2.1 Population data
2.2 Food Production Indices
2.3 The Food Balance Sheet Method (Dietary Energy Supply - (DES - kcals/caput/day))
2.4 Undernourished Population - percent and numbers with DES below 1.2 BMR2
2.5 Underweight Children - percent and numbers below minus 2 SD's weight for age
2.6 Infant Mortality Rates and Numbers
2.7 Child death rates and number of child deaths
Two data files were created for the estimates given in the First Report on the World Nutrition Situation (ACC/SCN, Nov. 1987, referred to here as me Report).
(i) A country file of basic data for each of the 94 countries included (given in table 3 of the Report). The data input into mis file is given in Table AI.
(ii) A survey data file containing data from countries with representative anthropometric survey data. The contents of this file are given in Table AIL
The data were processed mainly on IBM-compatible microcomputers, using DBase III and SPSS/PC.
The data used in the panels in the Report (i.e trends in indicators by country group) are given in Table Am. For each country group, data for 1960, 1970, and 1980 are available for most indicators; and yearly estimates from 1980. to 1985 are listed for population, per caput dietary energy supply (DES), and per caput food production indices only. Details of all indicators are discussed below. Distinctions are made between estimates obtained from other UN sources, and those derived by the SCN or the FAO Statistics Division. Sources arc summarized in the footnote to Table AIII. The number of decimal places shown in Table Am should not be taken to suggest undue accuracy. Numbers affected (expressed in millions) were usually reported to two decimal places because of the. very small populations in some country groups, particularly Middle America and the Caribbean.
1 Annual estimates of dietary energy supply (kcals/caput/day) at country level calculated by FAO are not published; either group means by year, or country estimates by three year periods are given. Similarly, country level estimates when derived by interpolation methods done by the SCN (i.e. prevalence of underweight children) are not given here, but are aggregated to country group level. This is because margins of error at country level are not known but arc probably substantial.A complete list of countries used in these calculations is given in Table 3 (p.45) of the Report, by country group. The number of countries with complete information on all indicators varies slightly; for example, not all 94 countries had available information on child deaths. However, this is unlikely to have biased the estimates because only countries with small populations tended to have missing data, and there were very few of these.
The estimates of total population and infant mortality rates (IMR) used in the Report were taken primarily from the UN Population Division statistics. These data were available on computer tape in FAO, and were retrieved by the FAO Statistics Division. These figures were then re-entered into the country data file. The data on total populations and IMR were then used exactly as given by the UN Population Division, with no further calculations except for aggregation into country groups.
This section gives a brief description of the origins of and assumptions underlying these figures, and is based on UN (1986), in which further details can be found. The description in this section therefore refers to calculations done before the data were retrieved, not to analyses done specifically for the Report, and is included as background information.
By country, the size of the total population and its age and sex structure is first determined for the chosen base year. In most cases, this information is obtained from national censuses (UN, (1986) p.3). The availability of population census data varies by region throughout the world, but during 1975-84, over 95% of me world's population was covered by a population census. Between 1975 and 1984, 50 of the 56 countries of Africa conducted a census; almost all countries of Middle and South America completed a census; and finally, 36 of 43 countries or areas in Asia were covered by population censuses. (Data from North America and Europe were not used in the Report). Where census data for the chosen base year are lacking, information from previous censuses, civil registration, demographic surveys, and other administrative records is used by the UN for population estimation.
Three aspects of population change are considered in making projections of population for a given year: fertility, mortality and migration. The base year population estimates (by age and sex) are adjusted for these as described in UN (1986) p.4. Estimates of age specific fertility and mortality rates are described in UN (1986) p.8. Where these are unreliable for a country, other methods are applied to obtain reasonable estimates. International migration is the most poorly documented; however, except for a few cases, net migration is considered very small and taken to be zero.
One basic underlying assumption in the estimation procedure is that during the projection period, rates of change are reasonably stable; thus catastrophes such as war, famines and epidemics are not considered. This assumption would certainly influence estimates in Africa, where a severe drought affected many countries and contributed to a greater than expected mortality rate. In this instance, mortality statistics might have been adjusted if data were available by country. However, owing to the lag-time in data processing, particularly in times of disaster, the effect of drought was not generally considered in the population and infant mortality estimates given in the Report.
In general, mortality trends assume a gain of 2.5 years in life expectancy for every five year estimation interval, until expectancy reaches 62.5 years when there is a slowdown in the gain. In Sub-Saharan Africa this gain was lowered to 2.0 years for every five year interval in many countries. Both models assume a built-in improvement in mortality rate which, as mentioned above, is assumed to be unaltered by disasters.
From UN (1986) p. 10: "In preparing fertility assumptions, past and current fertility trends for each country are evaluated and placed within me social, economic and political context of the country. Trends and anticipated changes in the socio-economic structure and cultural values of the society as well as policies and programmes directed towards family planning are considered vis-a-vis expected trends in fertility." For a more detailed discussion of assumptions in fertility estimation, including high, medium, and low variants in fertility rates, see UN (1986) p. 10.
International migration was considered to be a significant component of population growth in only a few countries. However, in these cases an accurate estimate is virtually impossible to obtain because migration is affected by many "unpredictable political, economic, and social circumstances" (UN (1986) p.10), which can alter migration patterns in very short periods of time. Caution should be used in considering these cases.
The index numbers of food production compare a volume of agricultural production in a given year with the base period, 1979-81. The definition is given by FAO definition (1987) page xi:
"They are based on the sum of price-weighted quantities of different agricultural commodities produced after deductions of quantities used as seed and feed weighted in a similar manner." The commodities included are considered edible and contain nutrients (e.g. coffee and tea are excluded). These indices are given per caput.
"The country indices are calculated by the Laspeyres formula. Production quantities of each commodity are weighted by 1979-81 average national producer prices and summed for each year. To obtain the index number, the aggregate for a given year is divided by the average aggregate for the base period 1979-81." (p. xi). National producer prices are expressed as "international commodity prices derived from the Geary-Khamis formula... This method assigns a single 'price' to each commodity (e.g. one ton of wheat has the same price in whatever country it was produced)."
Annual indices of food production by country group are listed in Table AIII for 1980 through 1985, and three year averages for 1960, 1970, 1980, and 1984. Like dietary energy supply (DES, see below), the 1960 value corresponds to 1962 (1961-63, because FAO annual series starts from 1961), the 1970 value to 1969-71, 1980 value to 1979-81, and 1984 value to 1983-85.
The food balance sheet calculated by FAO by country, gives dietary energy supply estimates used in the Report (sections B and C of the panels). "The amount of food potentially available for human consumption is derived after considering the sources of supply, and their utilization. The supply component considers the amount of food produced in a country, imports, and adjustments for changes in stocks. Utilization accounts for food exported, fed to livestock, used for seed, manufacturing for food and non-food uses, and losses during storage and manufacturing." (FAO (1984) p. viii).
Although the food balance sheet method provides estimates for the quantities of food reaching the customer, the amount of food actually consumed may be lower than the quantity shown in the food balance sheet, because losses of edible food and nutrients in the household (e.g. during storage, or in preparation and cooking pets feeding etc.) are not considered.
The Dietary Energy Supply (DES.) is thus an estimate of food availability at the national level. It is derived by applying food composition factors to all food commodities available for human consumption. Only the. total calories available for consumption was of interest. It is expressed on a per caput basis by dividing the total calories available by an estimate of the total population, then per day by dividing by 365, to give the unit kcals/caput/day (FAO, 1984).
The question of accuracy of food balance sheets is addressed thus by FAO (1987), page ix: "The accuracy of food balance sheets, which are in essence derived statistics, is of course dependent on the reliability of the underlying basic statistics of population, supply and utilization of foods and of their nutritive value. These vary a great deal between countries, both in terms of coverage as well as in accuracy. In fact, there are many gaps particularly in the statistics of utilization for non-food purposes, such as feed, seed and manufacture, as well as in those of farm, commercial and even government stocks. To overcome the former difficulty. estimates were prepared in FAO while the effect of the absence of statistics on stocks is considered to be reduced by preparing the food balance sheets as an average for a three-year period." However, it would appear, experience in compilation assures consistency and this permits trend assessment
²Basal Metabolic RateThe proportion of undernourished population was estimated by the Statistics Division of FAO, and provided by country group only. A discussion of this method of calculating both percent and numbers of undernourished is found in section 4.1.5 of the Report (p. 46) and in FAO's Fifth World Food Survey (FAO (1985) p.l8).
The proportion of underweight children 0 through 4 years of age was derived by the SCN using an interpolation model, as described in detail in section 3 of this supplement. Briefly, prevalence was predicted from a regression model using the sample of countries shown in Table AII. All factors in the model were taken from the same year as the prevalence estimates. The interpolation model thus derived was applied to the larger set of 94 countries for country group estimates (Table All). This model was based on three year averages for kcals/caput/day and IMR interpolated to the year of interest. Thus, the prevalence estimates listed for 1970 in Table AIII were based on three year average kcals for 1974/75/76 and IMR interpolated to 1975. The 1980 and 1984 estimates were exact midpoints of the three year averages. Numbers of underweight children were derived by country, and aggregated to the country group level as follows:
Number = % prevalence underweight x child population 0 through 4 years/I 00.Table. All lists the national data (36 countries, 45 surveys) used in the derivation of the inter-polation model. Only country surveys undertaken after 1975 and considered to be nationally representative on prevalence of low weight-for-age (WA) were included. Countries with more than one survey during this period were included, thus some countries were represented more than once. Representativeness was assessed on the basis of the survey design, usually obtained in the survey report, considering the sampling methodology used.
Sources of anthropometric survey data: Sources of anthropometric survey data are listed as part of Table An. The two major sources for survey data on prevalence of below - 2 standard deviations (SD's) WA were WHO, (1987) and Haaga, et al. (1985). Prevalences for five countries were obtained from other sources as follows: Botswana (Spafford Yohannes, 1986), Mauritius (correspondence between WHO Geneva and WHO Mauritius, November 4, 1986), Zimbabwe (K Test personal communication with Ministry of Health, Nutrition Unit), Peru (Rothe, Nieburg, personal communication), Papua New Guinea (correspondence with P Heywood).
Choice of nutritional status outcome measure: Weight-for-age (WA) was chosen as an indicator of nutritional status for primarily practical reasons, in that it was more often reported than either weight-for-height (WH) or height-for-age (HA). It combines both chronic (HA) and acute (WH) malnutrition, a distinction which may be important in individual child growth or for nutrition policies aimed at long-term development. However such a distinction is less crucial for the cross-national comparisons examined here.
The cutoff of -2 SD's WA for estimating prevalence of underweight was chosen because as a standardized measure it is more informative than percent median since it indicates exactly where the case lies relative to the mean. As such it is may be more appropriate for inter-country comparisons. This practice is recommended by WHO (1987).
Derivation of percent below -2 SD's Weight for Age from percent median: In some instances, only the prevalence below 80% median weight-for-age was available, so the prevalence below -2 SD's had to be calculated from this (for about 12 country-year cases). The method used was as follows:
1. From the NCHS data (WHO. 1983) a % median value equivalent to -2.0 SD's was calculated for several ages between 1 and 5 years, and for both sexes. These values were averaged across all ages and sexes to give a value of 78.5% median WA (the range of these values was 77 to 79.4%).The following is an example in the calculation of one point on the graph. To obtain other points, this method was repeated changing only the sample mean.2. The percent below 78.5% (i.e. -2.0 SD's), and the percent below 80% was calculated for several (theoretical) different distributions where only the mean was altered. The standard deviation was kept constant at 10.0, (in line with the SD observed in many surveys). All distributions were assumed to be normal, Z Tables were used to calculate these percentages (see example below).
3. These two percentages (or prevalences), below -2 SD's and 80% median were graphed for all scenarios. The result was linear with Y-axis = percent < -2 SD's; and X-axis = percent < 80% median WA.
Sample mean = 90.0 S.D. = 10.0 % median at -2.0 SD's = 78.5
Z score3 for 80% median = (80.0 - 90.0)/10 = -1.0.
From Z Table the % < -1.0 = 15.87%.
3 The z score calculation can be found in most statistical textbooks: z = (XI - X)/SD; where: XI is the value of interest, X is the mean of the sample, and SD is the standard deviation of the population.Z score for -2.0 S.D. = (78.5 - 90.0)/10 = -1.15.
Choice and sources of independent factors: The independent factors listed in Table All were selected for the year of the anthropometric survey. These factors were chosen as potentially useful predictors mainly because of their established association with nutritional status (see "purpose of interpolation model" in section 3.1 below). Also important was the availability and consistency of this information for all countries studied since 1975.
All independent factors except DES were obtained from data published by the World Bank (1978-86). DES was extracted from the Food Balance Sheet account in FAO (AGROSTAT). The purpose of restricting sources was to maintain consistency across countries and years. Data from the World Bank (1978-86) are updated and published annually, and are complete for all countries in this sample.
Data for the derivation of country group prevalences of underweight children: Table 3 of the Report lists all 94 countries used in the calculation of country group prevalence of underweight children for 1975, 1980, and 1984. The factors listed in Table AI are those required for the calculation of country level predicted prevalence of low WA for the three years; including kcals/caput/year (as three year averages), and IMR's (interpolated to the year of interest). Child populations of 0 through 4 years of age are needed to calculate the numbers of malnourished children by country. They were derived as described in section 4.1.6 (p. 50) of the Report, from total populations also listed in Table AI. The method used for estimating country group prevalence is discussed below in section 3.
Infant Mortality Rate (IMR) is the number of children under 1 year of age who died in a given year, expressed per 1,000 live births. Estimates reported in Table AIII are five year intervals calculated by the UN Population Division, and retrieved from FAO computer files (see section 2.1). In Table AIII, the IMR listed for 1960 is the average of 1960-'65, 1970 is the average of 1970-'75, 1980 is the average of 1980-'85. In the panels these estimates are graphed as the midpoints of the intervals (i.e. 1962.5, 1972.5 and 1982.5). (An additional point is included on the graph but not in this Table AIII, i.e. that for the period 1975-'80, which is plotted as 1977.5 in the panels).
The number of infant deaths was calculated as IMR multiplied by total number of live births per year/I.000. The actual calculation was done by the SCN, using the total number of live births in a given year and IMR's extracted from FAO files. The number of infant deaths are five year interval estimates comparable to IMR.
Child death rates are expressed as the number of deaths of children aged 1 through 4 years old, per 1,000 child population (1 through 4 years) in a given year. Child death rates were not readily available, probably because of poor reporting of deaths and child populations. Estimates reported here were taken from the World Bank (1978 - 86), and for some values, from UNICEF (1986).
The number of child deaths is equal to child death rate multiplied by child population of 1 through 4 years of age, in a given year. Basically, the population of 1-4 year olds was obtained by first estimating the proportion of the 0-4 year old age group in the population in question, since the latter is more readily available by country. The procedure is explained in more detail in the Report (section 4.1.8). Additional data listed below are the percentages of 1-4 year olds used in the calculation of child populations, by country group, derived by calculating from UN (1986).
|
Africa |
78.5 |
|
Middle America/Caribbean |
78.7 |
|
South America |
80.0 |
|
South Asia |
80.0 |
|
Southeast Asia |
78.9 |
|
Near East/North Africa |
79.7 |
|
China |
78.0 |
