The relationship between socioeconomic status (SES) and health is one of the most robust and well documented findings in social science. However, the reasons for the relationship are less clear since plausible causal mechanisms run in both directions. Anne Case et al.(2002) look at children in order to find the “origins of the gradient,” since the health of children may be assumed to have relatively little impact on their own socioeconomic status. They show that the well-known cross-sectional relationship between SES and health exists in childhood and is more pronounced among older than among younger children. Since poor health in childhood is likely to affect adult well-being directly through its effects on health, and indirectly through its effects on other forms of human capital accumulation, it is important to try to address the causes of SES-related gradients in health status among children. However, in a cross section it is not possible to distinguish between two different possible mechanisms underlying a steepening gradient. On the one hand, it is possible that low-SES children are less able to respond to a given health shock, so that the negative effects of health shocks persist and accumulate over time. This model would imply that low-SES children are in need of better medical care, or better access to care, so that they can respond to health shocks in the same way as higher-SES children. On the other hand, it is possible that low-SES children respond to health shocks in a way that is similar to high-SES children, but are just subject to more shocks. This model implies that SES-related gradients can be reduced by addressing the reasons why low-SES children are more likely to be subject to health shocks. For example, low-income children are more likely to have accidents, and to suffer from nutritionrelated disorders such as diabetes. This distinction is important for policy because it implies that it may be productive to spend social resources on measures designed to reduce the arrival rate of health shocks as well as continuing to improve children’s access to palliative medical care.

We examine these hypotheses using a panel of data on Canadian children from the National Survey of Children and Youth (NLSCY). We confirm that the results of Case et al.(2002) hold using a sample of Canadian children, despite the existence of universal health insurance coverage for doctor and hospital services in Canada. We find that the gradient steepens in cross section, and that this result is robust to controls for cohort effects. However, we find little evidence that the long-term effects of health shocks on future health are different for high-SES and low-SES children, even though in the short run, low-SES children suffer greater health losses than high-SES children after the arrival of a health shock. Instead, we provide evidence which suggests that the cross-sectional relationship between health, family income (or maternal education), and age arises primarily because low-income children are more likely to be subject to health shocks. The rest of the paper proceeds as follows: Section I provides some background regarding the relationship between SES and health. Sections II and III provide an overview of the data