The long-term association between child labour and cognitive development

ABSTRACT

Child labour can deprive children of the right to a normal childhood and impair their productivity and earning capacity in later life. The relationship between child labour and cognitive development is central to these effects but has not yet been a focus of empirical research. Using panel data from Ethiopia and applying an instrumental variables estimator, we find a strong association of cognitive development with the amount of time previously spent by children on income-generating work, and with the amount of time spent on household chores. Existing levels of child labour in Ethiopia are thus demonstrably harmful.

KEYWORDS:

• Child labour
• children
• cognitive development
• human capital
• Ethiopia

JEL CLASSIFICATION:

• D10
• D90
• I31
• J22
• J24

Geolocation information

The study is based on household survey data collected in the following provinces of Ethiopia: Addis Ababa, Amhara, Oromia, SNNPR, and Tigray.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. This relationship is not a foregone conclusion: see for example, Bhalotra (2007).

2. Note that a regression of $A_{t+1}^{ihc}$ on $A_t^{ihc}$ and period-t child labour is just a re-parameterization of a regression of $\mathrm{\Delta }{A}_{t+1}^{ihc}$ on $A_t^{ihc}$ and period-t child labour: the only difference will be in the coefficient on $A_t^{ihc}.$

3. The PPVT has a strong and positive association with other measures of cognitive capacity (Campbell et al., 2001; Gray et al., 1999).

4. For example, one of the arithmetic questions was ‘9.81 + 7.62 = □.’ One of the problem-solving questions was ‘[Name] has 6 red boxes. Each box has 4 pencils inside. She also has 3 blue boxes. Each blue box has 2 pencils inside. How many pencils does [name] have altogether?’ The data interpretation questions were based on an itemized receipt from a clothes store.

5. Examples of the words used include ‘lamp’, ‘exhausted’, and ‘adjustable’.

6. The following text from page 3 of the manual describes the conditions under which the tests were to be administered: ‘The instruments will be administered to each child individually inside or near to the household’s home or at school. Whenever possible the tests should be administered in a place that is as private as possible (e.g., far from interferences coming from street sounds, siblings, TV or radio) and has a chair (or similar for the child to sit on), a table or flat surface for the child to write on and for the field worker to show the items, it should not be excessively windy and must have good lighting. The test should not start before 7 in the morning or after 5 in the evening to ensure good lighting and to avoid the participant from being too tired and not able to give his/her best performance … Make sure the child has taken his/her meal and has enough time to take the tests in one go.’

7. Income-generating work includes activities such as street vending, work on the farm, or serving in the family store. Domestic chores include activities such as washing, cooking, cleaning and caregiving; these definitions are consistent with classifications used in the United Nations Systems of National Accounts (United Nations [UN], 2009).

8. The indicator variable for urban communities does not vary over time, so descriptive statistics for this variable are given only for 2009. Note that the standard deviations for the community-level variables in Table 1 are not identical to the standard deviations of the variables across communities, because there is some (small) variation in the number of children across communities. However, the values of the two sets of standard deviations are very similar. For example, the standard deviation of the men’s wage across communities in 2009 is 22.1 (compared with a value of 22.1 in Table 1), and the standard deviation of the women’s wage across communities in 2009 is 16.1 (compared with a value of 15.9 in Table 1).

9. These are averages across different occupations. The survey reports daily agricultural wage rates and monthly non-agricultural wage rates, and the latter need to be scaled when computing the averages. Assuming that there are 4⅓ weeks in a month and five working days in a week, we divide the monthly salaries by 21⅔.

10. The 2014 community-level variables are omitted from this table.

11. One Birr was worth about ten US cents in 2009.

12. The coefficient is −0.332 and the standard error is 0.094. Recall that the standard deviation of chore hours is about two, and the lower 95^th percentile of the normal distribution is at −1.96. The relevant calculation is therefore ((1.96 × 0.094) – 0.332) × 2 ≈ −0.3.

13. It is also possible to add an indicator variable for the test being in the child’s native language. The coefficient on this variable is insignificantly different from zero, and its addition makes no difference to the results.

Additional information

Notes on contributors

Yonatan Dinku

Yonatan Dinku is a Research Fellow at the Australian National University’s Centre for Aboriginal Economic Policy Research. He has a Ph.D. in Economics from the University of Otago. His research interests include child development and labour market outcomes, Aboriginal and Torres Strait Islander policy, impact evaluation, social capital and neighbourhood effects, and the economics of health and well-being.

David Fielding

David Fielding is Professor of Development Economics at the University of Manchester’s Global Development Institute and Professor of Economics at the University of Otago. He has a D.Phil. in Economics from the University of Oxford. His recent work focuses on ethnic/cultural diversity and economic development. In addition, he researches the drivers of altruistic behavior and attitudes towards foreign aid.