Abstract: This paper uses household data from Sudan to examine the factors which affect child mortality. Thus, the impact on child mortality of the education of the mother and the father, public health program provisions and household income per adult are examined. In examining the interaction between income and child mortality the former is instrumented on household assets, which are used as identifiers in the Two Stage Least Squares estimation of the mortality function. In Ordinary Least Squares estimates, parental education and income per adult are found to have a significantly negative impact on child mortality, and mother's education, in particular, is found to have a larger and more significant effect than that of the father. Public health Programmes are found to produce significant reductions in child mortality. However, Two Stage Least Squares estimates indicated that the most important factors influencing child mortality are the mother's age, household per capita income, area of residence and, to some extent, hospital services.
This paper examines the effect of parental education on child survival. The effect of income is also measured. Studies of other developing countries have found that education increases child survival (Cochrane 1982; Rosenzweig and Schultz 1982). The study of mortality is important since it affects population growth directly and indirectly through its association with fertility as they are found to be positively correlated. The relationship between child survival and fertility plays a crucial role in the mechanism of the demographic transition from a high fertility regime to a lower one, and has been investigated empirically for varied environments (Schultz 1981) but only for few African countries (Okojie 1991; Maglad 1994).
According to the 1993 Population Census (Population Census Office 1994), Sudan had a population of 25.6 million in 1993 and a rate of population growth of 2.6% per annum in the period 1983-1993. Of this total 29.2% was urban, 68.1% rural and 2.7% nomadic. Mortality rates have been falling over time, but they are still high compared with rates for North Africa. The crude death rates (CDR) have declined from 26 per thousand in 1955/56 to about 17 per thousand by 1983.
Infant mortality (IMR) was decreasing over time. It was 145 and 118 per 1000 live births in 1973 and 1983 respectively. According to the 1990 Sudan Demographic Health Survey (SDHS) and 1993 Maternal and Child Health Survey (MCHS) it is 69.5 and 69.9 respectively. However, estimates of IMR obtained by indirect methods from 1993 Census information is 108 for the period, which indicates that there has been an increase during the late 1990's. An analysis of the evaluation of IMR obtained from (MCHS) and 1993 Census suggests that the Census estimate might be more reliable (El-yamen 1995). The 1993 MCHS showed a rate of 49 per 1000 for children aged 1-5. It is shown that 113 die before reaching age 5.
Fertility levels in Sudan are high compared to the levels for Northern Africa. The crude birth rate in 1983 was 42.6 births per 1000 population compared with the higher rate of 51.7 in 1955/56. The estimated rate for Northern Africa is 38 per 1000. Total fertility has been declining according to the different population censuses and surveys. The 1993 Population Census gave a total fertility rate (TFR) of 4.4 whereas the PAPCHILD survey of 1993 indicated total fertility rate at 4.5 (both rates are unadjusted). The total fertility rate was 6.76 and 6.2 in the population censuses of 1973 and 1983 respectively.
The fertility rate showed variation between rural and urban areas. The Sudan Fertility Survey of 1979 gave adjusted TFR of 6.9 in rural areas and 5 in urban areas. The 1993 Population Census indicated that (unadjusted) TFR is 4.6 in rural areas and 4.5 in urban areas. Women's education has been spreading but females' school enrollment still lags behind males' school enrollment. In 1985/86 primary enrollment ratio for males, age 7-12, was 58% while it was 41% for females (Educational Statistics Section 1987). The 1993 Population Census showed that 42 per cent of females 10 years and above could read and write, compared to 31 per cent in 1983. The corresponding figures for males are 63.7 and 57 per cent respectively.
Adjustment Programmes aiming at reduction of the budget deficit have affected the provision of social services, especially health and education. For instance, Central Government expenditure on health over the period 1988/94 has declined, both as percentage of total Central Government expenditure and as a share of GDP. It has fallen from around 2.2per cent of total government spending during 1988-91 to around lper cent in the period 1991-94. As a result, actual per capita health expenditure has declined from L.S. 0.9 in 1988/89 to L.S. 0.24 in 1993/94 (Babiker 1995).
The rest of this paper proceeds as follows. In section 2 the empirical model is specified. In section 3 the data on which the analysis is based are discussed and in section 4 the empirical estimates are presented. A conclusion is given in section 5.
In this analysis child mortality, defined as death rate in age one to age five, is assumed to depend on the education level of the wife EW , husband's education Eh, household income per adult Y, wife's age AW, and public program variables, H, related to health. The mortality function M is thus specified,
M=_o+_1 Ew + _2Eh +_3Y + _4AW+ _5H+ _6R + µ
where p is an error term which captures the impact of all other unmeasurable factors on child death and is assumed to be normally distributed, and P's are parameters to be estimated. R are residence dummy variables introduced to capture the effect of regions on child mortality.
The income measure, which is used in the analysis, is permanent income as measured by annual consumption expenditure of food and non-food items (Deaton and Muellbauer 1980). One problem with this measure is its endogeneity. It is argued that women's' allocation of time between market and home production (upbringing and care of children) are jointly determined, resulting in correlation of the error term and income and hence in biased estimate of the income effect.1 A test of the erogeneity of this variable will be carried out.
The program variables which are used in the analysis are the availability of hospital beds per capita and services of the Blue Nile Health Project (B.N.I-LP)2. The B.N.H.P provides services in the area of sanitation and combats water-borne diseases like malaria and schistosomiasis that are encouraged by irrigated agriculture3.
Obviously the empirical specification used does not allow nutritional intake as a determinant of mortality, except indirectly through the effect of food purchasing power of income. Models of household behaviour based on maximisation of utility, where health enters as an argument, and the constraints include both production functions of health and household full income, yield reduced form equations for health demand where the prices of nutrients (health inputs), individual's wage and income are included among the determinant variables (Bhereman and Deolkakir 1988, Thomas and Strauss 1993). In this study we will not be able to measure the impact of the nutritional status (or health inputs), or health determinants due to lack of required data.
This analysis of the determinants of child survival uses a sample of 1400 Sudanese households resident in rural areas of the Central Region and one Rural Council in Kordofan. The rural sample included thirty-four villages located in four agricultural schemes that extend over most of the Central Region and some part of the Eastern State. The survey of the villages took place in 1991.
The households were selected by multi-stage stratified random sampling, where in each area villages are stratified according to the level of development, as indicated by the presence of services, with special emphasis on education, and a random village is chosen from each strata. In the second stage a random sample of households was chosen from the list of households in that village. For each household, two questionnaires, one for the household and one or more for all married women in the household, were completed. For purpose of this analysis, only households where both husband and wife were present, and with at least one birth, are analysed. This working sample includes 1187 households.
Table (1) defines each of the variables, and table (2) provides the sample statistics for the variables analysed below. As table (2) shows, the mean number for child morality is .11. The mean age of a wife is 37.2. The illiteracy rate is higher among women than men, where 66% of the women in the sample are illiterate compared with 45% of the men. In this sample the services of B.N.H.P benefited 43per cent of the women residing in the area covered by the project (81 per cent of the women in the sample are located in Gezira and Blue Nile Region- see table 2).
The OLS estimates of child mortality are presented in table (3). Three specifications are presented: A restricted form of the mortality function is estimated in column 2 of table 3, where the program variables and regions controls are excluded. The specification in column 2 of the table introduces health program variables. The regression in column 3 adds regions of residence as dummy variables. Wife's age is introduced as a five-year interval age dummy, with age cohort 15-24 as the reference category, in order to capture non-linearity.
First note that child mortality increases with the woman's age linearly. An old woman is more likely to fall in the high-order birth group, where the risk of child mortality is high, and hence to suffer more child loss. Child mortality and parental education are negatively associated. In the restricted form estimates, the negative effect of mother's education on mortality is more pronounced and is statistically significant compared with father's education. Mother's primary education produces a reduction in child death rate of 3 per cent while a primary education level of the father produces only 2 per cent.
Table 1. Description of Variables
|
Variable |
Definition | |
|
Endogenous Household |
Child Mortality |
Proportion of Live Birth Dead |
|
Exogenous Household |
Woman's Age |
Age of Wife in Years |
|
Wife's Education: |
· Dummy = 1 if Wife has complete or some Primary | |
|
Schooling | ||
|
· Dummy = 1 if Wife has complete or some Secondary | ||
|
Schooling | ||
|
· Dummy = 1 if Wife has above Secondary Schooling | ||
|
Husband's Education |
· Dummy =1 if Husband has complete or some Primary | |
|
Schooling | ||
|
· Dummy = 1 if Husband has complete or some | ||
|
Secondary Schooling | ||
|
· Dummy = 1 if Husband has above Secondary | ||
|
Schooling | ||
|
Log (lncome/adult) |
The value of annual consumption expenditure on food and | |
|
non food items, including the value of goods used for | ||
|
consumption from own farm production, in thousand | ||
|
pounds, divided by adults, 15 years and over, in household | ||
|
and expressed in natural logarithm. The variable is | ||
|
potentially endogenous. | ||
|
Exogenous Community Programmes |
Hospital Beds |
Number of Hospital Beds Per Ten Thousand in Area Council |
|
Blue Nile Health Project (B.N.H.P) |
Dummy = 1 if village is under Blue Nile Health Project | |
|
Regions |
· Gezira Main |
· Dummy = 1 if residence is Main Gezira |
|
· Gezira |
· Dummy = 1 if residence is Managil | |
|
· Extension |
· Dummy = 1 if residence is Rahad | |
|
· Eastern ezira |
· Dummy = 1 if residence is Elsuki | |
|
· Blue Nile |
· Dummy = 1 if residence is Kordofan | |
|
· Kordofan |
||
In column 3 when health program variables are introduced, the B.N.H.P. effect, though negative, is not statistically significant. Hospital services measured by bed availability produce a larger and statistically significant effect in reducing child mortality. This could be due to the limited coverage of the Blue Nile scheme and to the differential impact which these services might have in the different socio-economic groups.
|
Table 2. Means and Standard Deviations of Variables | |||
|
Variable |
Mean |
(s.d) | |
|
Endogenous Household |
|||
|
Child Mortality |
0.113 |
(0.167) | |
|
Exogenous Household |
|||
|
Woman's Age |
37.200 |
(11.9) | |
|
Wife's Education: |
0.223 |
(.416) | |
|
0.109 |
(.312) | ||
|
0.011 |
(.104) | ||
|
Husband's Education: |
0.238 |
(.426) | |
|
0.169 |
(.375) | ||
|
0 .047 |
(.212) | ||
|
Log (Income/adult) |
4.960 |
(1.05) | |
|
Exogenous Community |
|||
|
Hospital Beds* 102 |
0.280 |
(.297) | |
|
Blue Nile Health Project |
0.521 |
(.499) | |
|
Regions |
|||
|
Gezira Main |
0.482 |
(.500) | |
|
Gezira Extension |
0.069 |
(.254) | |
|
Eastern Gezira |
0.104 |
(.306) | |
|
Blue Nile |
0.163 |
(.370) | |
|
Kordofan |
0.181 |
(.385) | |
|
Sample Size |
1187 |
||
Note the reduction in the effect of mother's education on child mortality when public Programmes are introduced. Public program services of health and education in a country like Sudan tend to be made available together when they are provided. Main Gezira, where the project services are concentrated, is also an area where education facilities are widespread.
Studies of child mortality have found that the benefits from public health-sanitation services depend on mothers education (Schultz 1984; Rosenzweig and Schultz 1982). If the uneducated women are disproportionately using the hospital services, and since these are the groups which suffer most from child death, the effect of hospital service would be expected to be larger and more significant on child death reduction compared to that of the Blue Nile Health Scheme. One test of the interaction between education and program service (not reported) showed that the uneducated benefit more from hospitals in terms of child death reduction compared with the educated. It is also shown that the services of B. N. H. P program, which are largely of sanitary and protective medicine, are complementary to mother's higher levels of education.
Table 3. OLS Estimate of Child Mortality
|
Covariate |
1 2 3 | |||||||
|
Woman's Age |
||||||||
|
[15-24] |
0.048 |
(2.77)a |
0.050 |
(2.90) a |
0.051 |
(2.96) a | ||
|
25-29 |
0.054 |
(3.03) a |
0.058 |
(3.27) a |
0.056 |
(3.18) a | ||
|
30-34 |
0.044 |
(2.49) a |
0.050 |
(2.83) a |
0.048 |
(2.71) a | ||
|
35-39 |
0.061 |
(3.15) a |
0.071 |
(3.61) a |
0.072 |
(3.71) a | ||
|
40-44 |
0.062 |
(3.13) a |
0.073 |
(3.62) a |
0.073 |
(3.64) a | ||
|
45-49 |
0.085 |
(4.59) a |
0.098 |
(5.15) a |
0.099 |
(5.19) a | ||
|
50+ |
||||||||
|
Wife's Education |
||||||||
|
Primary |
-0.031 |
(2.30) a |
-0.020 |
(-1.46) a |
-0.024 |
(-1.74) a | ||
|
Secondary |
-0.047 |
(2.40) a |
-0.033 |
(-1.65) a |
-0.035 |
(-1.75) a | ||
|
Tertiary |
-0.080 |
(1.70) a |
-0.065 |
(-1.37) |
-0.067 |
(-1.41) a | ||
|
Husband's Education |
Primary |
-0.026 |
(2.07) a |
-0.022 |
(-1.73)b |
-0.016 |
(-1.24) | |
|
Secondary |
-0.045 |
(2.70) a |
-0.037 |
(-2.22) a |
-0.033 |
(-1.95)b | ||
|
Tertiary |
-0.047 |
(1.88) a |
-0.043 |
(-1.66)b |
-0.037 |
(-1.44)c | ||
|
Log (Income/adult) |
-0.013 |
(2.63) a |
-0.011 |
(-2.12) a |
-0.009 |
(-1.76)b | ||
|
Exogenous Community |
||||||||
|
Hospital |
-0.044 |
(-2.18) a |
-0.0034 |
(-1.55) | ||||
|
Beds, 10-2 |
||||||||
|
B.N.H.P |
-0.009 |
(-0.77) |
-0.036 |
(-1.03) | ||||
|
Regions |
||||||||
|
Gezira Main |
-0.014 |
(-0.35) | ||||||
|
Gezira Extension |
-0.006 |
(-0.22) | ||||||
|
Easstern Gezira |
-0.050 |
(-2.59)a | ||||||
|
Blue Nile |
-0.069 |
(-4.40)a | ||||||
|
[Kordofan] |
||||||||
|
Intercept |
0.150 |
(5.14) a |
0.175 |
(5.88) a | ||||
|
R2 |
0.11 |
0.11 |
0.13 |
|||||
|
F |
10.9 |
10.2 |
9.30 |
|||||
|
Joint F-test: |
||||||||
|
Wife's Age |
3.72 |
4.6 |
4.80 |
|||||
|
Wife's |
2.93 |
1.39 |
1.66 |
|||||
|
Education |
||||||||
|
Husband's Educ. |
2.92 |
2.10 |
1.49 |
|||||
|
Programmes |
4.58 |
1.73 |
||||||
|
Regions |
||||||||
|
Hausman Test: Log(income/adults) |
3.21 |
2.69 |
||||||
|
Sample Size |
1187 |
1187 |
||||||
Note: [1] Reference Category.
Figures in parenthesis are t-statistics.
a Coefficient Statistically significant at 5% significance level.
b Coefficient statistically significant at 10% significance level.
c Coefficient statistically significant at 20% significance level
The large magnitude and significance of the effect of father's education on child mortality, shown in specification 2 of table 2, may be over-estimated if education is correlated with some omitted variables that are themselves correlated negatively and significantly with mortality. If, for example, the educated are located in areas where the mortality rate is low, the estimated coefficients attached to husband's education will be biased upward. After controlling for regions of residence in specification (3), a reduction in the magnitude and significance of father's education is observed. Moreover, the geographical differences in child mortality are statistically significant and explain 2per cent of the variation in child mortality. The lowest mortality occurs in Blue Nile (Suki) and Eastern Gezira (Rahad). Irrigated agricultural extension in these areas is a recent development compared to Main Gezira and its Extension. Hence, an endemic disease like malaria might not have reached the level it has reached in Gezira. Rahad is also connected with an internal network of paved roads, which make accessibility to neighbouring urban centres easy in emergencies. Rural Kordofan was subject to desertification and drought in the last decade, and low provision of public services.
The estimated coefficient on the logarithm of permanent income per adult indicates the favourable effect of a rise in income on child survival, presumably because it can purchase better food and health inputs that reduce mortality, since in Sudan in the last decade, medical services have become increasingly purchased in the private market.
The estimated coefficient of the effect of income, however, may be biased and inconsistent if income is measured with error or is endogenous as we argued before. Also, the estimated income coefficient might not be measuring a pure income effect if there are regional or household differences in the prices of health inputs, which are correlated with measured income. The coefficient on income may thus be reflecting combinations of price and income effects. Since no account is taken of the cost of health inputs e.g. distance to the nearest health facility, the income estimate is expected to be underestimated.
Based on the Hausman (1978) specification test, the t statistics on the residual from predicted household expenditure (predicted as in table 4) is 3.2 and 2.7 in specification 2 and 3 respectively, as shown in the bottom of table (3). This is statistically significant at 5% level of significance. Household expenditure in the mortality function, therefore, appears to be endogenous, and other methods of estimation of income effects should be sought.
For this reason Two Stage Least Squares (TSLS) are used to estimate the mortality function where income is instrumental on some of the productive assets of the household as shown in column 1 of table 4. Log expenditure per adult is explained by the husband's education, wife's education, husband's age, wife's age, the regional dummies and assets. The variables measuring assets are categorical, based on ownership and are found to be significantly correlated with income per adult and not correlated with child mortality4. Thus they are used as identifiers of the income function. The results of estimation of child mortality by TSLS are presented in table (4), column 2.
Table 4. TSLS Estimate of Child Mortality
|
Covariate |
Log (income/adult) Mortality | ||||
|
Woman's Age |
|||||
|
25-2 9 |
0.077 |
(0.79) |
O.O53 |
(2.83)a | |
|
30-34 |
0.054 |
(O.51) |
O.O51 |
(2.64) a | |
|
35-39 |
0.093 |
(0.80) |
0.042 |
(2.18) a | |
|
40-44 |
-0.257 |
(-1.96)b |
0.035 |
(1.36)c | |
|
45-49 |
-0.476 |
(-3.43)a |
0.016 |
(-0.50) | |
|
50+ |
-0.464 |
(-3.35)a |
0.034 |
(1.00) | |
|
Wife's Education |
|||||
|
Primary |
0.041 |
(0.54) |
-0.023 |
(-1.56) c | |
|
Secondary |
0.126 |
(1.15) |
-0- |
(-1.22) | |
|
027 |
|||||
|
Tertiary |
0.484 |
(1.86 )b |
-0.027 |
(-0.50) | |
|
Husband's Education |
|||||
|
Primary |
0.258 |
(3.56) a |
0.012 |
((0.67) | |
|
Secondary |
0.190 |
(2.02 ) a |
-0.010 |
(-0.49) | |
|
Tertiary |
0.300 |
(2.10) a |
-0.004 |
(-0.13) | |
|
Husband's Age |
-0.013 |
(-0.94) |
|||
|
Husb.'s Age Square* 102 |
-0.001 |
(-0.05) |
|||
|
Ownership of Assets |
|||||
|
Commercial Vehicle Shop |
0.156 |
(1.87)b |
|||
|
Production Enterprise |
0.210 |
(2.83)a |
|||
|
Farm Machinery |
0.227 |
(1.67)b |
|||
|
Log (Income/adult) |
-0.085 |
(-2.72)a | |||
|
Exogenous Community |
|||||
|
Hospital Beds* 10-2 |
-0.031 |
(-1.30)c | |||
|
B. N. H.P. |
-0.035 |
(-0.93) | |||
|
Regions |
|||||
|
Gezira Main |
0.033 |
(0.72) | |||
|
Gezira Extension |
0.016 |
(0.49) | |||
|
Eastern Gezira |
-0.010 |
(-0.30) | |||
|
Blue Nile |
-0.059 |
(-3.37)a | |||
|
Intercept |
5.33 |
(15.6) |
0.528 |
(3.60) a | |
|
R 2 |
0.33 |
0.12 |
|||
|
F |
26.9 |
8.12 |
|||
|
Joint F-test: |
|||||
|
Wife's Age |
8. 32 |
1.86 |
|||
|
Wife's Education |
1.33 |
0.92 |
|||
|
Husband's Education |
4.58 |
0.56 |
|||
|
Programmes |
1.28 |
||||
|
Regions |
22.1 |
3.83 |
|||
|
Sample Size |
1187 |
1187 |
|||
Note: Figures in parentheses are t-statistics.
a Coefficient statistically significant at 5% significance level.
b Coefficient statistically significant at 10% significance level.
c Coefficient statistically significant at 20% significance level.
The TSLS for child mortality in table (4) are consistent with those obtained previously as far as how parental education affects child survival. This time, however, the effect of husband's primary education is only half the effect of primary maternal education. The magnitude and significance of husband's education is reduced considerably in these estimates. Income now has a larger effect on child mortality and is highly significant compared with OLS estimates. Note that while the health services have the expected effect on child survival, they are less significant. Husband's education is no longer significant as a determinant of reduction in child mortality. Since husband's education works through income the effect of father's education may be underestimated. The variables, which exert a significant influence on child mortality, are the woman's age, areas of residence, income and, to some extent, the hospital services.
The paper examined the impact of parent's education, health services and household standard of living measured by permanent income on child survival in rural Sudan.
Child mortality is found to be inversely associated with parental education. In regressions where only parental education and income are included, OLS estimates indicate that maternal primary education brings a reduction of 3% in average child mortality. With an average child mortality rate of .10 this implies a reduction of 3 per 1000. It is the mother's education, which is more important in influencing child mortality. When program and regional controls are introduced the impact of mother's education becomes statistically insignificant.
Government health services are indicated to improve the chances of child survival. Thus, the sample average hospital beds per capita (curative medicne) is shown to be associated with a reduction of 3-4 per cent in average child mortality. The provision of Blue Nile Health Project services (largely of preventive medicine) brings a reduction in child mortality, but the effect is not statistically significant. The services of the latter seem to be confined to areas which are relatively better endowed in terms of provision of infrastructure and social services. The estimates imply that they tend to benefit those with high income.
The computed average income elasticity of child death is - .l, indicating that doubling income from its sample mean would reduce child mortality by .01. Because of the endogeneity of income in the mortality function TSLS are sought to estimate the effect of income. TSLS overall estimates indicate that income produces a larger and more significant effect on child mortality. Based on TSLS estimates, an income elasticity of child death of - .7 is estimated. Thus a doubling of income would bring a reduction of .05 on average child mortality.
Also, TSLS estimates indicate that father education's effect is not statistically significant. The factors which significantly influence child mortality in TSLS regressions, are: wife's age, income, hospital services and areas of residence.
1. It was not possible to determine annual current income precisely. Although income from wages and salaries, income transfers and home production are observed, the value of services from durable goods and the imputed rent of an owner-occupant house could not be measured for all units. Imputed rent could be determined only for urban residents. Also for some households, where the head is retired or unemployed, none of the sources of current income are reported. The estimates of current income would probably suffer from sample selection bias.
2. Hospital beds availability is measured by dividing the number of beds in a particular rural council, obtained from Ministry of Health Surveys (Research and Health Statistics Department 1990), by total population in the area, using 1983 Population Census data.
3. The Blue Nile Health Project (BNHP) is a joint venture between the Sudan Government and the World Health organisation (WHO). The program was launched in 1979 and began its operations in 1980. The B.N.H.P. has bee successful in establishing improved sanitation and health education services. Safe water supplies, through the installation of deep bore wells, shallow wells with hand pumps, and construction of Horizontal Flow Roughening/Slow Sand Filters (HFR/SSF) with hand pumps, have been made available in all the villages covered by the project. In addition latrine slabs have been provided for all households in the covered areas (B.N.H.P.1989). In fact when the source of water for the community in rural areas is statistically controlled, the presence of the Blue Nile scheme becomes insignificant as a determinant of child survival.
4. The assets which are distinguished as identifiers of income are ownership of vehicles used for commercial purposes like pick-up trucks and lorries, ownership of a shop or grocery and ownership of small scale productive enterprises like bakeries, oil mills or flour mills. The farm machineries are things like tractors and harvesters. All these categories are used for productive purposes and they do not distinguish the household as being engaged in any one particular occupation e.g. farm jobs or commercial and services occupation. More often income from the main occupation is supplemented by engagement in secondary jobs through these activities.
Bhereman and Deolkakir. 1998. Health and Nutrition. In Handbook of Development Economics, vol. 1, edited by H. Chenery, and T. N. Srivassan. Amsterdam: Holland.
Babiker, M. A.B. 1995. The Impact of Liberalisation Policies on Health: Some Evidence from Sudan. In DSRC Seminar Paper Series. Khartoum: University of Khartoum.
Blue Nile Health Project. 1989. The Blue Nile Health Project Annual Report 1989. Ministry of Health, Republic of Sudan.
Cochrane, S. H. 1979. Fertility and Education: What Do We Really Know? World Bank Staff Occasional Papers, No. 26. Baltimore: Johns Hopkins University Press.
Cochrane, S. H., Rosenzwig and Schultz. 1982. Parental Education and Child Health: Intra-country Evidence. Health Policy and Education 2: 213-249.
Deaton, A. and J. Muellbauer. 1980. Economics and Consumer Behaviour. New York: Cambridge University Press.
EI-Yamen, S. A. 1995.Comparison of Child Mortality Rates Obtained from 1993 Advance Census Data and PAPCHILD Survey. Report of the Census Data Users' Meeting. Khartoum: Department of Statistics.
Educational Statistics Section. 1987. Educational Statistics: Academic Year 1985/86. Khartoum: Ministry of Education, Republic of Sudan.
Hausman, J. 1978. Specification Tests in Econometrics. Econometrica 47. (1): 153-162.
Okojie, C. E. E. 1991. Y'Fertility Response to Child Survival in Nigeria: An Analysis of Micro Data from Bendel State. In Research in Population Economics, edited by T. P. Schultz, vol. 7. Greenwich, CT: JAI Press.
Maglad, N. E. 1994. Fertility in Rural Sudan: The Effect of Landholding and Child Mortality. Economic Development and Cultural Change 42 (4): 761-72.
Rozenweig, M., and T. P. Schultz. 1982. Child Mortality and Fertility in Columbia: Individual and Community Effects. Health Policy and Education 2: 305-48.
Population Census Office. 1994. Fourth Population Census of the Sudan 1993. Khartoum: Department of Statistics, Republic of Sudan.
Schultz, T. P. 1976. Interrelationship between Mortality and Fertility. In Population and Development edited by R. G. Ridker. Baltimore: Johns Hopkins University Press.
________. 1981. Economics of Population. Addison-Wesley.
________. 1984. Studying the Impact of Household Economics and Community Variables on Child Mortality. In Child Survival, Population and Development Review, edited by W.H. Mosley and L.C. Chen. Supplement to vol. 10.
Thomas, T., and Strauss. J. 1993. Prices Infrastructure Household Characteristics and Child Height. New Haven, CT: Yale University.