Background Information

The effect of international trade on the employment of child labor has stirred a heated debate.

Opponents argue that market integration increases labor demand, expands the earnings potential of children and increases child labor

In the extreme stance, some have argued for trade sanctions against countries with high levels of child labor as a way to curb child labor

Proponents of globalization argue that free trade leads to income gains that may reduce child labor if parents view child labor as a bad

Overall, rice price increases can explain almost half the decline in child labor in rural Vietnam between 1993 and 1998 (403)

Theory

Trade theory predicts that an increase in the price of an exported good could either increase or decrease child labor depending on the magnitudes of the substitution and income effects. (401)

Rice is the dominant commodity in the Vietnamese economy and the staple food so these price changes should, in principle, affect child labor

Vietnam is a small net exporter of rice so the relaxation of trade barriers leads to an increase in the relative price of rice

We first assume that households both produce and consume rice - the product that will be exported after liberalization

Consequently, the response of child labor to rice price increases varies with household exposure to rice prices through consumption and production

There are two internal markets – agriculture (rice) and non-agriculture

Children can work in both markets

Child labor is a bad in the parents’ utility functions.

One channel through which trade theory predicts that product market integration will affect domestic labor markets is through changes in relative product prices

If agriculture is child labor intensive than relative rice price increases will raise the demand for child labor and lead to higher child wages.

If child labor is bad in parental preferences then increases in the relative price of rice will make consumption of child leisure cheaper and lead to a shift away from child labor

Another channel is through the income effect of rice

44% of food expenditure is on rice and 70% of households produce rice

Households that are net producers of rice should experience a net increase in household income as rice prices increase.

Moreover, given the large scope of rice cultivation, the price increase may also increase day laborer wages and increase incomes for families that are net consumers of rice as well

These increases in income may lower child labor if it is a bad in parental preferences

Alternatively, child labor may decrease if previously liquidity constraints resulted in higher levels of child labor than was efficient.

Consequently, the effect of an increase in relative rice prices is ambiguous and depends on the relative magnitude of these channels.

The goal of this paper is first to examine the overall relationship between child labor and rice prices and second to identify the income effect of rice prices on child labor through net production

Methodology (DATA)

The basic idea is to compare changes in child labor across communities that experience different changes in the relative price of rice over time

Description of Vietnam

The price of rice in Vietnam changed from 1993 to 1998 due to the integration of the national and international rice markets

The real price of rice increased, on average, across Vietnam by almost 30%

Starting in 1989, Vietnam instituted stringent export controls on rice in order to protect the domestic food supply.

Vietnam also had limits on the domestic flow of rice between regions

The government gradually liberalized its export quota starting in 1992, until it was nonbinding in 1998

In early 1997, the internal barriers to rice trade across regions were also lifted

The internal barriers create the exogenous variation in rice prices across regions that allow the regression analysis

The data comes from two panel surveys in 1992-1993 and 1997-1998

All prices are in 1998 Dong

The survey documents whether each child worked, in the past week,

Outside the house for pay

Works in agriculture for the household

Works in household business

Participates in household production activities such as collecting water and wood, building and maintaining the house, making or repairing tools, etc

They create an aggregation of these measures, which they call child labor

Namely, a child engages in child labor if it works for 7 or more hours per week in household production or if it works in agriculture, wage employment, or a family business

In addition to participation, the authors use data on total hours worked in all these categories

[I could include descriptions of the summary statistic graphs]

Methodology (Empirical Analysis)

They use a linear probability model

Y is the indicator for whether the child j living in community i at time t engages in child labor

RP is the natural logarithm of the real price of a kilogram of ordinary rice in community i at time t

We use community level rice prices rather than household-level rice prices because community-level; prices are more apt to be exogenous to the household and thus not affected by the household’s child labor decision

Second

The coefficient on rice prices beta one is the change in the probability that a child works associated with a one-percent increase in the price of rice in the child’s community.

They control for the gender and age differences in child labor using a third-order polynomial and all their interaction terms

They capture seasonal variations in rice pries and labor by including season indicators

They also control for harvesting time by including an indicator that is one if the interview took place during the harvest

Similarly, they control for the planting season

Third

Economy-wide differences in the probability that a child works are captured by a year indicator T that is one if the survey year is 1997/1998 (the omitted year is 1992/1993)

Fourth

Communities differ in many characteristics, including the availability of school, returns to education, labor market conditions, land and resources, etc

Lambda is a community-level fixed effect

Results – Table 2 – the Overall Relationship between Rice Prices and Child Labor

Column I

A positive and significant association between increases in rice prices and declines in child labor

A 30% increase in price leads to a 30% decline in child labor

Column II

Control for unobserved, region specific, time-varying shocks by including an interaction between each region indicator and a year indicator

Regions differ in types and scopes of reforms, and may differ in returns to schooling if labor mobility is limited across regions

Control for differences in child labor associated with accessibility by interacting a community’s accessibility with year indicators

More accessible communities might experience greater rice price changes because they are better integrated into regional and international markets. Likewise, children in more accessible communities may have better access to schools or unemployment opportunities.

The estimate is within 95% confidence interval of their first estimate

Column III

In this specification, the authors focus on other sources of time-varying factors that could bias their results

Infrastructure improvements because public infrastructure can dramatically improve living standards

Pest problems

Natural Disasters

High-yield seeds

Province-level child labor changes

The estimate is within 95% confidence interval of their first estimate

Results – Income Effects of Rice Prices on Child Labor Through Net Production

We predict that, all else equal, higher rice prices should be associated with declines in child labor in households that are large net producers of rice and increases in child labor in households that are large net consumers of rice

M is household net production in 1993

Rp*M is the interaction of rice prices with household net production in 1993

Mu is the household fixed effect

All the other notation is standard

Note that unlike in equation 1 it is important to control for household fixed effects in equation 2 because we are identifying income effects by looking at within community variation in household exposure to rice price changes. Unobserved household characteristics may be a concern

This identification requires that the 1993 household net production is exogenous to factors associated with changes in child labor between 1993 and 1998 (conditional of household fixed effects and the other controls such as season, age, gender patterns, etc). It seems unlikely that the household would choose its net production in 1993 in anticipation of changes in child labor five years hence.

[This approach limits their sample to children who were captured in both panel surveys – they check for this bias and attrition bias – it all seems ok]

Table 3

Column 1

The coefficient on rice price is -0.28 similar to the coefficients in table 1.

This suggests that the specification is reasonable

The coefficient on the interaction of rice prices with household net production is negative and significant.

This is exactly what the theory predicts

Column 2

This includes the interaction between rice prices and net production

The coefficient on rice prices grows smaller but remains negative

This could reflect that parents substitute towards the consumption of child leisure as rice becomes relatively more expensive

It could also reflect the income effects of rice prices on child labor associated with increased wages for day laborers

Alternatively, it could reflect a substitution effect away from child labor in production

Although agriculture is child labor intensive relative to non-farm business activities, the data suggest that agriculture is actually less child-labor intensive than home production. Thus the overall demand for child labor could decline after rice price increases

Columns 3 and 4

They estimate equation 2 separately for work in agriculture and in home production

Everything else equal, higher rice prices are associated with less child labor in agriculture and household production

Column 5

This column measures participation in market work – i.e. work excluding home production

Positive coefficient on prices negative coefficient on the interaction term

These results illustrate that even in sectors directly experiencing growth as a result of relative price movements (or product market integration) income effects can be large enough to affect a decline in child labor

Robustness Analysis

They run some robustness checks which all pan out

One of these tests includes a 2SLS test in order to use household land assignments as an instrument for net productivity.

These variables are really correlated

The find that their results do not change

Conclusions

Summary of Results

Increases in the relative price of rice result in declines in child labor

A 30% decline in the relative price of rice results in on average a 9% point decrease in child labor

Rice price increases can account for 45% of the decline in child labor experienced in rural Vietnam between 1993 and 1998

We see evidence that households which are net producers of rice experience larger declines in child labor when rice prices increase

Implications:

The increased earnings opportunities available to children working in export-oriented sectors do not necessarily lead to more child labor

Employing trade sanctions against poor countries in order to force them to eradicate child labor may instigate more child labor instead

The impact on child labor of punitive trade sanctions against a country’s exports depends on the distribution of the resources used in the production of the exported good.

Rice is so widespread that sanctions would affect everybody

The sanctions might work if they only hit a select minority

The sign of the effect of international trade on local markets is of great importance

Integration lowers the price of import competing goods

Integration may have a difference effect on child labor in households associated with the production of import-competing products

However, as in the present case, most child (and adult) labor in poor, relatively unskilled labor abundant economies occurs in either non-traded or export-oriented sectors.

1