New Evidence on Insurance and Health Behavior

Public health officials have long sought to improve health outcomes by influencing bad behaviors, like smoking or lack of exercise. Such behaviors are very clearly detrimental to health, so much so that it seems like no one would perform them—but people inevitably still do. Health policymakers are interested in improving these behaviors, not only to improve the health of the population, but also because they often lead to very harmful illnesses that increase the costs of and add strain to the healthcare system.

To understand how to improve, we need to know how having health insurance impacts negative behaviors like smoking. One thing we know for certain is that health insurance leads to more use of healthcare, which can have two possible and conflicting results on health behaviors. On the positive side, health insurance leads to more routine doctor’s visits, and evidence shows when doctors interact with patients, they can influence their health behavior for the better. On the other hand, the security of knowing one has insurance can also lead to worsened behavior, a phenomenon called “moral hazard”. Moral hazard can happen when insurance reduces the negative cost of bad behaviors (like how much money you have to spend to be treated), which leads to more participation in them. There is a significant amount of linking health insurance to negative behaviors such as smoking, so this moral hazard theory is well supported. These opposing effects make it difficult to isolate the exact impact of either effects, or to say exactly how large each of these effects is.

If health behaviors are in fact improved by using healthcare, then it stands to reason that the longer a person has insurance, the larger this effect should become. Dr. Aparna Soni tried to evaluate this effect by comparing health behaviors in states that expanded Medicaid in 2014 to those that never expanded Medicaid. The research focused on data from the fifth year after expansion, which could then be compared with studies from earlier years that had fewer data.

The data for this analysis comes from a CDC survey known as the Behavioral Risk Factor Surveillance System. The BRFSS is a nationally representative telephone survey which interviews more than 400 thousand adults each year. The researchers used four years of this survey pre-expansion and five years post-expansion, giving them a time range of 2010-2018. Rather than using the entire data set, they only focused on a specific subset of people who were demographically eligible for Medicaid: low-income, childless adults who were neither elderly nor disabled. This gives the researchers data on a wide range of health metrics for people who either had or would have had Medicaid depending on the state they lived in, including health insurance, access to a doctor, preventative care utilization and health behaviors.

They then tried to estimate the impact of Medicaid expansion on several of these health metrics by comparing the expansion states to non-expansion states both before and after the expansion happened. Assuming that these states are similar before Medicaid expansion, it is possible to measure the effect of expansion by comparing how different the states are after expansion to how different they were before expansion. This is known as the difference-in-differences model. The model also controlled for a wide range of demographic characteristics as well as the unemployment rate, which is important because these characteristics can have a big impact on health behaviors and outcomes.

The researchers found solid evidence that Medicaid expansion states had lower rates of negative health behaviors. They found strong evidence for a decrease in heavy drinking, and weak evidence for decrease in smoking and increase in exercise. These effects aren’t very big, but that is not surprising since Medicaid expansion only impacts a small portion of the population and much of this data comes from surveys which are not very precise. When the researchers only looked at the most recent year, they found very little significance, probably in part because people are so different and varied from each other. Using the same methods, the researchers also found that Medicaid expansion states had higher rates of insurance and healthcare utilization. While this is not surprising, it serves as a good litmus check for the researchers, since it would be necessary for their other findings to make any sense.

This research comes with several key limitations. Firstly, it uses survey data, which can be inaccurate because of who randomly gets surveyed, or could be biased by people who are not willing to be honest about their personal behavior. The bigger issue, however, is whether the non-expansion states serve as a reasonable comparison for the expansion states. Remember, difference-in-differences models rely on the two groups being similar, but a state’s decision to expand Medicaid is not random. For example, more liberal state governments might be more likely to expand Medicaid, and they may also be more likely to place more taxes on smoking. Differences like this would cause the model to be an inaccurate comparison, making the results of this study less meaningful than they appear.

This paper also raises an important question of what to do when a piece of research comes to a different conclusion from the existing body of evidence. It is possible that the results of this experiment were overstated. Yet it is also possible that this paper found evidence of long-term effects that went undetected in the short timeframes of previous research. This implies that having health insurance for a long period of time could lead to more doctor’s visits, which may indeed cause patients to eventually improve their health behaviors, but only over time.

This research is clearly a piece of evidence in favor of the theory that access to health insurance increases preventative care which leads to some positive changes in health behaviors. However, this is far from definitive given the significant contrary evidence from prior research. These findings do not mean that the moral hazard of health insurance doesn’t exist, but rather that the positive impact of medical coverage might be stronger—having insurance could still lead to increased negative behavior like smoking, but spending more time with a doctor could also convince people to stop or cut back. This could have many exciting implications not only for the long-term health of the American populace, but also for lowering America’s ballooning healthcare costs in the future.