The paper aims at identifying households who have their children work more because their access is constrained on the land and the labor markets. Data from the 2005 "Enquête auprès des Ménages" (EPM) collected in Madagascar provide information on the amount of hours worked by each household member along with measures for market imperfections. A simple theoretical model highlights that land should not influence the number of hours of child work when the household can fully access the land or the labor markets. When the access is limited in both markets, land may impact the amount of child work whereas the external wage should not. Using a switching regression model with unknown sample separation to classify households in the two regimes (constrained or not), this paper shows that not belonging to the largest ethnic group at the local level significantly decreases access to the market. The same result holds for religion, thereby highlighting the importance of the informal market. Abstract The paper aims at identifying households who have their children work more because their