Date post: | 23-Jan-2015 |
Category: |
Health & Medicine |
Upload: | leonard-davis-institute-of-health-economics |
View: | 127 times |
Download: | 4 times |
Understanding Overeatingand Obesity
By Christopher J. Ruhm
University of North Carolina at Greensboro, NBER and IZA
March 2010
Obesity: Major & Growing Health Problem
• Important Cause of Premature Death
• Linked to Many Health Problems
• Higher Medical/Economic Costs
• Rapidly Increasing
Contrasting Policy Perspectives
• Public Health – interventions to reduce weight
• Economics – individual choices optimal without market failures
Are Economic Models Missing Something Important?
Triune Brain
Key Elements of Brain Structure
New capabilities added to rather than replacing existing structures
• Affective (limbic) system rapidly responds to stimuli
• Deliberative system accounts for long-term consequences
• Systems operate in parallel; yield differences in perception/memory
• Emotional feelings separate from rational appraisals of them
Key Elements of Brain (continued)
Affective system often most powerful
• Affective system came first• Responds faster than deliberative system
• More neural connections from limbic system to neocortex than in opposite direction
Eating decisions involve interaction of:
• Higher order rational calculations &• Affective processes based on emotion, chemical
responses & feelings
Dual Decision Model of Eating and Weight
Utility-maximizing deliberative system
• Rational eating (Philipson & Posner, 2003)• Food influences utility directly & through body weight• “Ideal” weight (W0): weight chosen if costless to achieve• Most policy interventions reduce utility
Affective System (Loewenstein & O’Donoghue, 2004)
• Responds to external stimuli• Does not consider consequences of eating on weight
Both systems influence outcomes
Results in deviations from utility-maximizing food intake & weight
Deliberative System: Utility-Maximization
• maxf,c U(W(f),f,c) subject to c + pf = I
W = weight p = food prices
f = food intake I = income
c = other consumption
• First-Order Conditions: UWWf + Uf = pUc
– equalize MU of last $ on food & other consumption– direct utility of food + indirect utility through weight
Deliberative System (continued)
• Optimal food consumption
fd = argmaxf F U(W(f),f,z)
F = feasible food choicesz = preference shifters determining utility from calories
• Food innovations change characteristics (z) to increase utility from consumption
(taste, mouth-feel, texture, complexity, etc.)
Affective System
• Optimal food consumption
fm = argmaxf F M(f,s,z)
M = motivational functions = stimuli (do not affect deliberative system)
• Generally fm > fd
– Affective system does not consider future weight gain
– Food is engineered to appeal to affective system
Dual System Interactions
• Costly for deliberative system to override affective system
Loss Function: L = L(|f- fm|,R) = L(D,R)
D = |f - fm| = deviation from affective system optimum
R = willpower
Losses increase with D, decrease in R
Dual System Interactions (cont.)
• Food Consumption
fc = argmaxf F U(W(f),f,z) – L(D,R)
• FOC: UWWf + Uf + LD = pUc (if fd < fm)
• Therefore: fd < fc < fm
• Food consumption & weight exceed utility-maximizing levels
Dual System Characteristics
• Higher prices reduce eating– Deliberative system still operates– Time prices more powerful than money prices
• Explanation for Quasi-hyperbolic discounting
• Related to models where individuals voluntarily limit choices available– Planner-Doer (Sheffrin & Thaler, 1988)
– Long-run patient self, short-run impulsive selves (Fudenberg & Levine, 2006)
– Temptation Utility (Gul & Pesendorfer
– “Hot/Cool” States (Bernheim & Rangel, 2004)
Strategic behaviors to manage overeating
• Voluntarily increase time/money price of foods triggering affective system (don’t bring ice cream home)
• Limit exposure to stimuli triggering affect system (avoid parties/restaurants where overeating is likely)
• Strategies to increase willpower – stress management– formal diets (to reduce cognitive load)– Weight watchers/weight loss tournaments (public
announcement of weight affects loss function)
• Medical procedures (bariatric surgery) to reduce fm
Food Engineering
• Strategic manipulation of food to raise consumption– Increase utility of food (welfare-increasing)– Promote affective system responses (increase overeating)
• Components of food engineering (Kessler, 2009)– Balance of (high levels of) added fats, sugars & salt– Combinations of flavor, aroma, oral/visual texture, after-taste– Heightened complexity & multi-sensory effects
– Heavy use of processed products, chemical flavorings
– “Edible food-like substances” rather than food itself
– Environmental cues (for food outside home)
• Promoted by new technologies & agricultural policies
Predictions Common to Both Models
• Weight & food prices inversely related
• High BMI individuals more likely to report being above “ideal weight” (W0)
– But utility-maximizers are not above Wd
Dual Decision Model Predictions
• Eating mistakes common & positively related to BMI– High BMI: strong affective system responses, weak willpower
• Weight loss attempts frequent, often unsuccessful– positively correlated with BMI & increasing over time
• Weight increases over time, even with stable prices, due to advances in food engineering
• Largest weight increases occur in upper tail of BMI distribution
• High BMI individuals eat more engineered foods; pattern strengthens with time
Data
• National Health & Nutrition Examination Surveys
– NHES (1960-62); NHANES 2 (1976-80); NHANES 3 (1988-94); NHANES 99 (1999-2006)
– Measured height & weight (not self-reports)– 24-hour food diaries– Weight history & weight loss attempts (NHANES 99 best)– Inconsistent questions on weight loss attempts, so use
• Behavioral Risk Factor Surveillance System– 1991/94, 2000/03– Self-reported height & weight– Common questions on weight loss attempts
Data (continued)
• 25-60 year olds
• Demographics– Age (quadratic)– Race/ethnicity (black, Hispanic, other)– Education (<high school grad, some college, college grad)– Marital Status (married, widowed, separated/divorced)– Tobacco use (ever smoked, current smoker)– Additional covariates experimented with
•Sampling weights incorporated
Body Mass Index & Weight Classes
• Widely used, easy to measure
• Not a direct measure of body fat (or distribution)
• BMI Categories– Underweight <18.5– Healthy weight 18.5 – 24.9– Overweight 25.0 – 29.9– Mild Obesity 30.0 – 34.9– Severe Obesity ≥ 35.0
2
( )
( )
Weight kgBMI
Height m
Fig 1. Trends in Body Mass Index
0.0
2.0
4.0
6.0
8.1
Pro
bab
ility
15 20 25 30 35 40 45BMI: 25-60 Year Olds
1960-1962 1976-19801988-1994 1999-2006
Fig 2. Food Prices, Eating Time & Restaurants/Capita
80
90
10
011
012
013
0N
orm
aliz
ed V
alu
e (1
967
= 1
00
)
1967 1972 1977 1982 1987 1992 1997 2002 2007Year
Food Price Eating TimeRestaurants/Capita
Specific Hypothesis 1
Disproportionate weight growth (over time) in upper-tail of BMI distribution, even during periods of stable food prices
Table 1. Quantile Regression Estimates of Trends in BMI
PercentileBaseline
BMI(1976-80)
Change in BMI Since Baseline1988-94 1999-2006
(a) (b) (a) (b)
Males10 21.2 0.60 0.33 1.00 0.80
25 23.1 0.70 0.55 1.50 1.34
50 25.4 0.60 0.66 2.10 1.97
75 27.9 1.20 1.08 3.10 2.90
90 30.7 2.00 1.90 4.50 4.30
95 33.3 2.40 2.34 5.50 5.41
Additional Controls No Yes No Yes
Table 1. (continued)
PercentileBaseline
BMI(1976-80)
Change in BMI Since Baseline1988-94 1999-2006
(a) (b) (a) (b)
Females
10 19.5 0.40 0.37 1.20 0.84
25 21.3 0.50 0.61 1.80 1.62
50 23.9 1.20 1.35 3.10 2.92
75 28.0 2.30 2.32 4.90 4.75
90 33.1 2.80 3.04 5.50 5.74
95 36.5 2.90 2.60 5.90 5.51
Additional Controls No Yes No Yes
Specific Hypothesis 2
Weight mistakes are common, particularly at high BMI, & have increased over time
– Heavy individuals think they weigh too much
– Desired weight has not increased over time
– But actual weight & dieting have become more common
Table 2. Weight Perceptions & Dieting Behavior, 1999-2006
Full Sample
Under-weight
Healthy Weight
Over-weight
Mild Obesity
Severe Obesity
MalesWeight Self-Perception
Underweight 6.0 73.6 17.0 1.5 0.5 0.0
About Right 42.6 26.4 74.5 45.8 14.0 4.6
Overweight 51.4 0.0 8.5 52.7 85.5 95.4
Preferred Weight
More 8.9 83.4 24.9 3.0 0.8 0.0
About Same 33.4 16.6 60.9 34.9 10.3 3.3
Less 57.6 0.0 14.2 62.1 88.9 96.7
Attempted Weight Loss
25.1 0.0 7.5 27.1 37.3 41.1
Population Share
100.0 0.9 27.2 41.0 20.3 10.7
Table 2. (continued)
Full Sample
Under-weight
Healthy Weight
Over-weight
Mild Obesity
Severe Obesity
FemalesWeight Self-Perception
Underweight 2.5 58.0 3.3 0.2 0.2 0.0
About Right 28.0 40.6 60.2 17.7 4.0 1.4
Overweight 69.5 1.4 36.5 82.0 95.8 98.6
Preferred Weight
More 2.6 51.1 3.7 0.4 0.1 0.0
About Same 19.7 47.4 41.9 11.0 3.1 1.6
Less 77.7 1.4 54.4 88.6 96.9 98.4
Attempted Weight Loss
43.3 0.2 30.9 48.9 55.3 53.0
Population Share
100.0 2.2 35.7 24.4 17.8 17.9
Fig 3A. Desired and Attempted Weight Loss, 1999-2006
0.2
.4.6
.81
Pro
bab
ility
20 22 24 26 28 30 32 34 36 38 40BMI
Weigh Less, females Diet, femalesWeigh Less, males Diet, males
Fig 3B. Desired Weight Loss by Survey Year
0.2
.4.6
.81
Pro
bab
ility
20 22 24 26 28 30 32 34 36 38 40BMI
1988-1994, females 1999-2006, females1988-1994, males 1999-2006, males
Fig 3C. Attempted Weight Loss, BRFSS
0.2
.4.6
.8P
rob
ab
ility
20 22 24 26 28 30 32 34 36BMI
1991 & 1994, females 2000 & 2003, females1991 & 1994, males 2000 & 2003, males
Specific Hypothesis 3
Severe obesity results from uncontrolled eating & unintended weight gains
– Heavy individuals more likely to have gained weight in last year
– & particularly likely to have large unintended weight gains
Table 3. Weight Gain During Last Year, 1999-2006Weight Gain
Full Sample
Healthy Weight
Over-weight
Mild Obesity
Severe Obesity
MalesAny Gain 25.7 19.7 26.2 24.4 28.1
≥5 lbs 23.1 16.8 23.3 23.5 27.5
≥10 lbs 17.3 10.8 16.8 18.7 25.6
≥10 lbs & Unintended
10.7 4.0 11.5 13.4 21.9
FemalesAny Gain 28.2 28.2 29.3 25.2 26.3
≥5 lbs 24.6 21.5 27.1 23.4 24.7
≥10 lbs 18.6 14.3 20.4 18.2 22.4
≥10 lbs & Unintended
12.7 8.2 15.4 14.1 18.9
Table 4. Econometric Estimates of Weight Gain in Last Year, 1999-2006
Weight ClassUnconditional
Any Gain ≥5 lbs ≥10 lbs≥10 lbs &
Unintended
Males
Severe Obesity .090 .115 .159 .182
Mild Obesity .053 .074 .089 .099
Overweight .069 .071 .069 .080
Baseline .211 .178 .112 .040
Females
Severe Obesity -.033 .020 .073 .108
Mild Obesity -.040 .012 .035 .062
Overweight .004 .051 .059 .076
Baseline .297 .225 .149 .077
Table 4. (continued)
Weight ClassConditional on Some Weight Gain
≥5 lbs ≥10 lbs ≥10 lbs & Unintended
Males
Severe Obesity .134 .380 .579
Mild Obesity .118 .243 .357
Overweight .038 .106 .248
Baseline .842 .529 .179
Females
Severe Obesity .174 .346 .443
Mild Obesity .170 .229 .301
Overweight .160 .192 .262
Baseline .751 .496 .255
Specific Hypothesis 4
Obese individuals disproportionately eat engineered foods
– Particularly true in recent years as food engineering has advanced
– Focus on fat & sodium intake
Table 5. Econometric Estimates of Food Consumption, 1999-2006
Weight Class
% Calories From Sodium/Day(mg)
Calories/Day # Eating Occasions/
DayFatSaturated
Fat (a) (b)
MalesSevere Obesity
3.42 1.42 292.6 23.0 114.5 -0.20
Mild Obesity 2.26 0.97 334.5 68.5 41.9 -0.18
Overweight 0.46 0.30 -35.6 -5.0 21.2 0.01
Baseline 31.5 10.0 4026 2695 4.76
FemalesSevere Obesity
1.78 0.85 322.3 81.5 116.2 -0.26
Mild Obesity 1.73 0.66 159.4 22.1 47.0 -0.05
Overweight 0.84 0.35 89.1 -16.3 18.9 -0.07
Baseline 32.3 10.5 2834 1873 4.81
FIG 4A. Fat Consumption by BMI
01
23
4N
orm
aliz
ed
% C
alo
ries
fro
m F
at
20 22 24 26 28 30 32 34 36 38 40BMI
1976-1980, females 1999-2006, females1976-1980, males 1999-2006, males
FIG 4B. Sodium Consumption by BMI
-400
-200
02
00
400
600
Norm
aliz
ed
So
diu
m In
take
(m
g)
20 22 24 26 28 30 32 34 36 38 40BMI
1976-1980, females 1999-2006, females1976-1980, males 1999-2006, males
Conclusion
• Brain Structure suggests food consumption & weight often deviate from utility-maximizing levels
– Eating “mistakes” common & difficult to correct– Weight loss attempts increase with BMI– Food engineering increases overeating & obesity
• Empirical evidence more consistent with dual decision-making than utility-maximization
Expanded/different role for policy?
• Time costs more effective than money prices– Availability constraints not fat taxes
• Information less useful– Not used by affective system
• Default options to exploit systematic decision errors (nudging)– Put healthy menu choices first
• Do advances in food engineering increase scope for intervention?
Future Research Topics
• Role of physical activity to offset overeating
• More detailed analysis of food engineering– Consumption of specific foods & food products
• Heterogeneity in affective system effects– Across individuals & the lifecycle (e.g. children)
• Incorpore cue-conditioned responses (addiction)
• Dual decision-making & other types of consumption