Some studies on child labor have shown that greater land wealth leads to higher child labor, thereby casting doubt on the hypothesis that child labor is caused by poverty. This paper argues that the missing ingredient is an explicit modeling of the labor market. We develop a simple model which suggests an inverted-U relationship between land holdings and child labor. A unique data set from India that has child labor hours information confirms this hypothesis. It is shown that the turning point beyond which more land leads to a decline in child labor occurs at 3.6 acres of land per household, which is well below the observed maximum value of land-holding.