Volume 45 Number 178,
July-September 2014
Child Labor and its Causes in Mexico
Pedro Orraca *
Date received: September 4, 2013. Date accepted: January 31, 2014

This article studies the effect of birth order and low household economic resources on school attendance and labor market participation among the population of minors in Mexico. Drawing on data from the Child Labor Center, this work estimates a series of multivariate probit models with different specifications, aiming to minimize problems related to the endogeneity of the sample. The analysis reveals that first-born children are less likely to attend school and more likely to participate in the labor market than their siblings. This relationship is stronger in families where child labor is a response to a lack of economic resources, which suggests that it is the result of the greater income-earning capacity of first-born children.

Keywords: Mexico, birth order, child labor, children and teenagers, labor market.

Child labor is both a global and national problem. The International Labor Organization (ILO) estimated that in 2008, over 306 million individuals between the ages of 5 and 17 had participated in the labor market (ILO, 2010: 7). In Mexico alone, the total number of working children was calculated at 3 million in 2011 (INEGI, 2012).

This is an issue because child labor tends to prevent children and teenagers from achieving full development, by limiting their human capital accumulation and reducing long-term income levels throughout the labor life cycle. In Mexico, child labor is even more important because education tends to have high private returns, leaving the uneducated population at a severe disadvantage. Likewise, these effects will likely extend to future generations, in light of Mexico's low social mobility. On the aggregate level, child labor is generally undesirable, as it reduces the human capital stock in the medium and long term, limiting future economic growth.

Unlike what is generally believed, child labor is not merely the result of selfish parents. Rather, it emerges as a household strategy to survive and obtain a greater quantity of assets. Child and teen labor is a symptom of other problems in the country, such as deficient educational systems, discrimination within families and the lack of opportunities available to certain sectors of the population. This is evidenced by the fact that developed economies rarely have child labor (Basu and Van, 1998).

This article aims to study the effect of birth order and the lack of household economic resources on how much time minors in Mexico spend attending school and working remunerated or unremunerated jobs. This is an interesting case, because the country has medium-high income levels, extreme inequality, significant poverty and child labor and various wide-ranging social programs whose direct and indirect objectives are to promote human development.

This work therefore contributes to the scarce literature on the effect of birth order on school attendance and participation in the labor market in Mexico. Unlike much of the previous research, this study will not be limited to those 12 years of age and older, but will rather focus on the population between 6 and 16 years of age. Additionally, it will distinguish between remunerated and unremunerated work, while the econometric analysis will take into account how these labor decisions interact with school attendance. Finally, this research provides updated evidence on child labor on the national scale.

One of the most significant results reveals that the amount of time assigned to different activities varies greatly depending on birth order. Oldest children are less likely to attend school and more likely to enter the labor market as compared to their siblings. On the other hand, youngest children are more likely to attend school, and girls are less likely to engage in remunerated work. The relationships found here are even more extreme in households where children work due to a lack of resources. These findings contradict what is normally true of developed countries, where birth order tends to favor the older children. In Mexico, the effect of birth order on how much time is assigned to school versus labor activities tends to favor the younger children.


Article 123 of the Constitution of the United Mexican States stipulates that child labor under the age of 14 years is prohibited (CPEUM, Article 123, section A, clause III). In addition, the Federal Labor Law proclaims in Articles 22 and 23 that children under the age of 14 are not allowed to work in any case, and teens under age 16 that have not completed mandatory education are also prohibited from working, except in certain exceptional situations. However, despite the fact that there is a legal framework to prevent child labor, this cohort continues to work at significant levels. According to data from the Child Labor Module (MTI) (INEGI, 2012), 10.5% of the population aged 5 to 17 years worked in 2011. Of them, 70.9% were legally allowed to work, meaning that more than 882,000 individuals between the ages of 5 and 13 participated in the labor market.

Similarly, these figures reveal high heterogeneity by state of origin, gender, type of work performed and size of residential town. This is evidenced in the fact that while 4.2% of the minor population in Chihuahua works, 26.5% of children and teenagers in Guerrero participate in the labor market. On the national level, 67.9% of child workers are male, while 48.3% of minors that work do not receive remuneration. In addition, child labor tends to be more common in rural areas than urban, given that 37.5% of all child laborers live in towns with less than 2,500 residents.

Working in the labor market means less time assigned to school activities. Although the school attendance rate among children aged 5 to 17 years is 91.1%, this number is considerably lower among children that work, where only 60.9% of the population that works also studies. Finally, 4.8% of all minors neither work nor study.


Whether children or teens work is largely determined by the economic and demographic features of their households. These include age, gender and education of household members, the number of siblings living in the home, and costs and returns on time use (Bando, López-Calva et al., 2005: 4).

Studies analyzing child labor and school attendance tend to be based on neoclassical models of household decision-making or link these phenomena with the fertility choices of mothers (Bando, López-Calva et al., 2005; Dammert, 2010). In these models, altruistic parents determine how their children's time is assigned, and families work to obtain minimum subsistence before acquiring other assets or making investments (see Basu and Van, 1998: Baland and Robinson, 2000; Cigno and Rosati, 2005). When children are perceived as assets, parents invest in their education up to the point where the marginal cost is the same as the marginal benefit, taking into account the opportunity cost of not working (Dammert, 2010: 200). Because there are growing marginal costs linked to human capital investments, there is an inverse relationship between the demand for quality and quantity of children in the household (Becker, 1991).1

On the one hand, it could be said that in low-income households, first-born children are more likely to work, because their experience and human capital stock, as compared to others, would allow them to conduct more complex activities, achieving higher productivity and salaries and therefore more resources for the household. This additional income could take pressure off household budgetary restrictions, making it possible for the rest of the children to attend school (Emerson and Souza, 2008: 1649). As such, if child labor emerges out of the need for subsistence, then we would expect that first-born children would be more likely to work.

On the other hand, Dammert (2010: 200) argues that biological factors play an important role in how time is assigned within a household. Specifically, she argues that when a mother gives birth to her last child, because she is older, the children are born at a lower weight, which is correlated with lower skills in the medium and long term.

In light of that idea, it could be thought that work and attending school have nothing to do with experience, but rather with discrepancies in skills among children, meaning that the birth order effect may not be real. However, although greater skill levels reflect a greater opportunity cost of studying in the short term, it also means greater returns on education in the long term, and as such, if the child works, a greater economic loss than that of the other children (Emerson and Souza, 2008: 1648). Consequently, whether a child or teenager attends school or participates in the workforce depends on salary differences among the children and the value the households assigns to family consumption in the medium and long term, with respect to the child's wellness and human capital accumulation, among other factors (Emerson and Souza, 2008: 1649). This means that the relationship between birth order and how time is distributed between school and labor activities is, more than anything else, an empirical question.

Empirical Evidence

Many studies have focused on labor and educational decision-making among minors in Mexico. K. Córdova (2009) analyzes the effect of child birth order on years of schooling and how time is assigned to various activities, including recreation, household chores and caring for other family members. Based on the National Survey on Household Living Standards (ENNViH), the researcher used the two-stage method proposed by Heckman (1979) to estimate the effect of birth order on years of schooling. To understand its effect on how time is allocated, the study estimated a model using the ordinary least squares (OLS) method. The results did not reveal evidence that birth order is related to years of schooling. However, there was a significant effect of birth order and gender on the amount of time allocated for household activities and caring for other members. In another study, A. López (2005) jointly analyzed factors determining school and workforce participation among a population aged 6 to 17 years. Using the National Household Income and Spending Survey (ENIGH), A. López estimated an ordered probit model, where individuals choose between attending school, working part-time or working full-time. The results showed that income level and household assets increase the likelihood of children attending school, while credit restrictions on a household reduce it.

In addition, Levison; Moe et al. (2001) studied the factors determining school attendance and labor market participation among a population aged 12 to 17 years. To do so, they used information from the National Urban Employment Survey (ENEU), estimating a multinomial logistic model for the empirical exercise. The results revealed that girls are more likely to attend school; however, if the definition of labor includes household activities, girls are less likely to focus on studies than men. López-Calva and Freije (2001) analyzed the relationship between child labor and a series of variables, including: the poverty situation of a household, rate of unemployment among parents, salary and social acceptance for labor, among others. Using longitudinal data from the ENEU for Mexico and the Household Sample Survey (EHM) for Venezuela, the authors employ a series of models to conclude that poverty and incidence are robust determinants of child labor, while salary and unemployment are not.

Other studies have focused on analyzing the effect of the Progresa program, whose name has since been changed to Oportunidades, on school attendance and child labor. One piece of notable research by Skoufias and Parker (2001) uses data from various surveys to evaluate the effect of the program, finding that it had considerable repercussions for two variables by significantly increasing school attendance and reducing child labor. A similar study focused primarily on the indigenous population, by Bando; López-Calva et al. (2005) argues that the program had a positive effect on child labor, reducing its incidence by 8% between 1997 and 2000. It also observed that child labor is linked to fertility decisions made by the mother and the adult labor market, while the father's education is positively correlated with the likelihood that children attend school and negatively correlated with the likelihood of a minor working.

Looking to international literature, Dammert (2010) analyzed the effect of gender differences among siblings and birth order on work, domestic labor and school attendance in Guatemala and Nicaragua. The study found that first-born males dedicate more time to working and household activities, whereas female first-borns are more inclined to domestic labor as compared to the rest of their siblings. In a similar study for Ecuador, De Haan; Plug et al. (2012) observed that level of schooling increased with birth order but decreased with child labor. Finally, Emerson and Souza (2008) focused on Brazil, where they found that younger children were less likely to be working, while first-born children were most likely to participate in the labor market and less likely to attend school.


This study draws on data gathered by the MTI, collected in the fourth quarter of 2011 as an appendix to the National Employment and Occupation Survey (ENOE). The MTI complements normally collected information for the population aged 12 to 17 years in the ENOE and extends its analysis to include features related to schooling and working for individuals aged 5 to 11 years. This helps identify the factors that determine school attendance and labor market participation in the population of Mexican minors.

According to the MTI, a child laborer is defined as an individual between the age of 5 and 17 years, who, in the week prior to the interview, carried out some sort of economic activity or was about to start one imminently. Economic activity is understood as production for individual consumption or any actions intending to produce or provide goods and services for the market.2 The questionnaire was administered to over 96,000 individuals in the 32 states around the country. The module includes information on rural and urban areas and is nationally representative.

This analysis is limited to children age 6 to 16 years for better focus on mandatory school-age children. To minimize the effects of structural household differences, only homes where both the mother and father live were included. This study was also restricted to children born in Mexico to Mexican parents, to eliminate any cultural discrepancies between the national and foreign populations residing in the country.

Finally, child laborers were defined as minors that reported a positive number of hours worked, and individuals were categorized by gender and status as paid or unpaid workers. The final sample consisted of 46,613 individuals, of whom 23.765 were male and 22,848 were female.


Although children and teenagers distribute their time among numerous activities, this study exclusively focuses on time allocated for school attendance, remunerated work and unremunerated tasks. To do so, it estimates a multivariate probit model assuming that time allocation decisions are related to each other among these three options, even when these decisions are made individually or separately.

The multivariate probit model structure is similar to a system of apparently uncorrelated equations, where the principle difference is that in this analysis, the dependent variable is dichotomous. Consider a multivariate probit model with three equations:

and 0 in the other case

Where represents a latent underlying variable linked to which constitutes a dichotomous variable that takes the value of one if the event occurs and zero if it does not.3 Jointly, the assumption is that the errors follow a multivariate normal distribution, with mean zero and a variance-covariance matrix V that contains values of one on the diagonal and the correlations outside of it denotes an exogenous variable matrix containing the age of the children and parents and the level of schooling of the latter, among others included key regressors for this study, including a dichotomous variable that takes the value of one of the individual is the oldest child among household members and zero if not, and a dichotomous variable that takes the value of one if the child or teenager is the youngest member of the household and zero if not. In both cases, the reference group consisted of the other siblings living in the same household.

The multivariate normal structure of the error term indicated that even after controlling for a series of explanatory variables, the decisions made were connected. However, one weakness of the model is that it assumes the same correlation structure between any pair of decisions for all individuals, which is a highly restrictive assumption. On the other hand, one advantage of the model is that it does not require including the same set of regressors in each equation, although given the nature of this analysis, this study opted to do so. Finally, the model is estimated through simulated maximum verosimility, using the Geweke-Hajivassiliou-Keane simulator (see Börsch-Supan and Hajivassiliou, 1993; Keane, 1994).4

Descriptive Statistics

Table 1 presents the mean and standard deviations of the main variables used in the study, showing that 94.4% of boys and 95.4% of girls attend school. In addition, there are key differences in the proportions of workers by gender, where 9.9% of males in the study undertook some kind of labor activity, but only 4.6% of females did. This relationship remains even when the sample is divided among individuals that engage in remunerated and unremunerated work, or between those working more than 15 hours a week and those that do not. Both the fathers and mothers of these children were generally young, with an average age of 40.2 and 37.0 years, respectively. Finally, 28.9% of households analyzed belonged to rural communities.

To see if there were considerable differences among the population studied, school attendance and labor activities were analyzed by age range. Table 2 shows that as individuals grow up, fewer continue to study, where the peak number of children studying is between 6 and 10 years of age, and falls until age 16. This matches the natural trend of dropping out that increases as the grades go up. In terms of paid and unpaid work, the situation is the opposite, where those most likely to engage in labor activities are concentrated around age 15 and 16. This indicates a positive relationship between age of minors and likelihood of working.

Table 3 describes the reasons why children and teenagers participate in the labor market. While 18.4% of first-born males work due to a lack of resources, only 10.4% of youngest children work for the same reasons. This situation is true of both genders, as well as for paid and unpaid work, and provides tangential evidence that, as compared to other children, the choice of whether or not the first-born child works owes in large measure to a lack of household resources.

Econometric Results

To analyze the effect of birth order on school attendance and workforce participation, a three equation, multivariate probit model was estimated. The model includes the dichotomous variables of interest indicating if the individual is the youngest or oldest child, as well as a series of exogenous controls that influence the dependent variables, including age, level of schooling, parents' employment situation, number of children in the household, average age difference between siblings and a dichotomous variable indicating if the household is urban or rural.5

Table 4 introduces the results for the entire sample. The age of the child had a negative effect on school attendance and a positive influence on workforce participation. The effect exists for both genders, and is greater for remunerated work than unremunerated. This reflects the positive relationship between the productivity of minors and their age, while their inclusion helps estimate the net effects of birth order after controlling for age.

In terms of the main variables to analyze, first-born males were less likely to attend school than the rest of their siblings, although the discrepancy with respect to the youngest children was minimal. However, this difference is considerable for females, given that youngest daughters are most likely to attend school, whereas first-born daughters are substantially less likely to do so. In addition, the youngest sons are less likely to have a remunerated job and more likely to work in an unpaid position. For females, the situation is reversed, where being born last means they are more likely to have a remunerated job, which also reduces the likelihood of having an unremunerated job. This suggests that among females, there is a certain degree of substitution between remunerated and unpaid work, where the youngest and first-born daughters tend to be concentrated, respectively.

On the other hand, when parents have higher levels of schooling, there is a greater probability that both male and female children will attend school. In addition, although the parents' educational levels have a negative effect on likelihood to work, the effect of the mother's level of schooling is more ambiguous. These results are expected, given that the variable of education tends to be a good indicator of household income level, which generally reduces the proportion of children in the workforce. When the father or mother works, the likelihood of the child attending school decreases and the probability of the child working in a remunerated job increases. Although these effects are small, they do not show a direct relationship between unemployed parents and child labor, which suggests that children and teenagers entering the workforce is not necessarily a response to their parents being unemployed. However, it should be not be ruled out that child labor is the product of low parental income, a situation that can push households towards poverty.

By contrast, having more children is associated with a lower probability that they attend school and greater likelihood of workforce participation. This result reflects the fact that larger households tend to be poorer and face greater economic difficulties. Age difference between siblings has a positive effect on school attendance and workforce participation for females, while for males, the effect is only positive for unpaid work. Finally, living in a rural area reduces the likelihood of attending school and engaging in remunerated activities, and increases the probability of having an unpaid job. This is indicative of the low school attendance observed in rural communities, owing, in part, to decreased educational quality in rural zones and, in extreme cases, limited access to educational facilities. However, these results should be interpreted with caution, given the variety of endogeneity problems in this model.

Minimizing Endogeneity Problems

In estimating this model, the fertility variable (i.e. number of children) had a serious endogeneity problem that emerged due to a measuring error for the variable and the possibility of inverse causality, which generated biased and inefficient estimators that invalidated the empirical analysis.

The error in measuring the fertility variable arose because the regressor included both families that were no longer growing, and as such would not have additional children, as well as those still expanding (Emerson and Souza, 2008: 1651). Because the final number of children cannot be recorded for this latter group, a measurement error is produced. Similarly, because the variable counts two distinct phenomena under the same umbrella, it is highly likely that it is very correlated with the error term. In addition, the inverse causality issue comes up because decisions on how much to invest in children and how many children to have, are generally made at the same time (Dammert, 2010: 200).

Commonly, these problems can be solved by using instrumental variables. However, the lack of relevant tools available meant that this study chose a few alternative methods. Based on Emerson and Souza (2008), the first strategy consisted of limiting the analysis to households where fertility decisions had already been made and the families were no longer growing. To do so, the sample was restricted to households where the mother was over 40 years of age, and therefore the likelihood of adding more children was low. To control for inverse causality, the sample was restricted to families with the same number of children, in this case, families with three children. This guarantees that all households included in the sample are unbiased, given that there are no structural differences among them. If there are systemic discrepancies among families with three children and other households, there is a chance that the outcome cannot be generalized to the rest of the population.

The multivariate probit model estimates for this subset are found in Table 5, and the results are highly consistent with previous outcomes. Both for males and females, the youngest child is the most likely to attend school as compared to the other siblings, while for first-born sons, the situation is reversed, and for first-born daughters, there is no relationship to school attendance. In terms of paid and unpaid work, for males, it appears that the youngest son is less likely to work than the rest of the other siblings. Likewise, unpaid work is more common among the oldest sons in the household. On the other hand, first-born females are most likely to obtain remunerated work and less likely to work and not earn income.

In an attempt to solve the endogeneity problem, a second strategy based on the method proposed by Rosenzweig and Schultz (1987) could be pursued. This procedure involves estimating a fertility regression from which residuals are obtained, to later introduce them into the model as an exogenous variable to replace the number of children regressor.6 The logic behind this resides in the fact that residuals can be interpreted as a natural measure of fecundity, because they function as an estimator for the unexplained component of fertility. This is true because although the residuals are correlated with true fertility, they would not have a relationship with the error terms of the work or school attendance regression (Emerson and Souza, 2008: 1652).

Table 6 shows the results of this model, where the estimates are very similar to those introduced in Table 4. One interesting outcome is that the youngest male children are more likely to attend school than their first-born siblings, but less likely to attend as compared to the rest of their siblings. For females, the youngest daughters are more likely to attend school, whereas oldest daughters tend to attend school less. In terms of working, the youngest males are less likely to work than their first-born brothers, regardless of whether it is a paid or unpaid job. For females, first-born daughters tend to participate more in unremunerated labor, while the youngest daughters are more likely to obtain remunerated jobs.

The final empirical strategy involved reclassifying the sample, where only children or teenagers whose households required their work or economic contribution were considered child laborers. This suggests that these families introduce their children into the workforce when they face budgetary restraints. Because there are not enough resources to send the first-born to school, and because the first-born child has greater income capability and ease of finding an occupation, the expectation is that in these households facing poverty, the older children are more likely to work.

The estimates in Table 7 reveal similar results as described earlier. However, the magnitudes between the coefficients are now accentuated, such that there is a greater difference between the activities children carry out by birth order. This is suggestive of the idea that poverty and budgetary restraints play a significant role in decision-making regarding time allocation in households.


This article studied the effect of birth order on time allocated for attending school and participating in the remunerated and unpaid labor market in Mexico, using MTI data and estimating a multivariate probit model taking into account the ways in which these three situations are connected. The outcome revealed that first-born children are less likely to attend school and more likely to work, as compared to their siblings. On the other hand, younger children are more likely to attend school, and for females, less likely to engage in remunerated work as compared to their older siblings. The relationships found here are accentuated in households where children work due to a lack of resources. This study provides evidence for significant differences in the way time for various activities is assigned depending on the birth order of the children.

To mitigate child labor in Mexico, the laws that are currently in place and prohibit this practice must be applied more strictly. There must be strong regulation to guarantee favorable and flexible working conditions for minors within the legal age range to work. In addition, programs should be implemented to de-incentivize child labor and promote school attendance, taking into account not only the socioeconomic status of a household but also its makeup. However, more than anything, sustainable economic development will be required to raise real income levels so that households no longer feel the need to introduce their minors into the workplace and child labor will only exist as a choice, and not as the result of a lack of resources.

Finally, we must not forget that although child labor tends to be negative, there are exceptional cases in which it could be beneficial, as long as it does not interfere with attending school and doing homework, as it has the potential to promote the physical and mental development of children. Consequently, a poorly designed policy aiming to eliminate child labor altogether would not take into account the potential positive externalities the practice might imply, and could even raise the poverty levels of these households, further aggravating the initial situation.


Baland, Jean, and James Robinson (2000), “Is child labor inefficient?” Journal of Political Economy, vol. 108, no. 4, Chicago, University of Chicago Press, July, pp. 663–679.

Bando, Rosángela; Luis López-Calva et al. (2005), “Child labor, school atten-dance, and indigenous households: Evidence from Mexico”, World Bank, working paper, no. 3487, Washington.

Basu, Kaushik, and Pham Van (1998), “The economics of child labor”, American Economic Review, vol. 88, no. 3, Nashville, American Economic Associa¬tion, June, pp. 412-427.

Basu, Kaushik (1999), “Child labor: Cause, consequence, and cure, with re¬marks on international labor standards”, Journal of Economic Literature, Nashville, American Economic Association, septiembre, vol. 37, no. 3, pp. 1083-1119.

Becker, Gary (1991), A treatise on the family, Boston, Harvard University Press, pp. 304.

Börsch-Supan, Axel, and Vassilis Hajivassiliou (1993), “Unbiased multivariate probability simulators for maximum likelihood estimation of limited de¬pendent variable models”, Journal of Econometrics, vol. 58 no. 3, Amster¬dam, Elsevier, August, pp. 347–368.

Cappellari, Lorenzo, and Stephen Jenkins (2003), “Multivariate probit regres¬sion using simulated maximum likelihood”, Stata Journal, vol. 3, no. 3, College Station, Stata Press, first trimester, pp. 278-294.

Cigno, Alessandro, and Furio Rosati (2005), The economics of child labour, Oxford, Oxford University Press, pp. 264.

Córdova, Karina (2009), “The effect of birth order on education levels and time allocation decisions: Evidence from Mexican households”, mimeo. Tucson, University of Arizona.

Constitución Política de los Estados Unidos Mexicanos (2008), México, Tri¬bunal Electoral del Poder Judicial de la Federación.

Dammert, Ana (2010), “Siblings, child labor, and schooling in Nicaragua and Guatemala”, Journal of Population Economics, vol. 23, no. 1, Berlin, January, pp. 199-224.

De Haan, Monique; Erik Plug et al. (2012), “Birth order and human capital development: Evidence from Ecuador”, IZA, discussion paper, no. 6706, Bonn, IZA.

Emerson, Patrick, and André Souza (2008), “Birth order, child labor, and school attendance in Brazil”, World Development, vol. 36, no. 9, Amsterdam, Elsevier, August, pp. 1647-1664.

Hazan, Moshe, and Binyamin Berdugo (2002), “Child labour, fertility and economic growth”, Economic Journal, vol, 112, no. 482, London, RES, October, pp. 810-828.

Heckman, James (1979), “Sample selection bias as a specification error”, Eco-nometrica, vol. 47, no. 1, New Jersey, Wiley, January, pp. 153-161.

INEGI (Instituto Nacional de Geografía y Estadística) (2012), “Módulo de Trabajo Infantil 2011, Encuesta Nacional de Ocupación y Empleo”, Aguascalientes, available at: http://www.inegi.org.mx/est/contenidos/ proyectos/encuestas/hogares/modulos/mti/mti2011/default.aspx.

Keane, Michael (1994), “A computationally practical simulation estimator for panel data”, Econometrica, vol. 62, no. 1, New Jersey, Wiley, January, pp. 95–116.

Levison, Deborah; Karine Moe et al. (2001), “Youth education and work in Mexico”, World Development, vol. 29, no. 1, Amsterdam, Elsevier, January, pp. 167-188.

López-Calva, Luis, and Samuel Freije (2001), “Child labor, school attendance, and poverty in Mexico and Venezuela”, Centro de Estudios Económicos, documento de trabajo, no. 1, Mexico, Colegio de México.

López, Antonia (2005), “Schooling and child labour in Mexico: An empirical analysis”, Análisis Económico, vol. 20, no. 45, Mexico, UAM, third quarter, pp. 89-117.

OIT (Organización Internacional del Trabajo) (2010), Accelerating action against child labour, Ginebra, OIT, pp. 98. Ranjan, Priya (2001), “Credit constraints and the phenomenon of child la¬bor”, Journal of Development Economics, vol. 64, no. 1, Amsterdam, El¬sevier, February, pp. 81-102.

Rosenzweig, Mark, and Paul Schultz (1987), “Fertility and investments in hu¬man capital: Estimates of the consequence of imperfect fertility control in Malaysia”, Journal of Econometrics, vol. 36, núms. 1-2, Amsterdam, Else¬vier, September-October, pp. 163-184.

Skoufias, Emmanuel, and Susan Parker (2001), “Conditional cash transfers and their impact on child work and schooling: Evidence from the Progresa program in Mexico”, Economia, vol. 2, no. 1, Washington, Brookings Institution Press, Fall, pp. 45-86.

White, Halbert (1980), “A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity”, Econometrica, vol. 48, no. 4, New Jersey, Wiley, May, pp. 817-838.

*University of Sussex, United Kingdom. p.orraca-romano@sussex.ac.uk.

1 The literature explaining the reasons behind child labor is extensive, including Ranjan (2001), which explains child labor as the product of credit restrictions and unequal income distribution. On the other hand, Hazan and Berdugo (2002) claim that child labor is a common practice when an economy is rather underdeveloped. However, technology progress increases the salary gap between adult and child workers, meaning that adults substitute child labor for education. For a review of the literature, see Basu (1999).

2 This definition potentially excludes the most vulnerable members of society, like children, who offer their products and services on public roads or on public transportation (INEGI, 2012). It also leaves out children and teenagers living in households without parents.

3 Specifically, in this study Yi,1 = 1 if the minor attends school, 0 if not; Yi,2 = 1 if the minor engages in remunerated work, 0 if not and, finally ; Yi,3 = 1 if the minor engages in unremunerated work, 0 if not.

4 For a more detailed discussion of the estimation method and the statistical program used, see Cappellari and Jenkins (2003).

5 All of the regressions in this study included dichotomous variables indicating the state of residence, estimated using robust standard errors based on White (1980).

6 Concretely, a fertility regression is estimated using the LOS method with robust standard errors based on White (1980), where the dependent variable is given by the number of children. The exogenous variables vector is composed of the following regressors: father’s age, mother’s age, father’s level of schooling, mother’s level of schooling and a dichotomous variable indicating whether or not the household is rural.

Licencia de Creative Commons  Problemas del Desarrollo. Revista Latinoamericana de Economía by Instituto de Investigaciones Económicas, Universidad Nacional Autónoma de México
is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License

Published in Mexico, 2012-2018 © D.R. Universidad Nacional Autónoma de México (UNAM).
PROBLEMAS DEL DESARROLLO. REVISTA LATINOAMERICANA DE ECONOMÍA, Volume 49, Number 195 October-December 2018 is a quarterly publication by the Universidad Nacional Autónoma de México, Ciudad Universitaria, Coyoacán, CP 04510, México, D.F. by Instituto de Investigaciones Económicas, Circuito Mario de la Cueva, Ciudad Universitaria, Coyoacán,
CP 04510, México, D.F. Tel (52 55) 56 23 01 05 and (52 55) 56 24 23 39, fax (52 55) 56 23 00 97, www.probdes.iiec.unam.mx, revprode@unam.mx. Journal Editor: Moritz Cruz. Reservation of rights to exclusive use of the title: 04-2012-070613560300-203, ISSN: pending. Person responsible for the latest update of this issue: Minerva García, Circuito Maestro Mario de la Cueva s/n, Ciudad Universitaria, Coyoacán, CP 04510, México D.F., latest update: January 9th, 2019.
The opinions expressed by authors do not necessarily reflect those of the editor of the publication.
Permission to reproduce all or part of the published texts is granted provided the source is cited in full including the web address.
Credits | Contact

The online journal Problemas del Desarrollo. Revista Latinoamericana de Economía corresponds to the printed edition of the same title with ISSN 0301-7036