79
Cancer and Medical Debt: Evidence from the Panel Study of Income Dynamics
From a theoretical standpoint, this study starts with Becker’s household production model (1981) to hypothesize that household members are altruistic and jointly maximize utility given the household’s preferences, income, assets, and budget set. Faced with an idiosyncratic health shock such as cancer, households produce health in a manner that is consistent with the Grossman model (1972) by using a combination of market and non-market inputs. Depending on insurance coverage and treatment cost, the household may face high out-of-pocket costs (actual price of receiving care). Based on the permanent income hypothesis, households that experience high out-of-pocket costs may have to borrow, hence accumulating medical debt, to smooth consumption. This paper considers three potential mechanisms through which cancer may have an effect on medical debt: financial resources (employment, income and wealth), insurance and out-of-pocket costs.
This study uses data from the 2011 Panel Study of Income Dynamics (PSID), the first year in which unsecured debt was divided into several categories such as credit card debt, student loans, medical debt, legal bills and debt from relatives. The sample is restricted to 5,918 households with heads who are between 18 and 65 years old. An indicator of household cancer status is created if the head of the household or his/her spouse reports being diagnosed with cancer by a health professional.
I use several OLS and tobit model specifications and find a positive and significant association between cancer and medical debt. For instance, findings show a (semi) elasticity of cancer on medical debt ranging from 0.79 to 0.42 (p<0.001) for tobit models (expected log of medical debt at the means, given that debt has not been censored). Subsample analyses, conditional on households with any secured and unsecured debt, show similar results (0.86-0.48, p<0.001). Additionally, inverse probability weighting propensity score models show semi-elasticities of cancer on medical debt ranging from 0.40 to 0.32 (p<0.001). These models account for systematic differences on observables between households reporting a diagnosis of cancer compared to those with no cancer. Although it is unlikely that medical debt would cause cancer, this study plans to use instrumental variables to address potential omitted variables bias. Analyses stratified by income quartiles and insurance types will be conducted as well to address heterogeneity in findings. Finally, counterfactual policy simulation of the ACA Medicaid expansion in reducing medical debt associated with cancer will be completed.
Findings will contribute to the growing literature on medical debt and the broader literature of health and socioeconomic status (SES) while accounting for issues of endogeneity. These findings will help educate and guide policymakers, patients, physicians, and public health professionals in considering effective treatment options for cancer.