INTRODUCTION

Human development is always a central concern in any society. Human capital, on the one hand, constitutes an outcome of the social economic structure. On the other hand, human capital plays a positive and central role in economic growth and development. It is a direct input of the production process. For human capita itself, health is a very important component. Health care programs and policies are not only valuable in improving health itself, but also have positive impacts on output. For this reason, in most economic development programs, health improvement has been always taken into account.

     In Vietnam, health status in general and child health status in particular has been impressive in parallel to its economic development. Over the past 20 years, the prevalence of malnutrition has reduced significantly, for example, the stunting rate of children under five fell from nearly 60 percent in early 1980s to 38.2 percent in 1999(NIN, 2000). In comparison with the international level of economic development, Vietnam has low levels of mortality (Paul Gertler, 1994) with life expectancy at birth in Vietnam being 65 and 67 percent, respectively for males and  females (Human Development Report, UNDP, 2000).

     Vietnam, however, remains one of the world^’s poorest countries. Nearly 37 percent of its population still live in poverty. More than 38 percent of children under five are stunted and underweight (Human Development Report, UNDP, 2000). Indeed, Vietnamese children are among the most malnourished in the developing countries. The serious problem of child malnutrition in Vietnam leads the author of this thesis to be interested in the topic: “Explaining the malnutrition of preschool children in Vietnam”

      There have been a number of attempts over the past years to model the nutrition outcome in Vietnam. The Institute for Protection of Children^’s Health (Hanoi), in collaboration with some other Institutes, has done some multivariate analysis of maternal factor influencing the occurrence of low birth weights in northern Vietnam (WB,1999). The second study was done by Paul Glewwe using the VLSS97/98 data in a linear regression to analyze the income effect on children’s nutritional status in Vietnam, (Glewwe, 2001).

     This thesis differs from those studies in both scope and approach. Firstly,  children who are under five years of age will be the focus of this analysis. Secondly, instead of using child weight or Body Mass Index as dependent variables in the case of past studies, a more relevant criterion of height for age and weight for height Z score will be applied as dependent variables in regression analysis. Thirdly, more comprehensive variables of individual, household and environmental characteristics will be included in the empirical models.

Chapter I.        Theoretical framework

      I. Concepts and definitions

      1. The meaning of nutrition versus malnutrition.

      Nutritional status of one person is the condition of the body resulting from the intake, absorption and utilization of food (FAO, 1982). Nutritional status is thus the stock of energy that one person accumulates and absorbs over time.

      Malnutrition is conversely defined as nutritional disorder or condition resulting from faulty or inadequate nutrition. According to Jean Mayer (1976), malnutrition is classified into four types:

      -Over-nutrition: this is a condition of one person who consumes too many calories.

     -Dietary deficiency: this body condition results from consuming a diet that lacks sufficient amounts of a particular nutrient such as vitamin or mineral.

     -Secondary malnutrition: This is condition that an individual suffers when he or she is unable to digest or adsorb some food successfully.

     -Under-nutrition: this is a condition in which an individual does not consume enough food.

      2. Measuring nutritional status.

      To measure nutritional status of children, most studies use anthropometric indicators.

      The anthropometric indicators can be expressed in three different ways: Height for age, weight for height, and weight for age (Healy, 1986).

      Z – score is calculated as follow:

      In which:

                     Z_it is the actual value of a child i at the age of (month/ year) t,

                     (ex: height )

                    Z_rt is the average value of the reference children group of the same age t                  

                     SD is the standard deviation of the distribution of value for a reference population of healthy children of that age.

      According to World Health Organization (WHO), Z- score of – 2 or below signals malnutrition.

      This thesis uses height for age Z- score and weight for height Z- score to access Vietnamese preschool children’s nutritional status in long-run and short-run. Based on World Health Organization recommendations, the author will use the reference standard of U.S. National Center for Health Statistics (NCHS) in calculating Z- score.

      3. The consequences of malnutrition.

      Malnutrition causes a great deal of human suffering both physical and emotional. It is a violation of a child’s human rights. It is associated with more than half of all deaths of children worldwide (Smith, 1994). And it is a major waste of human energy. Adults who survive malnutrition as children are less physically and intellectually productive and suffer from more chronic illness and disability (UNICEF, 1998). The personal and social costs of continuing malnutrition on its current scale are enormous.

      II. Theories of health and nutrition

      1. Income- Health Linkages

      Many studies have shown a positive correlation between income and health status, both on cross – sectional and longitudinal bases. For example, Gertler and Vander Gang (1990) report cross – sectional analysis of the links between per capital gross national product (GNP) and health status for a sample of 34 countries in 1975. They find that, on average, a 10 percent increase in income is associated with one extra year of life expectancy, an 8.3 percent lower infant mortality rate, a 14.2 percent lower child mortality rate, and a 1.5 percent lower crude death rate.

The structural relationship between income and health status has shifted over time.  The correlation between income and nutritional status continued to be positive and become more stronger.

      There are two basic explanations for this shifting relationship between income and health. The first is that it has become easier, or cheaper, to attain and maintain given level of health over time because of technological innovations and investment in public infrastructure.

      The second possible explanation for the shift in the relationship is that individual’s preferences have changed over time, and that, for a given level of income, individuals have become more concerned about health.

      Clearly then, it is the use of income that is of importance in the determination of health status. As a general rule, we would expect that, as income increases consumption of health improving goods and services also increases.

      Perhaps the most obvious input into good health is medical service, including both curative and preventive treatments. Some studies have found that the level of education of individuals is a significant determinant of health status, and particularly that improvements in parent’s literacy are correlated with improvements in children’s health. (Leslie 1980, Strauss 1987).

      The final point to note with regard to this discussion of the production of health is that, despite the existence of some externalities (for example, related to immunization), one’s health depends primarily on one’s own consumption of medical care, food, shelter, clothing, water, sanitation, and so forth.

      2. Theory of intra-household resource allocation.

      Most studies rely on the theory of intra-household - resource allocation as a theoretical foundation to build the models of determinants of health. The underlying reason is that this theory can be used to explain the household demand function for various goods. Health-demand function has been frequently used based on this theory.

      Beginning with a simple static model of household behavior, we can draw the household welfare -function:

        W = w [U₁ (X, L,q, µ, e)…U_M (X, L, q, µ, e)]

      In which: U is individual utility function

                      X is commodity consumption

                      L is consumption of leisure

                      q is home produced goods such as education, health and the like.

                      e is unobsered heterogenity.         

      Based on collective model of household decision-making, we can draw the household demand function:

                     R = g (p, w, y₁…y_m, m, e)                  (Strauss, 1988)

      In which: p is price

                      W is wage

                      Y is non labor income                                    

      If instead, basing on unitary model of household, that is all household members have common preferences or that one member dictates all allocation decisions, the demand functions depend not on individual non-labor incomes but on their sum:

          R = g (p, w, _m y_m, m, e)                                     

      Typically, non-labor income represents only a small fraction of total resources available to a household for consumption and investment. Further more, non-labor income is unlikely to be measured without error. Thus, the demand function is usually tested in the case of total income:

                     R = g (p, _m Y_m,m, e)

      III. Nutrition models and empirical results

      1. A basic household anthropometric production function.

1.1.            Model

       Based on the model of intra-household resourse allocation, Thomas et al (1995) draw the health demand function:

          S_it = S (C_i, C_h, K, P_i, F_i, W_i, d_i)

In which:

          S_it  is health demand function

          C_i  is individual characteristics

          C_h is household characteristics

          K is environmental characteristics

          P_i is the set of prices of goods

          F_i is non- wage income

          W_i  is wage

         d_i is unobsered heterogeneity

       Assuming that expenditure may be a better indicator of the available resource in the long run and is regarded as a resource measure. To control household size, we can use household per capita expenditure. Thus, the function  now becomes:

         S_it = h(C_i, EP, C_h, K, P_i, d_it)                                   

      Where EP is household per capita expenditure.

      1.2. Empirical results.

      Thomas el at (1995) apply this model to test the health of children in Ghana. He find that, child health are positively associated with number of doctors, the availability of drugs in communities. There are no visible differences in the impact of health services on child health between the poor and non poor. The correlation between expenditure and child health is positive and significant. The children of better educated parents are healthier. There is a strong correlation between parental and child height which is attributable in part to genetic effects.  

      2. Intra-household resource allocation model for health.

      2.1. Model.

      This model is built under the assumption that in the current period, the household maximizes a quasi – concave utility function which depends on the consumption of commodities and services, X_t, the leisure, L_i, ,individual health status, q_t and household characteristics, Z_ht ( Lavy et al,. 1992) Then the household chooses to

          Max_xlq U (X_t, L_t, q_t, Z_ht, j_t)                                       (1)

      Where:  j_t represents unobserved heterogeneity in preferences.

      The reduced form anthropometric outcomes function is:

          H_it = g (Z_i, Z_h, Z_c, Y, e)                                            (2)

      Where :     Z_i : individual characteristics

                     Z_h : household characteristics

                     Z_c : community characteristics

         Y  : household income

         e : unobserved heterogeneity in anthropometric outcomes.

      To control household size, Thomas (1992) suggests to use per capita expenditure, PCE, instead of Y. Thus, the function (2) now becomes:

          H_it = g(Z_i, Z_h, Z_c, PCE , e_it)                                  

      2.2. Empirical results.

      The model is used to test child health in Code d’Ivoire

     In Thomas multivariate regression, children tend to be taller in communities with more doctors, the availability of drugs.

     With respect to local market prices, price elasticity is larger in the rural sector than in the urban sector. Price rises also tend to have negatively larger impact on the weight for height of children in poorer households.

      The household resource has a positive impact on child health. The height of the senior male and female in the household also has a positive impact on child height reflecting both genetic and family background influences. While child height is unrelated to the level of education of the senior female in the household. It is positively correlated with education of the senior male, especially in urban area.

      IV. CONCLUSION.

In this chapter, several conclusions can be drawn: Firstly, there exists evidence that health status and income are positively correlated. However, apart from income, health may very well be primarily affected by several other factors such as medical care, shelter, clothing, water, sanitation and so forth. Such factors are in turn determined by individual, household, environment characteristics. Secondly, the theory of intra-household resource allocation constitutes as the underlying foundation in building most health models. In those models, health is considered to be mainly determined by four factors: household income, individual, household and environment characteristics.

      With regard to empirical results, most empirical studies show that more convenient and available health facilities, increased education, increased quantity of food available... are significantly associated with health status (Thomas, 1992, 1995). However, it should be noted that the statistical significance of each coefficient is not equal to health status of male and female as well as to that of people in rural and in urban.

CHAPTER. II. CHILDREN^’S NUTRITIONAL STATUS IN VIETNAM

         I. CHILDREN^’S NUTRITIONAL STATUS IN VIETNAM

             1. General background

The renovation process initiated by the Vietnamese government has brought about noticeable achievements. These achievements in economy have indeed made remarkable progress in poverty reduction and hunger eradication.

      Vietnam, however, remains one of the world^’s poorest countries. More than 37 percent of its population still live in poverty. About 35 percent of children under five are stunted and underweight. Nearly 40 percent of adults aged 18 and older are chronic energy deficiency (NIN, 2000).

Table 1.1: The Prevalence of Malnutrition in Vietnam

         Source: VLSS97\98

 Note: Stunting, Wasting, and Underweight are calculated for all children under 155 months. (Under 15 years)

      For the whole nation, about 38.7%, 9.8%, and 36.7% of children under five are stunted, wasted, and underweight, respectively (NIN, 2000, p.23). The malnutrition in Vietnam is mostly in two forms, severe and moderate, that  are degree II and I. Malnutrition is most serious in Central Highlands (49.1%) and less acute in Southeast (29.6%), ( NIN, 1999, p.24,25).

2.      Malnutrition by age.

                                                Figure 2.1: Malnutrition rate by Age

Source: Based on VLSS97\98, (See appendix 2.3)

      Figure 2.1 shows the patterns of wasting, stunting and underweight of the Vietnamese children. Stunting increases rapidly from the birth to the age of two, after which it levels off at a high rate of incidence. Wasting peaks in the second year of life, then declines gradually. Both stunting and wasting patterns indicate that poor nutrition is most common during the first two years of life, especially from 6 months to 24 months, the time that children are usually weaned and thus exposed to infectious diseases through liquid and solid foods.

Source: Based on VLSS97\98 (See appendix 2.3)

      For all the age groups, malnutrition rate is far higher in rural than that in urban. This is explained in part by more availability and convenience of health services, higher income per capita, higher education level and the like.

      In general, the level of malnutrition is higher for male than for female, say, the stunting is 43.86%, 39.08% for male and female, respectively. 

Source: Based on VLSS97\98 (See appendix 2.3)

      Since the differences are small in magnitude, it is probably unlikely that it represents parental discrimination against girls in the intra-household allocation of food.

     With respect to gender of household head, malnutrition level of children is higher in male-headed households than that in female headed households. This may very well be explained by that headed females tend to spend a large share of resources on food or other good related to health improvement than do headed males.

      3. Malnutrition and income linkage.

      Together with social-economic development, the level of malnutrition in Vietnam has dropped significantly in recent years. In 1982, about 60 percent of the Vietnamese children under the age of five were malnourished. In 1992, the rate of stunting has declined to 53 percent and in 1998, this rate is 35.9 percent.   

      Source: VLSS1997\98, NIN 2000, GSO 1985, 1990, 1992-1999 (See appendix 2.5)

      Note: GDP/capita/year is calculated at constant prices of 1989 (thousand VND).

      The relationship between income and nutritional status can be examined by expenditure quintile. The poorest households have the highest malnutrition rate, and in contrast, the richest households have the lowest one. Obviously, the greater the income, the better the person’s nutritional status. This relationship is vividly illustrated in Figure 2.5

       Source: VLSS1997\98 (See appendix 2.7)

      Note: Expenditure quintile 1: poorest, 5: richest.

      The relationship between income and nutritional status can be examined by expenditure quintile. The poorest households have the highest malnutrition rate, and in contrast, the richest households have the lowest one. Obviously, the greater the income, the better the person’s nutritional status. This relationship is vividly illustrated in Figure 2.5

      4. Malnutrition and region linkage.

     Figure 2.6 shows the positive relationship between Food Poverty Line and Malnutrition. Among the seven regions, Central Highlands has the largest Food Poverty Line and stunting rate while Southeast has the smallest Food Poverty Line and stunting rate.

      Source: VLSS97\98; GSO, 2000 p,.116 & 189; Haughton J, 1999, p.6 (See appendix 2.8)

      Note: Region 1: Northern Uplands,    2: Red River Delta     3: North Central Coast,  

                          4: Central Coast            5: Central Highlands,  6: Southeast   7: Mekong Areas,

                         8: Rural Areas               9: Urban Areas,           10: Total Country

       Regarding child gender, Comparison of malnutrition by region and gender reveals that both in rural and urban, the rate of malnutrition is greater for male than for female, the rate of male and female stunting is 38 and 45 percent in urban and rural, respectively. In all seven regions, male exhibits a larger proportion of malnutrition than female does. The deviation of malnutrition between male and female is greatest in North Uplands and is smallest in Central Highlands.

In every region, the level of malnutrition of children is smaller in female-headed households than that in male headed households. This may be explained by the fact that food expenditure in female-headed households is generally higher than that in male-headed households.

       5. Malnutrition by mother education and mother age

Many studies in other developing countries have shown a positive association between child malnutrition status and mother education.

      Yet, in Vietnam, the situation seems to be contrary. Surprisingly, mother education seems to exert too little influence on child health.

     The negative effect at training school may be explained by that people who learn at training school are likely to have bad result when they were at secondary school. Thus, their perception about health care may not improve much even when they learn at higher level.

      The pattern of child malnutrition by mother age group can be illustrated in Figure 2.7. It can be seen that children tend to have higher nutritional status when mother age increases. Although the association of child health and mother age is not clear, this pattern implies that the older women may have better health than do the younger women, so children can benefit from healthy genetics of their parents. Furthermore, older women tend to have better living standard, so they are able to look after children better.

Figure 2.7: Malnutrition by mother age

                      Source: Author^’s calculation based on VLSS97\98 (See appendix 2.11)

      6. Malnutrition by water sources.

      The positive relation between child health and sanitation can be examined through water sources. The water sources can be classified into three levels:: water from tap, well and others. Obviously, people who are able to access to water from tap may be likely to have higher nutritional status, because of higher sanitation. In contrast, those who use water from well or other sources may have lower nutritional status, since they are easy to be infectious.

       II. COMPARISON OF MALNUTRITION BETWEEN VLSS92\93 AND VLSS97\98.     

1.      Comparison of malnutrition by age.

      Using data sets of VLSS92\93 and VLSS97\98, the patterns of children^’s malnutrition in Vietnam can be illustrated as in figure 2.9. On average, the proportion of children who are stunted is much less in 1997-98 compared with that in 1992-93 (the rate of children stunted is 41.5% and 54.6% in 1998 and 1993, respectively). The fluctuation of malnutrition between the VLSS92\93 and the VLSS97\98 is much similar. However, Figure 2.8 shows that the prevalence of malnutrition in 1997-98 has declined significantly for all age groups.

2.      Comparison of malnutrition by expenditure.

 The improvement of children health can be examined by comparing the relative change of children malnutrition and expenditure per capita in 1993 and 1998. On the whole nation, expenditure per capita in 1997-98 increased by 1.43 times relative to that in 192-93. At the same time, children’ malnutrition in 1997-98 reduced by 0.76 times compared with that in 1992-93. Relative change is greater in urban areas in comparison with that in rural areas (table 2.2). This suggests that, the difference of expenditure and malnutrition^’s incidence between urban and rural increases over time.

Table 2.2: Malnutrition and Expenditure

          Source: VLSS92\93, VLSS97\98; GSO, 2000, p.116 & 269

In conclusion, children^’s malnutrition has improved remarkably in 1997-98 compared with that in 1992-93. However, the difference of income, per capita expenditure as well as malnutrition between urban and rural is increasing day by day. This suggests that appropriate policies should be made to develop rural areas.

      IV. COMPARISON OF MALNUTRITION BETWEEN VIETNAM AND OTHER DEVELOPING COUNTRIES.

     To compare the child malnutrition between Vietnam and other developing countries, the author use GDP per capita and Human Development Index.

     In comparison based on GDP per capita with other countries with similar income level, it can be seen that,

Figure 2.9: Malnutrition by GDP per Capita

Source: Based on appendix 2.

higher income is generally associated with lower children malnutrition, but the deviations from this trend suggest that child malnutrition is affected by other important factors. Those factors may be the extent of public spending on health and nutrition, agricultural subsidies, women^’s literacy and fertility policies.    

      Using HDI criterion, in a similar HDI group, Vietnam has a surprisingly highest proportion of underweight children. What this means is that, among social indicators, such as education, life expectancy, the prevalence of Vietnamese children’ malnutrition is the most serious issue.

                                      Source: Based on appendix 2.15

      To summarize, the comparison between Vietnam and other developing countries using  GDP per capita and HDI shows that, in a similar income country group, the children malnutrition in Vietnam does not exhibit a better indicator. Furthermore, using HDI comparison, Vietnam seems to be the most serious malnourished nation. This forcefully suggests that urgent policies should be paid to reducing malnutrition level.

    V. CONCLUSION.

An important conclusion can be drawn in this chapter is that malnutrition reduction is associated with income increase. Nonetheless, malnutrition is also caused by many other factors such as, region, water source and so on. The previous presentation has shown clearly that, the rates of malnutrition are different across regions, and normally higher in rural compared with in urban. The prevalence of malnutrition is also more serious for male than for female, especially among the poor households. The pattern of children malnutrition is also not smooth by age group, gender of household head, expenditure quintile. Surprisingly, mother education and mother age seems not to exert much influence on child health. This is really contrary to other developing countries, which shows a positive relation between child health and mother education. Among the environment variables, water sources play considerable role. One crucial question is that how each of such factors affects child health given the rest constantly controlled? This question can be answered by using regression models in the next chapter.    

Chapter III.  ECONOMIC EVIDENCE OF MODEL SPECIFICATION

        I. MODEL SPECIFICATION, DATA SOURCE AND VARIABLE DESCRIPTION.

1.      Model specification

This chapter uses linear and logistic models to estimate the effect of household and environment characteristics on children nutritional status, as measured by height for age Z score. The linear regression of interest to be estimated is:

   HAZi = bo + byX_i^e + bcX_i^c + bhX_i^h + blX_i^l + ei                                  (1)

        Where:   HAZ_i is the height for age Z-score of a child i    

               X­_i^e   is the per capita expenditure of household i     

               X_i^c is a vector of characteristics of child i

                        X_i^h is a vector of the characteristics of the household in which child i lives

                        X_i^l is a vector of the characteristics of the local community in which child i lives

                        e_i  is an error term that measures unobserved characteristics of the child and of his or her household or local community that may affect height for age.

      The child is defined as malnourished if his or her HAZ <= -2, otherwise, he or she is called “normal”. Thus, the continuous model can be adjusted to be logistic model as follows:

                        P_i = E(Y=1/X _i) = 1/(1+e^-Zi)                                                      (2)

      Where:  Z _i = bo + byX_i^e + bcX_i^c + bhX_i^h + blX_i^l + e_i

                       Y_i = 1 if a child is called malnourished.

                       Y_i = 0 if a child is not called malnourished.

                       P_i is the probability of a child being malnourished.

      (2) can be written as follows:

                       L_i = Ln(P_i/(1-P_i) = Z_i                                                                    (3)

       Where: L_i is the log of the odd ratio in favor of being called malnourished.  

                    Other variables are defined as in the model (1).

      To estimate the model, we use the maximum likelihood method (Gujarati, D, 1992, pp. 556).

      2. Data source and description of variables.

      This thesis uses data extracted from the second Vietnam Living Standard Survey to estimate child health. Of the 6000 households surveyed, there are 2004 children under 6 years of age.

      The explanatory variables include: Expenditure per capita, Child age (Agecat), Child gender (Chgder11), Father^’s and mother^’s education (Fedu and Medu), Father^’s and mother^’s age (Fage and Mage), Gender of household head  (Sex12), Household size (Hhsize), Type of household (Farm), Region variable (Urban12), Ethnicity (EthD11), Religion (RelD11), Water sources (Water11, Water12, Water13), Toilet (Toilet11), Distance to nearest health center (Distance), Number of doctors, nurses available in nearest clinic or hospital (Doctors).

       II. Hypothesis

       Basing the results in chapters 1 and 2, it is expected that:

       Expenditure per capita, parental heights, parental education, mother age, sanitation, availability of health services are all positively associated with child health. In contrast, age of child, size of household, religion, rural area are negatively associated with child health.

      III. Methodology

       The methods of selecting variables and processing data are as follows:

-         Expenditure is regarded as a better indicator of permanent income (see, for example, Bouis 1994 & Lavy 1995), so per capita total expenditure is used as a proxy for income.

-         Heights of mother and father are used as a proxy for genetics. Water sources and types of toilet are used to present sanitation. Distance to nearest clinic and number of doctors, nurses constitute the availability of health services. Many variables are converted from continuous to indicator ones for comparison

-         Data is processed by STATA software. Model (1) is first run by ordinary least squared method, and is processed by three steps. In the first step, model (1) is run with all variables. In the second step, variables those are not statistically significant in the first step are dropped out; the model then is run with remaining significant variables. At this stage, some variables turn out to be insignificant, thus they should be dropped out in the next step. In the third step, the remaining variables are run and the results prove that all variables are now statistically significant.

-         Results from linear regression at the third step are tested for specification error and heterocedasticity. For survey linear and logistic regressions, testing of specification error is also used.

      III. Results and Interpretation

      1.  Linear regression.   

      Using model (1), the children^’s nutritional status is estimated for all children in the whole country and then is separately estimated for rural and urban areas, for males and females.

-         Expenditure impacts

Table 3.2.  Expenditure, child and environmental impacts