LDI Charles Leighton Memorial Lecture with Mark Chassin, MD 5_4_12
LDI Research Seminar 1_28_11- Brian Elbel, PhD, MPH
1. Obesity, Calorie Labeling and
Behavioral Economics
Brian Elbel, PhD, MPH
Assistant Professor of Medicine and Health Policy
NYU School of Medicine
NYU Wagner School of Public Service
University of Pennsylvania
January 28th, 2011
2. Outline of Presentation
• Overview of my work/approach
• Influence of calorie labeling: NYC
• Behavioral economics and obesity: Field experiment
• Conclusion/Next steps
3. My Work/Approach
• Broadly: How individuals make choices that
influence health and health care
– Consumers/patients choice
– Physician choice
• Behavioral Economics: Econ with psych
• Role and influence of pubic policy
• As well as broader influences on choice
4. My Work/Approach
• Methods
– Econometric analysis of secondary data
– Primary data collection, natural experiments
– Field Experiments
– Randomized trials
– “Lab-based” experiments
• Choice in three areas:
– Clinical
– Policy
– Obesity
5. Research Areas
• Clinical/Choice:
– Changing default for physician prescribing of
medication for high cholesterol (NIH/NIA/RC4)
• Policy/Choice
– Low-income use of quality information, behavioral
economics to improve use (RWJF)
– Influence of number of options and Medicare
supplemental choice (AHRQ/NSF)
– Role of cognitive ability in Medicare choices
6. Research Areas
• Obesity/Choice
– NYU Nutrition and Obesity Policy Research and
Evaluation Network (CDC)
– Evaluation of NYC initiative to bring supermarkets
to “food deserts” via tax credits (RWJF; Aetna
Foundation)
– “Bellevue Bodega”: Behavioral economics in a
“store” set up in a public hospital (CTSI)
– Influence of calorie labeling (NIH/NHLBI/R01,
RWJF)
– Behavioral economics and fast food choice
8. Obesity Trends* Among U.S. Adults
BRFSS, 1990, 1998, 2008
(*BMI 30, or about 30 lbs. overweight for 5’4” person)
1990 1998
2008
No Data <10% 10–14% 15–19% 20–24% 25–29% ≥30%
9. Attempts to Address Obesity
• Formerly, largely “education-based”
• Recent: attention to the Food Environment
• Environment considered by many to be “toxic”
– Little fresh food available
– Fast food ubiquitous
– Marketing of unhealthy foods – children
– Pricing of healthy versus unhealthy foods
• Altering the environment where consumers
make choices: public policy
• Consumer not always “rational” in food choice
10. Calorie Labeling
• First large-scale public policy to address obesity,
focusing on the food environment
• Combines information/education and
environmental-based approaches
Basic premise: Tell consumers before they
order the number of calories in each food item
11. Logic Behind Calorie Labeling
• Many people eat at fast food restaurants often:
Average 2x week
• Fast food eating associated with obesity, heart
disease
– Correlation data
– Some more rigorous studies
• Individuals currently poorly judge the caloric
content of food **
• Simply a “right” to the information
• Non-demand effect: restaurants changing menus
12. However…
• Food choice is incredibly multifaceted
• Nutrition is but one factor in consideration
• Everyone wants to start a diet or eat
“better”…starting tomorrow: time preferences
• Self control
• Labeling has a hard climb
13. Calorie Labeling in NYC
• First locality to institute calorie labeling
• Restaurants w/ more than 15 locations
nationally must:
– Post the caloric content of all “regular” menu items
– Post using the same font/format as the price
• Covers 10% of all restaurants in NYC
• 2,400 establishments
• Labeling began in July 2008
20. Menu Labeling Goes National
• Labeling was included in the federal health
reform bill
• FDA must propose national regulation by
March 2011.
• Includes all restaurants with 20 or more
locations nationally
• Federal law will preempt state and local laws
21. Our Study
• How does labeling influence the fast food
choices of individuals in low-income areas?
• Difference-in-Difference Design
– Natural experiment
– Examine calories purchased before and after labeling
• NYC
• Comparison city: Newark, NJ
22. Where we collected data
• Low-income areas of NYC matched w/ Newark
• NYC: Harlem, South Bronx, Central Brooklyn
• Why Newark?
– Matched our target communities in NYC
– Similar “urban” feel
• 2 restaurants in NYC for 1 in Newark
• Wendy’s, McDonald, Burger King, Kentucky Fried
Chicken
23. Timing of data collection
• 2-week period before labeling, July 2008
• Tuesday, Wednesday or Thursday
• Lunch (12:30 – 3) and dinner hours (4:30 – 7)
• Post-data collected 1-month after labeling
– Same restaurant, time, day of week
– Enough time for individuals to have exposure to labels
– Not enough time where:
• Seasons change
• Kids back in school
• Become desensitized to labels
• Other things could influence obesity
24. How data were collected
• RAs outside of fast food restaurants
• Ask subjects to bring back receipt, answer
survey in exchange for $2
• RAs confirmed what the subject ate on the
receipt, performed a short survey
• From receipt and survey calculated number of
calories purchased by each person
• Estimate a 50-60% response rate
26. Sample
Full Sample New York City Newark
Pre-Labeling Post-Labeling Pre-Labeling Post-Labeling
% N % N % N % N % N
Male 38% 426 38% 131 36% 161 40% 68 41% 66
Race/Ethnicity
Black 66% 760 58% 213 64% 286 74% 130 81% 131
Latino/a 20% 230 25% 91 22% 97 15% 26 10% 16
White/Other 14% 166 17% 64 15% 67 11% 20 9% 15
Mean S.D. Mean S.D. Mean S.D. Mean S.D. Mean S.D.
Age 38.2 14.1 38.3 14.7 37.9 14.7 39.7 12.6 37.4 12.6
Education
High School of Less 47% 205 48% 72
Not Not
Some College or AA 20% 86 16% 24
Collected Collected
BA or Above 33% 142 36% 54
27. 28% of those who saw labels
indicated the labels
influenced their choice
No change in calories purchased
825 before labeling v. 846 after labeling
28. No differences between those who noticed/indicated
responded to labels, demographic groups
29. Results from Children / Adolescents
• Focus on NYC Data (n=266)
• Adolescents 13-17 without caregiver were asked
the same questions as adults
• Parents were asked questions about what they
ordered for their children
• Separate results by:
– Adolescents 13+, by themselves (30%)
– Children with their caregiver (70%)
30. 16% of adolescents who saw
labels indicated they
influenced their choice
31. Children/Adolescents:
Calories Purchased
NYC
Sample (n=266)
Calories Purchased
Mean (sd)
n Pre Post
With Caretaker 183 610 (327) 595 (319)
Without Caregiver 83 730 (341) 755 (327)
32. Knowledge of Recommended
Calorie Allowance
• Clearly, labeling less effective if consumers don’t
know how many calories they “should” eat
• Federal law: requires that restaurants post a
statement on recommended allowance—FDA
now considering what this should look like
• We asked consumers how many calories they
thought adults should eat to maintain a healthy
weight
34. Change in Estimates of
Number of Calories Purchased?
• Prior work focused on hypothetical scenarios
• We asked for estimate of food and drink calories
• 23% indicated they didn’t know
NYC Newark
Pre Post Pre Post
Actual 810 842 829 843
Estimate 619 723 615 667
Difference 191 118 214 117
% Different 31% 16% 26% 21%
• Diff-in-Diff not statistically significant
35. Change in Estimates of
Number of Calories Purchased?
• Assume “correct” if estimates are within 100 calories of
actual
Food and Drink Pre Post
% Overestimating 24% 25%
Correct Estimate 15% 24% p<.01
% Underestimating 61% 51%
Food Only
% Overestimating 21% 23%
Correct Estimate 16% 21% ns
% Underestimating 63% 55%
Drink Only
% Overestimating 20% 23%
Correct Estimate 37% 45% p<.01
% Underestimating 43% 32%
Only those who know RDA
% Overestimating 34% 32%
Correct Estimate 23% 28% p<.05
% Underestimating 43% 40%
36. Who Noticed and
Responded to Labels?
• Just over half of individuals see labels, consistent
over time
• Of those who see, only about a quarter use the
information
• Who are these people?
– People who need it the most?
– People who are already ‘nutrition-inclined’?
37. Who Noticed and
Responded to Labels?
Of Those Who Saw:
% % Who Saw Labels % Who Were Influenced
Sex
Female 64% 55% 31%
Male 36% 53% 19% *
Race/Ethnicity
White/Other 16% 61% 45%
Black 63% 53% 24% **
Hispanic/Latino 21% 53% 22% **
NYC, Post-Labeling Sample, n=448; *p<.10; **p<.05
38. O f Tho se W ho Saw :
% % W ho Saw Labels % W ho Were Influenced
Which is More Im portant
N utritio n 17% 58% 54%
T aste 51% 51% 19% **
B oth 32% 59% 23% **
Lim it Food to Maintain W eight?
N ever 29% 51% 8%
Seldom/Sometimes 36% 54% 20% *
O ften 13% 61% 39% **
A lw ays 22% 53% 54% **
Education
HS or less 47% 47% 22%
Some College 20% 64% 34%
B A or h igher 33% 59% ** 27%
Tim es/W eek Eat Fast Food
0-2 30% 57% 32%
3-5 25% 66% 23% *
6-8 17% 49% 29%
9 or mo re 28% 45% ** 21%
Know RD A: b/w 1,500 and 2,500
No 58% 53% 25%
Yes 42% 61% * 31%
Adjusted M odels; ** < .05; * < .10 n=448 n=448 n=242
39. Limitations
• Non-random sample
• Limited geographic regions/sample
• Can’t observe:
– Amount of food eaten
– Food eaten later in the day
– Whether consumers are avoiding fast food
altogether
• Substituting to healthier foods?
• Substituting to less healthy/equally unhealthy foods?
• Single time period, one month post-labeling
• Not examining changes in menus
40. Consistent With Other Findings
• “Starbucks Study”: Bollinger et al.
• “Taco Time Study”: Finkelstein et al.
41. Next Steps: This Data
• Prices paid, change in prices paid
• Geographic/Community characteristics: location
of restaurant, location of residence
• Change in “type” of orders
• Racial/Ethnic Differences
42. Next Steps: Additional Collection
• Expansion to examine calorie labeling in two
cities, two comparison cities (R01, NHLBI)
• Recently collected data in Philadelphia,
Baltimore
• Receipt collection, larger cross-section of the
population
• Telephone Survey: capture substitution
behaviors, change in sit down restaurants,
greater data on food choice, can labeling be
made “better”
43. Conclusions
• Marked increase in noticing/responding to
calorie information in response to labeling
• Did not generally translate into decreases in
calories purchased for our study population in
our study period
44. Conclusions
• Those most likely to be influenced by labels:
potentially more ‘nutrition’ inclined; more
obese?
• Estimates of calories purchased improved; still
not perfect
• Many don’t know recommended calories; more
info better?
45. Other methods to influence food
choice at the “point of purchase”?
The Relative Influence of Behavioral Economic
Nudges and Calorie Labels in Altering Fast
Food Choice
Coauthors: Janet Schwartz (Duke), Jason Riis
(Harvard) and Dan Ariely (Duke)
46. Our Study
• Will customers “downsize” fast food meals?
• Self-control: Consumers will often not stop eating when
full
• Can we prime or initiate self-control at the beginning of
the meal, using a simple “nudge” or reminder
• Field Experiment, actual fast food restaurant
– Asian-Style chain restaurant
– College campus, near medical center
• All aspects of experiment performed by restaurant staff
47. Framework
• Study 1: Do customers downsize their order
with nudge, with or without financial incentive?
• Study 2: Relative influence calorie labeling v.
nudge
• Study 3: Are customers actually eating less?
48. Ordering a meal at
the fast food restaurant
Step 1
Order meal size
Step 2 Step 3 Step 4
Order side dish Order entrée(s) Pay
Fried
rice
Steamed
rice
Chicken
dish 1
Beef
dish 1
Pork
dish 1 …
$
Chow Mixed Veggie Etc. Etc.
mein veggies dish 1 …
48
49. How the nudge worked
Step 1
Order meal size
Step 2 Step 3 Step 4
Order side dish Order entrée(s) Pay
Fried
rice
Steamed
rice
Chicken
dish 1
Beef
dish 1
Pork
dish 1 …
$
Chow Mixed Veggie Etc. Etc.
mein veggies dish 1 …
“Would you like to cut more than 200 calories from your meal by
taking a half portion of your side dish?”
49
50. Study 1: Summer 2009
• N = 283
– Data gathered from receipt
• Side dish, entrée(s), downsize
– Gender (estimated by RA)
• 75% male
– Median age (estimated by RA)
• 26
– Summer, largely non-student population
51. Study 1
Sun Mon Tue Wed Thu Fri Sat
BASELINE BASELINE DOWNSIZE DOWNSIZE
no discount no discount
1% 35%
BASELINE BASELINE DOWNSIZE DOWNSIZE
25¢ discount 25¢ discount
4% 32%
•No difference in discount v not
•Taking “Nudge” resulted in 200+ calorie saving
52. Calories per Customer (N=283)
1200
Baseline Downsize
1010
1000 912* * p<.05
800
600 530 516
480
396*
400
200
0
Side Dish Entrée Total Cals
53. Summary
• Study 1
– People do not spontaneously ask for less food
– Customers will downsize meals if prompted (~90
calories averaging across all in condition)
• With and without a discount
– No compensation with higher calorie entrees
• Study 2
– Is downsizing more or less effective than calorie
labeling at reducing consumption? How do they
interact?
54. Study 2
• Fall 2009
• College students back in town
• N = 994
– Receipts
• Side dish, entrée, downsize
– Gender (estimated)
• 74% male
– Median age (estimated)
• 20
56. Study 2 Design
Sun Mon Tue Wed Thu Fri Sat
CALORIE LABELS GO UP (AND STAY UP)
56
57. Study 2 Design
Sun Mon Tue Wed Thu Fri Sat
baseline baseline baseline
NUDGE NUDGE NUDGE NUDGE
baseline baseline baseline
CALORIE LABELS GO UP (AND STAY UP)
baseline baseline baseline
NUDGE NUDGE NUDGE NUDGE
baseline baseline baseline
57
58. Study 2 Results
Sun Mon Tue Wed Thu % Accepting
Fri Sat
baseline baseline baseline 0%
NUDGE NUDGE NUDGE NUDGE 21%
baseline baseline baseline 1%
CALORIE LABELS GO UP (AND STAY UP)
baseline baseline baseline 0%
NUDGE NUDGE NUDGE NUDGE 14%
baseline baseline baseline 0%
Still accepting nudge, uptake of nudge is LOWER when labels are present
58
59. Calories per Customer (n=994)
1200
Baseline Downsize
1020 1033 1016
1000 945*
* p<.05
800
600 536 555 560
484 508
478 456*
437*
400
•Calorie labeling did not appear to alter food choice for this population
• 200 dish calories and total calories decreased with nudge in pre-labeling period
Side
•Still influence of nudge for side dish when labels are present, smaller
0
Side Dish Entrée Total Cals Side Dish Entrée Total Cals
Pre-label Post-label
60. Study 3
• Calorie labels still up
• Same procedure (baseline v. nudge)
– All customers were asked to take part in a survey for
$10, asked to bring any leftovers
– Leftovers were weighed
• Those in the baseline condition started with more
food….
• But, both conditions left the same amount of food
on the plate—almost nothing
• Not just ordering less, but eating less
61. Strengths / Weaknesses
• Strength:
– Field experiment
– Objective measure of purchasing
– Relative influence of two interventions
• Weaknesses:
– External validity
– Selected sample
– Mechanism
62. Conclusions
• Nudging did have an influence on food choice
and consumption; calorie labeling did not
• Labeling did not improve the influence of
“nudge”, potentially made it worse
• Policy: Greater attention to policy approaches
aside from simply information
– Defaults
– Other Nudges?
• Non-policy: model for employers, restaurants?
63. Overall Conclusions
• Labeling may be justified/helpful for other
reasons, but likely won’t alter food choice in a
large scale way by itself
• Behavioral economic approaches potentially
effective; public policy?
• Working with employers, others to implement
changes
• Multiple approaches, done at same time/place,
potentially more effective in influencing
obesity/food choice: labeling AND nudge?
64. Thanks
Collaborators Funders
• Rogan Kersh • NIH/NHLBI: R01HL095935
• Beth Dixon
• RWJF Healthy Eating
• Tori Brescoll
Research
• Tod Mijanovich
• Yale Rudd Center
• Gbenga Ogedegbe
• Beth Weitzman
Staff
• Dan Ariely
• Courtney Abrams
• Janet Schwartz
• Joyce Gyamfi
• Jason Riis
67. Noticed / Responded to Calorie Information
New York City post – labeling sample
Noticed Calorie Information?
N=450
YES No
N=242 (54%) N=208 (46%)
Info Influenced Choice?
YES No
N= 67 (28%) N=175 (72%)
Reported purchasing
fewer calories?
YES N=59 (88%)
68. Calorie Labeling in NYC
• The History
– First passed in 2006
– Restaurant industry sued
– Judge ruled against NYC DOHMH on “technical”
grounds
– Re-wrote the law in 2008, sued again
– Ruling went all the way to 2nd Circuit Court
– Ruled in favor of NYC DOHMH
69. What do we currently know about
labeling and its effectiveness
• When labels are not on the menu board:
– Only 0.1% to 10% of people “see” calorie information
• NYC DOHMH Study, AJPH:
– Looked at voluntary labeling at Subway
– Those who saw labels ate 50 fewer calories
• Experimental Work: Harnack et al., 2009
– Hotel Ballroom, mimicked fast food restaurant
– No reduction in calories ordered
– Males: increased calories ordered when labels present
• Experimental Work: Roberto, Brownell et al. 2010
– Randomized to: no labels, labels, labels w/ “2000” prompt
– Labeling decreased intake, with prompt