Modern Economy, 2013, 4, 1-8
http://dx.doi.org/10.4236/me.2013.410A001 Published Online October 2013 (http://www.scirp.org/journal/me)
The Impact of Meal Attributes and Nudging on
Healthy Meal Consumption
—Evidence from a Lunch Restaurant Field Experiment
Linda Thunström1,2, Jonas Nordström3,4
1HUI Research AB, Stockholm, Sweden
2Department of Economics and Finance, University of Wyoming, Laramie, USA
3Department of Economics, Lund University, Lund, Sweden
4Department of Food and Resource Economics, University of Copenhagen, Frederiksberg C, Denmark
Received July 9, 2013; revised August 10, 2013; accepted September 4, 2013
Copyright © 2013 Linda Thunström, Jonas Nordström. This is an open access article distributed under the Creative Commons Attri-
bution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
ABSTRACT
We use a field experiment in a lunch restaurant to analyze how meal attributes and a “nudge” impact healthy labeled
meal consumption. The nudge consists of increasing the salience of healthy labeled meals by placing them at the top of
the menu. We find that certain meal attributes (e.g. poultry and red meat) greatly increase both sales and the market
share of the healthy labeled meal. We conclude that a careful design of the healthy food supply may be efficient in en-
couraging healthier meal choices, e.g. supplying healthy labeled versions of popular conventional meals. We find no
impact on healthy labeled meal sales from the nudge.
Keywords: Healthy Food; Restaurant Meals; Food Supply; Meal Attributes; Nudging; Field Experiment
1. Introduction
The modern western diet is often high in calories while
low in healthy nutrients. In combination with a more
sedentary lifestyle, the characteristics of the modern diet
have been proven to be toxic: they have placed obesity,
overweight, and several serious diet related diseases (e.g.,
several types of cancer, diabetes, cardiovascular disease,
high blood pressure and osteoporosis) at the top of the
list on public health problems in many countries, both
developed and developing.
To encourage healthier food choices, policy reforms
that entail information, such as nutrition labelling, and
taxes on unhealthy food have been implemented, e.g.,
legislated menu labelling in many states in the US (start-
ing in New York City, 2008), and taxes on unhealthy
food and beverages in Denmark, Finland, France and
Hungary, introduced in 2011-2012.
However, field evidence of the impact on healthy food
consumption from information is at best mixed. Many
studies find no effect on the nutritional quality of con-
sumption from nutritional information, even when it is
the most visible, such as point-of-purchase menu label-
ling (e.g. [1-4]). Further, research on the impact of food
tax reforms implies that taxes of the magnitudes that are
politically feasible are likely to have little impact on food
consumption (e.g., [5-8]). In essence, information and
moderate price incentives do not seem to substantially
impact healthier food consumption1. This may be a result
of taste being one of the main determinants of food
choice, often found to dominate both health and price
(e.g. [11-13]). Taste is largely determined by attributes in
food.
Thunström and Nordström, [3], find that sales of con-
ventional meals substantially increase from certain at-
tributes, mainly poultry and red meat. They also find that
sales of conventional meals benefit from a “nudge” that
displays the meal at the top of the menu. Their results
render the question if designing healthy labelled meals
containing generally popular, or tasty, meal attributes
1Other policy initiatives are aimed at increasing the supply of healthy
foods in areas where availability has been limited (see e.g. the “Healthy
Corner Store Initiative” in Philadelphia, US). Research on the impact on
food consumption of increased availability to healthy food shows at
best limited effects (see e.g. [9,10]). Also, the prevalence of diet related
illnesses is high even in areas where healthy food is highly available;
suggesting that increased availability may not be a key in reducing the
overall prevalence of diet related illnesses.
C
opyright © 2013 SciRes. ME

L. THUNSTRÖM, J. NORDSTRÖM 2
(poultry and red meat) and nudging could be used to
successfully promote healthy eating, i.e. to encourage
healthy labelled meal consumption?
In this paper, we use data from a field experiment to
analyze the impact on healthy labelled meal sales from
manipulating meal attributes and a “nudge”. Nudging
entails make the preferred choice (from a policy perspec-
tive) more salient than other choice alternatives, thereby
encouraging consumers to make socially desirable choices
with minimal paternalism, i.e. without restricting the
consumer choice set (see e.g. [14]).
We summarize the aim of the paper in the following
hypotheses that we test empirically:
Hypothesis 1. “Popular meal attributes can be used in
healthy labelled meals to increase consumption of healthy
labelled meals.”
Hypothesis 2. “Nudging consumers by displaying the
healthy labelled meal at the top of the menu positively
impact consumption of healthy labelled meals.”
In our empirical analysis, we measure consumption
with sales data. To the best of our knowledge, no study
has previously analyzed the impact of meal design, in
terms of attributes, on healthy meal consumption. Menu
nudges have previously been shown to positively impact
healthy food choices [15,16]2.
2. The Field Experiment and Empirical
Analysis
Our analysis is based on a subset of data from the field
experiment reported in [3]. The experiment was con-
ducted in a lunch restaurant at an industry company in
southern Sweden. Collecting data from a field experi-
ment to perform our study has several important benefits.
The field experiment allows us to analyze the impact on
healthy labelled meal sales from meal attribute manipu-
lation and nudging, while holding constant prices, and
potentially also perceived nutritional content. Prices of
all meals (the healthy labelled meal and its non-labelled
substitute meals) are equal and constant throughout the
study period. The restaurant setting of the experiment is
also conducive to controlling for consumers’ perceived
nutritional content of meals: the nutritional content varies
over meals, but is likely to be non-transparent to con-
sumers. Evidence suggests consumers have difficulties in
making accurate estimates of the nutritional content of
prepared meals away from home [17]. Therefore, con-
sumers in a restaurant are likely to largely rely on the
healthy label to distinguish healthy meals from less healthy
meals.
Another benefit of the experiment data is that meal
prices and meal supply were not influenced by the au-
thors of this study. The restaurant is privately owned and
therefore guided by profit maximization. This ensures
that the empirical analysis is based on combinations of
meals, inputs and input costs that are part of a profit
maximizing strategy, while it for instance rules out
healthy meals or inputs that significantly may impact
sales, but would be too expensive to supply by a profit
maximizing entity.
The restaurant at which the field experiment was per-
formed is open to the general public, even if it primarily
serves contractor employees. There are a couple of other
lunch restaurants within walking distance. Restaurant
staff estimates that approximately 10 - 20 percent of
daily lunch eaters are civil servants, 80 - 90 percent are
blue-collar workers, and 30 percent are women. The staff
also estimates that the restaurant has an equal number of
potential customers each week day, despite the shorter
opening hours on Fridays. The lunch menu was posted
outside the restaurant every day and customers could also
get the menu via e-mail: the e-mail list contained ap-
proximately 50 - 60 people.
The restaurant introduced a healthy (Nordic “Keyhole”)
labelled meal on the menu on the 20th of April, and re-
ported meal sales for the following 6 weeks (27 business
days), i.e., until 29th of May 2010. The Keyhole label has
been a symbol for healthy food for 20 years in Sweden
and is well-known among the general public as an indi-
cator of healthy food choices. Meals eligible to carry the
Keyhole symbol must fulfill certain criteria. The general
criteria that applies for a Keyhole labeled meal are: the
meal should contain 400 - 750 calories (1.67 - 3.14 MJ),
max 30 energy percent from fat (more is allowed for
seafood), max 3 grams of sugar per 100 gram, max 1
gram salt per 100 gram, be well-balanced and contain at
least 100 gram of vegetables (excluding potatoes)3. The
Keyhole label was given to one of the alternatives on the
menu that fulfilled the Keyhole criteria. The same day,
the menu could, however, contain non-labeled alterna-
2The nudge by Downs, Loewenstein and Wisdom, [15], increased the
salience (and reduced the search cost) of low-calorie sandwiches, rela-
tive to higher calorie substitutes, by providing subjects with a menu that
contained low-calorie sandwiches on the front, and higher calorie
sandwiches at the back. They found that providing subjects with such a
menu significantly increased the percent of subjects choosing low-
calorie sandwiches, compared to when subjects were provided a menu
that contained a mixed of low and high calorie sandwiches at the front.
The nudge examined by Ellison, Lusk and Davis, [16], does not entail
re-arranging the order of meals on the menu—they use a field experi-
ment to examine the impact of the traffic label, versus numeric calorie
labels, in front of menu alternatives. They find that the traffic label may
be more efficient than numeric calorie labels in promoting healthy food
choices.
3See www.nyckelhalsrestaurang.se.
4In particular, on a large number of occasions (19 days), at least one o
f
the non-labeled alternatives on the menu contained less calories per
p
ortion than the Keyhole-labeled alternative served the same day
(sometimes even lower than the lower limit for the Keyhole calorie
criteria, i.e. 400 kcal/portion). Also, on 11 days, the fat content pe
r
p
ortion was lower for non-labeled alternatives than it was for the
healthy labeled alternative. It should, however, be noted that the
amount of calories and fat per portion is subject to uncertainty, as ex-
p
lained below.

L. THUNSTRÖM, J. NORDSTRÖM 3
tives that fully or partly fulfilled the criteria as well4.
Hereafter, the Keyhole labeled meal will be referred to as
the healthy labeled meal.
The order of the healthy labelled meal on the menu
was varied over the study period, and the data contains
information on where on the menu the healthy labelled
meal was displayed, the type of meals served each day,
and the amount sold of each meal. A nutritionist assigned
each meal its calorie and fat content per portion, using
the software Dietist XP5.
The restaurant was open all workdays, Monday to
Friday, and closed at 6 pm all weekdays, except Fridays,
when it closed at 3 pm. Every day one healthy labelled
meal and two non-labelled substitute meals were offered
on the menu, except April 30th, when only one non-la-
belled substitute meal was served. The price of all meals
was the same (SEK 63).
2.1. Data from the Experiment
We created 1) a set of dummy variables for the source of
protein in the healthy labelled meals (red meat, poultry,
fish or seafood or vegetarian; yes = 1; no = 0), 2) a set of
dummy variables for the source of protein in the substi-
tute meals (any of the substitute meals containing red
meat, poultry, seafood (including fish), or being vegetar-
ian: yes = 1; no = 0), 3) a couple of dummy variables
indicating the order of the healthy labelled meal on the
menu (first on the menu, versus second or last: yes = 1;
no = 0)6, and 4) a set of dummy variables indicating
weekday (Monday, Tuesday-Thursday, or Friday: yes = 1;
no = 0). We also include fat content in the analysis: fat in
meals can be both positively and negatively valued by
consumers—fat is taste increasing [18], but may pose a
health risk if over consumed. Table 1 shows descriptive
statistics of the variables included in the analysis.
Table 1 shows that the average number of portions
served of the healthy labelled meal per day was 153 dur-
ing the study period, and that the average market share of
the healthy labelled meal was 44 percent, where the
market share is equal to the number of portions sold of
the healthy labelled meal, divided by the total number of
portions sold that day. The highest number sold of the
healthy labelled meal was 232—a healthy labelled tradi-
tional Swedish dish: meatballs and mashed potatoes with
lingon berries, displayed at the top of the menu and
served on a Monday, and where the substitute meals con-
stituted of a meal with poultry and a meal with seafood.
The market share of this healthy labelled meal was 64
percent. The highest market share (68 percent) on any
day during the study period was held by a healthy la-
belled version of another traditional Swedish dish:
“Skansk kallops” (a beef stew from the Skane region)
with boiled potatoes and beetroot, displayed at the top of
the menu and served on a Friday, where the non-labelled
substitute meals contained seafood.
The lowest number sold per day of the healthy labelled
meal was 52, and the lowest market share of the healthy
labelled meal was 14 percent. The same meal holds both
these records—vegetarian spring rolls, with curry sauce
and rice, served on a Tuesday, Wednesday or Thursday,
displayed second or last on the menu, and where the
non-labelled substitute meals constituted of a meal with
red meat and a vegetarian meal.
2.2. Empirical Analysis
To analyze the factors that influence sales and the market
share of the healthy labeled meal, we use the above data
to estimate two different models represented by:
Keyhole



 SzD
In our first model, the content of vector Keyhole is
daily number of portions sold of healthy labelled meals at
t (t = 1,, 27). In our second model, Keyhole is a vector
of the daily market share of the healthy labeled meal at t.
The vector z contains grams of fat per portion in the
healthy labelled meal. D contains the dummy variables
indicating the source of protein of the healthy labelled
meal, and the dummy variables indicating the source of
protein in the non-healthy labelled substitute meals. D
also contains the weekday dummy variables, and the
dummy variable indicating if the healthy labelled meal
appears second or third on the menu. Note that the refer-
ence meal is a healthy labelled meal that contains sea-
food, is displayed at the top of the menu and is sold on a
Monday, with at least one of the substitute meals also
containing seafood7.
S
S
To test for autocorrelation in both models, we use
Durbin’s alternative test, which allows for non-normally
distributed residuals. The test implies that we cannot
onfirm the null hypothesis of no autocorrelation in the c
7In an initial specification of the model we also included fat content o
f
the non-labelled meal alternatives, as well as calorie content of both the
healthy labelled and non-labelled meal alternatives, as explanatory
variables. We did so based on the idea that both calories and fat could
p
ositively impact tas
t
e. Further, calorie and fat content can also be seen
as control variables for the healthiness of the non-labeled alternatives.
However, t-tests implied that in no case could we reject the null hy-
p
othesis that these variables had no impact on healthy labeled meal
sales. We used an F-test to examine if the variables as a group contrib-
uted to the explanatory power of the model, but could not reject the null
hypothesis that they did not; F(5, 11) = 0.45, p-value = 0.806. We de-
cided not to include calories and fat content of substitute meals in the
model.
5In Dietist XP, portion sizes are generally based on portions consumed,
not portions served. The nutritional values found in our data are there-
fore generally smaller than nutritional values calculated based on por-
tions served at restaurants. The nutritional content of the meals is also
subject to uncertainties, since the nutritional contents have been calcu-
lated based on meal descriptions as found on the menu, where cooking
p
rocedures, portion sizes, etc., are unknown.
6Only during a couple of days of the study period did the healthy la-
beled meal appear last on the menu. We therefore merged 2nd and 3rd on
the menu into a single dummy variable.

L. THUNSTRÖM, J. NORDSTRÖM 4
Table 1. Descriptive statistics.
Variable Mean Std. Dev Min Max No. Obs
Daily portions sold of the healthy labelled meal 152.852 41.544 52 232 27
Daily portions sold in total 349.259 55.318 82 381 27
Daily market share of the healthy labelled meal 0.444 0.108 0.140 0.683 27
Healthy labelled meal attributesa
Red meat, healthy labelled meal 0.444 0.506 0 1 27
Poultry, healthy labelled meal 0.148 0.362 0 1 27
Seafood, healthy labelled meal 0.370 0.492 0 1 27
Vegetarian, healthy labelled meal 0.074 0.267 0 1 27
1st on menu 0.444 0.506 0 1 27
2nd or 3rd on menu 0.556 0.506 0 1 27
Fat, grams, healthy labelled meal 14.441 3.060 10.2 21.7 27
Substitute (non-healthy labelled) meal attributesa
Any substitute contains red meat 0.593 0.500 0 1 27
Any substitute contains poultry 0.111 0.320 0 1 27
Any substitute contains seafood 0.296 0.465 0 1 27
Any substitute is vegetarian 0.815 0.396 0 1 27
Weekdaysb
Monday 0.222 0.424 0 1 27
Tuesday - Thursday 0.593 0.501 0 1 27
Friday 0.185 0.396 0 1 27
first model (dependent variable = daily units sold of the
healthy labeled meal): Chi2 = 4.906; Prob > Chi2 =
0.0268, and that we cannot reject the null in the second
model (dependent variable = daily market share of the
healthy labeled meal): Chi2 = 0.006; Prob > Chi2 =
0.9366. The first model was therefore estimated with
robust standard errors.
3. Results
The results from our empirical analysis are presented in
Tables 2 and 3.
3.1. Attributes of Healthy Labelled Meals
Tables 2 and 3 show that a healthy labelled meal that
contains poultry (chicken or turkey) is the big seller:
daily portions sold of the healthy labelled meal increases
by 55 meals if it contains poultry, and the market share
of the healthy labelled meal increases by 13 percent,
compared to if the healthy labelled meal contains seafood.
The healthy labelled meal also benefits from red meat: a
healthy labelled meal that contains red meat sells 33
more meals, and increases its market share by 11 percent,
compared to a healthy labelled meal that contains sea-
food. Vegetarian healthy labelled meals sell the worst: if
the healthy labelled meal is vegetarian, both sales and the
market share of the healthy labelled meal drop substan-
tially: daily sales decrease by 75 meals and the market
share drops by 31 percent, compared to if the healthy
labelled meal contains seafood.
The substantial positive impact of healthy labelled
meal sales from poultry and red meat lend support to our
Hypothesis 1.
Our finding of the impact from fat on sales of the
healthy labelled meal is mixed. Fat per portion seems to
positively impact the number of healthy labelled meals
sold per day, i.e. within the range of fat allowed in
healthy labelled meals, people seem to appreciate more
fat in health labelled meals: Table 2 shows that if the fat
content increases by 1 gram, 6 more healthy labelled
meals are sold per day. However, fat seems to have no
impact on the market share of the healthy labelled meal:
as shown by Table 3, the coefficient for fat content is
both small and not statistically significant.
3.2. The Nudge
Nudging, by displaying the healthy labelled meal on top
of the menu, does not seem to impact sales of the healthy
labelled meal or its market share: the coefficient for the
dummy variable that indicates the healthy labelled meal
being displayed second or third on the menu is both
small and not statistically significant, as shown by both
Tables 2 an d 3. Based on t-tests, we can therefore not
reject the null hypothesis that there is no difference in
sales of healthy labelled meals between those displayed
at the top of the menu and those displayed second or last
on the menu.
D
isplaying the healthy labelled meal at the top of the

L. THUNSTRÖM, J. NORDSTRÖM 5
Table 2. OLS regression results of determinants of healthy labelled meal sales.
Variable Coefficient s.e. p-value
Dependent variable: portions sold of healthy labelled meal
Constant 1.834 1.526 0.230
Healthy labelled meal characteristics
Red meat 32.620*** 13.891 0.032
Poultry 55.322*** 15.145 0.002
Vegetarian −75.942** 28.188 0.016
2nd or 3rd on menu 1.754 9.484 0.856
Fat content 6.311*** 2.012 0.006
Substitute (non-healthy) meal characteristics
Red meat 37.706** 14.943 0.023
Poultry 26.843* 14.852 0.090
Vegetarian 11.994 14.823 0.430
Weekdays
Tuesday - Thursday −34.653* 17.771 0.069
Friday −75.357*** 16.078 0.000
No of obs: 27, R-squared = 0.8360. Superscript “*” indicates the significance level at which the null hypothesis of a coefficient equal to zero can be rejected.
*significance level < 0.10, **significance level < 0.05, and ***significance level < 0.01.
Table 3. OLS regression results of determinants of the share of healthy labelled meal sales, of total meal sales.
Variable Coefficient s.e. p-value
Dependent variable: market share of healthy labelled meal
Constant 0.415*** 0.123 0.004
Healthy labelled meal characteristics
Red meat 0.109** 0.042 0.019
Poultry 0.132** 0.056 0.031
Vegetarian −0.305*** 0.084 0.002
2nd or 3rd on menu −0.010 0.081 0.904
Fat content 0.007 0.006 0.263
Substitute (non-healthy) meal characteristics
Red meat 0.022 0.039 0.585
Poultry −0.042 0.048 0.390
Vegetarian −0.136** 0.052 0.019
Weekdays
Tuesday - Thursday −0.001 0.092 0.992
Friday −0.012 0.048 0.800
menu does not seem to impact its sales. Our results
therefore do not lend support to our Hypothesis 2.
3.3. Control Variables
Sales of the healthy labelled meal seem to benefit from
its substitutes containing popular attributes as well, such
as red meat and poultry. However, the results in Table 3
imply that the healthy labelled meal does not gain market
shares if its substitute meals contain red meat or poultry,
suggesting that the increase in sales reported in Table 2
is a result from overall sales increasing due to non-la-
belled meals that contain red meat or poultry, compared
to if they contain seafood. The non-healthy labelled meals
that seem to compete the most with healthy labelled
meals are vegetarian meals. If the non-labelled meals con-
tain a vegetarian meal, the market share of the healthy
labelled meal drops by 13.6 percent.
4. Discussion
From Table 1, we know that the most commonly served
healthy labelled meal is a healthy labelled meal contain-
ing red meat, despite our finding that healthy labelled
meals with poultry sell better: 44 percent of healthy la-
belled meals contain red meat versus 15 percent that
contain poultry. Restaurant management is likely to
know that poultry meals are their best sellers, so why are

L. THUNSTRÖM, J. NORDSTRÖM 6
healthy labelled meals that contain red meat more com-
mon than those containing poultry? For one, overall
profit of the restaurant may not be maximized by maxi-
mizing sales of the healthy labelled meal. Also, poultry
in healthy labelled meals may be a more expensive input
than red meat. Vegetarian healthy labelled meals are un-
common, though, which is in line with our finding that
vegetarian healthy labelled meals are a hard sell. Table 1
shows that only 7 percent of the healthy labelled meals
served up during the study period were vegetarian.
How does meal attributes affect sales of healthy la-
belled meals compared to sales of conventional meals?
Comparing our findings to the results in [3] we find that
the top-selling sources of protein are the same for healthy
labelled meals and conventional meals, but the impact on
healthy labelled meal sales is greater. For instance,
Thunström and Nordström, [3], report that general meal
sales increase by 41 meals if the meal contains poultry
instead of seafood, and by 25 meals if the meal contains
red meat. The corresponding numbers for the healthy
labelled meal is 55 and 33 meals. This difference in sales
increases between conventional meals and healthy la-
belled meals from adding poultry or read meat to the
meals is substantial in real terms, but represents even
larger differences in percentage terms. Thunström and
Nordström, [3], do not find a drop in sales for meals in
general resulting from the meal being vegetarian, which
differs from our results on sales of healthy labelled
meals.
Our finding that the order of display on the menu has
no impact on sales seems contradictory to previous re-
search. Thunström and Nordström, [3], find that the same
nudge impacts sales of conventional meals. Downs,
Loewenstein and Wisdom, [15], find that nudging in-
creases sales of healthy sandwiches. The difference in
results may be due to differences in salience between this
study and Downs, Loewenstein and Wisdom: they show
that sales of healthy sandwiches increase if healthy
sandwiches are displayed on the front of a menu, while
regular/unhealthy sandwiches are displayed on subse-
quent pages. Menu nudging in their experiment therefore
imposes an additional search cost on regular/unhealthy
sandwiches (turning the page), compared to nudging in
our experiment where all meals are displayed together
with the healthy meal, with the healthy meal at the top of
the menu.
5. Conclusions
In this paper, we use a lunch restaurant field experiment
to analyze the impact on sales and the market share of a
healthy labeled meal from meal attributes, and from a
“nudge”—displaying the healthy labeled meal at the top
of the menu.
We find that attributes of the healthy labelled meal,
especially poultry and red meat, have a strong impact on
both sales of the healthy labelled meal and their market
share: by changing the composition of the healthy la-
belled meal, sales of the healthy labelled meal may in-
crease by 55 units (where 55 units is equal to 36 percent
of average healthy labelled meal sales), and the market
share of the healthy labelled meal may increase by 13
percentage points. We find no impact on sales or the
market share of the healthy labelled meal from the nudge
used in this study.
Our results imply that designing healthy labelled meals
to contain attributes generally preferred by consumers
(i.e. desired in both non-healthy and healthy meals) may
significantly impact healthy food choices. Regulating the
content of the healthy meal supply may therefore be im-
portant for agents that aim to create consumer incentives
to choose healthy meals, such as restaurant managers,
school board members or other policy makers. For our
sample, sales of healthy meals benefit from the same
ingredients as conventional meals—poultry and red meat
—and lean versions of traditional meals are the top sell-
ers. In other words, a successful strategy for increasing
healthy meal consumption may be to supply healthy
meals that mimic popular conventional meals, using
cooking techniques and ingredients that reduce the num-
ber of calories and nutrients often over consumed (un-
healthy fats, salt, sugar, etc.).
Our findings are encouraging, since results from pre-
vious research that evaluates alternate policy measures
designed to encourage healthy food choices are some-
what disappointing. Providing nutritional information
(e.g. menu labelling) seems to have limited impact on
food choices [2,3,15,19-23], as do politically feasible
food tax reforms [5-8,24].
The large impact on healthy labeled meal sales from
meal attribute manipulation that we find in this study is
especially encouraging given the context of the field ex-
periment. First, the food analyzed here is prepared lunch
meals away from home. Food away from home has been
found to be one of the main causes of the increase in
obesity and overweight [25-27], and of meals consumed
away from home, lunch meals have been found to have
the greatest impact on body weight [28]. Second, the
customer base of the field experiment restaurant consists
of consumer groups that generally show less of an inter-
est in healthy eating: primarily male and blue-collar
workers.
The food attributes that appeal the most to consumers
are likely to be context dependent, though, and may also
vary over consumer groups. A question for future re-
search is therefore how healthy meals and other foods
(e.g., snacks) can be composed in order to encourage
healthy food consumption in different contexts and over
consumer groups. We also encourage future research to

L. THUNSTRÖM, J. NORDSTRÖM 7
formally analyze the impact of policies that manipulate
food supply versus policy measures that entail informa-
tion provision or tax reforms, as well as the impact of
combinations of these measures, e.g. subsidies of healthy
meals that contain preferred meal attributes. Finally, the
nudge examined in this study seems to have no impact on
healthy labeled meal consumption, even though the same
nudge impacts consumption of conventional meals (see
[3]). Promotion of healthy meal choices may require
stronger nudges than simply placing the healthy labeled
meal at the top of the menu. Future research may analyze
how nudging can be designed to impact healthy labeled
meal choices. For instance, does effective nudging re-
quire that healthy meal alternatives are the “default op-
tion”, with a high level of salience and search costs asso-
ciated with finding non-healthy meal alternatives (e.g.,
turning the menu, or even asking for a separate, “non-
healthy”, menu—see [15])?
6. Acknowledgements
Financial support is gratefully acknowledged from the
Swedish Council for Working Life and Social Research.
We thank Eurest Dining Services for enabling this study
by providing data.
REFERENCES
[1] C. Lee, P. Linneman, B. Elbel, R. Kersh, V. L. Brescoll
and L. B. Dixon, “Calorie Labeling and Food Choices: A
First Look at the Effects on Low-Income People in New
York City,” Health Affairs, Vol. 28, No. 6, 2009, pp.
1110-1121.
http://dx.doi.org/10.1377/hlthaff.28.6.w1110
[2] L. J. Harnack, S. A. French, J. M. Oakes, M. T. Story, R.
W. Jeffery and S. A. Rydell, “Effects of Calorie Labeling
and Value Size Pricing on Fast Food Meal Choices: Re-
sults from an Experimental Trial,” International Journal
of Behavioral Nutrition and Physical Activity, Vol. 5, No.
63, 2008.
[3] L. Thunström and J. Nordström, “Does Easily Accessible
Nutritional Labelling Increase Consumption of Healthy
Meals Away from Home? A Field Experiment Measuring
the Impact of a Point-of-Purchase Healthy Symbol on
Lunch Sales,” Food Economics—Acta Agriculturae Scan-
dinavica, Vol. 8, No. 4, 2011, pp. 200-207.
[4] M. K. Vadiveloo, L. B. Dixon and B. Elbel, “Consumer
Purchasing Patterns in Response to Calorie Labeling Leg-
islation in New York City,” International Journal of Be-
havioral Nutrition and Physical Activity, Vol. 8, 2011, p.
51. http://dx.doi.org/10.1186/1479-5868-8-51
[5] H. H. Chouinard, D. E. Davis, J. T. LaFrance and J. M.
Perloff, “Fat Taxes: Big Money for Small Change,” Fo-
rum for Health Economics & Policy, Vol. 10, No. 2, 2007.
http://dx.doi.org/10.2202/1558-9544.1071
[6] J. Nordström and L. Thunström, “The Impact of Tax Re-
forms Designed to Encourage a Healthier Grain Con-
sumption,” Journal of Health Economics, Vol. 28, No. 3,
2009, pp. 622-634.
http://dx.doi.org/10.1016/j.jhealeco.2009.02.005
[7] L. M. Powell and F. J. Chaloupka, “Food Prices and Obe-
sity: Evidence and Policy Implications for Taxes and
Subsidies,” Milbank Quarterly, Vol. 87, No. 1, 2009, pp.
229-257.
http://dx.doi.org/10.1111/j.1468-0009.2009.00554.x
[8] S. Smed, J. D. Jensen and S. Denver, “Socio-Economic
Characteristics and the Effect of Taxation as a Health
Policy Instrument,” Food Policy, Vol. 32, No. 5-6, 2007,
pp. 624-639.
http://dx.doi.org/10.1016/j.foodpol.2007.03.002
[9] N. Wrigley, D. Warm and B. Margetts, “Deprivation, Diet,
and Food-Retail Access: Findings from the Leeds ‘Food
Deserts’ Study,” Environment and Planning A, Vol. 35,
No. 1, 2003, pp. 151-188.
http://dx.doi.org/10.1068/a35150
[10] H. Lee, “The Role of Local Food Availability in Explain-
ing Obesity Risk among Young School-Aged Children,”
Social Science & Medicine, Vol. 74, No. 8, 2012, pp. 1193-
1203. http://dx.doi.org/10.1016/j.socscimed.2011.12.036
[11] M. Lennernäs, C. Fjellström, W. Becker, I. Giachetti, A.
Schmitt, A. Remaut de Winter and M. Kearney, “Influ-
ences on Food Choice Perceived to Be Important by Na-
tionally-Representative Samples of Adults in the Euro-
pean Union,” European Journal of Clinical Nutrition, Vol.
51, Suppl. 2, 1997, pp. S8-S15.
[12] K. Glanz, M. Basil, E. Maibach, J. Goldberg and D. Sny-
der, “Why Americans Eat What They Do: Taste, Nutri-
tion, Cost, Convenience, and Weight Control Concerns as
Influences on Food Consumption,” Journal of the Ameri-
can Dietetic Association, Vol. 98, No. 10, 1998, pp. 1118-
1126. http://dx.doi.org/10.1016/S0002-8223(98)00260-0
[13] H. M. Blanck, A. L. Yaroch, A. A. Atienza, S. L. Yi, J.
Zhang and L. C. Mâsse, “Factors Influencing Lunchtime
Food Choices among Working Americans,” Health Edu-
cation & Behavior, Vol. 36, No. 2, 2009, pp. 289-301.
http://dx.doi.org/10.1177/1090198107303308
[14] R. H. Thaler and C. R. Sunstein, “Libertarian Paternal-
ism,” American Economic Review, Vol. 93, No. 2, 2003,
pp. 175-179.
http://dx.doi.org/10.1257/000282803321947001
[15] J. S. Downs, G. Loewenstein and J. Wisdom, “Strategies
for Promoting Healthier Food Choices,” American Eco-
nomic Review, Vol. 99, No. 2, 2009, pp. 159-164.
http://dx.doi.org/10.1257/aer.99.2.159
[16] B. D. Ellison, J. L. Lusk and D. W. Davis, “The Value
and Cost of Restaurant Calorie Labels: Results from a
Field Experiment,” 2012 AAEA/EAAE Food Environment
Symposium, 30-31 May 2012, Boston.
[17] S. Burton, E. Creyer, J. Kees and K. Huggins, “Attacking
the Obesity Epidemic: The Potential Health Benefits of
Providing Nutrition Information in Restaurants,” Ameri-
can Journal of Public Health, Vol. 96, No. 9, 2006, pp.
1669-1675. http://dx.doi.org/10.2105/AJPH.2004.054973
[18] A. Drewnowski, “Energy Density, Palatability, and Sati-
ety: Implications for Weight Control,” Nutrition Reviews,
Vol. 56, No. 12, 1998, pp. 347-353.

L. THUNSTRÖM, J. NORDSTRÖM 8
http://dx.doi.org/10.1111/j.1753-4887.1998.tb01677.x
[19] Y. H. Chu, E. A. Frongillo, S. J. Jones and G. L. Kaye,
“Improving Patron’s Meal Selection through the Use of
Point-of-Selection Nutrition Labels,” American Journal
of Public Health, Vol. 99, No. 11, 2009, pp. 2001-2005.
http://dx.doi.org/10.2105/AJPH.2008.153205
[20] L. J. Harnack and S. A. French, “Effect of Point-of Pur-
chase Calorie Labeling on Restaurant and Cafeteria Food
Choices: A Review of the Literature,” International Jour-
nal of Behavioral Nutrition and Physical Activity, Vol. 5,
2008, p. 51.
[21] C. A. Perlmutter, D. D. Canter and M. B. Gregoire, “Pro-
fitability and Acceptability of Fat- and Sodium-Modified
Hot Entrées in a Worksite Cafeteria,” Journal of the Ame-
rican Dietetic Association, Vol. 97, No. 4, 1997, pp. 391-
395. http://dx.doi.org/10.1016/S0002-8223(97)00097-7
[22] E. Pulos and K. Leng, “Evaluation of a Voluntary Menu-
Labeling Program in Full-Service Restaurants,” American
Journal of Public Health, Vol. 100, No. 6, 2010, pp. 1035-
1039. http://dx.doi.org/10.2105/AJPH.2009.174839
[23] C. A. Roberto, P. D. Larsen, H. Agnew, J. Baik and K. D.
Brownell, “Evaluating the Impact of Menu Labelling on
Food Choices and Intake,” American Journal of Public
Health, Vol. 100, No. 2, 2010, pp. 312-318.
http://dx.doi.org/10.2105/AJPH.2009.160226
[24] J. Nordström and L. Thunström, “Economic Policies for
Healthier Food Intake: The Impact on Different House-
hold Categories,” European Journal of Health Economics,
Vol. 12, No. 2, 2011, pp. 127-140.
http://dx.doi.org/10.1007/s10198-010-0234-6
[25] J. K. Binkley, “The Effect of Demographic, Economic
and Nutrition Factors on the Frequency of Food Away
from Home,” Journal of Consumer Affairs, Vol. 40, No. 2,
2006, pp. 372-391.
http://dx.doi.org/10.1111/j.1745-6606.2006.00062.x
[26] S.-Y. Chou, M. Grossman and H. Saffer, “An Economic
Analysis of Adult Obesity: Results from the Behavioural
Risk Factor Surveillance System,” Journal of Health Eco-
nomics, Vol. 23, No. 3, 2004, pp. 565-587.
http://dx.doi.org/10.1016/j.jhealeco.2003.10.003
[27] I. Rashad, “Structural Estimation of Caloric Intake, Exer-
cise, Smoking, and Obesity,” Quarterly Review of Eco-
nomics and Finance, Vol. 46, No. 2, 2006, pp. 268-283.
http://dx.doi.org/10.1016/j.qref.2005.11.002
[28] G. Kyureghian, R. M. Nayga, G. C. Davis and B.-H. Lin,
“Food away from Home Consumption and Obesity: An
Analysis by Service Type and by Meal Occasion,” 2007
Annual Meeting, 29 July-1 August 2007, Portland.