Are Migrant Children at Risk of Child Labour? Empirical Evidence from Pakistan

Abstract

Child labour is a universal concern as one child in every ten is engaged in child labour globally. Pakistan is no exception to this trend. This study attempts to identify the role of migration on children’s vulnerability to work. For that purpose, this study evaluates the demographic and household attributes, collected from the Labour Force Survey (LFS) (2017–2018), and Logit regression is used for empirical analysis. The findings reveal that migration is one of the decisive determinants of child labour, which is triggered by the high economic and social disparities existing across space in Pakistan. Therefore, balanced growth policies are recommended that require attention to the socio-economic development across all regions and provinces of the country. The study also emphasises the role of both supply-side and demand-side factors in controlling the menace; poverty alleviation reforms, development of social institutions in rural areas, and greater enforcement of child labour laws are strongly recommended.

KEYWORDS::

• Migration
• child labour
• demand-side factors
• supply-side factors

JEL Codes::

• J210
• J200
• J130
• J820

1. Introduction

152 million children (88 million boys; 64 million girls) are globally active as child labourers, which sums up to about 10% of all children in the world. Moreover, 73 million children are engaged in hazardous work (ILO, 2017). Child labour is a serious violation of the fundamental rights of children, which converts the country’s most prized asset into a liability (ILO, 2015).

During their childhood phase, children are allowed to experiment, make mistakes, and build their perspectives, based on their own experiences and understanding. However, those children who become economically active at an early age often miss these opportunities. Early engagement in work stunts their physical, emotional, and social development, and they remain unskilled, less productive, and incompetent throughout their life (Faria, 2010).

Child labour is a global phenomenon and Africa leads the world with the highest number of children (72 million) engaged in child labour, followed by Asia and the Pacific region, which ranks second for having the second largest population of children (62 million) working as child labourers. Africa and the Asia-Pacific region together account for nine out of every ten children engaged in child labour in the world (ILO, 2017).

According to Versik Maplecroft’s survey, Pakistan ranked among the top six out of 197 countries with the worst form of child labour (Sheer et al., 2018). In recent research, published by the Institute of Historical and Social Research (IHSR), it was highlighted that, during the year 2017-18, over 3 million children in Pakistan were active as child labourers which translates into one in every ten children in the country. Moreover, it was found that children, on average, worked for 43.79 hours per week, and most of the working children (66.67%) were part of hazardous work (Aman, 2022).

Migration is often considered a mitigating strategy to deal with poverty, promote employment, raise income, and enable access to improved healthcare and education facilities. However, the effects of migration may be experienced differently by people belonging to different socio-economic backgrounds. Day-to-day survival may present a challenge to several migrants, forcing them to work for low wages. Additionally, due to their weak socio-economic profile and inaccessibility to social and economic resources, migrant people often face un(der)employment, live in overpriced dwellings, and become subject to economic and social discrimination in host regions. Despite experiencing such a harsh environment, return migration is often not an option in South Asian and African countries, due to the higher cost associated with it. It also pushes migrants to find ways to survive in the host regions and send their children to participate in low-paid and dangerous work (Maddern, 2013).

Extensive literature has shed light on the determinants of child labour in both developed and developing economies. Nevertheless, only a few studies (Van de Glind, 2010; Van de Glind & Kou, 2013; Das & Mishra, 2013) have focused on migration as a decisive factor in promoting child labour. The International Labour Organization (ILO) is actively highlighting these problems in order to protect the rights of marginalised groups of children, such as child labourers, children engaged in hazardous work, and trafficked and migrant children. For instance, joint research conducted in 2012 by Child Helpline International (CHI) and ILO in Kenya, Nepal, and Peru concluded that migrant children who were engaged as child labourers were found to be working under harsh working conditions, compared to their non-migrant counterpart. The study found these enlightening results while considering the differences in work hours, incomes received, exposure to violence, engagement in hazardous work, exposure to captivity, living conditions, and level of education. In the case of Pakistan, Aman (2022) observed that migrant children worked for long hours and their educational attainment was less than the native children in urban areas. The study concluded that migrant children were generally more prone to work as child labourers than their native counterparts. To the best of our knowledge, no previous study has established this claim empirically. Therefore, the present study contributes to the literature by empirically testing this hypothesis in the context of Pakistan. Moreover, it also establishes a comprehensive policy framework in order to address the underlying factors that compel children generally, and migrant children particularly, to take up work responsibilities.

The present study is organised in the following manner. Section 2 presents the literature review encompassing the potential determinants of child labour in developed and developing economies. Section 3 demonstrates the current trends of migration in the context of Pakistan. In Section 4, the study constructs the empirical model while incorporating the role of demand and supply-side factors. The discussion on the methodology, sample size, and data source is provided in Section 5. Sections 6 and 7 discuss the general and major findings of the study, including policy implications specifically in the context of Pakistan. Finally, Section 8 gives a summary and concludes the study.

2. Review of Literature

In the literature, poverty is considered the most compelling reason for engaging in child labour (Goswami & Jain, 2006). This argument has gained worldwide empirical support; however, a substantial number of studies also acknowledge the role of other socio-economic factors in influencing the incidence of child labour. Ahmed (1999) analyzed the cross-sectional data of 74 countries and concluded that poverty is a minor determinant of child labour and the incidence is better explained by other socio-economic factors, such as income inequality, the share of agricultural output in GDP, adult literacy rate, primary enrolment rate, and the share of child population in a country. Similarly, Chauhan and Ul Ain (2019) provided evidence from 30 developing countries according to which child labour is influenced by poverty, and lack of social progress facilitates this relation. The study further revealed that adult unemployment directly impacts child labour incidence.

In the case of Pakistan, several researchers (Ali, 2011; Malik et al., 2012; Kazmi, 2015) believe that poverty is the prime cause of child labour; however, the role of other socio-economic factors is also considered to be substantial. Ali (2011) recognised a connection between the nuclear family system and working children in the Swabi district (Khyber Pakhtunkhwa). The study highlighted the role of other socio-economic factors apart from poverty and found that 73% of children engaged in child labour belonged to nuclear families; only 37% of children belonged to households with a household monthly income above Rs. 6000 (or $255.24)¹; 86% were unschooled; and 55% were not receiving pocket money. Also, this argument has gained countrywide empirical support in the context of Pakistan: Khan (2003) provided empirical evidence from two districts of Punjab (Pakpatan and Multan); Kazmi (2015) from Sahiwal district of Punjab; Malik et al. (2012) from the districts of Sukkur and Multan; and Ahmed et al. (2012) from Kohat and Hangu districts of Khyber Pakhtunkhwa.

The literacy attainment of a child’s parents is a factor that has a bearing on the likelihood of that child participating in work. This idea is also substantiated in a study conducted by Siddhanta and Nandy (2003) which empirically found a negative relationship between parents’ educational attainment and their children’s participation in work in the Howrah district of Indian West Bengal. Correspondingly, a linkage between adult literacy and the growth of human capital was identified by Chaudhary and Khan (2002). The study argued that an increase in the literacy level reduces economic pressure on children. Based on an extensive dataset of developed and developing countries, the study provided evidence of the negative association between adult literacy and child labour. Moreover, the importance of female literacy to combat child labour was highlighted by Krutikova (2009). The study claimed that education differences among parents give a relative bargaining/decision-making power and females’ negotiating power significantly affects the child labour incidence in Andhra Pardesh (India). Nevertheless, female literacy has the opposite effect on child labour across rural-urban regions; it discourages the incidence in urban regions, while encourages it in rural regions.

A distinct factor that could play a significant role in influencing child labour decisions is the gender of the household head. Although it is a general phenomenon that men are the providers in the family, in some cases, financial responsibilities are shared by women as well. Previous literature shows mixed evidence of the gender of the household head in influencing child labour incidence. For instance, Ray (2002) identified that children from female-headed households are more vulnerable to work than children who belong to male-headed households. Whereas Khan (2003) is among the few researchers who found contrasting results and argued that poor females usually lack the physical and social capital needed to send children to work. Therefore, children are sent to school instead of work.

Additionally, the age of the family head is also counted as a crucial determinant of child labour. With growing age, one’s physical ability to work decreases which causes a reduction in earning capacity; therefore, parents in older age rely upon their children for income. In other words, children with younger parents may be less likely to participate in child labour than those with older parents. Yet, parents’ age is rarely investigated in relation to child labour. Khan (2003) is among the few researchers who empirically tested the claim and found that the tendency to work at an early age is higher among households where the family head is older.

Moreover, other household characteristics, such as family size, number of children, household income, and the number of literate persons in the household are also well explored in influencing a child’s decision to participate in child labour. Siddhanta and Nandy (2003) observed that large family size and high dependency ratio of a household push children to work in the Zari industry in Indian West Bengal. While Ali (2011) shared this view in the context of Pakistan and found that for most of the working children, the average household size consisted of 10-13 persons compared to 1-3 employed persons who were the providers in these families in Pakistan. Also, empirical support has been provided for the positive linkage between household size and child vulnerability to work in the context of Pakistan (Khan, 2003; Ahmed et al., 2012).

Interestingly, pervasive literature has also focused on different aspects of child labour distinctively for developed and developing economies. Nevertheless, little research (Van de Glind, 2010; Van de Glind & Kou, 2013; Das & Mishra, 2013) engrossed the importance of migration in promoting child labour. In this regard, ILO has been promoting awareness and raising voices to discourage child labour and other marginalised groups of children, such as children employed in hazardous work, and trafficked and migrant children.

In the case of developing economies, Mishra (2017) identified rural-to-urban migration as a potential source of child labour. In this connection, Das and Mishra (2013) performed a comprehensive analysis to explore the phenomena of child migration, child trafficking, and child labour in India. Their study provides evidence that low-income families in rural areas often migrate to urban areas in search of an improved standard of living and push their children to participate in paid work activities to supplement their family income.

Similar research has been pursued jointly by ILO and CHI (2012) in Kenya, Nepal, and Peru to analyze child labour among migrant and local children. Findings revealed that migrant children are worse off than non-migrant children in terms of working hours, wages, exposure to violence, engagement in hazardous work, exposure to bondage, living conditions, and level of education.

Van de Glind and Kou (2013) analyzed the comprehensive literature reviewed by ILO-IPEC (2010) and identified that migrant children in the southern regions are highly vulnerable to child labour. Besides, migrant child labourers, being the least visible and more politically weak part of society, are worse off than native child workers in terms of working hours, income, and attending school. Aman (2022) further established this connection and indicated that migrant children work for longer hours, and that they are more vulnerable to the incidence than their native counterparts in Pakistan.

Conceptually similar work has been carried out by Van de Glind (2010). The correlation between migration and incidence of child labour was reviewed, and findings revealed that a family’s seasonal migration and a child’s independent migration increases the vulnerability of children to child labour incidence. The study also demonstrated that children who migrate on a seasonal basis along with their families during off-season activity often belong to the agricultural sector or brick kilns. During natural calamities, such as floods or earthquakes, they lose their shelter and bear damages to infrastructure and source of livelihood, which forces them to migrate at least temporarily. It becomes difficult for children who attend school to continue their studies when facing seasonal migration. The loss of education pushes these children to participate in labour. Predominantly, these independently migrated children become prey to being exploited by employers, as they suffer the limitation of none or little knowledge of legal channels.

Unlike prior research, which emphasises the role of supply-side factors, such as household poverty, household size, and parents’ illiteracy, in explaining child labour incidence, this study has adopted a balanced approach by focusing on supply and demand-side factors. In the case of Pakistan, only a few researchers, such as Fatima (2017), have focused on the role of demand-side factors, such as the size of the informal sector and the size of the agricultural sector in an area, in influencing child labour incidence. Therefore, this study could be a significant contribution to the literature on child labour.

3. Migration Trends in Pakistan

Pakistan is one of those South Asian countries where the trend of migration towards cities is intensifying swiftly; as a result, urban regions are growing faster in terms of population size. Figure 1 demonstrates the distribution of population across rural-urban regions. It is visible that the urban population’s share is constantly rising in Pakistan, and it reached a significant level of 36% in 2017. It is noted that the prime reason for this migration is to look for better employment opportunities. Specifically, the historical trends imply that the major share of this migration is from the agricultural sector to the manufacturing or services sectors. Moreover, as Karachi and Lahore are the two largest urban cities of Pakistan, therefore the majority of these migrant people move to these cities in Pakistan. Consequently, the size of these two cities is increasing exponentially. The United Nations projected that Karachi will maintain its rank among the top 12 largest cities, while Lahore is likely to be listed among the top 18 largest cities in the world by 2050 (World Urbanization Prospects, 2019). However, with growing urbanisation, the size of the informal sector is also massively expanding in Pakistan. Although this informal sector supports about 72% of total employment in the country, it is also the foremost perpetrator of escalating child labour incidence in Pakistan.

Figure 1. Population distribution across rural-urban regions in Pakistan (in percent).

Source: Author’s computations based on the data taken from Pakistan Economic Survey 2020-21.

Current migration trends in Pakistan, presented in Table 1, show that a large proportion of people (13.1%) live as migrants in Pakistan. Statistically, the incidence of migration is higher for females (56.16%) than males (43.84%). Moreover, a large share of migrated people (55.27%) resides in urban regions, indicating that migration to urban regions dominates in Pakistan. Across the whole region of Pakistan, Punjab hosts the largest population of migrants (62.54%), followed by Sindh (17.29%), Khyber Pakhtunkhwa (14.27%), Islamabad (5.04%) and Balochistan (0.86%).

Table 1. Migration trends in Pakistan*.

Status	Category	Frequency	Percent	Total
Migration	Native	115,555,824	86.9	132,974,929
	Migrants	17,419,105	13.1
Gender	Male	7,635,668	43.84	17,419,105
	Female	9,783,437	56.16
Region	Rural	7,792,184	44.73	17,419,105
	Urban	9,626,921	55.27
Province	Khyber Pakhtunkhwa	2,486,007	14.27	17,419,105
	Punjab	10,893,594	62.54
	Sindh	3,012,302	17.29
	Balochistan	149,582	0.86
	Islamabad	877,620	5.04

Source: Author’s computations based on the data taken from Labour Force Survey 2017–18; * Computations presented in this table have been adjusted with population-representative frequency weights, therefore computations can be generalised over the total population of the country.

Furthermore, the reasons for migration vary with respect to gender and age group. The study represents these reasons for migration among adult males, females, and children in Figure 2. For adult males, migration often takes place for economic reasons, classified as “job transfer” (5.9%); “found a job” (13.12%); “searching for a job” (11.91%); “business” (3.64%), which cumulatively account for 34.57% of total migration among them, indicating that some regions offer better employment prospects than others, generating the trend of migration among adult males in the country. “Return to home” (26%) is another main reason for migration among adult males, followed by “with parents” (22.09%). In contrast to adult males, most of the adult females migrate for non-economic reasons, such as marriage (62.6%); “with spouse” (20.69%); “with parents” (8.2%), which cumulatively account for 91.5% of the total migration incidence among adult females. On the other hand, children’s distribution by migration status indicates that child migration decisions mainly depend on parents’ migration status, as a significant proportion (83.62%) migrate with parents. In addition, the second noteworthy reason for migration among children is “education” (3.88%), indicating that some regions offer better education opportunities than others; thus, encouraging migration among children. Intriguingly, the distribution of migrant people by reason indicates that a substantial share of children and adult females’ migration eventually depends on the migration status of adult males who usually migrate for economic reasons. Lastly, it also indicates that child migration is primarily influenced by the socio-economic disparities existing across regions, as most children either move with their parents or migrate for better education.

Figure 2. Reason for migration in Pakistan.

Source: Author’s computations based on the data taken from Labour Force Survey 2017–18.

4. Theoretical Framework & Model Specification

Irrespective of its relative strength, poverty is a crucial determinant of child labour in developing countries. Therefore, the present study found its theoretical underpinnings from the “Luxury Axiom” of Basu (1999), which stresses the role of adult market wages as a substantial determinant of the supply of child labour. The Luxury Axiom assumes that households send their children to work if adults’ market wages are found below the minimum needed to finance expenses. Therefore, the poverty status of a child has been evaluated by household income per head, which is measured by the total income of adults in a household adjusted for the number of people in a single household. It is presumed that a higher per capita household income (HHINC) will reduce child labour incidence (CHLBR).

As literature has already emphasised the role of factors other than poverty in influencing child labour incidence, the present study controls the model for household factors, such as the age of the family head (HAGE), gender of the family head (HGEN), literacy status of the family head (HLIT), the number of literate adult males (ADMLIT) and the number of literate adult females (ADFMLIT) in the household. The role of household size (HHSIZ) is also well acknowledged in the literature in influencing a child’s decision to participate in work. However, the inclusion of the variable, due to its bi-directional causal relationship with child labour, may become a potential source of endogeneity in the model. Large family size may encourage child labour by lowering the per capita income of adult members in the household (Khan, 2003; Ahmed et al., 2012). Reverse causality arises if parents consider children a helping hand in increasing earnings, and thus expand the family size in order to supplement their family income (Fatima, 2017). As the inclusion of endogenous variables gives inconsistent estimates, the endogenous nature of household size has been tested by the correlation between household size and the residual. Ultimately, a weak correlation (0.007) between the two variables refutes any possibility of the household size’s endogenous nature and provides the statistical basis to include household size in the model.²

As far as a child’s attributes are concerned, they are also crucial in explaining the incidence of child labour. To reduce the biases of these attributes, the empirical model of this study controlled for child age, child age quadratic term, and gender effects. Besides, the study has considered child age (CHAGE) a continuous variable that measures child age in years, while child gender (CHGEN) is a discrete variable.

To test the effect of child migration status on a child’s vulnerability to engage in work, a dummy variable of child migration status (CHMIG) is incorporated into the model which takes a value of “1” if the child is a migrant and “0” otherwise. Given the feebler socio-economic profile of adult migrants in Pakistan, who usually migrate (to urban areas) in search of better livelihood, the present study posits that children’s migration stimulates child labour incidence in Pakistan. Through the incorporation of all the essential factors, the equation below evaluates the effect of child migration on his/her vulnerability to participate in child labour, where child labour is the dependent variable in the equation. $$\mathit{\text{CHLBR}}={\beta }_0+{\beta }_1\mathit{\text{HHINC}}+{\beta }_2\mathit{\text{HAGE}}+{\beta }_3{\mathit{\text{HA}}\mathrm{G}E}^2+{\beta }_4\mathit{\text{HGEN}}+{\beta }_5\mathit{\text{HLIT}}+{\beta }_6\mathit{\text{ADMLIT}}+{\beta }_7\mathit{\text{ADFMLIT}}+{\beta }_8\mathit{\text{HHSIZ}}+{\beta }_9\mathit{\text{CHAGE}}+{\beta }_{10}{\mathit{\text{CHAGE}}}^2+{\beta }_{11}\mathit{\text{CHGEN}}+{\beta }_{12}\mathit{\text{CHMIG}}+{\beta }_{13}\mathit{\text{REG}}+{\beta }_{14}\mathit{\text{PROV}}$$

As argued earlier, the effect of demand-side factors was rarely considered in literature. The present study contributes by filling this void through the inclusion of demand-side factors of child labour. For this purpose, regional (REG) and provincial (PROV) dummies have been introduced into the model. Characteristically, the regional dummy controls the model for the differences in labour market conditions existing across rural-urban regions; for instance, in some regions, there may exist industries that offer employment for children or where children can be substituted with adult workers, especially in case of unskilled work. Meanwhile, the provincial dummy controls the model for provincial heterogeneity having its distinct impact on child labour incidence in each province. Therefore, provincial heterogeneity might arise on account of labour market differences existing across provinces. For instance, the degree of enforcement of child labour laws may vary greatly from province to province or in some provinces, there may exist industries where children can work in.

5. Research Methodology

In this study, the nature of the dependent variable (Child Labour) is discrete; therefore, the present study utilises the Binomial Logit regression method (see, for instance, Studenmund & Johnson, 2017). Subsequently, this method allows the estimation of the probability of any binomial outcome to fall in the range of 0 and 1. Also, the true probability of the outcome variable taking value equal to 1 can be presented as follows: $$P=\frac{1}{1+e^{-\left({\beta }_0+{\beta }_i{X}_i\right)}}$$

As the true probability (P) cannot be observed, detected values of the outcome variable “Child Labour” support estimating the above logit equation. An estimated logit equation can be expressed as follows: $$p=\frac{1}{1+e^{-\left({\beta }_0+{\beta }_i{X}_i\right)}}$$

In addition, the Maximum Likelihood (ML) method is applied to estimate the Logit equation, as it comprises the estimation of non-linear parameters. ML method is the log of the odd ratios of the probability of the occurrence of an event (P) to the non-occurrence of an event (1-P) known as the “log of odds”. Besides, a few mathematical arrangements of the ML method allow us to write the standard Logit equation in terms of the “log of the odds”. Henceforth, the standard form of a simple Logit equation is: $$\mathit{\text{ln}}\left(\frac{P}{1-P}\right)={\beta }_0+{\beta }_i{X}_i$$

ML method follows large sample properties and offers consistent and efficient estimates. The discrete nature of our outcome variable of child-decision to participate in the labour force and the availability of a large sample allows us to utilise the Binomial Logit regression to serve the study’s requirements.

Unfortunately, no separate child labour survey has been conducted in Pakistan since 1996; due to this, the present study has collected the information on child employment status and the details on child demographics and household attributes from the Labour Force Survey (2017–18)³. It is a household survey that covers all the participants of the labour force in the country who are 10 years of age and older and gives detailed information on gender-disaggregated characteristics of the labour force at the national and provincial level with rural-urban breakdown. Information provided in the Labour Force Survey (2017–18) has been taken from a representative sample of 43,361 households in the country.

In the context of data limitation, employment details cannot be accessed for working children under 10 years of age from the source. Consequently, this study covers all the children who belong to the age group of 10-17 years. Hence, the study’s sample size is 52,269 children who fall in the age group of 10-17 years.

The definition of child labour has been adopted from Aman (2022), which comprises all children (below 12 years of age) who are economically active; children (12-14 years of age) who are economically active for at least 14 hours per week; children (15-17 years of age) who are economically active for at least 43 hours per week; children (10-17 years of age) working in hazardous processes or hazardous occupations.

As stylized facts, a brief demonstration of the socio-economic characteristics of the sample of this study is presented in Table 2. It indicates that 9.49% of the children were found participating in child labour. Among them, 53.87% were male children. Moreover, a substantial portion of the sample, 39.68%, falls in the age group of 12-14 years, followed by 33.02% of children falling in the age group of 15-17 years, and around 27.3% of children fall in the age group of 10-11 years in the sample. Furthermore, in the sample, 95.52% of children are native; the leading proportion of children, 66.77%, is residing in rural regions compared to 33.23% of children who inhabited urban regions. Besides, the provincial distribution of the study sample shows that an enormous proportion of children, 41.29%, reside in the province of Punjab, followed by 22.76% living in Sindh province, 20.8%, in Khyber Pakhtunkhwa and 14% in Balochistan.

Table 2. Frequency distribution of study sample*.

Child attributes	Category	Frequency	Percent	Total
Labour status	Not in child labour	47,311	90.51	52,269
	In child labour	4,958	9.49
Gender	Male	28,158	53.87	52,269
	Female	24,111	46.13
Age group	10-11 years	14,273	27.3	52,269
	12-14 years	20,739	39.68
	15-17 years	17,257	33.02
Migration status	Native	49,925	95.52	52,269
	Migrated	2,344	4.48
Region	Rural	34,902	66.77	52,269
	Urban	17,367	33.23
Province	Khyber Pakhtunkhwa	10,874	20.8	52,269
	Punjab	21,581	41.29
	Sindh	11,895	22.76
	Balochistan	7,319	14
	Islamabad	600	1.15

Source: Author’s computations based on the data taken from Labour Force Survey 2017–18; * This frequency distribution table is presented to provide details of the study sample and has not been adjusted for population-representative frequency weights; therefore, caution is required in generalising the computations in this table over the total population of the country.

6. Result Analysis

This section discusses the main findings of the study. The results obtained from the Logit regression method have been presented in Table 3. Pseudo R², a measure of goodness of fit of the model, is based on the Log-likelihood ratios of the estimated and null model which measures the improvement in the estimated model over the null model, is 0.166. Also, LR-Chi²-statistics, which measures the significance of the overall model, is found significant at the 1% level.

Table 3. Determinants of child labour in Pakistan (Logit Regression-Marginal Effect).

Independent variables	Marginal effect	Std. Err.	Descriptive data average
Household Adult Income Per person	−0.000002***	0.00	1809
Head Age	−0.002***	0.01	46.5
Square of Head Age	0.00001*	0.00	2267.3
Gender of Family Head (Male is reference category)	−0.05***	0.07	1
Literacy Status of Head (Illiteracy is reference category)	−0.10***	0.04	0.5
Number of literate adult males in the household	−0.01***	0.02	0.7
Number of literate adult females in the household	−0.02***	0.02	1.6
Household Size	0.01***	0.01	7.9
Child age	0.11***	0.10	13.2
Square of child age	−0.003***	0.00	180.4
Child gender (Male is reference category)	−0.09***	0.04	1.4
Child migration status (Native status is reference category)	0.02***	0.08	0.04
Region (Rural region is reference category)	−0.04***	0.04	1.3
Province (Khyber Pakhtunkhwa is reference category)
Punjab	0.06***	0.05	–
Sindh	0.05***	0.05	–
Balochistan	−0.002	0.07	–
Islamabad	−0.01	0.24
	Model Summary
Number of observations	52056
Log Likelihood	13590.03
Pseudo-R^2	0.1661
LR chi^2 (17)	5414.82
Prob > chi^2	0.00

Note: ***, ** and * represents significance at 10%, 5% and 1% significance level, respectively.

The findings of this study confirm the “Luxury Axiom” of Basu (1999). To be precise, one unit increase in per capita household adult income reduces the probability of children participating in work by 0.000002 points. Moreover, Child labour demonstrates a U-shape pattern over the household head’s life cycle. It implies that children with a younger household head are less likely to engage in child labour than children with an older household head. It also depicts that if the age of the family head increases by one year, the probability of child labour first decreases by 0.002 points, and after age reaches a specific limit, the probability elevates by 0.00001 points. In the previous literature, a similar pattern was observed by Khan (2003). Furthermore, the gender of the household head is statistically significant and shows that children from female-headed households are 0.05 times less vulnerable to child labour than their male counterparts. These findings, too, are consistent with Khan (2003).

Besides, the study found a statistically significant relationship between the literacy of household members and child labour incidence. To be precise, the probability of a child participating in labour is 0.1 times lower among children whose family head is literate. Likewise, it is found that the addition of one adult-literate male in the household decreases the probability of children participating as labourers by 0.01 points, while the addition of one adult female to the household reduces the incidence by 0.02 points in the country. These findings support the results of the previous research (Chaudhary & Khan, 2002; Khan, 2003; Ahmed et al., 2012).

Another important factor in the model, household size represents a statistically significant and positive association with child labour incidence in the country. Also, the results show that an increase in household size by one person is likely to increase child participation in work by 0.01 points. Previously, Ahmed et al. (2012) has also confirmed this finding.

When considering a child’s characteristics, it is found that a child’s age depicts an inverse U-shape (∩) and statistically significant association with his/her probability of taking up work. A one-year increase in age encourages a child’s probability to work by 0.11 points. However, a non-linear association between a child’s age and work indicates that child participation as a labourer increases with age at a decreasing rate. This might be due to the cost of hiring children (child wage-demand), which increases with child age and discourages firms from hiring older children. Child gender is also statistically significant in influencing a child’s tendency to participate in work. It appears that female children are 0.09 times less vulnerable to the incidence compared with their male counterparts. This happens due to the social norms of Pakistan, where labour force participation for females is not appreciated. This finding is aligned with Khalid and Shahnaz (2004) and Khan (2003).

Our primary variable, child migration status, is also statistically significant with an expected coefficient sign, implying that migrant children are more vulnerable to child labour incidence. To be precise, it shows that the tendency of participating in child labour is 0.02 times higher among migrant children than their native counterparts. One plausible reason for this phenomenon is that in Pakistan people usually migrate with weak socio-economic profiles and face difficulty in finding a job; this pushes them to engage in semi-skilled or unskilled occupations. In some cases, they work within the informal sector that does not even pay a minimum wage. Additionally, migrants face high settlement costs including high accommodation expenses; they usually live in slum areas or at the periphery of majorities, which raises their commute costs and hampers their access to basic facilities of health and education; this provokes the need of putting their children to work. Therefore, this study has found migration as a potential influencing determinant of child labour, which confirms the results of Mishra (2017).

In addition, the study controls for regional heterogeneity, which is found statistically significant, depicting that child employment is significantly higher in rural regions than in its urban counterpart. The most likely reason for a higher rate of child labour in rural regions is the existence of agricultural land and lack of access to educational facilities in rural regions compared to urban regions which leave children with the ultimate option of doing work. According to Aman (2022), the agricultural sector is the largest employer of child workers in Pakistan and “had to help with work” and “parents do not allow and school located too far away” are the fundamental reasons among male and female children, respectively for not attending a school in Pakistan. Sajid and Ahmad (2018) concluded the same results.

Likewise, the variable of provincial dummy is also found to be statistically significant for the provinces of Punjab and Sindh but statistically insignificant for Balochistan and the capital territory of Islamabad indicating that children living in the province of Punjab and Sindh are more vulnerable to participate in work than children living in other provinces of the country. It is due to the existence of large urban settings, informal sectors, and agricultural activities in both Punjab and Sindh provinces which are the largest employment sectors of working children in Pakistan. Nevertheless, despite the availability of better employment opportunities in Islamabad, the results suggest that child labour incidence is less prevalent in the region. Being a capital territory, this city offers sophisticated employment prospects in the formal sector that leads to disincentivising child labour in this region.

7. Discussion

Among all the variables, the prime contribution of this study is the identification of the impact of migration on child labour incidence. Correspondingly, the empirical findings of this study demonstrate that migration appears as a potential source of promoting child labour incidence in Pakistan. As discussed earlier, most children migrate with their parents for monetary reasons or for better educational opportunities, which specify that, in the case of Pakistan, child migration is largely influenced by the socioeconomic inequalities that exist across rural-urban regions and provinces in the country. It is presented in Table 1 that the provinces of Sindh and Punjab are the largest recipients of migrants in Pakistan; this is due to the existence of large urban centres in both provinces that offer improved access to employment and education. For instance, in the case of Sindh, most of the migrants choose Karachi as their ultimate destination, as it has plenty of work opportunities in its industrial zones, where both skilled and unskilled labourers may readily find work. Similarly, Lahore is another destination where migrants move to avail themselves of its ample employment prospects. During the last five decades, the Government of Punjab has invested profoundly in building Lahore’s infrastructure, however, it neglected to upgrade other parts of the province, creating visible disparities in economic opportunities across Punjab. Due to this, Lahore is a developed city with several government organisations, hospitals, educational institutions, and state-of-the-art infrastructure. The availability of all the above-mentioned facilities creates employment for both unskilled and skilled workers and encourages migration towards large cities. Additionally, both Karachi and Lahore also offer improved access to education facilities which is another important reason for moving towards these cities in Pakistan. Henceforth, this study recommends adopting balanced growth policies for controlling the high influx of migration to cities, where people move in search of employment, higher income, and better access to amenities, such as education and health-related facilities. However, the higher cost of settlement and other migration-related issues make their children vulnerable to work.

The regional and provincial dummy variables, which control the model for labour market differences across space, are found significant in affecting child labour incidence in Pakistan. Therefore, the present study acknowledges the role of demand-side factors in controlling child labour incidence and recommends adoption of a comprehensive approach to control the menace by focusing on both supply-side and demand-side sources of the incidence, instead of the current practice of targeting the supply-side factors alone. Given that the agricultural sector is the largest employment industry of children, and that child labour is growing under the expanding size of the informal sector in Sindh and Punjab provinces of Pakistan, there is a strong need to enforce child labour laws in these provinces on a priority basis and to encourage people to register their businesses under the formal sector so that labourers’ rights and privileges could be readily monitored and induction of underaged workers may be discouraged. Moreover, in rural regions, the reason “had to help with work” is the most common reason for not attending school among male children, and “parents do not allow and school located too far away” are more common among female children, indicating that rural regions offer poor access to education facilities and parents in these regions prefer child work over education. Thus, there is a strong need to establish more schools to give easy access to education in rural regions. On the other hand, there is a need to create awareness among people about the importance of education and the deteriorating effects of child labour on their physical, emotional, and mental health and abilities in the long run.

Also, the results of this study are consistent with the previous research in considering the role of household factors, such as household income, age, literacy status of the family head, the number of literate adult members in the household, and household size, namely supply-side factors, for influencing child labour incidence in Pakistan. Henceforth, the government should introduce poverty alleviation reforms and provide special attention to developing the social sector, including the education and health sectors.

8. Conclusion

Child labour is a worldwide phenomenon, and its intensity is continually on the rise in developing economies like Pakistan. The main objective of this study is to identify the role of migration on the vulnerability of children to participate in work, by employing the data from the Labour Force Survey (LFS) 2017–18 and Logit regression analyses. The study concludes that migration had a strong bearing on the incidence of child labour in Pakistan. The metropolitan cities absorb a major proportion of new migrants due to better employment opportunities, improved access to social services, and well-developed infrastructure, however, higher costs of settlement and other migration-related issues make migrant children vulnerable to work. Consequently, this study suggests adopting a more balanced approach in infrastructure development and in the provision of social services through promoting growth in all parts of the country rather than focusing only on a few regions, thus removing some of the underlying reasons why people choose to migrate.

Secondly, given the importance of labour market differences across regions and provinces, the study acknowledges the role of demand-side factors in affecting child labour incidence in Pakistan. To control this incidence, it suggests encouraging the growth of the formal sector and ensuring the enforcement of child labour laws, particularly in the provinces of Punjab and Sindh. Lastly, this study shares a general agreement on the importance of supply-side factors in influencing child labour decisions; therefore, it recommends the adoption of poverty alleviation reforms and social sector development reforms in this regard.

Acknowledgements

This is an independent study; hence it does not acknowledge any specific person or organisation.

Notes

1 Equivalency for Pakistani rupee (Rs.) to US dollar ($) has been calculated by using annual average data of US dollar to Pakistani Rupee ($-Rs.) exchange rate for the study year 2011; the data has been taken from the official website of State Bank of Pakistan www.sbp.org.pk.

2 The correlation was less than 1 for given observations of 52056.

3 Primary data was generated from + the Pakistan Bureau of Statistics database <http://www.pbs.gov.pk/content/lfs-2010-2018-microdata>. Derived data supporting the findings of this study are available from the corresponding author upon request.