Report on Surveys of the Nutritional Status of the Rural Population in the Sahel from 1960 to 1979’, in IDRC (1981).

Eicher, C. K., and Baker, D. C. (1982), ‘Research on Agricultural Development in Sub‐Saharan Africa: A Critical Survey’, MSU International Development Paper 1 (East Lansing, Mich.: Department of Agricultural Economics, Michigan State University).

Eksmyr, R. (1970), ‘Anthropometry in Privileged Ethiopian Pre‐school Children’, Acta paediat. scand. 59.

European Economic Commission (1986), ‘Conference Proceedings: Statistics in Support of African Food Strategies and Policies’, mimeo (Brussels, 13–16 May).

Eveleth, P. B., and Tanner, J. M. (1976), Worldwide Variations in Human Growth (Cambridge: Cambridge University Press).

FAO.(1980), Food Balance Sheets: 1975–1977 Average (Rome: FAO).

———(1984), Food Balance Sheets: 1979–1981 Average (Rome: FAO).

———(1985a), Fifth World Food Survey (Rome: FAO).

———(1985b), FAO Expert Consultation on Production Statistics of Subsistence Food Crops in Africa (Rome: FAO).

———(1987), ‘Daily Per Caput Calories’, computer outprint (Rome: FAO).

(p.195) ———(1988), Production Yearbook (Rome: FAO).

———/WHO.(1973), ‘Energy and Protein Requirements’, WHO Technical Report Series No. 522, Report No. 52 (Geneva: Report of a Joint FAO/WHO Ad Hoc Expert Committee, FAO Nutrition Meetings).

——— ———/UNU.(1985), ‘Energy and Protein Requirements’, Technical Report Series 724 (Geneva: WHO).

Floud, R. (1987), ‘Anthropometric Measures of Nutritional Status in Industrialized Societies: Europe and North America since 1750’, mimeo (Helsinki: WIDER).

Greene, M. H. (1974), ‘Impact of the Sahelian Drought in Mauritania, West Africa’, Lancet, 1 June.

Harriss, B. (1990), ‘The Intrafamily Distribution of Hunger in South Asia’, in the first volume of this book.

Hulse, J. H., and Pearson, O. (1981), ‘The Nutritional Status of the Population of the Semi‐arid Tropical Countries’, in IDRC (1981).

IBRD.(1986), Poverty and Hunger (Washington, DC: World Bank).

———(1987), World Development Report (Washington, DC: World Bank).

———(1989), Progress of Initiatives Benefiting Sub‐Saharan Africa (Washington, DC: World Bank).

International Development Research Centre (IDRC) (1981), Nutrition Status of the Rural Population of the Sahel (Ottawa: IDRC).

Kloth, T. I., et al. (1976), ‘Sahel Nutrition Survey’, American Journal of Epidemiology, 103.

Lawrence, M., et al. (ed.) (1985), ‘The Energy Cost of Common Daily Activities in African Women: Increased Expenditure in Pregnancy?’, American Journal of Clinical Nutrition, 42.

Lipton, M. (1983), ‘Poverty, Undernutrition and Hunger’, World Bank Staff Working Paper 597 (Washington, DC: World Bank).

———(1986), ‘Improving the Basic Data: Are Present Techniques Satisfactory?’, mimeo, European Community.

Maribe, T. (1984), ‘Up‐Date of the Nutrition Surveillance in Botswana’, background paper for workshop on Social and Nutritional Surveillance in Eastern and Southern Africa, Cornell Nutritional Surveillance Program.

Mayer, J. (1976), ‘The Dimensions of Human Hunger’, Scientific American, 235.

Oliver, R., and Crowder, M. (eds.) (1983), The Cambridge Encyclopedia of Africa (Cambridge: Cambridge University Press).

Osmani, S. (1990), ‘Nutrition and the Economics of Food: Implications of Some Recent Controversies’, in the first volume of this book.

Payne, F. (1987), ‘Undernutrition: Measurement and Implications’, mimeo (Helsinki: WIDER).

Poleman, T. T. (1975), ‘World Food: A Perspective’, Science, 188.

———(1977), ‘World Food: Myth and Reality’, World Development, 5.

Reutlinger, S., and Alderman, H. (1980), ‘The Prevalence of Calorie Deficient Diets in Developing Countries’, World Development, 8.

———and Selowsky, M. (1976), ‘Malnutrition and Poverty: Magnitude and Policy Options’, World Bank Staff Occasional Paper 23 (Washington, DC: World Bank).

Roberts, D. F. (1985), ‘Genetics and Nutritional Adaptation’, in Blaxter, K., and Waterlow, J. C. (eds.), Nutritional Adaptation in Man (London: John Libbey).

(p.196) Roberts, D. F., and Bainbridge, D. R. (1963), ‘Nilotic Physique’, American Journal of Physical Anthropology, 21.

Rosetta, L. (1986), ‘Sex Differences in Seasonal Variations of the Nutritional Status of Serere Adults in Senegal’, Ecology of Food and Nutrition, 18.

Schofield, S. (1979), Development and the Problems of Village Nutrition (London: Croom Helm).

Svedberg, P. (1987), ‘Undernutrition in Sub‐Saharan Africa: A Critical Assessment of the Evidence’, Working Paper 15 (Helsinki: WIDER).

———(1988), ‘A Model of Nutrition, Health and Economic Productivity’, Working Paper 46 (Helsinki: WIDER).

———(1990), ‘Undernutrition in Sub‐Saharan Africa: Is There a Gender Bias?’, Journal of Development Studies, 26.

———(1991a), ‘The Export Performance of Sub‐Saharan Africa’, Economic Development and Cultural Change, 39.

———(1991b), Poverty and Undernutrition in Sub‐Saharan Africa: Theory Evidence Policy, forthcoming.

Tomkins, A. M., et al. (1986), ‘Seasonal Variations in the Nutritional Status of Urban Gambian Children’, British Journal of Nutrition, 56.

UNCTAD.(1988), Handbook of International Trade and Development Statistics (Geneva).

USAID.(1976), ‘Liberia National Nutrition Survey’, mimeo (Washington, DC).

———(1977a), ‘Lesotho National Nutrition Survey’, mimeo (Washington, DC).

———(1977b), ‘Togo National Nutrition Survey’, mimeo (Washington, DC).

———(1978a), ‘United Republic of Cameroon National Nutrition Survey’, mimeo (Washington, DC).

———(1978b), ‘Sierra Leone National Nutrition Survey’, mimeo (Washington, DC).

———(1986), ‘Swaziland National Nutrition Survey 1983: Full Report’, mimeo (Washington, DC).

van Ginneken, W., and Park, J.‐G. (1984), Generating Internationally Comparable Income Distribution Estimates (Geneva: International Labour Office).

von Braun, J. (1988), ‘Effects of Technological Change in Agriculture on Food Consumption and Nutrition: Rice in a West African Setting’, World Development, 16.

Zarkovich, S. R. (1962), ‘Agricultural Statistics and Multisubject Household Surveys’, reprinted in FAO, Studies in Agricultural Economics and Statistics, 1952–77 (Rome: FAO).

Notes:

(1) The Bank's version of the model is described in some detail in several publications (Reutlinger and Selowsky 1976; Reutlinger and Alderman 1980; IBRD 1986). The FAO version is not as explicitly presented (FAO 1985a: 17–30 and Appendix 3), but there are no fundamental differences.

(2) The number of inhabitants per square km in sub‐Saharan Africa as a whole was 18 in 1985; the equivalent figure for South and South‐East Asia was 145 (UNCTAD 1988).

(3) See Harriss (1990) and the references cited there.

(4) The FAO/WHO/UNU (1985) expert committee recommended downward adjustments of the RDAs of 1973 on several accounts, but no new RDAs have been published by the FAO. In their annual publication, the State of Food and Agriculture, RDAs were formerly presented; in recent issues, this is no longer the case.

(5) On the general issue of ‘adaptation’, see Dasgupta and Ray (1990) and Osmani (1990).

(6) As we have seen, the organization claims that the per capita calorie requirement is 1,747 (1.4 BMR) and that the per capita availability is 1,856. With inter‐individual requirements perfectly correlated to actual intake, this would mean that everybody in the region has an intake 6% above requirement (which are too small to allow them to work, but that is another matter). With the IBRD assumption of a 0.70 correlation between personal requirements and intakes, and a higher requirement norm, some households would still be undernourished, but the percentage would be far below that shown in Table 5.2 above.

(7) People with a Nilotic ethnic origin make up about one‐quarter of the population in contemporary sub‐Saharan Africa (Oliver and Crowder 1983). They dominate in southern Sudan and in the northern parts of the Sahelian countries. They are also found in northern Kenya, Uganda, Tanzania, and in Burundi and Rwanda. The Nilo‐Hamites in Ethiopia and Somalia are usually considered a separate race.

(8) In the words of physical anthropologists, ‘the Nilotic group is particularly low in endomorphy and mesomorphy and shows an extreme degree of ectomorphic dominance’ (Roberts and Bainbridge 1963: 357).

(9) See Osmani (1990).