Working Paper Series
Health Halos: How Nutrition Claims Influence
Food Consumption for Overweight and Normal
Faculty & Research
Health Halos: How Nutrition Claims Influence Food Consumption for
Overweight and Normal Weight People
November 1, 2005
∗ Brian Wansink is the John S. Dyson Professor of Marketing and of Nutritional Science in the
Applied Economics and Management Department of Cornell University, 110 Warren Hall,
Ithaca NY 14853-7801, 607-255-5024 (Wansink@Cornell.edu). Pierre Chandon is Assistant
Professor of Marketing at INSEAD, Boulevard de Constance, 77300 Fontainebleau, France;
+33 1 60 72 49 87 (phone), +33 1 60 74 61 84 (fax), firstname.lastname@example.org. Thanks to
Erin Alexander at Kraft and Brian Zaff at Masterfoods for their cooperation and insights.
Thanks also to Karla Mohnke and Cristina Perez for their help with this research.
Health Halos: How Nutrition Claims Influence Food Consumption for
Overweight and Normal Weight People
How do the nutrition claims on packaged goods influence how much a person eats? In an
era of increasing obesity and increasing threats of legislation, regulation, and boycotts, this
question is a concern both to responsible packaged goods companies and to regulatory
agencies. To address this question, we develop and test a framework that shows how relative
nutrition claims (such as “low-fat”) can increase food intake by increasing perceptions of
appropriate serving size and decreasing anticipation of consumption guilt. Three studies show
that relative nutrition claims can lead all consumers to overeat, but this becomes more
exaggerated for overweight consumers than for those with a normal weight. Further results
show that providing objective serving size information eliminates the overeating that is
encouraged by low-fat nutrition claims, but only among normal weight consumers. With
consumer welfare and corporate profitability in mind, win-win labeling insights are suggested
for manufacturers and public policy officials.
“At one point, the most commonly asked question to the
Nabisco Consumer Hotline was whether [low-fat]
Snackwells could be purchased in larger packages,”
Food companies are on trial for contributing to the growing problem of obesity in the
United States and abroad. They have been threatened by taxes, fines, restrictions, legislation,
and the possibility of being “the tobacco industry of the new millennium” (Nestle 2002). In
particular, food labeling is of critical concern among regulators such as the U.S. Food and
Drug Administration (FDA). Although past research has effectively examined how health
claims and nutrition labels influence health beliefs and purchase intentions (Andrews et al.
1998; Balasubramanian and Cole 2002; Ford et al. 1996; Keller et al. 1997; Kozup et al. 2003;
Moorman 1990; Moorman 1996; Moorman et al. 2004), the pressing issue for the FDA is how
relative nutrition claims (such as “low-fat” or “reduced calories”) influence food intake on a
single eating occasion and on what companies can do about it (Food and Drug Administration
2003). In particular, the FDA is concerned that relative nutrition claims may lead to the over-
consumption of nutrient-poor and calorie-rich snack foods by the 65% of US consumers who
are already overweight1 (Hedley et al. 2004).
Although no food company would want to discourage consumers from purchasing their
products, it may be in their interest to use relative nutrition claims to help consumers better
control how much they consume on a single eating occasion (Wansink and Huckabee 2005).
Consider indulgent, hedonic foods, such as candies and snacks. Single occasion over-
consumption of these foods can not only lead to weight gain but can also lead to rapid satiation
and to delayed repurchasing (Inman 2001). Over the longer term, helping consumers better
control their consumption could also help promote more favorable attitudes toward the brand
and company. This may result in what Rothschild (1999) refers to as a “win-win” policy-
sensitive solution for both companies and consumers.
Because of these concerns with food labels and claims, the FDA raised three specific
questions for companies like Kraft Foods and M&M/Mars (now Masterfoods) to answer: 1)
How do relative nutrition claims influence how much people consume on a single eating
occasion? 2) Do relative nutrition claims influence overweight consumers differently than
consumers with a normal weight? and 3) Can serving size information eliminate this bias? To
help managers address these questions, we develop a general framework suggesting that
relative nutrition claims increase consumption because (1) they increase perceptions of the
appropriate serving size and (2) they reduce anticipated consumption guilt. This framework
also suggests that relative nutrition claims have less of an influence on consumption when
objective serving size information is provided and when there are high levels of consumption
guilt, which depends on both the food-type (hedonic vs. utilitarian) and the consumer
(overweight vs. normal weight).
We test this framework in three studies, in which we show the process by which low-fat
nutrition labels lead consumers to increase their snack food consumption. Study 1 establishes
the main finding of the research by showing that visitors to a university-wide open house eat
more chocolate candies when they are labeled as “low-fat” than when they are labeled as
“regular,” particularly if they are overweight. Study 2 investigates the two hypothesized
mediators and shows that low-fat nutrition claims lead all consumers to lower their anticipated
consumption guilt and to increase the amount they believe to be an appropriate serving size.
1 Following the guidelines of the World Health Organization, persons are classified as “normal weight” if their
body mass index (BMI) is between 18 kg/m² and 25 kg/m², as “overweight” if their BMI is greater than 25 kg/m²,
and as “obese” if their BMI is greater than 30 kg/m². BMI is computed as the ratio of weight, measured in
kilograms, to squared height, measured in meters.
Study 3 uses a movie theatre environment to show how relative nutrition claims and objective
serving size information jointly influence the consumption of granola by overweight and
normal weight consumers. In the last section, we discuss the implications of our findings for
nutrition research, responsible food manufacturers, and public policy officials.
How Relative Nutrition Claims Influence Consumption
When determining how much to eat, labels offer both objective and subjective
consumption cues. Objective consumption cues, such as those provided by the serving size
information on a label, explicitly suggest some guidance about how much should be consumed
on a single occasion (Caswell and Padberg 1992). Subjective consumption cues, such as those
provided by endorsed2 or relative nutrition claims, do not specify an exact serving size, but
they may influence serving size inferences (how much a person thinks is appropriate to eat)
and how guilty they should feel about it. Figure 1 shows the predicted relationship of these
variables on food consumption, and it foreshadows how these relationships may vary across
foods and across people.
--- Insert Figure 1 about here ---
Serving Sizes Inferences
Inferences made about serving sizes are in some ways similar to inferences made in daily
conversations. Because of conversational norms, consumers first assume that the information
communicated to them (such as in a conversation or on a label) is potentially informative and
relevant to their decisions (Grice 1975; Schwarz 1996). Consumers therefore use the
information provided to make inferences about missing attributes that are important for their
decision. For example, Broniarczyk and Alba (1994) found that people make inferences about
important missing attributes based on their intuitive beliefs about the relationship between the
attributes that are missing and those that are available, even in the face of contradictory cues.
Similar inferential mechanisms are well documented in several nutrition studies that have
shown that consumers inappropriately generalize health claims (Garretson and Burton 2000;
Ippolito and Mathios 1991; Kozup et al. 2003; Moorman and Matulich 1993). For example,
Andrews, Netemeyer, and Burton (1998) showed that consumers erroneously infer that foods
low on cholesterol are also low on fat. Similarly there is strong evidence that many consumers
erroneously believe that “low-fat” nutrition claims indicate fewer calories (National Institutes
of Health 2004). They do not realize that when determining whether “low-fat” nutrition claims
are appropriate, the FDA3 only looks at the amount of fat, not at the number of calories (see
---- Insert Table 1 about here ---
Supporting the argument that nutrition claims influence perceived serving sizes, studies
have shown that consumers’ perception of serving size are ambiguous and flexible. Of course,
with discretely packaged items—such as a 12-ounce can of a soft drink or a single-serving
candy bar—the intended serving size is obvious. In many other contexts—such as with large
bags of M&Ms, a large bag of granola, or a full 24-ounce serving bowl of macaroni and
2 Endorsed nutrition claims are those that have been tested, proven, and ratified by an endorsing entity such as the
Food and Drug Administration or the American Cancer Association. Endorsed claims specifically note the effect
a targeted ingredient has on health (Geiger 1998). These include links between soy and heart disease, between
folic acid and birth defects, and between fiber and cancer.
3 While it is true that fat contains more calories per grams than either carbohydrates or proteins, low-fat foods
typically compensate the reduction in fat by an increase in carbohydrate (Burros 2004). As a result, foods labeled
as “low-fat” do not, on average, contain significantly fewer calories per serving than foods without this label
(National Institutes of Health 2004). For example, the low-fat Snackwells cookies developed by Nabisco contain
less fat but not fewer calories than regular cookies because the fat had been replaced by high-calorie starches and
cheese—the appropriate serving size is more ambiguous. In the absence of salient,
unambiguous serving size information, an appropriate serving size is whatever amount a
person thinks should be the consumption norm for that particular food and situation (Wansink
Conversational norms, inferential mechanisms, and ambiguity about serving sizes lead us
to predict that relative nutrition claims could create misleading “health halos” that lead
consumers to believe that the appropriate amount to consume is higher when the food is
described as being lower in fat, calories, or carbohydrates. We therefore hypothesize that a
relative nutrition claim communicated by a “low-fat” label increases food intake per
consumption occasion because it increases consumers’ estimations of the appropriate
consumption volume (their perception of serving size) and because it decreases consumers’
estimations of the calorie density of the food (the number of calories per volume) and of their
associated consumption guilt.
Anticipated Consumption Guilt
Food decisions often involve a trade-off of two goals: The hedonic goal of short-term taste
gratification and the utilitarian goal of long-term health preservation and enhancement (Dhar
and Simonson 1999; Shiv and Fedorikhin 1999 ). Whichever of these two goals is most salient
will influence how consumers choose between hedonic and utilitarian foods and how they
determine how much they will eat. For example, Fishbach and Dhar (2005) showed that
making the hedonic goal more salient leads them to choose a candy bar over an apple.
For some consumers, fulfilling their hedonic goal rather than their utilitarian goals can lead
to strong feelings of guilt (Baumeister 2002; Kivetz and Keinan forthcoming). For example,
Kivetz and Keinan (forthcoming) found that consumers making hedonic choices (e.g., pleasure
rather than work) exhibit more guilt than consumers making utilitarian choices (e.g., work
rather than pleasure). Similarly, feelings of guilt occur when choosing taste over health.
Consider a restaurant’s dessert menu. Okada (2005) found that people eating at a restaurant
were more likely to order “Cheesecake deLite,” a low-fat dessert, than “Bailey’s Irish Cream
Cheesecake,” a high-fat dessert, when they were presented side-by-side on the menu, but they
preferred the high-fat dessert to the low-fat dessert when each item was presented alone. She
attributes these findings to the fact that presenting both options together increased the feelings
of guilt associated with the low-fat option.
We therefore expect that another way in which relative nutrition claims (such as low-fat
labels) increase consumption is by reducing a consumer’s anticipated consumption guilt. This
hypothesis is supported by several findings showing that low-fat products are considered more
hedonic than high-fat products. For example, Wertenbroch (1998) found that consumers
expect better taste when potato chips are labeled “25% fat” than when they are labeled as
“75% lean” (a frame which is known to reduce perception of fat). Similarly, Raghunathan,
Walker and Hoyer (2005) showed that increasing the perceived amount of “bad” fat increases
consumers’ expectations about the taste of a product.
People are more likely to feel guilty about overeating an indulgent, hedonic food (such as
chocolate) than they would about eating a food they saw as relatively more healthy (such as
granola). Yet the guilt of overeating is likely to be a more powerful motivator to some people
than others. One of the individual characteristics associated with feelings of consumption guilt
may be related to an individual’s weight. Studies have shown that one’s Body Mass Index
(BMI) is strongly associated with dietary disinhibition. That is, overweight people have a
greater tendency to lose control over eating and to have higher levels of hunger and lower
levels of consumption guilt (Hays et al. 2002). For example, the Three Factor Eating
Questionnaire, the main survey instrument for measuring dietary disinhibition, includes items
such as “life is too short to worry about dieting” (Stunkard and Messick 1985).
Implications for Reducing the Effects of Low-fat Labels on Consumption
Because of the ambiguity of sensory experience (Deighton 1984; Ha and Hoch 1989), it is
unlikely that consumers will realize that they are over-consuming foods with relative nutrition
claims. Indeed, a wide range of studies have shown that consumers are unable to monitor the
number of calories that they consume (Livingstone and Black 2003). As a result, we expect
that consumers will not be aware that their consumption has increased because of “low-fat”
Will serving size information reduce this tendency to over eat because of a health halo?
When objective serving size information is provided, consumers do not need to rely on low-fat
information to infer serving size. Support for this hypothesis comes again from studies of
conversational norms, which suggest that communication from institutional sources is
generally perceived as truthful in the absence of better information or disconfirming evidence
(Grice 1975; Schwarz 1996). Accordingly, although most consumers are skeptical of health
claims, they generally believe nutrition facts, labels, and nutrition claims (Wansink, 2004).
This hypothesis is also supported by self-generated validity research, which shows that inputs
that are perceived to be only moderately diagnostic are ignored when more diagnostic ones can
be retrieved (Feldman and Lynch 1988).
Yet the influence of serving size should vary across people and across foods. Recall that
for more utilitarian foods (such as granola), we expect that the tendency to consume more food
with low-fat labels will be similar for overweight and normal weight consumers. This is
because all consumers will anticipate the same low levels of guilt from consuming products
that are seen as generally good for them to eat. In contrast, low-fat labeling should have a
stronger influence on the consumption of hedonic foods (such as chocolate candies) for
overweight consumers than for normal weight consumers. This is because consumers with a
normal weight will anticipate higher feelings of guilt from consuming candies, regardless of
whether it is labeled as low-fat or not. As a result, objective serving size information should be
more effective in reducing the effects of relative nutrition claims for normal weight consumers
than for overweight ones.
Study 1: Do “Low-Fat” Nutrition Labels Increase Consumption?
Study 1 examines whether low-fat nutrition claims increase the actual and estimated
consumption of hedonic chocolate candies by overweight and normal weight consumers. To
achieve this, we asked adults participating in a university open-house to serve themselves
unusual colors of M&M’s (gold, teal, purple, and white) which were very clearly labeled as
either “New Colors of Regular M&Ms” (regular label condition) or as “New Low-Fat” M&Ms
(low-fat label condition), and we measured how many calories of M&Ms they served
themselves and how many they thought they served.
Participants were incoming students and their families who were visiting a university
open-house to look at information, videos, and interactive displays related to food science and
human nutrition. The open-house operated from 9:00 to 4:00 on a Friday and Saturday and
sign-in records indicated that at least 361 people visited the area in which these displays were
located. As families entered the display area, they were greeted by a research assistant who
welcomed them and provided a brief overview of the display area. Following this, each family
was taken to one of two gallon-size serving bowls of M&Ms that had been placed on either
side of the entrance but out of sight of each other. Each family member was given a 16-ounce
bowl and sanitary gloves and told they could help themselves to the M&Ms. The gallon bowls
were placed on separate tables and participants could only see the nutrition labels on the M&M
bowl to which they had been led. To ensure that participants would pay attention to the
nutrition label, bowls were filled with unusual colors of M&Ms (gold, teal, purple, and white).
Participants in the regular label condition saw a gallon bowl with a professionally-designed
8x5-inch label which read “New Colors of Regular M&M’s.” Participants in the low-fat label
condition saw a gallon bowl with a similar label which read “New Low-fat M&M’s” (no such
product is currently available on the market).
Immediately after these open-house guests had selected the M&Ms they had wanted, the
research assistants asked them if they wanted to be involved in a series of demonstrations and
short surveys about how consumers make choices and decisions. Of the 361 visitors, 293
claimed to be of legal age (18 and over) and 269 of these eligible adults agreed to participate
(91.8%). The research assistant then asked permission to weigh their plastic bowls (which
contained their M&Ms) and handed them a brief survey which asked their age, gender, height,
weight, nutrition knowledge, and familiarity with M&M’s. Nutrition knowledge and product
familiarity were self-assessed on 3-point scales (from low to high). After they completed the
questionnaire, the research assistant asked each visitor to estimate how many total calories of
M&Ms they had served themselves. At this time, each person was told that the M&Ms they
had selected were actually regular (full calorie) M&Ms. They were then thanked, given a
bookmark, a refrigerator magnet, and a nutrition tip-sheet. Most people then took 15 to 20
minutes to read the materials and watch the videos before exiting out of a far door. Because
they had been told that they could not eat food outside of this display area, all but seven
(97.3%) finished their M&Ms in the display area.
Following the analysis guidelines of the World Health Organization, we classified
participants as overweight (n = 103) or normal weight (n = 166) depending on whether their
BMI was above or below 25 kg/m². To facilitate the comparison between actual and estimated
consumption, we converted the weight measures of the M&Ms into calories using the
information available in the manufacturer’s website. We used ANCOVAs to analyze estimated
and actual consumption using the nutritional label (low-fat versus regular), the individual’s
body mass (below vs. above 25 kg/m²), and their interaction. Each participant’s gender, age,
self-assessed nutrition knowledge, and familiarity with M&Ms were included as covariates.
--- Insert Figure 2 about here ---
Actual consumption. As predicted, low-fat labels increased consumption (F(1, 251) = 13.1,
p < .001). Participants ate 28.4% more M&M’s when they were labeled as “low-fat” (M = 244
cal) than when they were labeled as “regular” (M = 190 cal). Furthermore, overweight
participants took 16.7% more M&M’s (M = 237 cal) than normal weight participants (M = 203
cal), F(1, 251) = 4.3, p < .05. As expected, the interaction between low-fat labeling and body
mass was statistically significant (F(1, 251) = 3.9, p < .05). Low-fat labeling had a stronger
impact among overweight participants, whose intake increased by 90 calories (a 47%
increase), than among normal weight participants, whose intake increased by only 30 calories
(a 16% increase). Contrast tests show that the calorie increase due to low-fat labeling was
statistically significant among overweight participants (F(1, 91) = 9.2, p < .001) but not among
normal weight participants (F(1, 156) = 2.2, p = .13). None of the effects of the covariates
were statistically significant (p > .10).
Consumption estimation bias. As illustrated in Figure 2, calorie estimates were not
influenced by the nutrition labels and were only marginally influenced by a person’s body
mass. To directly examine the dissociation between actual and estimated calories, we
conducted an ANCOVA using consumption bias (the estimated minus the actual number of
calories) as the dependent variable and using the same factors and covariates as in the analysis
of actual consumption. On average, participants underestimated the number of calories of
M&M’s by 48% (F(1, 251) = 42.5, p < .001). Overweight and normal weight participants
similarly underestimated the number of calories contained in the M&M’s that they were about
to consume (F(1, 251) = 1.7, p = .21). The main effect of nutrition labels (F(1, 251) = 23.9, p <
.001) and its interaction with body mass (F(1, 251) = 4.5, p < .05) were both statistically
significant. That is, people who saw low-fat labels were more biased in their calorie estimates
(M = –132 cal) than those in the regular label condition (M = –81 cal), and overweight
participants were even more biased than normal weight participants (see Figure 2). The effects
of gender, self-assessed nutrition knowledge, and familiarity were not statistically significant
(respectively, F(1, 251) < .1, p = .71, F(1, 251) = 1.3, p = .25 ,and F(1, 251) = .5, p = 0.48),
but the magnitude of the consumption estimation bias decreased by 0.7 calories per year of age
(F(1, 251) = 3.9, p < .05), leading older participants (who tended to take smaller amounts) to
be more accurate than younger ones.
There are three key results from Study 1. First, it shows strong support for a health halo:
Participants consumed 28% more M&M’s when they were described as “low-fat” than when
they were described as “regular.” Second, the influence of low-fat labeling was stronger
among those who were overweight. Third, all participants strongly underestimated the number
of calories they consumed, and they were unaware that low-fat labeling influenced their
consumption. The magnitude of this calorie underestimation was particularly strong among
overweight participants, the people for whom underestimating calories is potentially the most
Because low-fat foods contain slightly fewer calories than regular versions, it would be
reasonable for a hungry calorie-counter to consume more candy when it is described as low-fat
than when it is not. The key question for public health is whether relative nutrition claims,
such as “low-fat,” lead people to increase consumption so much that it offsets the lower calorie
density of low-fat foods.
To determine this, we conducted a market survey of all brands of chocolate candies, bars,
cookies, milk drinks, and muffins with at least a 5% market share. We found 17 brands sold
with both a regular and a low-fat version. Serving sizes were nearly identical across versions (t
= 1.08, p = .30). On average, low-fat versions contained 59% less fat per serving than regular
versions (3.2 vs. 6.7 grams, t = 7.43, p < .001). However, they contained only 15% fewer
calories per serving than regular versions (140 vs. 170 cal, t = 4.79, p < .01). Assuming that
participants in Study 1 had eaten real “low-fat” M&M’s, and that they had an average level of
reduced fat and calories, participants would have consumed 48% less fat from the low-fat
M&M’s but would have consumed 9% more total calories. In reality, the increase in calories is
likely to be even higher because the ingredients used to replace fat tend to make people
hungrier (Nestle 2002).
Study 2: Do “Low-Fat” Nutrition Labels Influence Perceptions of Serving Size and
Study 1 established that relative nutrition claims, such as “low-fat” labels, can lead people
to increase their food consumption without realizing they are doing so. These results do not,
however, establish why this might occur, or why this tendency is so much stronger among
overweight people than among those with normal weight. In Study 2, we examine the role of
two possible mediators of this label-consumption relationship (perceived serving size and
anticipated consumption guilt), and we will do so for the hedonic foods (M&Ms) as well as for
the more utilitarian foods (granola).
We recruited 74 consumers on a major university campus for a study that they were told
dealt with visual illusions and volume perceptions. In exchange for free movie coupons, each
participant was told that we would ask them to rate familiar snack foods displayed in
transparent containers. Two transparent cylindrical measuring cups (20-ounce capacity) were
placed in each of two separate rooms. In each room, one cup contained 10 ounces of M&M’s
(1,380 cal) and the other contained 10 ounces of regular granola (1,330 cal). Half the
participants were assigned to the first room and saw measuring cups that were labeled as
“Regular M&M’s” or “Regular granola.” The other half of the participants were assigned to
the other room and saw measuring cups labeled as “Low-fat M&M’s” or “Low-fat granola.”
The two snacks were chosen based on pre-tests which indicated that, although both foods have
the similar calorie density (135-140cal/oz), granola is perceived as less hedonic, more
nutritious, and more healthy than M&Ms.
Each participant was given a two-page questionnaire. On the first page, they were asked to
evaluate serving sizes by a) estimating the number of ounces of the snack that would be
appropriate for a typical person to eat during a 90-minute movie, and by b) estimating the
number of ounces that would be appropriate for them to eat in the same situation. Calibrating
such estimates was aided by detailed ounce markings that were on the side of each measuring
cup. They were also asked to estimate the total number of calories contained in each container,
and to rate how guilty (1 = not guilty; 9 = guilty) they would feel after consuming a two-ounce
bowl of each food. The sequence of these four questions was systematically rotated across the
participants to avoid an order bias. On the back of the questionnaire, participants recorded
their gender, height, weight, and were asked to guess the purpose of the study. We then
thanked the participants and debriefed them. No one guessed the purpose of the study.
To increase the reliability of the serving size estimates, the two measures of serving sizes
were averaged and analyzed with a repeated-measure ANOVA using the two different foods as
a within-subject factor and the nutritional label, each individual’s BMI, and their interaction as
between-subjects factors. Using the World Health Organization BMI cutoff of 25 kg/m², 52
participants were classified as being normal weight and 16 were classified as being
overweight. As shown in Table 2, all participants who saw a food that was labeled as “low-fat”
believed that an appropriate serving size was 25.1% larger (F(1, 60) = 6.0, p < .01). There
were no differences between overweight and normal weight participants (F(1, 60) = .4, p =
.55). The effect of low-fat labels was consistent regardless of one’s BMI (F(1, 60) = .4, p =
.55). None of the interactions between food type (granola or M&Ms) and any of the between-
subjects factors was statistically significant (p > .20). On average, participants estimated that
the appropriate consumption amount was 1 oz higher in the low-fat label condition than in the
regular label condition. As shown in Table 2, the effects of low-fat labels was similar across
snacks (1.1 oz for M&M’s and 0.9 oz for granola) and across BMI groups (1.5 oz for normal
weight participants and 0.9 oz for overweight participants).
Estimates of calorie density followed the opposite pattern than serving size estimates.
Participants who saw a food that was labeled as “low-fat” believed it to be much lower in
calories (F(1, 69) = 4.1, p < .05). There were no differences between overweight and normal
weight participants F(1, 69) = 2.9, p = .10). The effect of low-fat labels was consistent
regardless of one’s BMI (F(1, 69) = 2.9, p = .10). None of the interactions between food type
(granola or M&Ms) and any of the between-subjects factors was statistically significant (p >
.20). As shown in Table 2, low-fat labels decreased calories estimates by an average of 260
calories and the amount of the reduction was similar across both M&Ms and granola (–292
versus –229 calories) and across both low- and high-BMI groups (–254 versus –272 calories).
---- Insert Table 2 about here ---
Did low-fat labeling lessen one’s anticipation of consumption guilt? In general, all
participants anticipated they would feel less guilty (M = 3.6 on a nine-point scale anchored at
1= “not guilty” and 9 = “guilty) in the low-fat label condition than in the regular label
condition (M = 4.6, F(1, 69) = 3.8, p < .05), but this was stronger for overweight participants.
As shown in Table 2, the interaction between food type and BMI was statistically significant
(F(1, 69) = 5.0, p < .05). M&M’s elicited more guilt among normal weight participants (M =
4.6) than among overweight participants (M = 3.2), F(1, 69) = 5.4, p < .05. In contrast, granola
elicited little guilt by either normal weight participants (M = 2.9) or overweight participants (M
= 2.5), F(1, 69) = .6, p = .44. Overall, these results show that low-fat labels reduced the guilt
everyone associated with consuming granola. With M&Ms, low-fat labels only made the
overweight participants feel less guilty.
Low-fat labels increase perceived serving sizes and anticipated guilt even when no snack is
consumed. This provides important evidence as to the mechanism of how low-fat foods might
influence food intake (the mediation analysis itself will be tested in Study 3). Low-fat labels
decrease the perceived calorie density of both foods, and they help explain why low-fat labels
make larger portions more appropriate: Consumers expect that low-fat M&M’s will contain
20% fewer calories per ounce than their regular counterparts and that low-fat granola will
contain 25% fewer calories per ounce than their regular counterparts. They therefore view that
comparable increases in serving size are justified for both foods when they are labeled as “low
fat” (21% for M&M’s and 18% for granola).
The patterns of results regarding feelings of guilt help explain why overweight participants
in Study 1 ate three times more M&Ms when they were labeled as low-fat. It suggests that
normal weight participants in Study 1 responded less strongly to low-fat labeling than
overweight participants because of their higher guilt—they knew it was still indulgently
hedonic. In addition, the social nature of Study 1 (in which participants served themselves in
the presence of the research assistant and of family members) may have exacerbated the guilt
of normal weight consumers, further reducing the amount of M&M’s they took and consumed.
If anticipated guilt influences how much one consumes of a product, we should find little
difference between overweight and normal weight consumers when they are eating a food they
both feel low levels of guilt about eating. We will examine this in Study 3 using low-fat
One purposeful limitation of Study 2 is that we did not measure consumption. This was
necessary in order to avoid contamination between the measures of consumption and measures
of the mediators. For example, a consumer might exaggerate serving sizes to justify high
consumption. This raises the question of whether providing objective serving size information
can help prevent people from overeating when they see a low-fat label. We also address this
question in Study 3.
Study 3: Can Objective Serving Size Information Reduce the Effects of “Low-Fat”
Nutrition Labels on Consumption?
Study 3 tests the hypothesis that “low-fat” nutrition labels increase consumption because
they increase consumers’ perception of what is the appropriate serving size. By doing this, we
also test whether providing objective serving size information can prevent people from
overeating a food with relative nutrition claims. To further explore the role of anticipated
consumption guilt, Study 3 also examines the consumption of granola (which is less hedonic
than M&M’s) and uses a procedure in which product quantity is exogenous and in which
consumption is unobserved by others. This enables us to test the prediction that low-fat
nutrition labels had a stronger influence on overweight consumers because they felt less guilty
about over-serving themselves.
Study 3 uses a 2 (“Regular” vs. “Low-fat” label) by 3 (no serving label, “Contains 1
Serving” label, “Contains 2 Servings” label) between-subject design. Two hundred and seven
university staff, undergraduates, and graduate students were recruited from across a large
university campus to be part of a study asking them to evaluate a pilot episode for a television
show. In exchange, they were given a free ticket to an upcoming movie, a coupon for a free
dessert at a local cafeteria, four chances to win $100 gift certificates to a campus bookstore,
and free snacks to eat during the viewing. The study was conducted over ten sessions that
lasted from 3:30 to 5:00 on each of ten days (Tuesdays and Thursdays for five nonconsecutive
weeks). Upon arriving to the dimly-lit theatre, participants were seated at every other seat and
asked to watch and rate a series of made-for-TV movie previews and a 60 minute pilot show
called “Hazard County.” They were also told that since it was late in the afternoon, they would
be given a cold 24-ounce bottle of water and a bag of granola from a respected campus
restaurant called the “Spice Box.” They were told to enjoy as much or as little of it as they
Based on a pre-test involving 79 consumers similar to those participating in the main
study, we selected a brand of granola mix that was well-liked, and we chose a realistic amount
(160 gm) that was sufficient to ensure that no one would eat all the granola within a 90-minute
time period. Each participant received 640 calories (160 grams) of granola in zip-lock bags
that were labeled with an attractive 3.25 x 4 inch color label. Depending on the condition, the
granola was either described as “Regular Rocky Mountain Granola” or “Low-fat Rocky
Mountain Granola.” Below this, the label either indicated that it “Contains 1 Serving”,
“Contains 2 Servings,” or it provided no serving size information.
Following the completion of the previews and the show (approximately 75 minutes),
participants were given an evaluation sheet that was consistent with the cover story. Because
there was concern of contamination between the ten sessions if too many questions were asked
about the granola, such questions were limited to estimations of how many calories they
believed they ate and to measures of weight, height, and gender. As they left the theatre, they
were allowed to select a free coupon for an upcoming movie along with a dessert coupon
while the amount of granola remaining in their bag was weighed. As the bag was being
weighed out of their sight, participants were asked how many serving sizes they believed their
bag contained. Following this, they were thanked and dismissed. Because of fears of
contamination across subsequent sessions, there was no debriefing at that time. Those who
were interested in learning more about the study signed an e-mail roster and were e-mailed a
debriefing later that semester after all the sessions were completed.
Seven individuals were eliminated from the study because they did not stay until the end of
the show. Three individuals were eliminated from the study because they spilled their granola
on the floor, and three more were eliminated because they transferred the granola to another
container while standing in line waiting to have their bag weighed. Of those who successfully
completed the study and returned their granola, four were eliminated because dietary
restrictions did not allow them to eat any granola, and fourteen were eliminated because they
did not provide complete weight and height information.
Effects of low-fat labels on consumption (no serving size information condition). No
participant consumed all of the granola in their bag while they were watching the show. As in
Study 1, we converted consumption volume into calories in order to facilitate comparison with
their calorie estimates. To benchmark with Study 1, an analysis of the no serving size
condition used an ANCOVA with three factors (nutrition label, body mass group, and their
interaction) and gender as a covariate. As in the earlier studies, the determination of whether
one was overweight was made by the BMI cut-off of 25 kg/m² established by the World
Health Organization. In this study, 110 had a normal weight (BMI < 25 kg/m²) and 69 were
overweight (BMI > 25 kg/m²).
Consistent with our earlier findings, people in the “Low-fat” label condition consumed
50.1% more granola than people who were given granola labeled as “Regular” (249 vs. 165
calories, F(1, 62) = 7.8, p < .01). As in Study 1, overweight participants consumed more
granola (M = 256 cal) than those with a normal weight (M = 174 cal), F(1, 62) = 8.5, p < .01.
As expected based on the guilt results from Study 2, the interaction between label and BMI
was not statistically significant (F(1, 62) = .3, p = .59), and the increase in consumption was
similar for low- and high-BMI consumers (see Figure 3).
To test our hypothesis that perceived serving size mediates the influence of low-fat
labeling on consumption, we conducted a mediation analysis using data from the control (no
serving size information) condition (Baron and Kenny 1986). We first regressed the number of
calories eaten onto the same three factors (nutrition label, body mass group, and their
interaction) and covariate (gender). We found that low-fat labels increased consumption by
73.9 calories after controlling for the effects of body mass and gender (B = 73.9, t = 3.0, p <
.01). We then regressed the mediator (the estimated number of servings in the bag) on the
same variables and found that low-fat labeling reduced the estimated number of servings
contained in the bag (B = –.61, t = –3.3, p < .01). None of the other coefficients were
statistically significant (p > .10). We then regressed the dependent variable onto the mediator
and the same control variables. The coefficient of perceived number of servings was negative
and statistically significant (B = –65.9, t = –4.3, p < .01), indicating higher consumption
among participants with lower estimates of the number of servings in the bag (and thus higher
estimations of serving size).
In the last regression of the mediation analysis, we regressed calories eaten on low-fat
labeling, estimated number of servings, and the control variables. When the mediator was
included in the regression, the coefficient of low-fat labeling became statistically insignificant
(B = 40.5, t = 1.6, p = .11) whereas the influence of estimated number of servings was
significant (B = -55.1, t = –3.4, p < .01). A Sobel test (1982) indicated that the mediation effect
was statistically significant (z = 2.69, p < .01). Overall, these results suggest that the estimated
number of servings mediates the effects of low-fat labeling on consumption. However,
because serving size estimations were measured at the end of the study, it is possible that they
were contaminated by consumption decisions. To better test the moderating role of serving
size information, we now turn to the analysis of the experimental manipulation of objective
serving size information.
Moderating effects of objective serving size information. As a manipulation check, we
examined participant’s estimates of the number of servings contained in the bag in the two
conditions where serving size information was available. People’s estimations were accurate
when given a bag with a “Contains 1 Serving” label (M = 1.05, t = 1.7, p = .08) or when given
one with a “Contains 2 Servings” label (M = 1.92, t = –1.6, p = .10). The estimated number of
servings in these two conditions was uninfluenced by low-fat labeling (F(1, 114) = .5, p = .48)
or by the participant’s body mass (F(1, 114) = .1, p = .76). These results show that the
manipulation of serving size information was successful in blocking inferences about serving
size from nutrition labeling. Participants who were provided with serving size information
clearly knew the number of servings purported to be in their bag and were uninfluenced by
To examine the moderating effects of serving size information on consumption, we used
an ANCOVA with four factors (“Contains 1 Serving,” “Contains 2 Servings,” body mass
group, and serving size information), all two- and three-way interactions, and one control
variable (gender). The two factors “Contains 1 Serving” and “Contains 2 Servings” were
necessary to estimate the effects of each of the two serving size labels. As expected, the main
effects of low-fat labeling, “Contains 2 Servings” condition, and body mass group were all
statistically significant (respectively, F(1, 167) = 12.1, p < .001; F(1, 167) = 11.8, p < .001;
and F(1, 167) = 13.4, p < .001), as was the three-way interaction between them (F(1, 167) =
4.6, p < .05). No other interactions were statistically significant, and gender was not significant
as a covariate. To facilitate the discussion of the results, we conducted separate ANCOVAs for
normal weight and overweight participants and discuss their results separately.
--- Insert Figure 3 about here ---
First, consider participants with a normal weight. When bags of granola were labeled as
“Contains 2 Servings,” consumption was reduced by 50 calories compared to the control
condition in which no serving size information was provided (F(1, 103) = 4.7, p < .05). When
the bags were labeled as “Contains 1 Serving,” however, there were no differences compared
to the control condition (F(1, 103) = .6, p = .44). With these normal weight participants, the
main effect of labeling the granola as “Low-fat” was not statistically significant (F(1, 103) =
2.2, p = .14), but its interaction with the “Contains 2 Servings” manipulation was significant
(F(1, 103) = 4.3, p < .05), and its interaction with the “Contains 1 Serving” manipulation was
directionally significant (F(1, 103) = 3.1, p = .08). These important interactions are illustrated
in Figure 3A. In the control condition, low-fat labels increased consumption by 62% (F(1, 36)
= 6.8, p < .01) but this effect disappeared when granola was labeled as “Contains 1 Serving”
(F(1, 36) = .3, p = .85) or as “Contains 2 Servings” (F(1, 36) = .13, p = .72). There was no
influence of gender (F(1, 103) = .6, p = .43). Providing objective serving size information was
therefore effective in eliminating the effects of low-fat labeling.
Now let us consider overweight participants. Again, labeling the bag of granola as
containing two servings reduced consumption by 74 calories compared to the control condition
(F(1, 62) = 7.0, p < .01) but labeling them as containing one serving had no effect (F(1, 62) <
.1, p = .88). In contrast to normal weight participants however, the main effect of low-fat
labeling was statistically significant (F(1, 62) = 11.4, p < .001) and there was no interaction
with the “Contains 1 Serving” manipulation (F(1, 62) < .1, p = .96) or with the “Contains 2
Servings” manipulation (F(1, 62) = 1.4, p = .23). There was no influence of gender (F(1, 62) =
.32, p = .57). Overweight consumers ate more granola when it was labeled as low-fat,
regardless of the serving size information that was provided to them
Biases in consumption estimation. Consistent with Study 1, we examined the dissociation
between actual and estimated calorie consumption by conducting an ANCOVA with the
difference between the estimated and actual number of calories as the dependent variable and
the same factors and covariates as in the analysis of actual consumption. As in Study 1,
participants underestimated their consumption (by 45 calories or 18%, F(1, 167) = 11.4, p <
.001) and this underestimation was larger when they saw a “Low-fat” label than when they
saw a “Regular” label (–69 vs. –21 calories, F(1, 167) = 38.5, p < .001). This calorie
underestimation bias was also larger among overweight participants than among participants
with a normal weight (–61 vs. –30 calories, F(1, 167) = 5.5, p < .05). The interaction between
low-fat labels and body mass indicated that the biased influence of low-fat labels was even
stronger among overweight participants (F(1, 167) = 3.8, p < .05). Finally, calorie
underestimation was also stronger in the control condition (M = –55 cal) than in the “Contains
2 Servings” condition (M = –26 calories, F(1, 167) = 5.5, p < .05). None of the other two-way
or three-way interactions were statistically significant.
When no serving size information was given to participants, the results of Study 3 replicate
those of Study 1. Low-fat nutrition labels increased granola consumption by 48% (80 calories).
Again, an important question is whether this increase in consumption volume would
overcompensate for the calorie savings in the lower-fat granola.
To determine this, we conducted a market survey of the nutrient content of regular and
low-fat granola bars and cereals, and we found 14 brands sold with both a “Low-fat” and a
“Regular” version. Serving sizes were almost identical across both versions (t = .99, p = .34).
Low-fat granola contained 56% less fat per serving than regular versions (2.3 vs. 5.9 grams, t
= 4.0, p < .001). However, they contained only 10% fewer calories per serving than regular
versions (196 vs. 173 calories, t = 3.3, p < .01). Assuming that participants in Study 3 had
eaten real low-fat granola, and that these low-fat granola had the average level of fat and
calories for the category, participants would have consumed 35% less fat from the low-fat
granola but would have consumed 33% more total calories. The increase in calorie would have
probably been even higher than that because the ingredients used to replace fat tend to make
people hungrier (Nestle 2002).
The calorie estimation results also replicate the findings of Study 1 that consumers are
unaware that low-fat labeling increases their consumption. As a result, whereas they were able
to accurately estimate their calorie consumption when the granola were labeled as “regular”
(they only underestimated consumption by 17 calories), they strongly underestimated actual
consumption (by 94 calories or 35%) when the granola was labeled as “low-fat” and when
they ate more of them. The convergence between the results of Study 1 (which were obtained
pre-intake) and Study 3 (which were obtained post-intake) shows the robustness of our finding
that low-fat labeling increases the calorie underestimation bias.
By measuring and manipulating serving sizes, Study 3 further contributes to our
understanding of why low-fat labeling increases consumption and what can be done to better
control it. The mediation analyses show that when no objective serving size information is
available, consumers' estimate of how many serving sizes are in a bag determined how much
they would eat. This was consistent with all consumers regardless of body mass. While this
would lead us to believe that putting salient serving size information on packaging would
eliminate the biasing influence of low-fat labeling, this only occurred with normal weight
participants. Overweight participants increased their consumption in response to low-fat
labeling regardless of whether information about serving sizes was made available.
Further analyses show that this difference cannot be explained by the fact that overweight
participants did not pay attention to serving size information. First, the manipulation checks
showed that all participants accurately identified the number of servings in the two conditions
where serving size information was provided on the label. Second, overweight participants
reduced their overall consumption when the granola bags mentioned “Contains 2 Servings”
compared to the control condition when no serving size information was available. One
explanation may be found in the low levels of guilt triggered by consuming granola and in the
fact that participants could not serve themselves less food, a common self-control mechanism
(Wertenbroch 1998). Even in the “Contains 2 Servings” label condition, when overweight
participants knew that the quantity of granola in the bag was higher than what most people
consume, they did not want to restrain their consumption. This is consistent with the findings
from Study 2 showing that consumption creates less guilt among overweight individuals than
among individuals with a normal weight.
In our obesity-inducing environment, a priority of the FDA is to examine whether relative
nutrition claims, such as “Low-fat,” lead to the over-consumption of food, especially high-
calorie snack foods consumed by “at-risk” (overweight) consumers. This issue of relative
nutrition claims and consumption is also an important for food companies that are battling
accusations that they contribute to the obesity epidemic. Yet, research on health claims and
nutrition labels has generally been focused on perceptions and purchase intention, not on
single occasion consumption.
We address this consumption issue by developing a framework of how relative nutrition
claims influence how much a person consumes. The framework predicts that relative nutrition
claims increase food intake per consumption occasion by increasing perceived serving sizes
(the perception of how much is appropriate to eat) and by reducing anticipated consumption
guilt. This helps explain why the influence of relative nutrition claims differ according to the
factors associated with guilt, such whether an individual is overweight or whether the food is
hedonic. We test the predictions of the framework in one lab study and two field experiments
involving actual food consumption. Our key results are as follows:
Labeling snacks as “low-fat” increases food intake during a single consumption occasion
by up to 50%. This is robust across hedonic and utilitarian snacks, young and old consumers,
self-reported nutrition experts or novices, in a public or private consumption setting, and
regardless of whether consumers serve themselves or not.
For guilt-inducing hedonic snacks and consumption settings, the influence of low-fat
labeling is higher for overweight consumers than for normal weight consumers.
For low-guilt snacks and consumption settings, the influence of low-fat labeling is the
same for overweight and normal weight consumers.
Providing objective serving size information eliminates the influence of low-fat labeling,
but only for consumers with a normal weight.
Implications for Public Policy
Agencies such as the U.S. FDA have created specific guidelines for nutrition labels and
claims (recall Table 1). These rules have guaranteed that nutrition information and claims are
consistently accurate, and they have resulted in high public trust in nutrition information. Yet
because of the robust influence of health halos, our findings suggest that truthful labels and
claims may not be sufficient to significantly improve food behavior. In light of this, what can
the FDA do to prevent consumers from making erroneous inferences that lead them to
One solution is to make serving size information more salient. This would be useful for
packaged goods, whose serving size labels are not always inspected (Balasubramanian and
Cole 2002). It would be even more useful for the fast-food industry, where relative nutrition
claims are widespread and nutrition labels are not mandatory. For example, one Subway ad
shows that a Subway Sweet Onion Teriyaki chicken foot-long sub has only 10 grams of fat
whereas a Big Mac contains 33 grams of fat. In the ad, the Subway spokesman, Jared Fogle,
then states that this means that “you can eat another and another over the course of three
different meals and still not equal the fat content of one Big Mac.” What the ad does not
mention, however, is that the Onion Teriyaki chicken foot-long sub contains 740 calories and
100 milligrams of cholesterol, compared to 600 calories and 85 milligrams of cholesterol for
the Big Mac. Hence, eating three Subway Subs would provide 1,620 more calories and 215
more milligrams of cholesterol than eating one Big Mac.
Another solution for regulators would be to increase the threshold for relative nutrition
claims. For example, to claim “reduced calories,” the food might have to contain 33% fewer
calories than the reference food instead of only 25% less (which is currently the case). This
would increase the likelihood that calorie intake is really lower for those eating “low-fat”
foods than for those eating “regular” (higher fat) foods, even after accounting for differences
in consumption volume. More generally, it would be important to account for health halos
when forming future guidelines or endorsements for foods (such as for low carbohydrates or
The most thorough reform would be to change the definition of serving sizes so that they
are higher for foods making relative nutrition claims. Currently, the FDA defines serving sizes
as “an amount of food customarily consumed per eating occasion by persons 4 years of age or
older, which is expressed in a common household measure that is appropriate to the food.”
Measures of the amount of food customarily eaten per eating occasion are obtained from the
Nationwide Food Consumption Surveys conducted by the U.S. Department of Agriculture.
The key problem is that these reference amounts are measured for all the foods in a category,
regardless of their nutrition claims, and they do not reflect how much is over eaten because of
a relative nutrition claim. Increasing serving sizes in the case of relative nutrition claims would
increase the number of calories per serving mentioned in the label for these foods. This would
have the double advantage of more accurately describing the actual amount of food consumed
and in deterring consumers from eating more than the serving size.
Implications for Food Manufacturers
Our findings also have implications for responsible food manufacturers (see Table 3).
First, manufacturers could consider being more explicit when defining something as low-fat or
as reduced-calorie. This could be in the form of the percentage reduction over the regular
version along with an absolute calorie level per serving or container. Because people infer the
appropriate serving size of a food from a variety of external cues, a manufacturer could also
help consumers better control their consumption by making packaging changes that alter their
perception of the appropriate serving size. One potentially profitable approach may be to
manufacture multi-packs of products with smaller individual servings. This would provide
break-points at which a person would have to reassess whether they were going to continue
--- Insert Table 3 about here ---
Another option for food manufacturers would be to consider manufacturing smaller,
premium-priced packages. Although they would not be priced competitively (per unit)
compared to the larger packages, they would satisfy the person who was willing to trade-off
value for self-control (Wertenbroch 1998). Although such packaging would increase
production costs, the $12 billion per year diet industry would suggest there is a portion-
predisposed segment that would be willing to pay a premium for packaging that enabled them
to eat less of a food in a single serving and to enjoy it more. For example, a loyalty program
survey of current consumers of a Kraft product indicated that 57% of these will be willing to
pay up to 15% more for portion-controlled packaging (Zaff 2004).
By showing that relative nutrition claims influence consumption our findings open up new
areas for nutrition research. Because of the interest of the FDA and of Kraft, we focused on
nutrition labels involving “low-fat.” Pilot studies with other nutrition labels (including
“reduced calorie”) showed similar results. It would be interesting to study whether the same
health halo can be caused by simple nutrition facts (such as “100-calorie” portions) or even by
general health claims (such as those made by Pepsico’s “Smart Spot” and Unilever’s “Better-
Another extension would be to study how these claims influence what else a person eats.
For example, consuming a food with a relative health claim (such as a low-fat sandwich) may
lead a consumer to subsequently indulge in a calorie-rich dessert. As a result, consumers may
end up eating more calories when going to “healthy” restaurants than to restaurants which are
not making these claims. Because such unintended overeating might even operate with vague
claims (such as Subway’s “Eat fresh” campaign), it would be difficult for the FDA to regulate.
Yet, the health halos that are inferred by relative claims on labels may offer an important
conceptual tool to help investigate the single occasion consumption of non-food products of
which the FDA is also concerned, such as medicines, supplements, and personal care products.
Another important area for future research is to understand whether the influence that
relative nutrition claims have on consumption might also be related to the influence they have
on the taste of a food. Raghunathan, Walker, and Hoyer (2005) found that labeling food as
“healthy” reduces consumers’ taste expectations and even reduces their post-intake taste
experience. Could this mean that relative nutrition claims increase consumption because
consumers trade off taste reductions with increased consumption? When health perceptions are
manipulated but the actual food remains the same, one study found that consumers actually
consumed more when the product seemed to taste better because it was labeled as “unhealthy”
(Raghunathan et al. 2005). In reality however, the low-fat ingredients used to replace the fat
may leave more of a watered-down taste, which consumers may try to offset by consuming
more. It would therefore be important for future research to study the joint effects of nutrition
claims and actual nutrition content.
Finally, an important area for future research would be to extend these results to additional
consumer segments, additional product categories, and additional consumption contexts. For
example, the packages in Study 3 made the labeling information highly salient. Yet people do
not always eat out of clearly marked packages. As with nutrition information in general, the
format of information on the label should influence its effectiveness (Levy and Fein 1996).
The effects of the message are also certainly influenced by consumers’ motivation to process
nutrition information and by their objective and subjective nutrition knowledge (Andrews et al.
1998; Moorman et al. 2004). Not all consumers have nutrition or weight-loss as an objective in
their life, and heavy-handed regulation could increase the current consumer backlash against
diet and nutrition messages (Patterson et al. 2001). To make matters even more difficult, recent
research on willful ignorance suggests that even highly-involved consumers may actively
ignore nutrition information in order to avoid the negative emotions that may arise, should
nutrition information be below expectations (Ehrich and Irwin 2005).
We are at a point of development where much of the incremental improvement in our life
span—and especially in our quality of life—is likely to come more from behavioral changes in
our lifestyle than from new medical treatments. When it comes to contributing most to the life-
span and quality of life in the next couple of generations, smart, well-intentioned marketers
may be well-suited to effectively help lead the movement toward behavior change. Obesity is a
good place to start.
FDA Definitions and Restrictions for Selected Relative Nutrition Claims
Claim Free Low Reduced/
Less/ Fewer Comments
Less than .5 g
For meals and
less than .5 g
3 g or less per
(or per 50 g if
Meals and main
dishes: 3 g or less
per 100 g and not
more than 30% of
calories from fat.
At least 25%
less fat per
food may not
be “Low Fat”.
“__% Fat Free”: OK if meets the
requirements for “Low Fat”.
“100% Fat Free”: Food must be
For dietary supplements: fat
claims cannot be made for
products that are 40 calories or
less per serving.
Less than 5
for meals or
40 cal or less per
(and per 50 g if
Meals and main
dishes: 120 cal or
less per 100 g.
At least 25%
“Light” or “Lite”: if 50% or
more of the calories are from fat,
fat must be reduced by at least
50% per reference amount. If
less than 50% of calories are
from fat, fat must be reduced at
least 50% or calories reduced at
least 1/3 per reference amount.
“Light” or “Lite” meal or main
dish product meets definition for
“Low Calorie” or “Low Fat”
meal and indicates which
definition is met.
hydrates No Current
Use of the term
“low” only in
reference to a diet
or lifestyle, but
not to describe a
Any notation of carbohydrate
content must describe how the
calculations were made. Any
method is acceptable as long as
enough information is provided
on the labeling.
Note: The information on calories and fat have been principally adopted from U.S. Food and
Drug Administration Center for Food Safety and Applied Nutrition (Stehlin 1993). The FDA has
yet to publish definitions characterizing the carbohydrate content of food, but are soon believed
to (1) establish definitions for “carbohydrate free” and “low-carbohydrate,” (2) confirm the use
of the relative claims “reduced,” “fewer” or “less” in relation to a reference food; and (3)
confirm the use of the term “modified carbohydrate” (Burros 2004).
Study 2: How Low-fat Labels Influence Perceived Serving Size, Calorie Density, and Guilt
(Means and Standard Deviations)
Normal weight participants
(BMI < 25) Overweight participants
(BMI > 25)
Perceived serving size
Note: Perceived serving size was measured as the number of ounces appropriate to eat during a
90-minute movie. Perceived calorie density was measured as the estimated number of calories in
a 10 oz cup. Anticipated consumption guilt was measured by asking respondents to rate how
they would feel after eating two ounces of the product on a nine-point scale (1 = not guilty, 9 =
Potential Implications of Health Halos for Responsible Packaged Goods Companies
Food categories Labels that could
Impact of health halos on
normal weight consumers Impact of health halos on
overweight consumers Possible
- Main meals
- Side dishes
- “Better for You”
- “Smart Spot”
- Labeling decreases
feelings of guilt
- Labeling increases
appropriate serving sizes
- People consumed 61%
- People consumed 45%
- Labeling decreases feelings
- Labeling increases
appropriate serving sizes
- People consumed 24% more
- People consumed 12% more
- Consider providing
sub-packaging on some
- Make serving size
- Offer re-sealable
- Potato chips
- Ice creams
- Low fat
- Low carbs
- Reduced calorie
- Vitamin fortified
- Labeling does not greatly
- Labeling does not greatly
reduce serving sizes
- People consumed 16%
- People consumed no
- Labeling decreases feelings
- Labeling increases
appropriate serving sizes
- People consumed 47% more
- People consumed 25% more
moderation, such as
“Eat less, enjoy more
- Underscore calorie
reduction as a
percentage and in
terms of total calories
- Reduce calories in
Note: Illustrations of consumption volume increases are based on the increases observed in Study 3 for utilitarian foods (control
condition/no serving size information) and on the increases observed in Study 1 for hedonic foods. Illustrations of calorie increases are
based on industry-wide calorie averages for low-fat and regular versions of granola and chocolate candies.
A Framework of How Relative Nutrition Claims Influence Consumption
(Food intake per
Study 1: How Low-Fat Labels Influence Actual and Estimated Calorie M&M’s Consumption
for Normal Weight and Overweight Consumers
107 101 119 128
"Regular" label "Low-fat" label "Regular" label "Low-fat" label
Normal weight consumers (BMI < 25) Overweight consumers (BMI > 25)
Estimated calorie consumption Actual calorie consumption
Study 3: How Objective Serving Size Information and Low-Fat Labels Influence Granola
Consumption for Normal Weight Consumers (Panel A) and Overweight Consumers (Panel B)
119 137 169 141 130 110
No serving label "Contains 1 serving" label "Contains 2 servings" label
Actual calorie consumption
Estimated calorie consumption
Normal weight consumers (BMI < 25)
213 173 180 173
No serving label "Contains 1 serving" label "Contains 2 servings" label
Actual calorie consumption
Estimated calorie consumption
Overweight consumers (BMI > 25)
Alexander, Erin (2004). Personal Correspondence: Kraft Foods Inc., Northfield, IL, September
Andrews, J. Craig, Richard G. Netemeyer, and Scot Burton (1998), "Consumer Generalization of
Nutrient Content Claims in Advertising," Journal of Marketing, 62 (4), 62-75.
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.
Baron, Reuben M. and David A. Kenny (1986), "The Moderator-Mediator Variable Distinction in
Social Psychological Research: Conceptual, Strategic, and Statistical Considerations,"
Journal of Personality and Social Psychology, 51 (6), 1173-82.
Baumeister, Roy F. (2002), "Yielding to Temptation: Self-Control Failure, Impulsive Purchasing,
and Consumer Behavior," Journal of Consumer Research, 28 (4), 670-76.
Broniarczyk, Susan M. and Joseph W. Alba (1994), "The Role of Consumers' Intuitions in
Inference Making," Journal of Consumer Research, 21 (3), 393-407.
Burros, Marian (2004), "Eating Well; New 'Low-Carb' Foods Aren't All-You-Can-Eat," The New
York Times (April 14, 2004).
Caswell, Julie A. and Daniel I. Padberg (1992), "Toward a more comprehensive theory of food
labels," American Journal of Agricultural Economics, 74 (2), 460-68.
Deighton, John (1984), "The Interaction of Advertising and Evidence," Journal of Consumer
Research, 11 (3), 763.
Dhar, Ravi and Itamar Simonson (1999), "Making Complementary Choices in Consumption
Episodes: Highlighting Versus Balancing," Journal of Marketing Research, 36 (1), 29-44.
Ehrich, Kristine R. and Julie R. Irwin (2005), "Willful Ignorance in the Request for Product
Attribute Information," Journal of Marketing Research, 42 (3), 266-77.
Feldman, Jack M. and John G. Lynch, Jr. (1988), "Self-Generated Validity and Other Effects of
Measurement on Belief, Attitude, Intention, and Behavior," Journal of Applied
Psychology, 73 (3), 421-35.
Fishbach, Ayelet and Ravi Dhar (2005), "Goals as Excuses or Guides: The Liberating Effect of
Perceived Goal Progress on Choice," Journal of Consumer Research, December.
Food and Drug Administration (2003), "Exploring the Link Between Weight Management and
Food Labels and Packaging," Public Workshop held at the National Institutes of Health,
Bethesda, Maryland, November 20, 2003. Transcript available at:
Ford, Gary T., Manoj Hastak, Anusree Mitra, and Debra Jones Ringold (1996), "Can Consumers
Interpret Nutrition Information in the Presence of a Health Claim? A Laboratory
Investigation," Journal of Public Policy & Marketing, 15 (1), 16-27.
Garretson, Judith A. and Scot Burton (2000), "Effects of Nutrition Facts Panel Values, Nutrition
Claims, and Health Claims on Consumer Attitudes, Perceptions of Disease-Related Risks,
and Trust," Journal of Public Policy & Marketing, 19 (2), 213.
Geiger, Constance J. (1998), "Health Claims: History, Current Regulatory Status, and Consumer
Research," Journal of the American Dietetic Association, 98 (11), 1312-22.
Grice, H. P. (1975), "Logic and Conversation," in Syntax and Semantics: Speech Acts, P. Cole
and JL. Morgan, Eds. Vol. 3. New York: Academic.
Ha, Young-Won and Stephen J. Hoch (1989), "Ambiguity, Processing Strategy, and Advertising-
Evidence Interactions," Journal of Consumer Research, 16 (3), 354.
Hays, Nicholas P, Gaston P Bathalon, Megan A McCrory, Ronenn Roubenoff, Ruth Lipman, and
Susan B Roberts (2002), "Eating behavior correlates of adult weight gain and obesity in
healthy women aged 55-65 y," American Journal of Clinical Nutrition, 75 (3), 476-83.
Hedley, Allison A., Cynthia L. Ogden, Clifford L. Johnson, Margaret D. Carroll, Lester R.
Curtin, and Katherine M. Flegal (2004), "Prevalence of Overweight and Obesity Among
US Children, Adolescents, and Adults, 1999-2002," The Journal of the American Medical
Association, 291 (23), 2847-50.
Inman, J. Jeffrey (2001), "The Role of Sensory-Specific Satiety in Attribute-Level Variety
Seeking," Journal of Consumer Research, 28 (1), 105-20.
Ippolito, Pauline M. and Alan D. Mathios (1991), "Health Claims in Food Marketing: Evidence
on Knowledge and Behavior in the Cereal Market," Journal of Public Policy &
Marketing, 10 (1), 15.
Keller, Scott B., Mike Landry, Jeanne Olson, Anne M. Velliquette, Scot Burton, and J. Craig
Andrews (1997), "The Effects of Nutrition Package Claims, Nutrition Facts Panels, and
Motivation to Process Nutrition Information on Consumer Product Evaluations," Journal
of Public Policy & Marketing, 16 (2), 256.
Kivetz, Ran and Anat Keinan (forthcoming), "Repenting Hyperopia: An Analysis of Self-Control
Regrets," Journal of Consumer Research.
Kozup, John C., Elizabeth H. Creyer, and Scot Burton (2003), "Making Healthful Food Choices:
The Influence of Health Claims and Nutrition Information on Consumers' Evaluations of
Packaged Food Products and Restaurant Menu Items," Journal of Marketing, 67 (2), 19.
Levy, Alan S. and Sara B. Fein (1996), "Performance characteristics of seven nutrition label
formats," Journal of Public Policy & Marketing, 15 (1), 1-15.
Livingstone, M. Barbara E. and Alison E. Black (2003), "Markers of the Validity of Reported
Energy Intake," Journal of Nutrition, 133 (3), 895S-920S.
Moorman, Christine (1990), "The Effects of Stimulus and Consumer Characteristics on the
Utilization of Nutrition Information," Journal of Consumer Research, 17 (3), 362-74.
---- (1996), "A quasi experiment to assess the consumer and informational determinants of
nutrition information," Journal of Public Policy & Marketing, 15 (1), 28.
Moorman, Christine, Kristin Diehl, David Brinberg, and Blair Kidwell (2004), "Subjective
Knowledge, Search Locations, and Consumer Choice," Journal of Consumer Research,
31 (3), 673-80.
Moorman, Christine and Erika Matulich (1993), "A model of consumers' preventive health
behaviors: The role of health motivation and health ability," Journal of Consumer
Research, 20 (2), 208-28.
National Institutes of Health (2004), "Diet Myths." Bethesda, MD: NIH Publication No. 04-4561,
Nestle, Marion (2002), Food Politics: How the Food Industry Influences Nutrition and Health.
Berkeley: University of California Press.
Okada, Erica Mina (2005), "Justification Effects on Consumer Choice of Hedonic and Utilitarian
Goods," Journal of Marketing Research, 42 (1), 43.
Patterson, Ruth E., Jessie A. Satia, Alan R. Kristal, Marian L. Neuhouser, and Adam
Drewnowski (2001), "Is There a Consumer Backlash Against the Diet and Health
Message?," Journal of the American Dietetic Association, 101 (1), 37-41.
Raghunathan, Rajagopal, Rebecca E. Walker, and Wayne D. Hoyer (2005), "The Unhealthy =
Tasty Intuition and its Effects on Taste Inferences, Enjoyment, and Choice of Food
Products." University of Texas at Austin: Working paper.
Rothschild, Michael L. (1999), "Carrots, Sticks, and Promises: A Conceptual Framework for the
Management of Public Health and Social Issue Behaviors," Journal of Marketing, 63 (4),
Schwarz, Norbert (1996), Cognition and Communication: Judgmental Biases, Research Methods
and the Logic of Conversation. Hillsdale, NJ: Erlbaum.
Shiv, Baba and Alexander Fedorikhin (1999), "Heart and Mind in Conflict: The Interplay of
Affect and Cognition in Consumer Decision Making," Journal of Consumer Research, 26
Sobel, Michael E. (1982), "Asymptotic Confidence Intervals for Indirect Effects in Structural
Equation Models," Sociological Methodology, 290-312.
Stehlin, Dori (1993), "A Little 'Lite' Reading," Food and Drug Administration,
Stunkard, Albert J. and Samuel Messick (1985), "The Three-Factor Eating Questionnaire to
Measure Dietary Restraint, Disinhibition and Hunger," Journal of Psychosomatic
Research, 29 (71-83).
Wansink, Brian (2004), "Environmental Factors that Increase the Food Intake and Consumption
Volume of Unknowing Consumers," Annual Review of Nutrition, 24, 455-79.
Wansink, Brian and Mike Huckabee (2005), "De-Marketing Obesity," California Management
Review, 47 (4), 6-18.
Wertenbroch, Klaus (1998), "Consumption Self-Control via Purchase Quantity Rationing,"
Marketing Science, 17 (4), 317-37.
Zaff, Brian D. (2004), "Personal Correspondence." M&M/Masterfoods, Hackensack, NY.