Short Paper
Working in Garment Industry and Child Education and Work in Cambodia
2022 Volume 58 Issue 2 Pages 82-89
Details
2022 Volume 58 Issue 2 Pages 82-89
Working in the garment industry may positively affect children’s education due to income effect. It may also decrease children’s time in school and increase time at work when the parent(s) and/or other adult household members work outside the home. This study explored the effects of parental and other household members’ work in that industry on the education and work of school-age children in Cambodia. The estimated results of the coarsened exact matching analysis revealed that parental work in the garment industry significantly impacts child education and work; in particular, female children are more likely to be in school and less likely to work if their parents are working. In contrast, an adverse effect was found when adult household members worked there; children were less likely to stay in school and more likely to work longer hours. Rural children seem less affected by adult members’ work.
Cambodia has maintained a high economic growth rate since 2000 and is witnessing improvements in both national GDP and living standards. The manufacturing sector in Cambodia is the mainstay of its economy, and garment and shoe industries (hereafter, garment industry) are particularly important as they continue to expand. Cambodia is endeavoring to become an upper-middle income country by 2030; however, it is still one of the least developed countries and continues to face the problem of poor children constituting human capital. The effect of the manufacturing sector’s growth on child education and labor has important implications in accumulating child human capital and reducing poverty, and it is therefore worth investigating.
Many garment and shoe factories (hereafter, garment factories) are located in the capital city, Phnom Penh, and in the Kandal province. The establishment of garment factories in provinces has been a recent trend in Cambodia, driven by the increasing export of garment and footwear products and travel bags since 2010 1. This leads to consistent economic growth and provides opportunities for work outside the home in both urban and rural areas. According to the Cambodia Labour Force Survey 2019, 16.7% of the employed population is in the manufacturing sector, and females are more frequently employed in manufacturing (22.3%).
Several studies have shown that the expansion of the industrial sector (including manufacturing) leads to increased household income, consumption, and reduced poverty (e.g., Nicita and Razzaz, 2003; Getahun and Villanger, 2018). Increased household income (wage) translates into a reduction in child labor and gain in education through the income effect (Soares et al., 2012).
However, from the perspective of the employed sector, maternal labor force participation is negatively associated with their children’s schooling and work in developing countries (Francavilla and Giannelli, 2007; Degraff and Levison, 2009; Self, 2011; Francavilla et al., 2013). On the other hand, Krauss (2013) showed that if one or both parents are employed in the non-agricultural sector; their children are less likely to participate in labor and more likely to attend school in Ghana.
Francavilla and Giannelli (2007) showed that fathers’ work increases the probability of studying in school and reduces the probability of children working. Yokying and Floro (2020) found that the effect of parental labor force participation depends on the sector in which a parent works. There is a negative correlation between children’s work and parents’ employment when a parent is an employee or worker in the formal sector, while the opposite is true for parents’ informal employment.
Degraff and Levison (2009) also considered the effect of the employment characteristics of the family (adult male members, but not fathers of children) and concluded that improvement of labor market opportunities for other family members decreases the probability of children’s employment, mainly due to the income effect. Similarly, Abang Ali and Arabsheibani (2016) showed that the probability of children working decreases in the presence of employed adults in the household.
Overall, it is clear that the effects of parental and other household members’ work on child education and work are mixed, and results from existing studies are conflicting.
In a study from Cambodia, Miwa (2019) showed that maternal work in the non-agricultural sector, especially in a garment factory, has a significantly positive effect on the nutrition of Cambodian children living in rural areas. However, it was not a nationwide study and it only focused on mothers’ work and child nutrition (health), with no consideration of child education, work, and other household members’ work effects.
The present study explores the effects of parental and other household members’ work in the garment industry on school-age children’s education and work, based on datasets from Cambodia Socio-Economic Survey (CSES): 2014 and 2016, respectively. To examine this, the coarsened exact matching (CEM) method was employed in the empirical analysis.
The data from the CSES 2014 and 2016 was used for the analysis. The CSES is a national representative survey conducted by the National Institute of Statistics, which is a part of the Ministry of Planning in Cambodia. The CSES is a household survey including questions for households and their household members about housing conditions, education, economic activities, household production and income, household level and structure of consumption, health, victimization, vulnerability, etc. CSES 2014 is the thirteenth CSES with 12,096 sample households and CSES 2016 is the fifteenth CSES with 3,840 sample households 2.
School-age children (aged 6–17 years) are the main focus of this study. This study includes children aged 15 to 17, who are beyond the age of compulsory education. In Cambodia, the further spread of secondary education is needed to ensure sustainable growth by accumulating child human capital. After finishing lower-secondary school, children have to decide to either continue studying in an upper secondary school or to enter the labor market. Therefore, it is worth investigating their decision. Together with CSES 2014 and 2016, information on education and work (economic activity) is available for 16,218 children. Of these 70.9% of children reside in the rural area.
Table 1 presents child education and work status of school-age children in Cambodia. Overall, 93.4% of children were enrolled in elementary school, and 81.5% among them were still in school (current student) at different school levels. Rural children were less likely to be in school than urban children; the student ratio was 79.2% and 87.1% in rural and urban areas, respectively.
| Variable | Definition | All | Rural | ||
|---|---|---|---|---|---|
| Mean | SD | Mean | SD | ||
| Student | 1 if the child is a student, 0 otherwise | 0.81 | 0.79 | ||
| Work | 1 if the child worked (engaged in an economic activity) in the last 7 days, 0 otherwise | 0.18 | 0.21 | ||
| Work hours | Average hours of working per day (in the last 7 days) | 0.93 | 2.27 | 1.06 | 2.37 |
| Parent work | 1 if child’s mother or father (or both) work in the garment industry, 0 otherwise | 0.09 | 0.08 | ||
| Adult work | 1 if at least one adult member (expect parent of child), who lives with the child, works in the garment industry, 0 otherwise | 0.13 | 0.12 | ||
| Observable covariates | |||||
| Child’s age | Child’s age in years | 11.45 | 3.45 | 11.38 | 3.43 |
| Khmer | 1 if the child is Khmer, 0 otherwise1) | 0.96 | 0.96 | ||
| Parent | 1 if the child is living with one or both parents, 0 otherwise | 0.92 | 0.91 | ||
| Head’s age | 1 if the household head’s (HH’s) age is 10s or 20s, 2 if 30s, 3 if 40s, 4 if 50s, 5 if 60s and over | 3.08 | 1.06 | 3.01 | 1.06 |
| Head’s educ | 1 if HH has no education, 2 if has 1 to 6 years of education, 3 if 7 years of education and more | 2.16 | 0.74 | 2.04 | 0.72 |
| Own land | 1 if the household owns agricultural land, 0 otherwise | 0.60 | 0.77 | ||
| Wealth | Categorized into quartiles based on the level (value) of durable goods holdings (1 if the first quartile, the lowest group) | 2.49 | 1.09 | 2.33 | 1.06 |
| Location | 1 if the household is in an urban area, 0 otherwise | 0.29 | 0.45 | 1.00 | |
| Factory | 1 if there is at least one garment factory in the same commune in one year before the survey, 0 otherwise2) | 0.09 | 0.29 | 0.04 | 0.20 |
| Other children’s characteristics | |||||
| School enrolment ratio (%, those who have ever enrolled in the elementary school) | 93.38 | 24.87 | 92.32 | 26.63 | |
| Average completed years of school (yrs) | 5.42 | 2.98 | 5.30 | 2.90 | |
| Years delayed relative to age (yrs) | 1.04 | 2.14 | 1.17 | 2.28 | |
| Average days of work in the last month | 4.17 | 9.53 | 4.81 | 10.04 | |
| Average hours of work in the last 7 days (those who worked) | 35.79 | 18.43 | 24.79 | 18.33 | |
| Number of children | 16,218 | 11,495 | |||
| Ratio of female (%) | 48.93 | 48.69 | |||
Source: Calculated from the CSES 2014 and 2016.
1) Includes Cham and other ethnic minorities, Chinese, Vietnamese, Thai, etc.
2) Commune refers to the administrative unit which consists of several villages. Bulletins published by the Garment Manufacturers Association in Cambodia were checked to identify the location of the factory.
Further, 18.2% of the children worked (engaged in an economic activity) at least one hour in the last seven days prior to the survey 3. In this study, if children worked even one hour in the past seven days, it is considered child work. In total, 61.2% and 11.7% of working children were in the agricultural and garment industries, respectively. Children reside in the rural area work more and longer (working ratio is 21.3% and working hours are 1.06 per day on the average) than the national average.
The CEM method was employed in this study to explore the causal inference of parental and adult household members’ work in the garment industry on child education and work.
PSM (Propensity Score Matching) is an enormously popular method for matching analysis. However, it often increases imbalance, inefficiency, model dependence, and bias. On the other hand, CEM works well compared to other matching methods in terms of reducing imbalance, model dependence, estimation error, bias, and other criteria (Iacus et al., 2011; King and Nielsen, 2019). The concept of the CEM method is as follows: while exact matching provides perfect balance, it typically produces few matches because of the curse of dimensionality issues. CEM temporarily coarsens each variable (observable covariate X) into substantively meaningful groups, exactly matches on these coarsened data, and then retains only the original (uncoarsened) values of the matched data.
Let
The observed outcome variable is Yi=TiYi(1)+(1−Ti)Yi(0), where Yi(0) is the potential outcome for unit (child) i if the child does not receive treatment, and Yi(1) is the potential outcome if the child i receives treatment.
TEi=Yi(1)−Yi(0), the treatment effect (TE) for child i is unobservable, as usual. Causal quantities of interest are the average of TEi over different subsets of units in the sample and must be estimated. We focus on the sample average treatment effect on treated (SATT), written as
Three outcome variables are used: (A) a dummy variable 1 if child is in school (current student) now, 0 otherwise; (B) dummy variable 1 if child is working, 0 otherwise; and (C) child’s working hours per day (which is 0 if the child is not working).
Robust standard errors clustered at the household level are employed for estimating SATT to deal with the bias since there is more than one school-age child in the same household. The definitions and summary statistics of all the variables used in the empirical analyses are listed in Table 1. The number of garment factories is recently rising in the rural area, and the expanding of the garment industry may impact children more in rural rather than urban area, where the agriculture used to be the mainstay of people’s life. Therefore, the treatment effects for rural children are also explored in this paper. Female and male children are separately estimated to examine whether the impacts are different or similar among them.
Table 2 presents the results of CEM (SATT), where the treatment variable is: one or both parents of the child work in the garment industry. L1 shows the overall imbalance in covariates. Before the CEM, the L1 value for unmatched data is 0.738 and 0.727 for the analysis of all female and male children, respectively; and those values become very small after CEM, implying that it becomes less imbalanced.
| (A) All Children | (B) Rural Children | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Student | Work | Work Hours | Student | Work | Work Hours | |||||||
| Female | Male | Female | Male | Female | Male | Female | Male | Female | Male | Female | Male | |
| SATT | 0.031* (0.018) | 0.015 (0.019) | −0.005 (0.018) | 0.001 (0.017) | −0.056 (0.109) | −0.005 (0.089) | 0.045** (0.023) | 0.031 (0.024) | −0.014 (0.024) | −0.007 (0.023) | −0.131 (0.137) | −0.034 (0.122) |
| Treated (Matched) | 534 | 522 | 533 | 522 | 533 | 522 | 356 | 338 | 355 | 338 | 355 | 338 |
| Controls (Used) | 2070 | 2211 | 2068 | 2210 | 2068 | 2210 | 1582 | 1698 | 1580 | 1698 | 1580 | 1698 |
| L1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| By Age Group (SATT) | ||||||||||||
| Aged 6–11 | 0.031* (0.018) | 0.017 (0.022) | 0.0001 (0.011) | −0.009 (0.012) | −0.016 (0.032) | −0.011 (0.029) | 0.033 (0.023) | 0.035 (0.026) | 0.001 (0.016) | −0.005 (0.017) | −0.021 (0.047) | 0.004 (0.040) |
| Aged 12–14 | 0.071*** (0.025) | 0.018 (0.032) | −0.034 (0.039) | 0.006 (0.038) | −0.515*** (0.158) | −0.136 (0.130) | 0.083** (0.036) | 0.031 (0.040) | −0.035 (0.056) | −0.016 (0.052) | −0.597*** (0.206) | −0.207 (0.193) |
| Aged 15–17 | −0.023 (0.063) | 0.003 (0.061) | 0.013 (0.064) | 0.032 (0.065) | 0.388 (0.483) | 0.200 (0.429) | 0.039 (0.079) | 0.011 (0.079) | −0.046 (0.079) | −0.004 (0.082) | 0.038 (0.603) | 0.059 (0.557) |
1) Statistically significant at the *10%, **5%, and ***1% levels. Robust standard errors clustered at the household level are in parentheses.
From Table 2 (A), overall, female children with parents who work in the garment industry are more likely to be in school. In particular, parental work has a positive and significant impact on younger children’s schooling. On the other hand, parental work in the garment industry does not have a significant effect on child work, overall; however, it significantly reduces the working hours of female children aged 12 to 14 years. Since girl’s average working hours at that age group are 0.74 hours (44 minutes) per day, parental work may reduce it approximately 70% (31 minutes).
Female children living in the rural area are more likely to be in school if parents work in the garment industry (right column of Table 2). The coefficient is 0.045, which means that parental work increases 5.7% of probability of female children being school in rural area; which is higher than the analysis of all children. Note that urban children’s schooling is not significantly affected by parents’ work (results have not been stated due to space constraint). Rural children’s work is also not significantly affected by parental work status overall; although, it significantly reduces 66.6% (36 minutes) of working hours of female children aged 12 to 14 years. These facts imply that expansion of the garment industry will help accumulate child human capital (especially, education) in rural Cambodia.
Here, the simple comparison (without matching) of education and work status for both female and male children (no matter where they reside) are significantly different when checking the simple average difference (by T-test); more children with parents who work in the garment industry are in school and work less than others. On the one hand, the coefficients of SATT show the same sign but are insignificant, especially for boys’ education and work. This implies that by using CEM method, which improves the estimation of causal effects by reducing imbalance in covariates between treated and control groups, it is appropriate to conclude that girls’ education and work hours (at the age range of 12 to 14 years) would be positively affected by parental work in the garment industry in Cambodia.
The garment industry creates the opportunity to obtain formal employment, which brings home a stable (and sometimes enough) monthly income for people in Cambodia. This fact leads to increased educational expenses and contributes toward a child’s continuing education, even after graduating from elementary school. Women, especially, have greater chances of getting jobs in the garment industry. Since being a garment factory worker requires a certain level of education, parents may send their girl children to school to obtain formal employment in the future.
When adult household members’ work in the garment industry is considered as the treatment variable, an adverse effect can be found. From the estimation results shown in Table 3(A), it can be inferred that, in contrast to parental work, adult members’ work has negative and significant effects on girls’ education, and it positively increases the probability of female children working, in addition to working longer hours.
| (A) All Children | (B) Rural Children | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Student | Work | Work Hours | Student | Work | Work Hours | |||||||
| Female | Male | Female | Male | Female | Male | Female | Male | Female | Male | Female | Male | |
| SATT | −0.050** (0.020) | −0.014 (0.020) | 0.047** (0.020) | −0.010 (0.021) | 0.615*** (0.140) | 0.065 (0.120) | −0.029 (0.023) | −0.011 (0.024) | 0.035 (0.023) | −0.024 (0.025) | 0.557*** (0.157) | −0.043 (0.146) |
| Treated (Matched) | 777 | 748 | 776 | 748 | 776 | 748 | 602 | 559 | 601 | 559 | 601 | 559 |
| Controls (Used) | 2618 | 2787 | 2616 | 2787 | 2616 | 2787 | 2141 | 2279 | 2139 | 2279 | 2139 | 2279 |
| L1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
| By Age Group (SATT) | ||||||||||||
| Aged 6–11 | 0.032* (0.019) | 0.025 (0.019) | −0.012 (0.011) | −0.010 (0.012) | −0.062* (0.034) | −0.018 (0.035) | 0.036 (0.021) | 0.026 (0.024) | −0.012 (0.015) | −0.011 (0.016) | −0.076* (0.044) | −0.013 (0.048) |
| Aged 12–14 | −0.010 (0.027) | −0.040 (0.033) | 0.012 (0.031) | 0.008 (0.037) | 0.274 (0.176) | 0.039 (0.176) | 0.001 (0.031) | −0.054 (0.039) | 0.010 (0.037) | 0.018 (0.044) | 0.275 (0.204) | −0.045 (0.210) |
| Aged 15–17 | −0.180*** (0.041) | −0.042 (0.041) | 0.145*** (0.040) | −0.026 (0.042) | 1.699*** (0.309) | 0.190 (0.278) | −0.134*** (0.046) | −0.017 (0.047) | 0.115** (0.035) | −0.072 (0.047) | 1.572*** (0.341) | −0.075 (0.315) |
1) The L1 value for unmatched data was 0.675 and 0.674 for the analysis of all female and male children, respectively; and the overall imbalance is improved after matching.
2) Statistically significant at the *10%, **5%, and ***1% levels. Robust standard errors clustered at the household level are in parentheses.
These results imply that female children are more susceptible to the changes. For example, the probability of female children’s work increases by 26.4% if adult household members work in the garment industry (since the ratio of female children’s labor is 17.9%).
When focusing on rural children, as shown in Table 3(B), the same adverse effects could be found for children’s work hours. Adult member’s work enforces girls to work 0.56 more hours (34 more minutes) per day, which is equal to 52% increase of work hours. On the other hand, overall, child’s schooling and work status are not significantly affected by adult member’s work in the rural area. Only female children aged 15 to 17 years face the negative impacts; adult member’s work in the garment industry significantly reduces the probability of girls being in school by 27.3% and increases working by 23.4% and work hours by 48.0%. The SATT coefficients are smaller than all children’s analysis; however, these still lead to the view that there is a huge impact on child human capital accumulation.
Although, parental work had insignificant effect for children aged 15 to 17 years, the negative impacts of adult member’s work are mainly for girls at that age range both in the nationwide and rural area, but not for other age groups.
The adverse impacts of adult members’ work on child education and work are as follows: when some household members work in the garment industry (i.e., they work outside the home) school-age children may need to work on farms or family businesses to replace the adult members. It is also possible that school-age children might work in the garment industry. In fact, it is easier to get a job in a garment factory if some family members or relatives already work there, and it is not uncommon for children aged 16 years and above to be employed in a garment factory. These factors lead to reduced time spent in school and increased work hours, especially for children aged 15 to 17.
Among the working children, 52.2% and 37.8% are unpaid family workers and paid employees, respectively (these are 54.3% and 34.7% in rural area, respectively). When the employment status of child work is considered, as shown in Table 4, adult members’ work in the garment industry significantly reduces the probability of being an unpaid family worker for children (mainly in the agricultural sector); however, it increases the probability of being a paid employee. Rural children also face these impacts. As 30.3% of children who are paid employees are working in the garment industry and 25.8% are in the agricultural sector5, these data support the interpretation that school-age children may work with or instead of adult household members.
| (A) All Children | (B) Rural Children | |||
|---|---|---|---|---|
| Female | Male | Female | Male | |
| Unpaid Family Worker1) | ||||
| Adult Work | −0.042*** (0.012) | −0.021 (0.016) | −0.050*** (0.015) | −0.029 (0.020) |
| Paid Employee1) | ||||
| Adult Work | 0.104*** (0.017) | 0.030** (0.014) | 0.099*** (0.019) | 0.030* (0.017) |
1) These variables take 1 if children are working as unpaid family worker/paid employee, and 0 otherwise.
2) Statistically significant at the *10%, **5%, and ***1% levels. Robust standard errors clustered at the household level are in parentheses.
With the expansion of the garment industry in Cambodia, an increasing number of workers are engaging in garment work. This study explored the impact of parental and adult household members’ work in the garment industry on school-age children’s education and work.
The estimation results show that parental work in this industry has a significant and positive impact on child education. On the other hand, when adult members (e.g., older sibling or aunt/uncle of school-age children) are working in the garment industry, an adverse effect can be found; children are less likely to continue to study in school and more likely to work longer hours. Note that children residing in the rural area seem to be less affected by adult member’s work; it significantly increases work hours, but, has insignificant effects on schooling and work status.
This implies that the impact of expanding garment industry on children’s schooling and working is mixed in whole of Cambodia including rural regions.
This study considers the effect of household members’ work in the garment industry on children’s school attendance. However, other aspects of education, such as attendance rate, school performance, and school dropout, could not be considered, mostly due to data constraints. The type and content of children’s work also need to be considered. In addition, it is necessary to examine the trade-off between schooling and work. These topics will be reserved for future research.
