Content uploaded by George Loewenstein
Author content
All content in this area was uploaded by George Loewenstein on May 17, 2015
Content may be subject to copyright.
164
American Economic Journal: Applied Economics 2 (April 2010): 164–178
http://www.aeaweb.org/articles.php?doi
=
10.1257/app.2.2.164
During the past 20 years, the United States has seen a dramatic increase in obe-
sity. In 1991, only four states had obesity prevalence rates as high as 15 percent,
and not a single state had a rate above 20 percent. By 2005, only ve states reported
rates below 20 percent, with 17 states registering rates equal to or above 25 per-
cent (H. M. Blanck et al. 2006). Economic analyses of this trend have implicated
a variety of potential causal factors (e.g., Shin-Yi Chou, Michael Grossman, and
Henry Saffer 2004; Eric A. Finkelstein, Christopher J. Ruhm, and Katherine M.
Kosa 2005), but much of the rise in obesity can be attributed to an increase in caloric
intake as opposed to a change in energy expenditure (David M. Cutler, Edward L.
Glaeser, and Jesse M. Shapiro 2003).
In this paper, we compare the efcacy of two different types of interventions
intended to change the food intake of fast food restaurant patrons. The rst inter-
vention provides calorie information to consumers and is intended to mimic recent
legislation requiring chain restaurants to display calorie information prominently on
their menus, as recommended by the Center for Science in the Public Interest (2003).
The other intervention is based on insights from the eld of behavioral economics
* Wisdom: Depar tment of Social and Decision Sciences, Carnegie Mellon University, 208 Porter Hall,
Pittsburgh, PA 15213 (email: jwisdom@cmu.edu); Downs: Depa rtment of Social a nd Decision Sciences,
Carnegie Mellon University, 208 Porter Hall, Pittsburgh, PA 15213 (email: downs@cmu.edu); Loewenstein:
Department of Social and Decision Sciences, Carnegie Mellon University, 208 Porter Hall, Pittsburgh, PA 15213
(email: gL20@andrew.cmu.edu). We thank the United States Food and Drug Administration (USDA) Economic
Research Service (grant numbers 58400060114 and 59400080077) and the Center for Behavioral Decision
Research at Carnegie Mellon University for nancial support, and Howard Seltman, Jay Variyam, and Roberto
Weber for numerous helpful suggestions on the design and analysis of our results. We also thank Michael Benisch,
Lauren Burakowski, Aya Chaoka, Charlotte Fitzgerald, Nathaniel Gales, Lizzie Haldane, Min Young Park, Eric
Tang, and Victoria Vargo for help with data collection.
†
To comment on this article in the online discussion forum, or to view additional materials, visit the articles
page at http://www.aeaweb.org/articles.php?doi=10.1257/app.2.2.164.
Promoting Healthy Choices:
Information versus Convenience†
By J W, J S. D, G L*
Success in slowing obesity trends would benet from policies aimed
at reducing calorie consumption. In a eld experiment at a fast-food
sandwich chain, we address the effects of providing calorie informa-
tion, mimicking recent legislation, and test an alternative approach
that makes ordering healthier slightly more convenient. We nd that
calorie information reduces calorie intake. Providing a daily calorie
target does as well, but only for non-overweight individuals. Making
healthy choices convenient reduces intake when the intervention is
strong. However, a milder implementation reduces sandwich calo-
ries, but does not reduce total calories due to compensatory effects
on side orders and drinks. (JEL I12, I18, L81)
Contents
Promoti ng Healthy Choic es:
Information versus Convenience† 164
I. Prior Attempts to Encourage Healthy Behavior 165
A. Information Provision 165
B. Asymmetr ic Paternalism 165
II. The Cu rrent Studies 166
A. Methods 167
B. Participants 168
III. Results 168
A. Informational Effects on Total Meal Calor ies 168
B. Asymmetr ically Paternalistic Effects on Total Meal Calories 168
C. Mechanisms 169
IV. Discussion 175
References 177
VOL. 2 NO. 2 165
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
and strives to make healthier meal choices marginally more convenient. We nd that
providing either calorie information for menu items or a recommendation for daily
caloric intake decreases total calories ordered. We also nd suggestive evidence
that daily calorie recommendations may not be effective for overweight people who,
arguably, could benet most from reduced calorie intake. Results for convenience
are mixed. A relatively heavy-handed intervention that makes unhealthy sandwiches
harder to choose reduces total caloric intake, but a milder version of the intervention
does not. Additional analyses reveal the importance of addressing compensatory
behavior, as the milder version of the convenience intervention did lead to increased
choice of lower calorie sandwiches, but the effect was undermined by an increase in
the caloric content of side dishes and drinks. Overall, neither approach (providing
calorie information or making lower calorie sandwiches more convenient) appeared
to interfere with the other, so an additive approach of both classes of intervention
may be promising for future policy consideration.
I. Prior Attempts to Encourage Healthy Behavior
A. Information Provision
The Nutrition Labeling and Education Act (NLEA), requiring consistent nutritional
information for packaged foods, was implemented in 1994 (USDA 1994). Research
on the impact of the NLEA suggests that it had some benecial effects, including
directing consumers’ attention to negative nutrition attributes such as sodium levels
(Siva K. Balasubramanian and Catherine Cole 2002); reducing fat intake (Marian
L. Neuhouser, Alan R. Kristal, and Ruth E. Patterson 1999; Alan D. Mathios 2000);
reducing fat calories, cholesterol, and sodium intake (Sung-Yong Kim, Rodolfo M.
Nayga, Jr., and Oral Capps, Jr. 2000); and even decreasing body weight—albeit only
among some groups (Jayachandran N. Variyam and John Cawley 2006). However,
these benecial effects vary according to individual characteristics, such as age and
cognitive ability (Cole and Gary J. Gaeth 1990), motivation (Christine Moorman
1996), and self-control (Finkelstein, Ruhm, and Kosa 2005).
One reason for the limited effects of labeling may be that information alone can-
not overcome other forces, such as the high cost of healthy foods (Kelli K. Garcia
2007) or the delayed and intangible nature of benets from dieting (John G. Lynch,
Jr. and Gal Zauberman 2006; Scott Rick and Loewenstein 2008). Furthermore,
choices made in restaurants may be even less amenable to informational interven-
tions. Restaurant consumers tend to be hungry and in a hurry, and thus may be
more short-sighted and less motivated to process nutritional information. Our stud-
ies explore whether a similar pattern of partially benecial effects, as found for the
NLEA, results from providing calorie information on a fast-food menu, and examine
how these effects interact with demographic characteristics.
B. Asymmetric Paternalism
The second class of approach we test is what behavioral economists refer to
as an “asymmetrically paternalistic” intervention (Colin Camerer et al. 2003;
166 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
Richard H. Thaler and Cass R. Sunstein 2003), which seeks to steer consumers
toward “better” behaviors without limiting their freedom of choice. Many such
interventions exploit biases that usually detract from the quality of decision mak-
ing to, instead, change behavior in benecial ways (Peter Kooreman and Henriëtte
Prast 2007; Loewenstein, Troyen Brennen, and Kevin G. Volpp 2007; Rebecca
K. Ratner et al. 2008). The intervention in our studies plays on two such biases.
The rst is present-biased preferences, whereby individuals place disproportion-
ate weight on immediate costs and benets at the expense of delayed outcomes
(David Laibson 1997; Ted O’Donoghue and Matthew Rabin 1999). To offset the
immediate, calorie-promoting allure of a large and tasty meal, our intervention
aimed to make healthier options slightly more convenient, thereby introducing
an immediate cost—the cost of the extra effort required to order a less healthy
meal—to choosing the unhealthy option. Avoidance of this small immediate cost,
accentuated by present-biased preferences, weighed in favor of healthy selections.
The second bias is the tendency for people to stick with the default option, even
if superior options are available (William Samuelson and Richard Zeckhauser
1988). Policies that set the desired behavior as the default have been shown to
increase retirement put-asides (Brigitte C. Madrian and Dennis F. Shea 2001) and
organ donation (Eric J. Johnson and Daniel Goldstein 2003). The default bias was
invoked in our experiment by making healthy options the implicit default.
Biases such as these have been explored in interventions aimed at diet. For exam-
ple, changes in convenience have been shown to reduce snacking (James E. Painter,
Brian Wansink, and Julie B. Hieggelke 2002) and alter decisions about peripheral
meal items such as chips and candy (Herbert L. Meiselman et al. 1994), typically
without consumer awareness (Brian Wansink 2006). The intervention in this paper
examines the effectiveness of a similar strategy exploiting such biases applied to the
main entrée of a fast-food meal.
II. The Current Studies
This paper reports ndings from two studies designed to assess the effects of
informational and asymmetrically paternalistic approaches to encouraging low-cal-
orie meal choices. The two studies included identical informational manipulations,
but differed in the implementation of the asymmetrically paternalistic manipulation,
with the second study employing a somewhat weaker intervention than the rst.
The informational manipulations were: (1) providing a daily calorie recommenda-
tion, and (2) providing specic information about the caloric content of menu options
(so as to mimic much recent legislation). The asymmetrically paternalistic interven-
tion in both studies made healthy sandwich options slightly more convenient. Based
on the mixed effects of information provision observed in prior research, coupled with
the great success of default manipulations applied to savings behavior, we anticipated
that the convenience manipulation would have more robust effects than the informa-
tional manipulations. Although the main dependent measure in the two studies is the
caloric content of diners’ selections, we also had subjects complete a short survey that
elicited items that we thought might interact with the experimental interventions or
help to explain variance in food choice that was unrelated to the interventions.
VOL. 2 NO. 2 167
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
A. Methods
During lunch hours, we approached customers entering a fast-food sandwich res-
taurant and offered them a free meal of their choice in exchange for completing a
survey. Patrons who agreed to participate were instructed to pick their meal from
the provided menu, rst selecting a sandwich, then a side dish and drink. Next,
participants completed the survey, after which they were handed a gift card and a
coupon with their order to give to the restaurant. To minimize subjects’ concern that
they would be judged on the basis of their food choice, the setup was designed to
give the impression that the meal choice was incidental—merely compensation for
completing the survey.
The menus varied in a 2 (daily calorie recommendation offered or not) × 2 (calo-
rie information for menu items shown or not) × 3 (convenience of healthy options)
design. Daily calorie recommendations, when offered, were presented for men and
women with sedentary versus active lifestyles. Calorie information, when provided,
was listed prominently next to each menu item, including sandwiches, side dishes,
and drinks. The two informational interventions were identical in both studies,
which permits us to aggregate the data from both studies in analyzing their impact,
maximizing statistical power.
The convenience intervention was implemented only for choice of sandwiches,
not for side dishes or drinks, with two slightly different manipulations in each of
the studies. In both studies, the rst page after the instructions listed ve of the
ten sandwich options as “featured” sandwiches. This page contained either the ve
most caloric, least caloric, or a mix of high- and low-calorie sandwich options, as
presumably would be seen in a real-world eating environment. Study 1 implemented
a stronger version of the manipulation. Participants were informed that they could
choose from the featured menu page or, as noted in large print at the bottom of the
page, that they could choose from the complete menu by opening a packet with addi-
tional options. That packet listed the ve remaining, non-featured, sandwiches and
was sealed with a small round paper sticker. Participants were asked to indicate their
sandwich choice, whether from the featured or non-featured list. Using the paper
seal in Study 1 enabled us to record whether the second menu had been opened, pro-
viding insight into the mechanism driving any effect of the manipulation but added
a small extra component of difculty to choosing a sandwich off the featured menu
page, thus making the intervention slightly more heavy-handed.
Study 2 implemented the convenience manipulation in a subtler manner by listing
the second set of sandwiches on the next page. To request a sandwich from the rst
page, the participants merely checked their choice on a form. To request a sandwich
from the second page, the participant was required to write it out, similar to some
sushi restaurants, where ordering something other than a standard combination plat-
ter requires writing down one’s specic selection of sushi.
In both studies, after indicating their sandwich choice, participants chose their
drink (e.g., diet or non-diet soda, juice, or water) and side dish (e.g., potato chips
or fruit), with calorie information listed for each option in the conditions in which
calorie information was provided. The presentation of drinks and side dishes was
identical across convenience manipulation treatments.
168 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
The survey, which was completed following meal selection, asked subjects to
estimate the caloric content of their chosen meal as well as their recommended daily
calorie intake. In addition, the survey asked subjects to rate, on a seven-point scale,
their hunger, anticipated enjoyment of the meal, the extent to which they carefully
considered what to order, whether they ordered less than usual, ordered healthier
than usual, were usually careful about what they ate, considered calories when
ordering, how often they ate at that fast-food chain, whether they were currently
dieting, their height and weight, and other demographic information.
B. Participants
A total of 638 diners participated (292 in Study 1 and 346 in Study 2). Across both
studies more than half of the customers who were approached agreed to participate.
The sample was 61 percent male, 54 percent white, 11 percent African American, 30
percent Indian/Asian, 2 percent Hispanic, and 3 percent other. Participants were 29
years old on average (range 18 to 86). The average body mass index (BMI, calculated
as the ratio of self-reported weight in kilograms to squared height in meters) was
25 (range 16 to 44). Forty-one percent of participants were overweight by conven-
tional standards (BMI ≥ 25). Twenty-one percent of participants reported that they
were currently dieting. Participants reported a mean hunger level of 5.1 and a mean
anticipated meal enjoyment of 5.7 (both on 1-to-7 scales). On average, participants
reported that they visited the restaurant chain where the study was conducted about
twice a month.
III. Results
A. Informational Effects on Total Meal Calories
The impact of the manipulations on total calorie intake was estimated with OLS
regression, with controls for demographic characteristics (Table 1). Aggregating the
data from the two studies (column 3), providing specic calorie information led
participants to order signicantly fewer calories (B = −60.7, t(621) = −3.20, p <
0.001), as did the daily calorie recommendation (B = −37.8, t(621) = −2.01, p <
0.05). Although we had anticipated that the recommendation might help people to
use the specic information more effectively, the interaction between these variables
did not approach signicance. Rather, the effects of the informational interventions
appear to be additive, with the combination of the two reducing meals by almost 100
calories (Figure 1).
B. Asymmetrically Paternalistic Effects on Total Meal Calories
Two dummy variables were included in the regression (Table 1) to test for the
effects of the convenient menu comprising healthy or unhealthy sandwiches, both
compared to a convenient menu with a mix of both types of sandwiches. In Study 1
(column 1), the healthy featured menu had a large and signicant negative impact on
total meal calories (B = −76.65, t(282) = −2.25, p < 0.03), despite being applied
VOL. 2 NO. 2 169
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
only to the sandwich menu and not to the side dishes or drinks. In Study 2 (column 2),
however, the subtler convenience manipulation did not reduce total meal calories (B
= 22.23, t(334) = 0.72, p = 0.47). Column 3 of Table 1 allows for an explicit com-
parison of the impact of the two interventions. The signicant positive interaction
between the dummy for Study 2 and convenience indicates that Study 1’s stronger
manipulation had a signicantly larger impact on total calories than did Study 2’s
weaker manipulation.
C. Mechanisms
Sandwich versus Non-Sandwich Calories and Compensatory Behavior.— Clues
about how and why the different interventions did or did not reduce total meal calo-
ries are provided in separate OLS regressions examining the determinants of sand-
wich choice (Table 2) and non-sandwich (i.e., side dish and drink) calories (Table 3).
The informational manipulations achieved their effects purely through lowering
non-sandwich calories. Table 2 shows that there was no signicant impact of the
informational interventions on sandwich choice, and Table 3 shows that a signicant
decrease in non-sandwich calories resulted from both the specic calorie informa-
tion (B = −48.57, t(621) = −3.25, p < 0.001) and the daily calorie recommenda-
tion (B = −35.91, t(621) = −2.42, p < 0.02). Although participants randomized to
receive information had seen it at the time when they made their sandwich decision,
their choice of sandwich does not seem to have been affected. This suggests that par-
ticipants may have found it easier to cut calories by changing side dishes or drinks,
or that doing so detracted less from their anticipated enjoyment of the meal than
changing the sandwich would have.
The convenience manipulation, on the other hand, had its strongest effect on
sandwich choice (Figure 2), which is not surprising since this manipulation changed
F 1. E M M T M C C, B S,
F P S C I D C R
Note: Bars indicate two standard errors.
!
Study 1
Study 2
Total meal calories
Calorie information
900
850
800
750
700
650
600
90
80
70
60
50
40
30
20
10
Percentage of participants
ordering a low-calorie
sandwich
Featured menu
Healthy Mixed Unhealtthy
Not provided Provided
Study 1
Study 2
Calorie recommendation
Not provided
Provided
170 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
only the convenience of ordering different types of sandwiches, but not drinks or
side orders. The effect of the convenience manipulation was stronger in Study 1,
where participants were 44 percent more likely (an increase of 23 percentage points)
to choose a low-calorie sandwich when the healthy menu was made convenient (B
= 0.23, t(282) = 3.47, p < 0.001), and 44 percent less likely to do so when the
unhealthy menu was convenient (B = −0.23, t(282) = −3.35, p < 0.001) (Table 2).
The weaker manipulation of Study 2 resulted in a smaller but still signicant impact
of the healthy menu on sandwich choice. When healthy sandwiches were made more
convenient, subjects were 35 percent more likely to order a low-calorie sandwich (B
= 0.15, t(335) = 2.22, p < 0.03), but making unhealthy sandwiches more conve-
nient had no effect. However, in Study 2, the convenience manipulation appeared to
produce a compensatory effect on non-sandwich calories (Table 3), which increased
sufciently in the healthy sandwich condition (B = 57.81, t (334) = 2.40, p < 0.02)
T 1—T C C
Study 1 Study 2 Combined stud ies
Constant 843.62**
(46.13)
902.88**
(41.82)
850.85**
(35.67)
Calorie in formation provided −48.05
(28.82)
−71.73**
(25.29)
−60.69**
(18.95)
Daily calorie recommendation provided −37.46
(28.61)
−38.94
(25.14)
−37.81*
(18.83)
Healthy featured menua−76.65*
(34.00)
22.23
(30.70)
−76.66*
(33.19)
Unhealthy featured menu 16.00
(35.62)
4.40
(30.84)
15.11
(34.73)
Female −69.47*
(30.06)
−88.37**
(26.06)
−79.90**
(19.64)
Age (in years) −0.75
(1.12)
−0.36
(1.07)
−0.54
(0.77)
Africa n American 130.47**
(49.09)
93.04*
(44.14)
112.07**
(32.67)
Study 2 46.00
(32.50)
Study 2 × healthy menu 98.65*
(45.53)
Study 2 × unhealthy menu −10.98
(46.71)
N = 290 N = 342 N = 632
F(7, 282) = 3.12,
p = 0.003
F(7, 334) = 3.64,
p = 0.001
F(10, 621) = 6.42,
p < 0.001
R2 = 0.07 R2 = 0.07 R2 = 0.09
Note: Standard errors in parentheses.
a
Combined effect of the healthy menu across studies without interaction effects: B = −24.63, t(623) = −1.08,
p = 0.28.
** Signicant at the 1 percent level.
* Signicant at the 5 percent level.
VOL. 2 NO. 2 171
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
to completely offset the impact of the manipulation on sandwich calories, leaving
no effect of the convenience manipulation on total calorie consumption (Table 1,
column 2).
This compensatory effect may result from the attention drawn to the forgone items,
which was heightened in the second study. In Study 1, only 38 percent of people
(similar in all three conditions, χ2 = 0.08, p = 0.96) opened the packet to see alter-
native options. In the second study, however, all subjects are likely to have seen the
additional options when turning the menu page, before choosing their side dish and
drink. Choosing from the healthy menu may have led to a sense of deservingness
upon seeing the unhealthy sandwiches that were passed up, leading people to reward
themselves with higher-calorie side dishes and drinks. Regardless of the mechanism
driving these results, they point to the fact that compensatory behavior can be critical
in determining the overall impact of interventions aimed at behavior change.
Perceptions of Caloric Content.—A likely mechanism for the effect of informa-
tion on eating behavior is improved knowledge about, and consideration of, caloric
content and guidelines. To explore these effects, we performed a series of regressions
T 2—C L-C S
Study 1 Study 2 Combined stud ies
Constant 0.52**
(0.09)
0.43**
(0.09)
0.52**
(0.07)
Calorie in formation provided (versus not)0.08
(0.06)
0.03
(0.05)
0.05
(0.04)
Daily calorie recommendation provided
(versus not)
0.01
(0.06)
0.04
(0.05)
0.02
(0.04)
Healthy featured menua0.23**
(0.07)
0.15*
(0.07)
0.23**
(0.07)
Unhealthy featured menu −0.23**
(0.07)
0.00
(0.07)
−0.23**
(0.07)
Female 0.01
(0.06)
0.07
(0.06)
0.05
(0.04)
Age (in years) −0.001
(0.002)
−0.002
(0.002)
−0.002
(0.002)
Africa n American −0.17
(0.10)
−0.20*
(0.10)
−0.18**
(0.07)
Study 2 −0.09
(0.07)
Study 2 × healthy menu −0.08
(0.09)
Study 2 × unhealthy menu 0.23*
(0.10)
N = 290 N = 343 N = 633
F(7, 282) = 7.12,
p < 0.001
F(7, 335) = 1.99,
p = 0.06
F(10, 622) = 6.25,
p < 0.001
R2 = 0.15 R2 = 0.04 R2 = 0.09
Notes: Standa rd errors in parentheses. OLS regressions; logistic regression produced virtually identical results.
a
Combined effect of the healthy menu across studies without interaction effects: B = 0.19, t(624) = 4.00, p
< 0.001.
** Signicant at the 1 percent level.
* Signicant at the 5 percent level.
172 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
predicting the difference between participants’ estimates of recommended daily cal-
ories and actual recommended values (Table 4, column 1), the absolute value of this
difference (column 2), the difference between participants’ estimates of the calories
in their chosen meal and their meal’s actual calories (column 3), the absolute value
of this estimation error (column 4), and the extent to which participants reported
considering calories when deciding on their meal (column 5).
Participants’ knowledge of recommended daily caloric intake and of the caloric
content of their meal choices was relatively poor. Overall, participants greatly under-
estimated daily recommended calorie intake (mean difference between estimated
and actual recommendation = −547.2, p < 0.001) and the calories in their meal
(mean difference between estimate and actual meal calories = −119.2, p < 0.001).
Receiving calorie information signicantly reduced the magnitude of the absolute
error in estimating meal calories (Table 4, column 4; B = −67.70, t(595) = −2.82,
p < 0.01). The daily calorie recommendation marginally increased estimates of
T 3—N-S (Side Dish and Drink) C C
Study 1 Study 2 Combined studies
Constant 445.59**
(36.54)
472.55**
(32.87)
447.30**
(28.17)
Calorie in formation provided (versus not) −34.13
(22.83)
−61.79**
(19.88)
−48.57**
(14.97)
Daily calorie recommendation provided (versus not) −40.52
(22.66)
−33.60
(19.76)
−35.91*
(14.87)
Healthy featured menua−22.58
(26.93)
57.81*
(24.13)
−22.70
(26.21)
Unhealthy featured menu −37.13
(28.21)
4.21
(24.24)
−38.72
(27.43)
Female −68.97**
(23.81)
−63.53**
(20.48)
−65.85**
(15.51)
Age (in years) −1.19
(0.89)
−0.78
(0.84)
−0.98
(0.61)
Africa n American 90.06*
(38.89)
29.93
(34.69)
58.96*
(25.80)
Study 2 23.72
(25.67)
Study 2 × healthy menu 80.54*
(35.96)
Study 2 × unhealthy menu 42.55
(36.88)
N = 290 N = 342 N = 632
F(7, 282) = 2.73,
p < 0.01
F(7, 334) = 4.48,
p < 0.001
F(10, 621) = 6.79,
p < 0.001
R2 = 0.06 R2 = 0.09 R2 = 0.10
Note: Standa rd errors in parentheses.
a Combined effect of the healthy menu across studies without interaction effects: B = 20.13, t(623) = 1.12, p
= 0.26
** Signicant at the 1 percent level.
* Signicant at the 5 percent level.
VOL. 2 NO. 2 173
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
recommended daily allowance of calories, moving them closer to the recommenda-
tions that we provided (from 1,680 to 1,838 for women, and from 2,072 to 2,090 for
men; Table 4, column 1; B = 127.45, p = 0.056), and signicantly increased esti-
mates of meal calories to a corresponding degree (Table 4, column 3), but did not
reduce error.
It may seem contradictory that the daily calorie recommendation would increase
estimates of how much one should eat but reduce the actual amount that one does
eat. These results suggest that the daily recommendation worked not by improv-
ing knowledge, but, perhaps, by raising the salience of calorie considerations.
This account is supported by the fact that those who received the recommendation
were more likely to report that they considered calories when ordering (Table 4,
column 5), even though they were no more accurate in reporting the calories in their
meal (Table 4, column 4).
Effects among Overweight Participants.—The analyses above assess the degree
to which these three interventions reduce calorie intake among the population as a
whole. However, people who are not overweight are not the intended audience of
such campaigns because they have no special reason to reduce their calories. Thus,
it is of particular importance to measure the impact of these interventions on the
people who most need to change behavior—those who are overweight or obese.
Table 5 presents regressions including only the subsample of overweight (BMI ≥ 25)
participants (n = 262; 41 percent of sample), predicting total meal calories (column
1), choice of a low-calorie sandwich (column 2), and non-sandwich calories (column
3) using the same independent variables as the main analyses.
F 2. T P P W C L-C S E S
F “F” M
Note: Bars indicate two standard errors.
!
Study 1
Study 2
Total meal calories
Calorie information
900
850
800
750
700
650
600
90
80
70
60
50
40
30
20
10
Percentage of participants
ordering a low-calorie
sandwich
Featured menu
Healthy Mixed Unhealtthy
Not provided Provided
Study 1
Study 2
Calorie recommendation
Not provided
Provided
174 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
Although the smaller sample size diminishes statistical power and reduces most
effects to being non-signicant, comparison of effect sizes reveals no evidence that
provision of specic calorie information, or the convenience manipulation, has a dif-
ferent effect on total calories consumed among overweight individuals compared to
the broader sample. The calorie recommendation, on the other hand, seems to have
no benet for this population (indeed, the direction of the effect for total calories is
reversed from that found in the full sample, although it is not signicantly different
from zero). If the effect in the overall population is due to increasing salience of
calorie information rather than providing usable information, as proposed above,
perhaps this nding speaks to the greater attention already paid to calorie consider-
ations by overweight people. Although this analysis is merely exploratory, it speaks
T 4—E I C M K
P
Dependent var iable
Estimated–
actual
recommendation
Absolute
recommendation
error
Estimated–
actual meal
calories
aAbsolute meal
calorie erroraConsidered
caloriesb
Constant –280.38**
(127.19)
507.99**
(105.39)
–277.70**
(63.06)
324.28**
(45.25)
1.79**
(0.28)
Calorie
information
–74.47
(67.10)
–17.76
(55.60)
47.66
(33.49)
–67.70**
(24.03)
0.28
(0.15)
Calorie
recommendation
127.45†
(66.62)
–67.31
(55.20)
143.14**
(33.28)
0.85
(23.88)
0.39**
(0.15)
Healthy
featured menu
–131.67
(117.45)
25.57
(97.32)
35.70
(58.10)
1.62
(41.69)
0.29
(0.26)
Unhealthy
featured menu
42.94
(123.39)
–48.66
(102.24)
35.83
(61.26)
13.10
(43.96)
0.08
(0.27)
Female –288.57**
(69.49)
–148.07**
(57.58)
47.89
(34.63)
–55.17*
(24.85)
0.41**
(0.15)
Age –8.53**
(2.77)
7.52**
(2.30)
1.81
(1.34)
1.39
(0.99)
0.03**
(0.01)
Africa n American –362.60**
(116.45)
368.22**
(96.49)
–220.54**
(57.84)
186.95**
(41.51)
–0.51*
(0.26)
Study 2 –180.18
(114.99)
73.01
(95.28)
–44.67
(57.70)
–10.12
(41.41)
0.03
(0.25)
Study 2 × healthy
featured menu
252.15
(161.21)
108.23
(133.58)
–66.52
(80.62)
53.10
(57.85)
–0.43
(0.36)
Study 2 × unhealthy
featured menu
147.54
(165.32)
60.11
(136.99)
–49.51
(82.45)
–27.23
(59.16)
–0.10
(0.36)
N = 622 N = 622 N = 606 N = 606 N = 621
F(10, 611) = 6.75,
p < 0.001
F(10, 611) = 4.37,
p < 0.001
F(10, 595) = 4.13,
p < 0.001
F(10, 595) = 3.66,
p < 0.001
F(10, 610) = 5.48
p < 0.001
R2 = 0.10 R2 = 0.07 R2 = 0.07 R2 = 0.06 R2 = 0.08
Note: Standa rd errors in parentheses.
a Eight extreme outliers (with estimate errors > 1700) were removed from these analyses.
b Endorsement of “I considered calor ies when ordering”, on a 1-to-7 scale.
** Signicant at the 1 percent level.
* Signicant at the 5 percent level.
† Sign icant at the 10 percent level.
VOL. 2 NO. 2 175
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
to the need to examine target populations more closely, so as to ensure that the
intended effects are affecting the behavior of those most in need of behavior change.
IV. Discussion
The results of this study indicate that both information and convenience can
affect the food choices of fast-food restaurant patrons. Averaging across all par-
ticipants, both calorie information and a calorie recommendation decreased total
calories ordered. The more heavy-handed convenience manipulation also reduced
total calories signicantly, whereas the lighter intervention of Study 2 inuenced
sandwich choice but not total calories.
Although the daily calorie recommendation appears, on average, to decrease cal-
orie intake, there is a disturbing suggestion in the data that it may not be helpful for
those who need to cut back on calories—those who are overweight. The potentially
different, and less impressive, results for people who are overweight have important
implications for policies targeted at restaurant, and specically fast food, dining.
Before expanding the implementation of policies that involve provision of informa-
tion, it would be very helpful to understand more about when information helps and
T 5—E I C M O
(BMI ≥ 25) P
Dependent var iable Total calories
Choice of a low-
calorie sandwich
Non-sandwich
calories
Constant 779.31**
(54.96)
0.56**
(0.11)
379.51**
(42.98)
Calorie in formation provided (versus not)–41.27
(29.71)
0.04
(0.06)
–33.78
(23.24)
Daily calorie recommendation provided
(versus not)
8.48
(29.36)
–0.10
(0.06)
–19.17
(22.96)
Healthy featured menu –93.37
(48.97)
0.18
(0.10)
–50.74
(38.30)
Unhealthy featured menu –16.94
(50.79)
–0.17
(0.11)
–52.97
(39.72)
Female –43.37
(32.64)
0.06
(0.07)
–29.26
(25.53)
Age (in years)–0.26
(1.04)
0.00
(0.002)
–0.12
(0.81)
Africa n American 126.48**
(43.03)
–0.30**
(0.09)
46.67
(33.65)
Study 2 73.12
(49.31)
–0.16
(0.10)
35.29
(38.57)
Study 2 × healthy menu 80.56
(69.62)
0.14
(0.14)
111.66*
(54.45)
Study 2 × unhealthy menu 2.92
(72.04)
0.32*
(0.15)
70.80
(56.34)
N = 259 N = 260 N = 259
F(10, 248) = 2.61,
p < 0.01
F(10, 249) = 3.27,
p = 0.001
F(10, 248) = 2.80
p < 0.01
R2 = 0.10 R2 = 0.12 R2 = 0.10
Note: Standa rd errors are in pa rentheses.
** Signicant at the 1 percent level.
* Signicant at the 5 percent level.
176 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
when it does not, and why these patterns emerge. In addition, although the provision
of calorie information reduced non-sandwich calories, it appeared to have no effect
on the choice between high- and low-calorie sandwiches, which was the rst choice
made by participants, and is arguably the main course of the meal. Further research
is needed to understand this pattern, and specically to determine whether calorie
information might have differential effects on central versus peripheral components
of a meal, or whether its effects might be greater in a situation in which sandwich
calories were not bimodally distributed, as they were constructed to be for this study.
Because informational strategies do not appear to be a panacea for the obesity
epidemic, future interventions should consider additional methods for changing eat-
ing behavior. In these studies, simply making it easier for consumers to choose a
low-calorie option signicantly reduced sandwich calories. However, the impact of
the weaker convenience manipulation was undermined by compensatory choices of
drinks and side orders. This raises the question of whether a similarly weak conve-
nience manipulation would have been more effective if it had encompassed drinks
and side orders (e.g., made bottled water and fruit part of the featured menu).
The compensatory effect on non-sandwich items highlights the importance of
including and analyzing choices that are not directly targeted by interventions. The
convenience manipulation was applied only to sandwich choice, allowing us the
opportunity to measure compensatory effects on nontargeted choices within the
entire meal, and these proved to be important. However, all participants had the
option of purchasing additional items (e.g., cookies) when they collected their meal,
or purchasing other items at nearby establishments selling food, and were also likely
to make further food choices later in the day. Those who choose lower calorie meals,
whether due to information or convenience, may feel hungrier later in the day, or
more entitled to indulge, and may end up consuming more calories later. We were
unable to measure such purchases, so we cannot rule out that they occurred or, if so,
assess what impact they had on total calories consumed in the different conditions.
Note that a similar criticism applies to the studies examining the impact of changing
defaults on retirement contributions. Although these studies have revealed uniformly
positive effects of high defaults, without information about other nancial activities
there is no way of knowing whether the net impact of such changes increases sav-
ing. It is possible, for example, that those contributing more to retirement plans may
incur credit card debt, or even take out payday loans to compensate for their lower
paychecks.
It is, moreover, unclear whether either type of intervention could produce sus-
tained changes in behavior. People might learn to work around the interventions, for
example, if they discover that their preferred options were always reserved for later
in the menu, or they might come to ignore calorie information. Conversely, and more
optimistically, if either of these kinds of manipulations work for some period, they
might prove to be habit-forming, thereby creating long-term changes in diet even
if the interventions were removed. In addition, restaurants may change their menu
options and nutritional content, for better or worse, as a result of the legislation, thus
either aiding or undermining individual behavior.
In this study, we aimed to test for effects of information and convenience inde-
pendently. However, it is possible that our subjects interpreted the composition of the
VOL. 2 NO. 2 177
WISDOM ET AL.: PROMOTING HEALTHY CHOICES
featured menu as conveying information, and even more likely that this would be true
in real-world settings. People could assume that items made more convenient were
selected based on health concerns, as would likely be the case if such a presentation
were mandated by legislation. Alternatively, selection for menu prominence might
be interpreted as conveying other information about the selected options, such as
that they are more popular, fresher, or tastier. Thus, the context in which consumers
understand the menu may lead them to glean different kinds of information from
the fact that some options are made especially prominent, potentially affecting their
behavioral response.
The current study explores the promise of different kinds of interventions, includ-
ing both the standard approach of providing more information, as well as subtler
attempts to nudge people in a healthier direction. Both kinds of intervention showed
some promise, and they did not interact with one another, suggesting that there
would be no downside to applying them simultaneously. However, before imple-
menting either type of interventions on a broader scale, more research is needed to
understand and guard against both the potential for perverse effects on subsets of the
population and the possibility of compensatory behaviors.
REFERENCES
Balasubramanian, Siva K., and Catherine Cole. 2002. “Consumers’ Search and Use of Nutrition
Information: The Challenge and Promise of the Nutrition Labeling and Education Act.” Journal
of Marketing, 66(3): 112–27.
Blanck, H. M., W. H. Dietz, D. A. Galuska, C. Gillespie, R. Hamre, L. Kettel Khan, M. K. Serdula, et
al. 2006. “State-Specic Prevalence of Obesity among Adults—United States, 2005.” Morbidity
and Mortality Weekly Report (MM WR), 55(36): 985–88.
Camerer, Colin, Samuel Issacharoff, George Loewenstein, Ted O’Donoghue, and Matthew Rabin.
2003. “Regulation for Conservatives: Behavioral Economics and the Case for ‘Asymmetric Pater-
nalism.’ ” University of Pennsylvania Law Review, 151(3): 1211–54.
Center for Science in Public Interest. 2003. “Anyone’s Guess: The Need for Nutrition Labeling at
Fast-Food and Other Chain Restaurants.” http://www.cspinet.org/new/pdf/anyone_s_guess_nal_
web.pdf. (accessed June 28, 2008).
Chou, Shin-Yi, Michael Grossman, and Henry Saffer. 2004. “An Economic Analysis of Adult Obe-
sity: Results from the Behavioral Risk Factor Surveillance System.” Journal of Health Econom-
ics, 23(3): 565–87.
Cole, Catherine A., and Gary J. Gaeth. 1990. “Cognitive and Age-Related Differences in the Abil-
ity to Use Nutritional Information in a Complex Environment.” Journal of Marketing Research,
27(2): 175–84.
Cutler, David M., Edward L. Glaeser, and Jesse M. Shapiro. 2003. “Why Have Americans Become
More Obese?” Journal of Economic Perspectives, 17(3): 93–118.
Finkelstein, Eric A., Christopher J. Ruhm, and Katherine M. Kosa. 2005. “Economic Causes and
Consequences of Obesity.” Annual Review of Public Health, 26: 239–57.
Garcia, Kelli K. 2007. “The Fat Fight: The Risks and Consequences of the Federal Government’s
Failing Public Health Campaign.” Penn State Law Review, 112(2): 529–86.
Johnson, Eric J., and Daniel Goldstein. 2003. “Do Defaults Save Lives?” Science, 302(5649): 1338–
39.
Kim, Sung-Yong, Rodolfo M. Nayga, Jr., and Oral Capps, Jr. 2000. “The Effect of Food Label Use
on Nutrient Intakes: An Endogenous Switching Regression Analysis.” Journal of Agricultural and
Resource Economics, 25(1): 215–31.
Kooreman, Peter, and Henriëtte Prast. 2007. “What Does Behavioral Economics Mean for Policy?
Challenges to Savings and Health Policies in the Netherlands.” http://members.ziggo.nl/peter-
kooreman/pibe.pdf.
Laibson, David. 1997. “Golden Eggs and Hyperbolic Discounting.” Quarterly Journal of Econom-
ics, 112(2): 443–77.
178 AMERICAN ECONOMIC JOURNAL: APPLIED ECONOMICS APRIL 2010
Loewenstein, George, Troyen Brennan, and Kevin G. Volpp. 2007. “Asymmetric Paternalism to
Improve Health Behaviors.” Journal of the American Medical Association, 298(20): 2415–17.
Lynch, John G., Jr., and Gal Zauberman. 2006. “When Do You Want It? Time, Decisions, and Public
Policy.” Journal of Public Policy and Marketing, 25(1): 67–78.
Madrian, Brigitte C., and Dennis F. Shea. 2001. “The Power of Suggestion: Inertia in 401(K) Partici-
pation and Savings Behavior.” Quarterly Journal of Economics, 116(4): 1149–87.
Mathios, Alan D. 2000. “The Impact of Mandatory Disclosure Laws on Product Choices: An Analysis
of the Salad Dressing Market.” Journal of Law and Economics, 43(2): 651–77.
Meiselman, Herbert L., Duncan Hedderley, Sarah L. Staddon, Barry J. Pierson, and Catherine R.
Symonds. 1994. “Effect of Effort on Meal Selection and Meal Acceptability in a Student Cafete-
ria.” Appetite, 23(1): 43–55.
Moorman, Christine. 1996. “A Quasi Exeperiment to Assess the Consumer and Informational Deter-
minants of Nutrition Information Processing Activities: The Case of the Nutrition Labeling and
Education Act.” Journal of Public Policy and Marketing, 15(1): 28–44.
Neuhouser, Marian L., Alan R. Kristal, and Ruth E. Patterson. 1999. “Use of Food Labels Is Associ-
ated with Lower Fat Intake.” Journal of the American Dietetic Association, 99(1): 45–53.
O’Donoghue, Ted, and Matthew Rabin. 1999. “Doing It Now or Later.” American Economic Review,
89(1): 103–24.
Painter, James E., Brian Wansink, and Julie B. Hieggelke. 2002. “How Visibility and Convenience
Inuence Candy Consumption.” Appetite, 38(3): 237–38.
Ratner, Rebecca K., Dilip Soman, Gal Zauberman, Dan Ariely, Ziv Carmon, Punam A. Keller, B.
Kyu Kim, et al. 2008. “How Behavioral Decision Research Can Enhance Consumer Welfare:
From Freedom of Choice to Paternalistic Intervention.” Marketing Letters, 19(3–4): 383–97.
Rick, Scott, and George Loewenstein. 2008. “Intangibility in Intertemporal Choice.” Philosophical
Transactions of the Royal Society B: Biological Sciences, 363(1511): 3813–24.
Samuelson, William, and Richard Zeckhauser. 1988. “Status Quo Bias in Decision Making.” Journal
of Risk and Uncertainty, 1(1): 7–59.
Thaler, Richard H., and Cass R. Sunstein. 2003. “Libertarian Paternalism.” American Economic
Review, 93(2): 175–79.
United States Food and Drug Administration. 1994. “Food Labeling Guide.” http://www.fda.gov/
Food /Guida nceComplianceRegulatoryInfor mation /GuidanceDocu ments/ FoodLabelingNutri-
tion/FoodLabelingGuide/default.htm.
Variyam, Jayachandran N., and John Cawley. 2006. “Nutrition Labels and Obesity.” National Bureau
of Economic Research Working Paper 11956.
Wansink, Brian. 2006. Mindless Eating: Why We Eat More Than We Think. New York: Bantam
Books.