Introduction

In the previously central-planned economy of Vietnam, there were almost no child laborers. At that period of time, a great proportion of Vietnamese families took part in cooperatives. The families often fulfilled their assigned amount of work with not much difficulty. As a result, they did not have to make their children work early. More over, the education system was subsidized so children went to school with almost no cost. Children spent most of their time studying and the main economic activity that children took part in was to do the housework. In general, the private sector had not developed and the economic activities of families were in small scale, so the was no demand as well as supply of child labor.

The renovation process initiated since 1986 has brought about crucial socioeconomic changes in Vietnam. The "Doi Moi" process started with the transfer of land to the families. The old kind of cooperatives has been replaced by the confirmation of economic role of families as independent economic unit. Trade and price liberalization has encouraged peasants to develop production. However, most of the agriculture activities are still labor intensive, so the demand for child labor appears. In addition, with the elimination of the subsidized education system, the cost of education becomes expensive for many families. Market mechanism has also made the gap between the rich and the poor widen. All the said reasons have contributed to the relatively swift development of child labor in Vietnam.

To take care of, to educate and protect children have always been the profound concern of the Vietnamese State, the families and the societies as a whole. Vietnam is one of the first nations in the world that ratified the 1989 United Convention of the Rights of the Child. The Law on Labor of Vietnam, as well as other under law documents has had conditions strictly regulating the use of non-adult labor, especially the children aged under 15. The government has also committed to allocate the budget and ODA to basic social services, especially education development, heath care and reduce malnutrition. However, despite these undertakings, child labor is still an integral part of recent Vietnamese economic activity.

   To date, there have been some researches considering children's work and education in Vietnam including: Jere R. Berhman and James C. Knowles (1999), Pete R. Moock, Harry Anthony Patrinos and Meera Venkatamara (1998), World Bank (1997), Vu Ngoc Binh (1997), Truong Sy Anh (1998). However, most of the previous studies were not carried out in a comprehensive and systematic way on the child labor topic. Therefore, our understanding on child labor issue is still limited. In attempting to fill this gap, our research take a study of children's participation in economic activities, focusing on 6 to 17 year old children. The core objective of this study is to find an answer for the central research question.

                What are the major determinants of child labor in Vietnam?

    The study firstly focus on reviewing theoretical and empirical issues relating to factors affecting child labor. Then we shall try to identify determinants of and quantify their effects to child labor in Vietnam based on data from Vietnam Living Standard Surveys. Finally, some policy recommendations to reduce child labor in Vietnam in the coming years are provided based on the analysis results. The scope of this research is to examine determinants of child labor and in Vietnam in 1997/1998. The unit of analysis is the child at home ages 6 to 17 who reports a direct or indirect relationship with the household head or his spouse.

      In addressing research question, qualitative method is used, including statistical and descriptive analysis, review of historical trends, and comparative methods. Besides that, quantitative method is extensively used.

                  Data used in the analysis extracts from the VLSS that was conducted by General Statistics Office with the technical assistance of World Bank.

       The research includes an introduction and four chapters. Chapter 1 represents an analytical framework. Chapter 2 provides an overview of child labor and schooling in Vietnam. Chapter 3 presents an econometric model of child labor. Chapter 4 provides some conclusions and implications of our findings.

Chapter 1: Analytical Framework

      1.1 Major concepts

      Childhood. The term “child” generally encompasses young people dependent on others for survival and subsistence. Childhood is often defined in relation to adulthood and includes a set of roles and expectations, in addition to age boundaries, distinguishing children from adults (Hull, 1981). While childhood seems like a simple, almost self-defining concept, the range of ages used to delineate children in research varies considerably. Further, a number of researchers point to problems in using age to mark childhood, particularly in a developing country context where transitions to adulthood may be based more on the life stage of an individual or needs of the family than biological age (Hull, 1981, Silva, 1981). In Vietnam, the Labor code (1994) stimulates 15 years of age is the minimum age for children to take part in economic activities (article 120). Also according to this code, a minor laborer is one under 18 years of age (article 119). In the Law on the protection, care and education of children (1991), children are defined as citizens under 16 years of age (article 1).

      Child labor. In this study the concept of “child labor” will be defined as children aged 6-17 years who are economically active, or in other works: they are involved in economic activities. Economic activities can include works for pay outside of the household, works for the household in agriculture and works for the household in self-employment or a household run business. There are some reasons for this selection. Firstly, it dues to the difficulty of making separations between different kinds of child labor – harmful and harmless, for example - from available data (VLSS 1997/1998). Secondly, time allocated to work during childhood should reduce potential time to education and leisure and so may have harmful consequences, regardless of types of child labor. Thirdly, the economic and social benefits of children's economic activities, in general, may not exceed the negative effects on children (i.e. child labor may interfere with a child's education, in turn this would reduce the productivity in the future).

     1.2. Economic development and the contribution of children to their families

      Process of economic development, such as industrialization, mechanization of agriculture, urbanization, compulsory educational laws, development of social insurance for elderly or parental investment in non-human economic assets might undermine the importance of children's economic contribution to their families. First, technical changes which substitute capital for labor reduces the demand for unskilled labor both in agricultural production and in labor-intensive industries. In other words, the sources of demand for children's labor tend to diminish with technical change in agricultural production or industrialization (Barnhouse & Briggs, 1993; Barnhouse & James, 1992). Additionally, rates of return to schooling increase in urban areas with more industrialization while the same occurs in rural areas with technical change in agriculture. Then, in the trade-off between quantity and quality, altruistic parents decrease child labor and increase the time devoted by their children to school (Becker, 1993). Second, compulsory educational laws might raise the costs of children by imposing a given quality for children (Becker, 1993).

                1.3. Theoretical model

             The static neoclassical model of resource allocation as extended by Gary Becker (1965) provides the theoretical motivation of the reduced form model estimated in this study. The Becker time allocation model is applied to the labor supply of the family and to parental decisions about children's work. Household resource allocation and consumption decisions are the result of utility maximization subject to constraints. The model implicitly assumes (i) that all resources are pooled and (ii) allocation of all family resources follows some common rules, in this case, presumably, determined by one parent or both parents acting jointly.

                The Becker household allocation model is used to derive reduced form functions. Parental utility depends upon the educational attainment of each child, care for preschool-aged children, and other goods. Each family produces and consumes these commodities, subject to time and budget constraints, in such a way to maximize parental utility and thus family well being. A reduced form approach is adopted here in recognition of the prevalence of corner solutions - when children work in the labor force, for example - as well as the absence of variables which identify a structural model. Inspire of certain limitations, the model provides us with a coherent picture of household decision making.  

In this model, family is considered to function like a small firm, allocating their time and resources in order to maximize family well being, which is increased by the consumption of market-purchased goods, home-produced-goods and leisure. A more specific specification is deduced from the general model and is used to derive reduced form functions. A reduced form approach is adopted here in recognition of the prevalence of corner solutions – when children work at home and not in the labor force and vice versa, for example – as well as the absence of variables which could identify a structural model. Consider a simple model of a family with a specific demographic composition: two parents and three children. For this example, assume child 1 is 14 years old, child 2 is 10 years old, and child 3 is 5 years old. Characteristics of the different individuals who are components of the household unit are expected to influence child activities in consistent and predictable patterns, holding constant all other child and family characteristics. Family member i, some child between ages 7 and 14, is our focus and the unit of observation. Call this child the “subject” for convenience, since the word “child” can refer both to a type of relationship and to an individual young person. The model can help to see how (i.) gender of the subject, (ii.) gender of the subject's siblings, (iii.) age of the subject's siblings, (iv.) family size, (v.) the mother activities, (vi.) the father and other adult male and (vii.) family income affect the allocation of time of children between household work and market work.

                1.4 Empirical evidence

                Firstly, the study of Malathy Duraisamy (2000) on child schooling and child work in rural areas of India based on a household survey that covers all rural India. She finds that children's participation in work increase at a diminishing rate with the age of the child. Boys are likely to work than girls. Father's and mother's education levels have negative effects on work over the reference group, illiterate, and the effects are statistically significant at the 5 percent level. The children of the educated parents are less likely to be working compared to children of illiterate parents. One of the interesting findings is that mother's education has bigger and stronger effects than father's education.

                Secondly, the study of Harry Anthorny and George Psacharopoulos (1997) on child labor using the 1991 Peru Living Standard Measurement Survey, focusing on students who also work. The results are that being indigenous and rural residences both have a great impact on the likelihood that the students also work. Father's schooling is also significant. Being male is strongly and positively associated with child employment. Age, private school attendance and number of rooms are not statistically significant. In their study, a specific attention is paid to the issue of family size and sibling age and activities. The results show that the coefficient of number of siblings not in school is insignificant, but the number of siblings between 0 and 6 year of age is significant, having a positive effect on child labor. This implies that older siblings work to support younger siblings. It may also imply household specialization such as older siblings work to support younger siblings. While the number of siblings may not have an effect on participation in the labor force (all the other coefficients that denote number of siblings are insignificant), the number of young siblings does.

                Chapter 2: Child labor in Vietnam

                2.1. Economic reform and emerging of child labor

                The centerpiece of the Vietnamese government's socioeconomic development strategy since 1986 has been “Doi Moi”, the reform process aimed at transforming the Vietnamese economy from a centrally planned system to one that is dynamic and market-based. “Doi Moi” brought dramatic changes in the general socioeconomic situation in Vietnam. The reform process started first in agriculture with the reallocation of land to households as autonomous economic units. Combines with price and trade liberalization, these given great incentives to farmers to produce. With limited initial capital and underdeveloped market institutions, households often have to rely on their own resources and substitute inaccessible inputs by labor. This creates ground for child work and child labor to grow rapidly.

                Despite remarkable improvement of macroeconomic, the lack of an effective social safety net also exposed many Vietnamese, especially rural dwellers, to various excessive risks. As a result, a number of problems emerge such as the widening disparity between urban and rural areas, increasing inequality among population groups and geographical regions, the degradation of education and health services. The drop-out rates among school children, including primary school pupils increased because their families could not afford their schooling. All these problems also contributed to the increased number of working children and the flow of children migrated to earn a living.

              2.2. Trends in child labor in Vietnam

        The Vietnamese economy is undergoing a rapid transformation from planned to market economy. This transition has been associated in more recent years with a large reduction of child participation in the labor force. To analyze the Vietnamese case, as mentioned above, we use the two surveys (VLSS 1992/1993 and VLSS 1997/1998) carried out by the Vietnamese Government and the World Bank. After making in-depth analysises, a briefly summarization of the main findings of changes in the child labor supply in Vietnam is as flows:

                . In recent years, We observed a large decrease in the rates of child labor in Vietnam. This change has been brought about by a reduction both in the number of children working (full-time and part-time) and in the average time spent at work by these children. Fertility has also been declining and this also has contributed to the declining share of child labor.

                2.3. Child labor and schooling in 1997/1998

                Regarding the child labor and schooling in 1998, non-enrolled children were firstly mentioned. At all ages (6-17), enrollment rate is higher in urban areas and boys' school enrollment is higher than that of girls. These regional and gender differences result in the difference in education achievement at age of 17. On average, a rural child has received 6.67 years of education whereas an urban child has received that of 8.43 years. Boys and girls have almost the same years of education (7 years) due to the gender difference is not large.

                Child labor rates are quite different in 7 regions of Vietnam. The rates are highest in Northern Uplands and North Central Coast Regions and lowest in South Central Coast and Southeast regions. The regional different might be explained by the difference in socioeconomic development of each region.

                In term of gender, the proportion of girls who participating in work is higher than that of boys both in rural and urban areas, however the gender difference is not large. As regards rural-urban difference, child labor rate in rural areas is much higher than that of urban areas. This indicates that child labor in Vietnam could be considered largely as a rural issue. As children get older they are more likely to work. There is a sharp increase in child labor rate when we move from 6-10 to 11-14 and 15-17 age ranges and the increase reaches a larger extend in rural areas. The fact indicates that child labor is not a significant problem among 6-10 years old children. However, the attention should be given to age group of 11-17.            

                Concerning the relation between child labor and poverty, child labor rates are highest in the two poorest quintiles of per capita expenditures. Against, there is evidence for the negative relationship between poverty level and child labor in case of Vietnam.

                Two most important parents' characteristics of child laborers are mentioned, those are: parents' education and parents' employment status. Parents' education seems to have a positive relation with percentage of school-only children and a negative relation with that of work-only children. Mother's education shows larger effects. Regarding the parent's employment status, children whose fathers are employed are less likely to work. However children whose mother are unemployed are more likely to attend school and less likely to work.

                Finally, a portrait of full-time child workers is described. Children in this group have a relatively high average age (15 years old) and lower average years of education compared with all 6-17 years old children. Full-time child workers are found more often in 40% poorest households and girls take a larger share in this group. Rural children account for 90% of this group and they also have a higher average working per week.

                Chapter 3: Determinants of child labor in Vietnam

                3.1 Methodological approach

                The working yes-no choices are limited dependent variables and will be estimated using probit technique.  The overall strategy is to study the determinants of the parental decision that a child should take part in working. Although it will be more interesting to include an equation of hours worked, a comprehensive analysis of hours worked of children is beyond the scope of this thesis. In stead we focus on differences between results for the participation decision which may have implications for policies intended to encourage working children remain in school, for example. Much of the empirical works in this chapter is concerned with estimating the probability that a child will work given his own and his family's characteristics (or estimating the differences between probabilities for children with marginally different vectors of characteristics). At this point it is useful to make explicit the connection between the reduced-form theoretical equations of chapter I and the observable, estimable equations of this chapter. Consider the issue of child labor force participation. We start out with a deterministic theoretical model, but the fact that some determinants of labor force participation cannot be observed implies that we must estimate a probabilistic model. In other words, there is a set of characteristics, which (in theory) fully determines the child's labor force supply. We do not, however observe their joint distribution because we cannot observe all the characteristics; we know only the marginal distributions of the observable characteristics. From these we want to estimate the probability that the child take part in economic activities (i.e. reported working last week).

                3.2. Model specification

                Let Y_i denotes the working status of the child in the past week (worked or not worked), X_i is a set a child characteristics, H_i represents a set of variables proxied for the household economic position, D_i indicates the household composition, and R_i represents the household geographic location.

                Y _i = F(X_i, H_i, D_i, R_i, u_i)                                                                            (3.1)

                The regression model (3.1) will be estimated by maximum likelihood technique in the STATA software. Data used for estimated (Vietnam Living Standard Survey 1997/1998) was described in detailed in Chapter II. There are 8501 observations in the whole sample, but due to missing information, only 8470 observations were used in estimation.

                3.3. Results of estimation

                Table 3.1 presents probit result for determinants of working last week for 6-17 year olds in Vietnam. The table displays the two specifications: the second of which includes the same variables as in the first model and plus 4 standard of living variables in addition to the household expenditure per capita variable.

                The three characteristics of the child  (age, sex and ethnicity) are all significant at the ten percent level. Older children and females are more likely to work, all else constant, than are younger children and males. A quadratic term for age resulted in insignificant coefficients for both age and age-squared and was excluded. On average, an additional year old make the children's probability of working increase by 7% whereas being males would reduce the probability of working by nearly 3%. Besides that, children who are belonged to ethnic minority groups are more likely to work in compared with their Kinh majority peers.

                Parental characteristics are seen to be important determinants of the probability that a given child will work. Firstly, the head's age was found to have a significant non-linear effect on child work. In the first specification: children of older parents are less likely to work, and this effect increases with the head's age. However in the second specification when the living standard variables were included, the coefficients of the two head's age variables is no longer statistically significant at ten percent level. Secondly, relative to children in families with male head, children in families with female head are more likely to work. Thirdly, as regards the parent's education, the two variables of mother's and father's completed schooling levels both have important negative effects on the probability that a child will work. Father's education has a stronger negative effect on the probability that a child will work. Keep all other things unchanged, if father has one additional year of education, the working probability of his child would reduce by 7%, whereas one additional year of mother's education only make the probability fall by more than 4%. In term of the two variables concerning the father's employment status: whether or not the father is working and the father is working but not a salaried one. Children whose father is working are less likely to work; however the coefficient of this variable is not statistically significant at ten percent level. The second variable is statistically significant and has a positive effect on the probability that a child will work, indicating that children whose father is both working and a salary earner are less likely to work than the ones whose father is working but not a salaried one.  

                The household's expenditure per capita variable is, as expected, negative and significant. Children in families with a higher level of per capita expenditure are less likely to work. The size of the coefficient diminishes somewhat but remains significant at the one percent level when the four other standard of living variables are included. All the four standard of living variables likewise have a negative effect on the probability that a child will work: children living in residences having private tap or well with pump, flush toilet, the concrete outside walls and the family residence is private are less likely to be employed.          As regards characteristic of the child, age is significant at the one- percent level. Older boys are more likely to work, all else constant, than younger boys. A quadratic term for age resulted in insignificant coefficients for both age and age-squared and was excluded. On average, an additional year old makes the probability to work of a boy increase by 6.8%. Ethnicity is also an important determinant of the probability that a given boy is working. Being in the Kinh majority groups reduces the probability of working for a boy by 5.4%, other things being constant.

  Parental characteristics are seen to be not very important determinants of the probability that a given boy is working. Of the parents' completed schooling levels, only mother's completed schooling levels have a significant effect on a boy's probability of working. All other parental variables, including: father is working, father is working but not a salaried one, although had expected signs, were all found to have an insignificant effect on a boy's probability of working.

  The heads' ages were found to have an insignificant nonlinear effect on boys' working status although having the expected signs. Similarly, variables indicating the sex of the household head and the status that only the head or the spouse is present are all insignificant despite having the expected signs.

  The expenditure per capita variable is, as expected, negative and significant. Boys in families with a higher level of expenditure per capita are less likely to work. The size of the coefficient diminishes somewhat but remains significant at the one percent level when the three other standard of living variables are included. Two of the four of living standard variables likewise have a negative effect on the boys' probability of working: boys in household that have private house, private tap or deep drill well with pump are less likely to work.

  As regards household composition variables, these variables have less explanatory power with respect to child labor than was expected, although three of 10 age-sex categories are significant. The presence of 0-6 children and 7-9 year old boys makes the working probability of a boy increase by 3% and 5% respectively. The coefficient for number of 20-64 year old female is also statistically significant but displays a negative relation with child labor. One additional 20-64 year old female reduces the probability of working of a boy by 3%, indicating substitutability between child labor and female labor supply. As a result, boys of the families with presence of 20-64 year old female are less likely to work.

                Finally consider the regional location variables. Examining first the dummy variables indicating 7 regions. Relative to the Northern Uplands (reference region for the regional dummy), the average probability of working of children in the South Central Coast, the Central HighLands, the South East and the Mekong Delta are significantly lower. The children' probability of working is lowest in the Southeast. Oppositely, the probability in the North Central Coast is higher compared with that in the Northern Uplands. At ten percent level, the average probability of working of a given boy in the Red River Delta does not appear to differ significantly from that of the Northern Uplands. Turning to the urban-rural dummy variable, this variable has significantly negative effect on the children' probability of working, being urban children would reduce the probability of working by ten percent.       

Table 3.1: Determinants of the probability of working in the past week, Vietnamese children aged 6-17, 1997/1998 (Probit)

Source: Author's calculations based on data of Vietnam Living Standard Survey 1997-1998

Table 3.2: Determinants of the probability of working in the past week, Vietnamese boys aged 6-17, 1997/1998 (Probit)

Source: Author's calculations based on data of Vietnam Living Standard Survey 1997-1998

Table 3.3: Determinants of the probability of working in the past week, Vietnamese girls aged 6-17, 1997/1998 (Probit)

Source: Author's calculations based on data of Vietnam Living Standard Survey 1997-1998

Table 3.4: Definitions of variables

                3.4. Sex differences in children's probability of working

                The gender differences in children's probability of work, which was seen in the bivariate analysis in chapter II, and in the derivatives discussed above can also be seen in the multivariate analyses of the probability that a child will work. Table 3.2 and Table 3.3 present probit results of working last week for girls and boys respectively.

                First, examining the age and ethnicity of child. Results for boys and for girls are the same, both variables were statistically significant and had negative effect on the probability that a child will work. The effect of age was stronger in model for girls, while the effect of ethnicity was slightly stronger in model for boys.

                Turning to the parental characteristics. In the model for boys, only two of seven parental variables are statistically significant at ten percent level. Those are variables concerning mother's education and whether or not father is a salaried one. Unlike girls, boys' probability of working does not appear to be affected by the head's age, head's sex and father's education. Whereas the head's age, head's sex and father education appear to be important determinants of probability that a girl will work.  

                Concerning the expenditure per capita, this variable is statistically significant and has negative effect on the probability of working both in models for boys and for girls, implying that increase in household expenditure can reduce the use of child labor. However, this can not be achieved in short run but in a quite long-run development and depend on many other socioeconomic factors.

                As regards the household composition variables, only two in eleven age-sex categories are significant at ten percent level in model for girls while the number was five in model for boys. The presence of 20-64 males and females has the similar negative effect on the household's propensity to put its boys and girls into work. While the presence of children aged 0-6 and that of boys and girls aged 7-9 have no statistically significant effects on the probability that a girl will work, they does have statistically significant and positive effect on the probability that a boy will work.

                In term of regional variables, in both models for boys and girls, they show the same pattern with the general model. That is relative to the Northern Upland region, the probability that a boy or a girl will work is statistically lower in five regions: the South Central Coast, the Central High Lands, the South East, the Mekong Delta and statistically higher in the North Central Coast. However, against it is the same as in the general model, the probability that a boy or a girl will work in the Red River Delta does not appear to differ significantly from that of the Northern Uplands. Concerning the urban-rural dummy variable, in models for boys and for girls, being urban have 8% and 12% less probably to work respectively.

            Chapter 4: Conclusions and policy implications

                4.1. Conclusions

                This research has explored the determinants of child labor in Vietnam. The central thesis is that children's participation in economic activities is governed to a large extend by the economic position of the household, although the demand for child labor may also play important roles. The economic position of the household is determined by a complex set of interactions; theoretical issues in determining children's participation in economic activities are explored theoretically in chapter I.  In particular, measures of income and wealth and the demographic structure of the household are focal points of the analysis. As demonstrated in Chapter I, changes in household composition such as the addition of a child under age 7 will have income and substitution effects on the employment and enrollment behavior of an older sibling. In general, effects will be ambiguous but the theoretical results give some basis for interpretation of the empirical results.

                The data of Vietnam Living Standard Survey 1997/1998 is used for the multivariate analysis of this study. The survey was designed to nationally representative, covering 5999 households. The VLSS provide labor information on labor status of person aged 6 and above, so this is can be a rather good source of information to study child labor. There are 8501 children aged 6-17 in the survey, however due to missing information problem, only 8470 observation were used in the multivariate analysis.

                The first stage of the empirical study, in chapter II, involved an in-depth examination of the available information on children's work and schooling available in Vietnam, combines with a descriptive examination of the data. Overall, there has been significant improvement in child labor situation between 1993 and 1998. The change is shown in the reduction of both number of children working and in average time spent at work by these children. Regarding the child labor situation in 1998, there are some noticeable point should be noted. First, child labor rates are quite different in 7 regions of Vietnam. Second, the proportion of girls participating in work is higher than that of boys. Third, the child labor is mainly a rural issue as the number of working children in rural areas is much higher than that in urban areas. Fourth, child labor issue is not serious problem with respect to children aged 6-10; proper attention should be paid to children aged 11-17. Finally, descriptive analysis has shown that parent's education, parents' employment status and the household's economic status might play important role in determining child labor

                In the introductory part we posed a list of questions to be addressed by the thesis. These will now be addressed in light of the results from the econometric results first. Results of this study show that children's participation in economic activities is responsive to economic and demographic factors. Moreover, the responses are usually, but not always, in the expected directions. Their magnitude are often not large but some combinations of child and household characteristics can have substantial effects on the likelihood that a child will work.

                The characteristics of children that consistently had effect on children's probability of working were age, sex and ethnicity, with older children, girls and ethnic minority being substantially more likely to work than younger children, boys and Kinh majority.

                Household expenditure per capita is used to reflect the socioeconomic status of the family because they provide a more reliable estimate of permanent income than household income. This variable always had a significant effect: children in families with higher expenditure per capita were less likely to work. However, this effect was small in magnitude. The variables which proxied the family's wealth and standard of living had effects in the same directions as household's expenditure per capita – better living conditions were associated with lower probability of working – and those effects tend to be larger than the effect of expenditure. Age of the household head may also be proxying for wealth, since younger parents have had less time to accumulate assets and are likely to have greater number of dependents than at the other parts of the family cycle.

                The results concerning family composition are among the most interesting and important findings of this study. As the number of children in a family increase, the likelihood that any particular child will take part in economic activity also increases. This effect is especially robust and large in the case of infants, preschoolers and 7-9 year old children. An increase in the numbers of 0-6 as well as in the numbers of 7-9 year olds consistently increase the probability that a child will work. In contrast the presence of males or females aged 20-64 decreases the children's probability of working.

                4.2. Policy implications

                So far we have identified five key factors which affect the household's decision to supply child labor: the age and gender of the child, the education and employment status of the parents, the household's economic status (especially the poverty status) and the household's geographic location.  Base on the above findings, following is a set of policy implications to reduce child labor:

                Our results have shown that low education levels of parents, both father and mother, increases the probability of child labors. This is primarily an income effect, since low education leads to low earnings, but parents who themselves received little education may be less able to perceive the benefits of education for their children. More educated parents might have a better knowledge of the returns to education and/or be in a position to assist for their children in exploiting the earning potential acquired through education. Through the mechanism of the association between parents' education and child labor is not clear understood, educated parents are more likely to respond to information about the benefits of education or cost of child labor. However, parental education in itself is not a short-term policy variable hence low parental education can be used as a targeting variable for interventions.

In the case of Vietnam, it is found that the poor, children from rural areas and children belonged to ethnic minority are more likely to work than others. Since child labor is concentrated in specific disadvantaged regions, it makes sense to concentrate policy interventions in those areas. Because those areas also tend to suffer from supply constraints in education and health services, part of the policy package needs to be the buildup of the education and health infrastructure. Income support measures to the households as an important instrument for reducing child labor will be most effective when targeted to these groups, especially poor households in areas where access to capital markets or to other forms of transferable assets is limited.

It indicates, first, that a future development path which puts equitable growth and poverty reduction at its core (such as the Government of Vietnam has recently articulated in its Socioeconomic Development Strategy 2001-2010) is likely to generate further reductions in child labor. Secondly, it demonstrates that at the household level, there should be concern surrounding the hardship that could confront poor families – including their children – if child labor were eliminated without due consideration to the consequences of household income. Initially, interventions should aim to make possible the combination of work and schooling, rather than to eliminate immediately all child work. Flexibility of school hours and vacation periods in rural areas that coincide with harvest times are two potentially measures to facilitate the work-school combination.

The VLSS 1997/1998 has shown that most Vietnamese are self-employed farmers, household farming is also the kind of economic activity that most children take part in. The second kind of economic activity attract many children is household business. As in other developing countries, Vietnamese people rely on home enterprises, including farming and other kinds, for bulk of their income. Thus policy measures to support the development of such enterprises (such as the provision of credit and technical and marketing assistance) are necessary. There is a danger that promoting household enterprises will increase parents' demand for their children to work in these enterprises. Although in the long run the income effect will reduce the need for child labor, any negative effects in short-run can be counteracted by also providing school enrollment incentives to households. Such incentives are clearly most critical at the primary education level, but they should not be limited to that level. The evidence suggests that child labor increase with age, especially at the secondary school age. Children of poor households may therefore need subsidies for school fees, books, and uniforms at the primary and secondary level.