Estimation is a means of simplification. It is a practical way to quickly determine the order of magnitude of a particular item, and is useful when taking an exact measure would be too complex, time-consuming or costly.
Of course, the problem with estimates is that they are by definition WRONG. And the “accuracy” of an estimate is judged by how far the estimate is from the actual, correct answer. Unfortunately, we tend to be really bad at estimating in certain circumstances, as illustrated by problems with estimates related to obesity in the U.S.
Statistical House of Cards: Obesity Level Estimates
If estimates are by definition incorrect, then estimates based on other estimates have the potential to be wildly incorrect, as errors are compounded. Such is the case with estimates of obesity levels in the U.S., which currently place obesity at about 30% of the population. The commonly quoted CDC population obesity estimate is based on the Body Mass Index, or BMI – a calculation that is itself an estimate of individual body composition. Using BMI calculations that are derived from sample survey data, the government extrapolates population level estimates of obesity. Bad news — a recently released analysis of BMI data in comparison to actual body fat analyses shows that BMI understates obesity in women by 50%, while overstating obesity for men by 25%. Based on this correction, population level obesity is potentially upwards of 50%-60% rather than 30%.
Guessing the Unknown: Obesity Cost Estimates
When trying to ascertain the size of something that is unknown or not calculable, “estimate” is really just another word for “guess”. While there can be more educated or more sophisticated guesses, this does not change the fact that you will not know how good your estimate is (since you don’t have any way to get the correct answer) until someone comes along with a more compelling estimate that shows your initial estimate to be lacking. This is exactly what happened when Cornell and Lehigh economists recently conducted a new type of analysis to better calculate the medical costs of obesity, using family histories to leverage the genetic tendencies toward obesity that exist within families. What they found – average annual health care costs per person related to obesity are double the previous estimates — $2,800 per person versus $1,400. This indicates a population level cost of obesity equal to 17% of all health care expenditures in the U.S.
Relying On Truthfulness: Exercise Participation Estimates
One of the most widely cited government sources of information on lifestyle is the Bureau of Labor Statistics. The BLS conducts surveys about how people spend their time, including the amount of time spent in exercise and sports activities. In the most recent 2010 BLS data, you’ll see that only 18.5% of Americans are active on a daily basis, with that group averaging 1.66 hours per day of exercise or sports — meaning that over 80% of the population does not exercise regularly. This already seems pretty bad, however an NIH study suggests that the BLS data likely greatly overstates activity levels due to the inaccuracy of self-reported information. The NIH study, which used measurement devices worn by participants to record actual activity levels, found that less than 5% of adults were meeting the recommendation of 30 minutes per day of exercise (dramatically lower than their self-reported exercise levels). This means that BLS estimates of exercise time could be too high by a factor of 4.
So What Do We Really Know?
For one, we know that our estimates of obesity, related health care costs and sedentary lifestyle are still wrong, even with the modifications noted above. And these modified estimates likely continue to understate our problems. We appear to have placed too much faith in an imperfect process of estimation for understanding the issue of obesity in the U.S., and thereby have failed to recognize an even more dire health situation than is popularly believed.
The takeaway — projections of 50% obesity levels in the U.S. by 2020 may in fact have already been achieved, and we just haven’t realized or come to grips with it yet.
Paul
Paul Amoruso, CEO
WeightNot
Does this lack of information and incorrect data benefit insurance companies?
Hi Randolph — Great question. I think that this is a bad situation for insurance companies insofar as they cannot accurately predict their business costs, but since they pass through their costs to consumers, it will ultimately be the consumer who pays the price for this problem (directly in premiums, through reduced wages that are redirected toward premiums by employers, and taxes that support health care). Insurer models of lifetime health care costs underestimate the steepness of the trend line in obesity, and therefore underestimate the increase in the liabilities that they face. This means that they are not as proactive in wellness and preventive programs, lifestyle interventions, improved biometric monitoring or other steps that might mitigate their long term costs and improve population health.