The figures from the United States on increases in obesity and BMI (body mass index) are startling. The percentage of obese Americans has increased from around 14 percent in the late 1970s to 28 percent by the late 1990s. Over 60 percent of Americans have a BMI (weight in kilograms divided by height in meters squared) above that considered most healthy. Although BMI increased slowly throughout the twentieth century, most of the increases were towards the most healthy BMI, while recent changes have been away from optimal BMI. In addition, obesity rates (a BMI over 30) exploded, particularly in the 1980s and early 1990s. Between 1978 and 1991, for example, obesity rates increased by 55 percent (Chou, Grossman and Saffer 2002).
Economists have only recently begun to research the causes of increasing obesity and BMI. Bibliographic references to three recent papers discussed here are given below. In evaluating this research three things should be born in mind. First, the individual papers mentioned below are all working papers not yet published. Second, the economic literature as a whole is new. Third, aside from these papers, there is very little solid empirical work in any field on the causes of increasing obesity.
Given these constraints we should not expect, and do not find, that any of these papers is definitive. Instead, what these papers do is introduce some key questions, suggest some hypotheses and offer some evidence in support of these hypotheses. All of the papers succeed in this task. It will take some time, however, to sort out competing hypotheses and arrive at consensus conclusions.
One of the difficulties with a new field of research is obtaining good data, and that is an issue here. There are several surveys of BMI in the U.S., but these surveys are not typically linked to other information such as wages, income, type of employment, food prices, food consumption, prevalence of fast food restaurants, smoking or other factors that might causally explain increases in BMI and obesity. Linking health data to other data is not always easy, but the authors use some clever techniques. Chou, Grossman and Saffer (2002) and Lakdawalla and Philipson (2002), for example, use a dataset that has variables they are interested in like income and wages but that only has self-reported weight and height. Self-reports are likely to be inaccurate. To correct for this they make use of a different dataset that doesnt have the income and other information that they want but does have both self-reported and measured height and weight. The information in the second dataset is used to build a statistical model that estimates the relationship between reported height and weight and true height and weight allowing for different relationships for men, women, the overweight, underweight, etc. The results from this model are then used to predict true height and weight in the first dataset. This point is just illustrative of some of the tricky issues that are involved in drawing together data from a wide sets of sources that were produced for different purposes.
Although there are very few papers explaining changes in BMI and obesity we do know quite a bit (from epidemiologists and others) about the cross-sectional or demographic correlates of obesity. We know, for example, that on average, blacks and Hispanics have higher BMI than whites, women have higher BMI than men, and the less educated have higher BMI than the more educated. These cross-sectional or demographic facts cannot explain, however, why BMI and obesity rates have increased in the U.S. because the fraction of the population that is black, Hispanic and female has not changed much, and education rates have been rising not falling.
The Cutler, Glaeser and Shapiro (2003) paper is the weakest of the three but contains a number of important points and interesting hypotheses. A good point to bear in mind in all this research is that, according to calculations by CGS, even small increases in daily caloric intake can explain large weight gains over a sustained period. The data on increasing obesity over the past two decades, for example, could be explained by an increase of 100-150 calories per dayequivalent to a can of Pepsi or three Oreo cookies. Aggregate data on the food supply indicate that calories per person have increased by more than enough so that in principle this could explain increasing obesity rates.
CGS focus specifically on the implications of lower-cost mass-prepared food. Technology has now made it possible to produce much more food away from the home with consequent savings in preparation costs for the family. The reduction in the full-price of food, that is in the money plus time price, has caused people to eat more and also more frequently and in greater variety.
An implication of the CGS theory is that increases in BMI should be highest in people and nations that have experienced the largest price declines for mass-produced food. CGS show that obesity has increased the most in married females. Since married females have seen the greatest time-savings in food production this is consistent with their theory, although CGS cannot explain why the families of married women have not similarly increased in weight. It is not clear whether the weight gain in married females is due primarily to increased consumption of food or to reduced expenditure of energy in home production.
CGS also show that countries with higher levels of food regulation have lower levels of obesity with their explanation being that regulation prevents technology that lowers the cost of food from being put into place. The analysis, however, is done using very aggregate data, variables that only loosely relate to the theory and without many controls for other factors. Everything in the CGS paper needs to be taken with a grain of salt, but future papers will surely develop and test many of the hypotheses that they introduced.
Increased food consumption is not necessarily the explanation of higher obesity rates. Decreased energy expenditure could also explain higher obesity rates even holding food consumption constant. This appears to have occurred during most of the twentieth century (1900-1960). Decreased energy expenditure has occurred as home and work life have both become more sedentary. The corollary of the CGS results, mentioned above, that a small increase in food consumption alone could explain increasing BMI, is that a small decrease in energy expenditure alone could explain an increase in BMI. Lakdawalla and Philipson (2002) merge data on job strenuousness with data on height and weight and show that, holding other factors constant, a worker who spends a career in a sedentary job has a BMI 3.3 units higher than someone in a more active job. Changes of this magnitude can explain the entire increase in weight over the last century.
Lakdawalla and Philipson create an economic model that determines income, hours worked, food consumption and weight. The model leads to some important insights. If work becomes more sedentary, for example, then energy expenditure declines, but so does food consumption (active workers demand more food)what then is the effect on weight? LP are able to prove in their model that the net effect is always positive, weight will increase.
As noted above, LP find that people with sedentary jobs weigh more but this could be due to a selection effectperhaps people who weigh more choose sedentary jobs. Using a time-series analysis, however, they find that the more years one spends in a sedentary job the greater the weight gain, which is inconsistent with a pure selection effect. LP also perform some other tests that also suggest the direction of causality is from sedentary jobs to weight increases.
Lakdawalla and Philipson also examine the effects of changes in food prices. Using their model they estimate that 40% of the weight gain from 1976 to 1994 is due to decreases in food prices resulting in higher food consumption, and 60% is due to decreases in energy expenditure.
The best of the recent economics papers on obesity, in my view, is by Chou, Grossman and Saffer (2002). CGS have assembled detailed data on food prices, both at home and in fast-food and full-service restaurants, the price of cigarettes, as well as income, wages, hours of work, and other similar factors.
CGS find two novel findings. First, that the increase in the number of per-capita restaurants is positively associated with the increase in obesity since 1978. Also, they find that stability in this variable is associated with the stability of obesity between 1960 and 1978an important finding that suggests a non-spurious association. Note, however, that CGS do not say that fast-food restaurants cause obesity. Instead, their interpretation is that the growth in fast-food and other restaurants reflects an increased demand for these restaurants stemming from an increase in the value of non-market time. This finding is consistent with other aspects of the model because CGS also find that greater hours worked, holding income constant, increases obesity. Thus CGS find that a large part of the increase in obesity is associated with higher values of time that encourage people to eat at fast-food and other restaurants.
The second novel result in CGS is that obesity has increased in places where taxes and other interventions have raised the price of cigarettes. Its well known that individuals that quit smoking tend to increase in weight, but it is somewhat surprising to find that this effect is also clear in the aggregate data. The war on cigarettes has made the war on fat more difficult to fightanother story of unintended consequences.
As noted above, the economic literature on obesity is new (and recall that this is essentially the only literature that tries to identify the causes of increasing obesity). In some places the literature is inconsistent, and much needs to be done to clarify results. To illustrate some of the difficulties, consider some of the channels through which income can affect BMI. At low levels of income, higher income increases BMI and improves health, but as income per se increases even further, the demand for health may come to dominate, and higher income reduces (or no longer increases BMI). Yet higher income may be associated with longer hours, which suggests more energy expenditure, unless these hours come at the expense of more active leisure. Longer hours and higher wages also suggest a higher value of time, thus increasing the demand for fast food. Sorting out these effects will require some time and better data.
Something of a puzzle is why exactly is it that fast-food is higher-calorie food? Would it not be possible to have cheap, fast, healthy food? Or is it that cheap, fast, healthy food is not demanded? Humans evolved during times when fats and sugar were rare and highly valuable for survival. Its not surprising, therefore, that evolution would have provided us with a demand for these foods. The difficulty occurs when fats and sugars are no longer rare or valuablecan our rational minds overcome the tastes that evolution has given us? Cutler, Glaeser and Shapiro deal with these issues briefly, but more work would be valuable.
Another issue that has yet to be examined is the form that calories take. Carbohydrates versus fat. Much skepticism now exists over the governments promotion of the food pyramid and its war on fat (Taubes 2002). Some have suggested that the war on fat is itself a cause of greater obesity. Only a few clinical trials have tested this idea, and, as of yet, no time-series evidence is available.
In summary, while not definite, the new economic literature is an important advance in our understanding of the causes of increasing obesity in the United States.
Chou, S.-Y., M. Grossman, and H. Saffer. 2002. An Economic Analysis of Adult Obesity: Results from the Behavioral Risk Factor Surveillance System. NBER Working Paper Series 9247.
Cutler, D. M., E. L. Glaeser, and J. M. Shapiro. 2003. Why Have Americans Become More Obese?, Harvard Department of Economics, Working Paper.
Lakdawalla, D., and T. Philipson. 2002. The Growth of Obesity and Technological Change: A Theoretical and Empirical Analysis. NBER Working Paper Series 8946.
Taubes, Gary. 2002. What if Its All Been a Big Fat Lie?, New York Times Magazine (July 7, 2002).
© 2003 The Independent Institute. Permission is granted to reprint or broadcast this article if credit is given to the authors and to the Independent Institute. Nothing in this article should be interpreted as necessarily reflecting the views of the Independent Institute or as an attempt to aid or hinder the passage of any legislation.
|Alexander Tabarrok is Senior Fellow at the Independent Institute, Assistant Editor of The Independent Review, and Associate Professor of Economics at George Mason University. He received his Ph.D. in economics from George Mason University, and he has taught at the University of Virginia and Ball State University. Dr. Tabarrok is the editor of the Independent Institute books, Entrepreneurial Economics (Oxford University Press), The Voluntary City, and Changing the Guard.|
The fear of litigation reduces innovation, drive physicians and manufacturers out of lawsuit-prone specialties, and increase manufacturing and consumer costs. In the courts, data from thousands of cases all over the country demonstrate that tort system awards are driven by political factors such as judicial elections, jury compositions, and the location of courts themselves.