The determinants of diabetes using state-level panel data.

Author:Assane, Djeto

    There is general consensus among healthcare professionals and economists that diabetes is a major health problem that is primarily spurred by two key factors: obesity and physical inactivity (e.g., Kleinfield, 2006; Sullivan et al., 2005). There is also considerable evidence suggesting that these key variables are closely linked to socioeconomic conditions and state-specific factors (Lynch et al., 1997; Williams, 1990). For example, at the individual level, the prevalence of diabetes in the United States is observed in adverse health practices that bear relationships with less education and high incidence of poverty, whereas at the state level the incidence of diabetes tends to be associated with population composition, state laws and health policies that discourage high-risk behavior along with state mandates that require insurance providers to cover diabetes care. What is less clear, as evidenced by sparse empirical research on the issue, is the extent to which this relationship holds at various levels of obesity and physical inactivity. For example, using the Centers for Disease Control (CDC)'s Behavioral Risk Factor Surveillance System (BRFSS) state-level data, we find an unexpected inverted U-shaped relationship between diabetes and physical inactivity indicating that the incidence of diabetes among states increases with physical inactivity up to a point and then declines at higher levels.

    The objective of this paper is to explore and analyze empirically this issue motivated by the state-level data. The paper assesses the relative importance of obesity and physical inactivity as key factors leading to diabetes in a framework in which socioeconomic conditions and state-specific factors play a significant role in the dynamic of the disease. Our empirical model uses state-level data for the 50 U.S. states and District of Columbia from 1994 to 2002 primarily from the CDC's BRFSS to analyze the determinants and prevalence of diabetes in the U.S., controlling for observable and unobservable state-specific effects. Our findings are striking and easy to summarize. We find that diabetes cases increased at an annual rate of 4% between 1994 and 2002 and are spurred by obesity and physical inactivity as expected. However, we find a surprisingly strong inverted U-shaped correlation between physical inactivity and diabetes cases which may reveal the interplay of socioeconomic and state-specific factors on the dynamics of the disease. Overall, the findings prove robust to alternative static or dynamic specification.


    Diabetes is a growing disease among all ages, races, and genders, and has become a major health problem throughout the world. In the United States over 18 million individuals are diagnosed with diabetes, which is the fifth leading cause of death by disease. While this fact indicates the dangers of diabetes, diabetics are also at higher risk for other diseases and serious medical conditions, such as heart disease, blindness, kidney failure, extremity amputations, and other chronic conditions.

    Although diabetes is usually described as a single disease, there are three main types of diabetes. First, Type-1 diabetes, or juvenile-onset diabetes, occurs when the body destroys pancreatic cells, which are the only cells that regulate insulin, which in turn regulates glucose levels. Next, Type2 diabetes, or adult onset diabetes, occurs when cells in the body do not use insulin correctly, leading to a greater need for insulin, and, eventually, causing the pancreas to lose the insulin-producing capability. Finally, the third type of diabetes, which is the least prevalent, is known as gestational diabetes, which is a glucose intolerance that often occurs in pregnant women. Of the three distinct forms of the disease, Type-2 diabetes is by far the most prevalent, accounting for 90% to 95% of all cases (CDC, 2003). Hence, in this paper, by diabetes we refer to Type-2 diabetes for ease of discussion.

    The rising costs of diabetes resulting from unhealthy diets and inactive lifestyles are a cause for particular concern in the United States. The American Diabetes Association estimated the total cost of diabetes exceeded $174 billion in 2008. A caveat worth emphasizing is that given the strong association between physical inactivity and obesity, it may be difficult to specifically disentangle the costs of diabetes attributable to physical inactivity and those pertaining to obesity.


    We analyze the incidence and prevalence of diabetes along two dimensions of variables. The first group of variables is described as key variables that include the two significant causes of diabetes: obesity and physical inactivity. The second group of variables control for socioeconomic and state-specific factors. To evaluate the effects of these groups of variables on the prevalence of the disease, we consider the following reduced-form specification for state i (i = 1,..., 51) in period t (t = 1994, ..., 2002).

    Log [] = [X'.sub.i,t] [beta] + [W'.sub.i,t] [gamma] + [Z'.sub.i,t] [zeta] + [[PHI].sub.i] + [[epsilon].sub.i,t] (1)

    where the dependent variable, Log D, is the natural log of the prevalence of diabetes, X is the vector of key variables, W and Z are the vector of socioeconomic and state-specific variables, respectively, the coefficients [beta], [gamma], and [zeta] are vectors of parameters to be estimated, [epsilon] is the error term and [PHI] allows for state-specific effects, which include factors such as public health policies or health-related infrastructures that are not captured by X, W and Z but can vary across states. Our empirical results suggest that these effects are significant, and the Hausman test concurs.

    The dependent variable describes cases of diagnosed diabetes per 100 of the adult population. The explanatory variable OBESITY is the percentage of the state adult population that is considered obese. Specifically, OBESITY is defined as those with a Body Mass Index (BMI) that is 30.0 or greater. BMI is derived by dividing weight in kilograms by height in meters squared. The high incidence of obesity has been associated with changes in lifestyle choices resulting in energy intake that exceeds energy expenditure. Energy intake imbalance is particularly evident in unhealthy diets and inactive lifestyles. The variable INACTIVITY is the percentage of the state population that reports no leisure-time physical activity during the past month. The incidence of mortality due to chronic diseases, including diabetes, is significantly high among individuals with inactive lifestyles. Hence, physical inactivity and obesity are key factors in the incidence and prevalence of diabetes.

    Turning to observable state control variables, there are indications that diabetes can be fostered by socioeconomic factors such as income, age, and race. A number of studies have documented a strong inverse correlation between socioeconomic status and diabetes-related disease or mortality (Kleinfield, 2006; Lynch et al., 1997; Williams, 1990). For example, residents of low income communities typically receive less education and healthcare services, leading to inactive and less healthy lifestyles. Turning to race...

To continue reading