In the past 40 years there have been substantial changes to how we transport ourselves from home to work, school, and social activities. Partially through demand of users and partly because of system-induced demand–the way we have built our cities and towns–we are driving more, walking less, and living at longer distances from our daily activities than we were before.
In addition, highway design standards have evolved to become the defacto design standard against which all other roads are judged. This has been due to the perceived safety and efficiency of a road designed to highway standards: usually divided multiple lanes in each direction, reduced crossings though substantial turning lanes at major intersections in combination with “free” rights (a single lane with a radius that allows turning without waiting for a signal change), specific grade changes and curve radii, among other features. Originally developed for the Eisenhower Interstate System for mainly rural areas, these design standards have been applied at some level to all means of road systems, often without context in mind. At the same time, we have seen an ever increasing level of public debt to pay for the new infrastructure. Together with the vast increase in disconnected suburban housing divisions that have a sole-reliance on the automobile commuting travel, these changes have been marked by an exponential increase in vehicle miles traveled. In perceived correlation, we have also seen obesity rates rise, and since the mid-1990s, rise at faster rates than before.
To begin to examine these changes, I have chosen one year in the recent past where all the necessary data is available, 2008. To further sort the data, I’ve created two groups: states that were listed as obese in that year and those that were not (where 0=not obese and 1=obese). Those states are shown in the Obesity Trends (figure 1) here as Alabama, Mississippi, Oklahoma, South Carolina, Tennessee, and West Virginia.
My research hypothesis is that there is a positive correlation between the levels of higher highway debt as a percent of total expenditures, and higher vehicle miles traveled per licensed driver, and higher obesity rates.
This hypothesis comes from two assumptions. First, there are limited functions of government and also limited resources. A state that spends more on highways often does so as an economic investment though highways tend to serve automobile-centric sprawl in the suburbs and tends to reduce walkability in urban areas. Could it be that the bonding of highways takes resources away from programs and policies that might encourage walkable neighborhoods and discourage driving? Secondly, the hypothesis assumes that sedentary lifestyles—driving is essentially sitting—contributes to weight gain. There is some research that validates both of these assumptions; it is beyond the scope of this study to examine these any more than anecdotally.
Interestingly, while the level of the population that is measured as overweight has seen little change in the past 50 years (figure 2), while the level of obesity—those with more than 30% body fat (BMI) has risen since the late 1970’s and this roughly correlates with the 150% percent change since then in vehicle miles traveled.
I’ve also looked an additional variable beyond automobile use and infrastructure. When a road is built to highway standards it often promotes the development of specific businesses designed to perform best in automobile-centric environments; parking lots easy to access, large signs, long expanses of street frontage. One of the most successful business models has been the convenience store, often with an attached gas station. Their growth rate has risen dramatically over the past forty years and the business model is largely made possible by roads designed to highway standards. Convenience stores usually sell heavily processed food products that nutritionists claim lead to obesity. I compare the amount of convenience stores per licensed driver, the amount of vehicle miles traveled per licensed driver, to the number of lane miles available (where one lane three miles long would be the same as three lanes one mile long). As a secondary hypothesis, I propose there is a positive correlation between these variables.
All but one of the variables included in this study comes from the federal government and primarily from the Federal Highway Administration (FHWA). Additionally, I have captured a few variables from the Census Bureau and the Center for Disease Control (CDC).
The FHWA and the CDC do not actively engage in data collection. Rather their role is to work with the states — their highway and health departments respectively — to establish collection guidelines and to combine and publish the data. The data from the FHWA comes from the annual Highway Statistics. Data from the CDC comes from their also annually published Behavioral Risk Factor Surveillance System, a measure of behavioral risk factors of the adult population.
To find the annual expenditures of each state, I have turned to the professional membership organization National Association of State Budget Officers’ (NASBO) annual State Expenditures Report. Its data is gathered directly from the budget officer of each state and all budget officers are members and report.
The Variables, in detail
1) Amount of Highway Debt. From Highway Statistics, sheet SB-1 “State Obligations for Highways – 2008, Obligations Issued or Assumed During Year”. This number describes the combined debt currently owed by the state for all years up to and including the study year. The table can be found at http://www.fhwa.dot.gov/policyinformation/statistics/2008/sb2.cfm.
2) Total Expenditures. From the National Association of State Budget Officer’s annual State Expenditures Report. The number includes nearly everything the state spends money on, no matter the source of funds. The report can be found at http://www.nasbo.org/sites/default/files/2009-State-Expenditure-Report.pdf.
3) Vehicle Miles Traveled (VMT). Vehicle miles traveled are the number of miles driven within the study area. Data included in this report comes from Highway Statistics, table VM-2, “Functional System Travel, Annual Vehicle – Miles” found at http://www.fhwa.dot.gov/policyinformation/statistics/2008/vm2.cfm.
4) Total Lane Miles. Defines the hypothetical length of all the lanes in the study area if they were connected in as single continuous lane. For example, a four lane one mile road would be four functional lane miles. The data comes from the table HM-60“Functional System Lane-Length – 2008 Lane-Miles”, found at http://www.fhwa.dot.gov/policyinformation/statistics/2008/hm60.cfm.
5) Obesity Rate. The Center for Disease Control classifies adults as obese is they have a Body Mass Index (BMI) of over 30. BMI is a calculated score created by an individual’s weight divided by the square of their height and multiplied by 703. Those classified as obese are statistically prone to higher health issues. The numbers used in this study are for the population of the state as published in the 2008 Behavioral Risk Factor Surveillance System.
6) Number of Licensed Drivers. This number is the population of all licensed drivers, therefore a reduced population number that includes only those with the ability to be the primary user of the infrastructure. This number is taken from Highway Statistics table DL-22 “Licensed Total Drivers” found at http://www.fhwa.dot.gov/policyinformation/statistics/2008/dl22.cfm.
7) Number of Convenience Stores. This number by counting the convenience stores in each state, using data from the American Fact Finder, based the North American Industry Classification System (NAICS) codes for convenience stores (445120) and convenience stores with gas station (447110).
To develop my dataset, I took these variables and combined them into a single spreadsheet. Further, I did calculations to break several down on a per licensed driver level before bringing it into Stata.
My research hypothesis is that there is a positive correlation between the levels of higher highway debt as a percent of total expenditures, and higher vehicle miles traveled (VMT) per licensed driver, and higher obesity rates.
Using multiple regression analysis, with the obesity rate (ObesRt) as the dependent variable, we can see that vehicle miles traveled strongly correlates to obesity but we would reject that highway debt has correlation.
If nobody drove, the data is reporting that the obesity rate would be about 19%. The data shows that there is a very strong correlation between the amount a population drives and the overall obesity rate. (figure 3)
Interestingly, this correlates well with the measured rise in national obesity levels describes in the first section of this study. However, how much highway debt a state has has little significance to its rate of obesity. In fact, this dataset shows a negative correlation; the more highway debt the lower rate of obesity in a state.
(figure 4) Additionally, to compensate for the potential “red flags” of outliers in both of these graphs, the same anayslis was performed with removing them but achieveing the same general outcomes. Further, each comparison has different outliers and comparing all states was important to the integrety of the dataset.
When looking at vehicle miles traveled broken down by states that are labeled as obese (group 1) against states that are labeled as not obese (group 0), the two groups show as be less than or equal confirming my hypothesis between VMT and obesity rates . (Stata command: ttest VMTperLD, unpaired unequal by(ObesState))
So what is it that does correlate? I took all the variables and created a correlation matrix (see attachment at the end of this paper). What it has found is that most of the correlations are market demand related (highlighted in yellow). For example, expenditures of a state correlates to the amount of highway debt; vehicle miles traveled per licensed driver correlates to the amount of lane miles a state has per licensed driver; the number of convenience stores correlates to the number of licensed drivers in the state. Among those correlations though three were less than obvious (highlighted in orange):
- Obesity rate to total lane miles
- Obesity rate to vehicle miles traveled per licensed driver
- Obesity rate to the number of convenience stores per licensed driver
I think there is a more complex relationship between obesity and road infrastructure—without regard to how its construction has been paid for, which is why the convenience store statistics in the dataset (though I now find them to be too limited) have been included. There are many factors at the individual-choice level that define the level of obesity in the overall population. For example, scientists often point to the intake of added sugars from soft drinks as one of many major correlations with obesity. (http://www.hsph.harvard.edu/nutritionsource/sugary-drinks-fact-sheet/). Connecting that, soft drinks are the number one selling item at convience stores, according to Convienence Store News. (http://www.examiner.com/article/the-cost-of-top-4-sellers-convenience-stores).
This dataset shows a strong correlation between the more road infrastructure and the amount of convenience stores, even when controlling for the amount of licensed drivers. (figure 5)
The more roads available, the more your population has the opportunity to drive and visit a convenience store to purchase soft drinks. This study used a combination of standalone convenience stores and those that also have a gas station, so some of this correlation is simple market response of providing fuel to drivers. However, gasoline sales have low profit margins (figure 6) and convenience store sales augment this.
The connection between roadway infrastructure and obesity illustrates a point: obesity rates are partially reflective of an infrastructure—physical and economic—that induces behavior. For example, by doing a regression analysis of just lane mile to number of convenience stores, the data shows a very high significance level.
Large-scale travel by automobile and consumption of processed food products—both of which would only occur when it is perceived as the “easiest” option and the availability of processed food also seen as “easiest” (or rather “convenient”?)—could lead to obesity. Again, this dataset shows a disconnect when trying to over-generalize. Correlation between lane miles per licensed drivers and obesity is weak. States that have the highest lane miles per licensed driver have no higher obesity rates than states with the lowest. (figure 7)
What is interesting though is that we can make connections between the amount of road infrastructure and convenience stores, and we can make connections between obesity rates and convenience stores, and between obesity rates and the amount the road infrastructure is used (via vehicle miles traveled). Connections are possible yes, but statistical correlations using the dataset given are inconclusive, and all further discussions all becomes anecdotal without further study.
One takeaway is that the tradeoffs of land use planning and government spending become apparent using the dataset. States that design their infrastructure and land use to require more use of the automobile—the sprawl model of growth—the less it spends per person. In essence, it is an inexpensive upfront cost to grow through sprawl for states. Two correlations show up in a kind of counterbalance: states that spend more per person have less convenience stores overall. When there are less convenience stores per licensed driver, the obesity rate falls in conjunction. Basically: States that spend less per person force their populations to use the automobile as their primary means of transportation (because they are not paying for the complex coordination and development of walkable neighborhoods with transit that requires a larger government role). The more automobile travel, the higher the obesity rate and the higher the obesity rate, the more convenience stores there are.
There is a bit of murkiness to the dataset and another takeaway is that my dataset is too small and/or random. To start to speak to questions about state spending would need more economic variables in the dataset—GDP, or a measure of inequality, or simply income levels set into groups.
My final takeaway was found by breaking the states into two groups: those labeled as obese and those not. The dataset seemed even more insufficient when doing analysis in this manner: states were not “obese” had significant correlations with regard to vehicle miles traveled (VMT) and their overall obesity rate, with an intercept of 20% if VMT were zero.
In the obese group of states, VMT had little significance to the obesity rate, however their intercept was 28% obese if VMT were zero.
Summarized: there is something else at play in states labeled as obese.
This brings the analysis back to the number of convenience stores. Running through multiple regression of the rates of obesity to vehicle miles traveled, lane miles, and the number of convenience stores (all by the number of licensed drivers in the state), the grouping of obese states and not obese states shows an interesting result: the number of convenience stores is the most significant finding to the obesity rate among the obese states.
Final summary correlation (but not causal) takeaway: states become overweight (heading to obesity) if they drive a lot, but only to a certain level. To become obese, they need to have a lot of convenience stores. To put another way and to put very broadly: inactivity will make you gain weight, but inactivity plus poor food choices will make you obese.
 From the executive summary of the FHWA report Highway Statistics, prepared by The Office of Highway Policy Information: “The vast majority of highway data are submitted by the States. Each State is analyzed for consistency against its own past years of data and also against other State and Federal data. Major issues are resolved with the help of the data provider.”
 From the 2008 BRFSS Overview document: “The Behavioral Risk Factor Surveillance System (BRFSS) is a collaborative project of the Centers for Disease Control and Prevention (CDC) and U.S. states and territories. The BRFSS, administered and supported by CDC’s Behavioral Surveillance Branch, is an ongoing data collection program designed to measure behavioral risk factors for the adult population (18 years of age or older) living in households. BRFSS field operations are managed by state health departments that follow guidelines provided by the CDC. These health departments participate in developing the survey instrument and conduct the interviews either in‑house or by using contractors. The data are transmitted to the CDC’s National Center for Chronic Disease Prevention and Health Promotion’s Behavioral Surveillance Branch for editing, processing, weighting, and analysis. An edited and weighted data file is provided to each participating health department for each year of data collection, and summary reports of state‑specific data are prepared by CDC.”
 From the appendix of the report: “The Fiscal Year 2009 State Expenditure Report reflects three years of data: actual fiscal year 2008, actual fiscal year 2009, and estimated fiscal year 2010.The text of this report focuses on actual fiscal year 2009 data, with a secondary focus on estimated fiscal 2010. This report documents state expenditures in six functional categories: elementary and secondary education, higher education, public assistance including Temporary Assistance for Needy Families and other cash assistance, Medicaid, corrections, and transportation. All other expenditures make up a seventh category. The report includes expenditures from four fund sources, including general funds, federal funds, other state funds, and bonds. States were asked to include spending from the American Recovery and Reinvestment Act of 2009 (ARRA) in the federal funds totals for the seven categories. Data for each category typically include employer contributions to current employees’ pensions and to employee health benefits for employees.”