Determinants of Household Choice of Breakfast Cereals:

Healthy or Unhealthy?

Alla Golub
Department of Agricultural Economics
Purdue University

James Binkley
Department of Agricultural Economics
Purdue University

Selected paper prepared for presentation at the American Agricultural Economics Association Annual
Meeting, Providence, Rhode Island, July 24-27, 2005

This research is conducted under the Cooperative Agreement “Purchase Patterns of Foods with Salient
Nutrition Characteristics” (Purdue 596 1145-0699/0YT80)

Copyright 2005 by Alla Golub, James Binkley. All rights reserved. Readers may make verbatim copies of this
document for non-commercial purposes by any means, provided that this copyright notice appears on all
such copies.
 Determinants of Household Choice of Breakfast Cereals:

Healthy or Unhealthy?

Abstract

We studied consumer demand for more and less healthy breakfast cereals. Using ACNielsen Homescan

database and USDA food nutrition data, we developed three cereal nutrition indexes for each household in

the data. In addition to the standard demographic characteristics of households and prices, we included

variables representing differences between private labels and national brands. We found that the structure of

the industry, through its effect on the product mix produced, affects consumer choice of nutritious foods.

Some households buy fewer healthy cereals simply through reluctance to trust private labels. Among all

factors expected to influence consumer purchases, the prices appear to have the strongest effect on the

healthiness of the choice of breakfast cereals, which is a relatively inexpensive product. Households with

children and teens buy less healthy cereals, while older and more educated households make healthier

choices.

Keywords: consumer demand, healthy and unhealthy food, breakfast cereals.

2
Introduction

The U.S. Food, Drug and Cosmetic Act, passed in 1938, made it illegal to use disease-prevention claims in

promoting food products. This law was enforced until 1984, when Kellogg began making such claims for

their All-Bran breakfast cereals. A list of “Preventive Health Tips from the National Cancer Institute” (NCI)

appeared on the All-Bran box and the NCI was also mentioned in All-Bran TV and print ads (Consumer

reports, October 1986). The Food and Drug Administration (FDA) decided to permit this promotion.

Subsequently, the Nutrition Labeling and Education Act of 1990 provided FDA with specific authority to

require nutrition labeling of most foods regulated by the agency and to require that all nutrient content claims

(i.e., 'high fiber', 'low fat', etc.) and health claims be consistent with agency regulations. The regulations for

nutrition labeling become effective in 1994. The goals of this regulation were to reduce the negative effects

of untruthful and exaggerated nutrition claims and to provide consumers with nutrition information at the

point of sale. As a consequence of these developments, nutrition claims have became a standard marketing

tool for many food products.

However, obesity and other nutrition-related health problem have worsened. According to the

American Heart Association, the prevalence of overweight in children ages 6-11 increased from 4.2 percent

in 1963-65 to 15.8 percent in 1999-2002. The prevalence of overweight in adolescents ages 12-19 increased

from 4.6 percent to 16.1 percent over same period. The obesity in Americans in the age 20-74 increased from

20.6% in men and 25.9% in women during 1988-1994 to 27.6% in men and 33.2% in women during 1999-

2002. This suggests that nutrition information provided by food labels and advertising may have limited

impact on actual food choice. One reason may be costs, in terms of time and efforts, of gathering and

processing the information (Stigler and Becker). Also, it may be that nutrition cannot compete with other

factors, such as taste and convenience to which food companies are catering with an increasing array of food

products.

Because of these trends, the question of consumer demand for healthy/unhealthy foods is a topic of

growing interest in agricultural economics. One of the issues addressed is consumer responsiveness to

3
nutrition and health information. Most studies have used aggregate commodity data or household food

surveys. Brown and Schrader constructed a measure of information on the links between cholesterol and

heart disease available to physicians to investigate how this information had affected consumer demand for

shell eggs. They analyzed aggregate national consumption and price data and found that the information had

decreased per capita shell egg consumption and changed shell egg’s own price and income elasticities.

Chern, Loehman and Yen applied Brown and Schrader’s index to the FDA Health and Diet Survey data and

found that the health information resulted in decreased consumption of butter and lard and increased

consumption of vegetable oils with less saturated fat. Kinnucan et al. updated Brown and Schrader’s index to

look at the effects of health information together with advertising on the shifts in U.S. aggregate meat

demand from beef to poultry. The health information appeared to be important: the health-information

elasticities were larger than price elasticites, while effects of generic advertising were found to be small. The

conclusion from these studies is that consumer demand, at least at the aggregate level, is responsive to the

health and nutrition information.

A study at the individual level was conducted by Variyam, Blaylock and Smallwood. They used

Continuing Survey of Food Intake of Individuals (CSFII) and the companion Diet and Health Knowledge

Survey (DHKS) conducted by U.S. Department of Agriculture (USDA) to estimate the effects of fiber-

specific information on dietary fiber intake. Fiber helps ward off heart disease, diabetes and may help to

lessen chances of developing colon cancer. Variyam, Blaylock and Smallwood measured information using

survey questions on fiber content of foods, attitude toward consuming fiber-rich foods, and awareness of

fiber-health links. Their results confirmed the positive influence of nutrition information on fiber intake and

highlighted the enhancement effect of the education on the level of information.

In this study we use a commercial data set to examine household demand for more healthy/less

healthy breakfast cereals. For a focus on nutrition, cereal is an excellent product to study. It is one of the

largest grocery categories, purchased by nearly all households. More important, it contains products of

widely varying nutritional quality. On the one hand, many cereals are important sources of whole grains and

4
fiber, which are generally agreed to be a preventive for digestive cancers and chronic heart diseases. On the

other, there are highly sweetened cereals, which many view as junk food and as potentially contributing to

obesity and type-II diabetes, especially among children. These characteristics are well-publicized: cereal

marketing makes especially frequent use of health claims. Thus, consumers have a relatively high knowledge

of the health aspects of various cereals.

In this study we examine the effect of prices, household structure, and purchase behavior on the

healthiness of household cereal choice. We used the ACNielsen Homescan database, consisting of all retail

food purchases and prices paid by 7195 US households during 1999. The database also contains households’

demographic characteristics. As explained below, we developed three cereal nutrition indexes for each

ACNielsen household based on their cereal purchases and nutrition contents of various cereal brands. This

was used as the dependent variable in a regression on prices, measures of household structure, income,

education and purchase behavior represented by the percent of household’s purchases in all retail food

categories that were (i) private label, and (ii) bought on a deal. The last two variables were included to

capture a unique aspect of the study. We consider the possibility that the structure of industry, through its

(possibly incidental) effect on the product mix produced, can affect consumer choice of nutritious foods.

The cereal industry is composed of a small number of national brand producers, a fringe of small producers,

and a very large number of private labels. While each of these groups produces cereals of all types, there are

some possibly important differences in the mix of healthy/unhealthy cereals. For example, the fringe group

has several firms specializing in “natural” cereals, a large number of which would be classed as healthy by

most criteria.

What is of particular interest here is differences between private labels and national brands. Cereals

that are viewed as healthy tend to be basic whole grain types, with simple formulas and non-proprietary

names, like raisin bran, oatmeal, and shredded wheat. Less healthy cereals tend to be sweetened and made

with more complex formulas using refined grains and added flavorings. They also have proprietary names.

5
These product differences makes private label more competitive for healthy cereals, which will affect their

product mix.

If we classify brands in ACNielsen data by the percent of sugar and fiber in their weight, the sugar

content of “other” (other than major brands) branded products is the lowest, reflecting their concentration in

adult-type whole grain products. But this group is closely followed by private labels, with average sugar

content much below that of all the major manufacturers. For fiber, private label actually has a higher content

than does the “other” group.

By this classification, then, we find that private label cereals as a group are among the healthiest, and

generally superior to the national brands. As a consequence, consumers who are not reluctant to buy private

label products will for that reason tend to purchase healthier cereals. This is the reason we included the

private label variable. The “bought on a deal” variable is used to measure the national brand prone

consumer. Price reductions, especially with coupons, are used extensively by national cereal brands, but

seldom by private labels. Thus, heavy coupon users will tend to buy national brands, and hence, on average,

less healthy cereals. These two consumer types are evidently reasonably distinct, for we found the

correlation between private label and bought on a deal variables to be virtually zero.

The paper is organized into six sections. Section two provides review of the literature devoted to

different aspects of demand for breakfast cereals and links between consumer health and breakfast cereals

consumption. Section three describes methodology and datasets used in the analysis. Section four is devoted

to explanatory variables used in the analysis. Section five presents results, and section six is the conclusion.

Previous breakfast cereals studies1

1
Breakfast cereals is a very interesting food category. It attracts not only demand economist, but also the attention of industrial
organization economists. See, for example, Schmalensee, Connor, Price and Connor, and Reimer.

6
Several hedonic pricing models have studied cereals. These include Morgan, Metzen and Johnson; Shi and

Price; Stanley and Tschirhart. The empirical content of hedonic models is under debate because the content

may be just a consequence of arbitrary functional forms (Ekeland, Heckman and Nesheim) and may be

subject to endogeneity problems. This may explain why these cereal studies often yielded implausible

results, such as negative valuation of fiber and positive valuation of sugar content by consumers.

Binkley and Eales conducted an exploratory study on how prices of breakfast cereals, demographic

variables and health characteristics of consumers affect the consumer choice between high fiber and low

fiber breakfast cereals. Specifically, they looked at relationship between high fiber cereal consumption and

incidents of mortality by digestive cancer. Their work was motivated by Armstrong and Doll who found that

cereal consumption is negatively associated with every type of cancer considered, with the association

especially pronounced for digestive cancer. Several data sources were used, including: Sales Area

Marketing, Inc (SAMI) data on product sales, National County Mortality Data File of the National Center for

Health Statistics, and food composition data from USDA Continuing Household Survey data base. Five

Kellogg’s cereals were considered. All-Bran, a product rich in fiber, was least price and expenditure elastic,

implying that All-Bran consumers viewed this product as something “necessary”. Healthy life style,

measured by consumption of total and high fiber breakfast cereals, had a negative relationship with incidents

of mortality from digestive cancer. Finally, there was evidence that consumers in markets with high mortality

rates due to digestive cancer were motivated to buy cereals with very high fiber content. Overall, the

conclusion of this study is that choice of food is governed not only by price considerations, but also dietary

characteristics and it is important that consumers are aware of these characteristics and their effects on

health.

Aside from these studies is the work by Ippolito and Mathios. They evaluated whether the policy

change in 1985, that allowed food manufacturers to explicitly link diet to disease risks in advertising and

labeling, led to improved consumers food choices, or had “…confused consumers sufficiently to slow

improvements in diet that would otherwise occur…” During the 1978 -1984 period the evidence of a link

7
between reduced cancer rates and high fiber diets and corresponding educational programs was growing. At

the same time health claims by manufacturers were prohibited and, as a result, Ippolito and Mathios found

there were no increases in high-fiber cereals consumption. However, when health claims through labeling

and advertising by manufacturers were allowed in 1985, the consumption of high in fiber cereals increased,

as well as number of new high fiber cereals produced. An analysis of food consumption patterns revealed

that large differences in high fiber cereals consumption among demographic groups, existing prior to the use

of health claims, diminished after the change in information policy. This leads to the conclusion that

producers’ claims are more effective in reaching consumers and influencing their behavior than government

educational campaigns and other information sources. The authors attributed this to the fact that government

information is usually distributed in very general form, while producers’ advertising ties nutrition

characteristics of specific product to consumer health.

Data and methodology

The Homescan database contains many product categories, four of which are cereals: ready-to-eat, hot,

natural and granola, and wheat germ. In the database, there were 1 888 635 purchases from those categories,

conducted by 6 998 households. 197 households bought no breakfast cereal. Using primarily the USDA food

nutrition data base but also other sources (including cereal boxes), nutritional contents for each brand

purchased were obtained. Major cereals appear directly in the USDA data. For others we either used

alternative sources or matched them with USDA cereals. The total number of cereals identified as different

by their nutrition content is 451.

We are interested in “healthiness” of the breakfast cereal consumption of each household and in

drivers of the choice between healthy and unhealthy breakfast cereals. Knowing the nutrition content of each

breakfast cereal purchased and recorded in the ACNielsen database, we calculated the nutrition content of

the breakfast cereal “bundle” bought by each household during 1999. We separated “healthy” and

“unhealthy” cereals “bundles” by three criteria: (1) sugar content; (2) fiber content; and (3) by a healthiness

8
index based on rating developed by Consumer Reports Magazine which is based on protein, sugar, fiber, fat,

and sodium. The Consumer Reports Magazine rating of most popular in the US breakfast cereals is provided

in table 1. Using these ratings and regression analysis, we developed a CU index for every household in the

ACNielsen data. Specifically, we regressed the ratings in table 1 on the protein, sugar, fiber, fat, and sodium

content of each breakfast cereals. The results of the regression are shown in table 2. The weights that we

used to calculate the index for each household are shown in the coefficient column. Note that high (positive)

marks are assigned to protein and fiber and low marks (negative) to fat, sugar and sodium.

The distribution of the healthiness of breakfast cereals consumed during 1999 is shown in table 3 and

Figures 1-3. Sugar and fiber indices are calculated as percent of weight of all breakfast cereals bought during

1999. The distributions of CU and sugar indices are pretty symmetric, while fiber index is skewed to the left

suggesting that some households are biased toward rich in fiber breakfast cereals compare to the rest of the

sample.

The three measures of the healthiness of consumer choice of breakfast cereals were used as the

dependent variable in a regression on measures of income, education, household structure, purchase behavior

and prices. That is, we implemented three regressions where independent variables are the same, but the

dependent variables are different: fiber index, sugar index and Consumer Reports index.

Explanatory variables

All variables definitions and expected effects on the dependent variables are provided in the table 4.

The first group of the explanatory variables is demographic characteristics constructed from the information

in the ACNielsen database. Among demographic variables, only income per household member is treated as

continuous and the others are dummy variables. 2 Our expectation about the effect of income variable on the

healthiness of consumed breakfast cereals is mixed. Higher income may be a proxy for better access to and

knowledge about the nutrition content of food. However, Variyam, Blaylock and Smallwood found that as

2
In the ACNielson database, the household income variable is given by intervals. For example, $12 000 - $14 999, $15 000 –
$19 999 and so on. Income per household member is calculated as the mean point of the interval divided by size of household.
9
income increased, households reduce fiber consumption. This may be because higher-income individuals

may view rich in fiber products as inferior, or high income individuals have high time costs of obtaining the

nutrition information. We expect education to have a positive effect on the healthiness of breakfast cereals

because it should give a better ability to gather and process information about nutrition and to distinguish

between truthful and exaggerated nutrition claims.

We expect older people to choose cereals rich in fiber because of negative association between high-

fiber foods consumption and cancer. We expect that younger people are more taste than nutrition oriented

and choose less healthy cereals. The presence of children in the age of 6-17 should reduce healthiness of the

household breakfast cereal bundle because they tend to judge foods solely on taste, and for many brands

targeted towards children half the cereal is sugar. The presence of female head in the household should

improve healthiness of the choice because females have been found to be more concerned about nutrition

and health issues. If female head does not work, she may spend more time on household menu planning and

pay even more attention to nutrition. We included race and Hispanic variables to capture possible cultural

differences (Variyam, Blaylock and Smallwood). Regional dummy variables are included to capture

variations due to similar reasons, but also to overcome consequences of the error component specification of

the model. We will address this issue below.

To calculate prices, we first defined “most healthy” and “least healthy” cereals as cereals in the upper

quartile and lower quartile, respectively, of the distribution of CU or fiber index. For the sugar index the

definition is the reverse. ACNielsen lists all prices paid by households and we could simply take these prices.

However, consumers with low search costs and access to many outlets can exercise some control on prices

paid, generating potential endogeneity. To avoid this endogeneity problem, we proceeded as follows. In the

ACNielsen data, all US territory is divided into 52 markets. For each breakfast cereal in each market we

calculated volume-weighted average price. That is, households in the same market are assigned the same

price for a particular breakfast cereal. Such structure of the data may lead to error-component model

(Moulton and Randolph). If errors actually follow an error components specification, then the use of ordinary

10
least square can lead to seriously biased standard errors and test statistics. Therefore, it is important to test

for the presence of error components. We conducted this test and found no evidence of error-component

problem. However, this was true only when the models included the regional variables discussed above. The

private label and “deal” variables were calculated as the percent of all purchases in the entire data base (not

just cereals) that were private label and deal, respectively. This was computed for every household.

Results

The results of the regression analysis are presented in table 5. The results for the three measures of

healthiness are very similar for all variables. The higher income per household member has significant

positive effect on the healthiness of breakfast cereals choice: income variable enters the CU index equation

and fiber equation with a positive sign and the sugar equation with a negative sign. Given that education

variable is also included in the model, the income is most likely a proxy for better access to nutrition

information. The education variable is also highly significant and has a positive effect on CU and fiber

indices and negative effect on sugar index, as expected. Compared to young households, older households

buy healthier breakfast cereal.3 The same is true for middle age household, but the difference is smaller.

Households with children in the age from 6 to 12 and from 13 to 17 buy less healthy breakfast cereals, but

presence of very young children have no effect. The 6 to 12 age group is likely to be the group most

susceptible to advertising, and highly advertised cereals are often very high in sugar. We tested if better

education in the families with children influences the choice of breakfast cereals, expecting to find that

educated parents would buy more healthy food for themselves and especially for their children. For this test,

we added interactions of education variable with babe, youth and teen variables to the model. However, these

variables did not have any significant effect on the healthiness of breakfast cereals.

The presence of a female head does not have any effect on the choice in all three models. However,

households with only a male head choose less healthy cereals compared to households where only female

3
White race young household living in central region is our reference group.
11
head or both heads are present, at least in terms of CU index. As expected, female heads not employed

outside the home make better choice of breakfast cereals. Interestingly, effect of the male head being not

employed is also positive and even stronger. This may suggest that males, if they have time to shop, are more

careful shoppers than females. The effect of Hispanic ethnicity is not significant. Black and oriental

households choose lower in fiber breakfast cereals. The results for Hispanic and black households is very

similar to what is found by Variyam, Blaylock and Smallwood.

The private label variable enters all three models as expected: the effects on CU index and fiber index

are positive, and the effect on sugar index is negative. However, the effect on fiber index is not significant,

suggesting that the health advantage of private labels mainly involves low sugar. The special price or deal

variable has a significant negative effect on the healthiness of the choice as expected. Branded cereals in the

aggregate tend to be less healthy, perhaps because of an emphasis on taste features.

The sign and large magnitude of the price of more and less healthy breakfast cereals reveal that the

choice of breakfast cereal, which is a relatively inexpensive product, is sensitive to price. In the markets

where less healthy cereals are more expensive, consumers choose more healthy cereals. That is, there is

significant substitution effect. In fact, consumers pay extra when the manufacturer adds the sweetener. Sugar

costs about 35 cents a pound in the store. In the box of cereal such as Honey Smacks, the cost of sugar is four

times larger (Consumer Reports). We found that the average prices are higher for unhealthy cereals than for

healthy in all markets, when the healthiness is measured by CU index, and in 47 of 52 markets when the

healthiness is measured by the fiber or sugar indexes.

Conclusions

In this study we examined the effect of prices, household structure, and purchase behavior on the healthiness

of household breakfast cereal choice. We created three measures of the healthiness: an overall nutritional

quality index, a fiber index and a sugar index. Households with children and teens buy less healthy cereals,

while older households make healthier choices. More educated households and higher income households

12
also choose healthier cereals. Reasons for the latter are not obvious, since some of the cheapest cereals fall

in the healthy group. (However, many natural and multigrain cereals tend to be high-priced.) Our

explanation for this positive relationship between income and healthiness of the choice is the higher income

serves as a proxy for better access to nutrition information. The private label and bought on deal variables

were both significant with expected signs. This suggests that nutrition profiles can differ simply due to

different shopping behavior. Evidently some households buy fewer healthy cereals simply through

reluctance to trust private labels.

Strikingly, among all factors expected to influence consumer purchases, the prices appear to have the

strongest effect on the healthiness of the choice of breakfast cereals, relatively inexpensive product. As price

of less healthy cereals increases, consumers choose more healthy breakfast cereals. In the same time, the

price increase of more healthy cereals leads to less healthy choice. These suggest that prices are important: as

more healthy cereals become less expensive, more households would buy them. This is especially important,

for one oft-mentioned policy choice is to tax less healthy food products and /or to subsidize those viewed as

healthier. Our results provide evidence that such policies may be effective.

References

American Heart Association: http://www.americanheart.org

Binkley, James and James Eales, 2000. “Demand for High Fiber and Low Fiber Cereals”, A Selected Paper

Presentation at the American Agricultural Economics Association Annual Meeting in Tampa,

Florida.

Brown, D. J. and L. F. Schrader, 1990. Cholesterol Information and Shell Egg Consumption.

13
 Amer. J. Ag. Econ., 72:548-555.

Chern, W. E., E. T. Loehman, and S. T. Yen. (1995) Information, Health Risk Beliefs, and the

Demand for Fats and Oils. Rev. Econ. Stat., 77:555-564.

Connor, J.M. “Breakfast Cereals: The Extreme Food Industry,” Agribusiness, 15(1999): 247-259.

Consumer Repotrs, October 1986: 628-642.

Ekeland, Ivar, Jemas J. Heckman and Lars Nesheim, 2001. “Identifying Hedonic Models,” Cemmap

Working Paper CWP06/02, The Institute for Fiscal Studies, Department of Economics, UCL.

Ippolito, Pauline M. and Alan D. Mathois, 1989. Health Claims in Advertising and Labeling, A Study if the

Cereal Market, Bureau of Economics Staff Report, Federal Trade Commission, August 1989.

Kinnucan, Henry W., Hui Xiao, Chung-Jen Hsia, and John D. Jackson. “Effects of Health Information and

Generic Advertising on U.S. Meat Demand.” Amer. J. Agr. Econ. 79(February 1997):13-23.

Mojduszka, Eliza M. and Rachel M. Everett, 2003. “Endogenous Consumer Preferences and Knowledge

About Nutrition”, A Selected Paper Presentation at the American Agricultural Economics

Association Annual Meeting in Montreal, Canada.

Morgan, Karen J., Edward J. Metzen and S.R. Johnson, 1979. “An Hedonic Index for Breakfast Cereals,”

The Journal of Consumer Research, Vol. 6, No. 1 (June), 67-75.

Moulton, Brent R. and William C. Randolph, 1989. “Alternative Tests of the Error Components Model,”

Econometrica, Vol. 57, No. 3 (May), 685 – 693.

Price, Gregory K. and John M. Connor, 2003. “Modeling coupon values for ready-to-eat breakfast cereals”,

Agribusiness, Vol. 19, Issue 2, 223-243.

Reimer, Jeffrey J., 2004. “Market Conduct in the U.S. Ready-to-Eat Cereal Industry.” Journal of

Agricultural & Food Industrial Organization, Vol. 2, Article 9.

Schmalensee, R. “Entry Deterrence in the Ready-to-Eat Breakfast Cereal Industry,” Bell Journal of

Economics. 9(1978): 305-327.

14
Shi, Hongqi and David W. Price, 1998. “Impacts of Sociodemographic Variables on the Implicit Values of

Breakfast Cereal Characteristics”, Journal of Agricultural and Resource Economics 23 (1), 126 – 139.

Stanley, Linda R. and John Tschirhart, 1991. “Hedonic Prices for a Nondurable Good: The Case of Breakfast

Cereals,” The Review of Economics and Statistics, Vol. 73, No. 3 (August), 537 – 541.

Stigler, George J., and Gary S. Becker, “DE Gustibus Non Est Disputandum,” American Economic Review,

67(2), March 1977, pp. 76-90.

Variyam, J. N., Blaylock J., and D. Smallwood, 1996. A Probit Latent Variable Model of

Nutrition Information and Dietary Fiber Intake, Amer. J. Agr. Econ. 78(August):628-63

Weaver, C.M., 2003. Atwater Memorial Lecture: Defining nutrient requirements from a perspective of bone-

related nutrients. J. Nutr. 133:4063 – 4066.

Figure 1. Distribution of the constructed CU index of the breakfast cereal consumption in the ACNielsen
data, 1999.

15
 12

10

P 8
e
r
c 6
e
n
t
4

0
23 27 31 35 39 43 47 51 55 59 63 67 71 75

CU I ndex

Figure 2. Distribution of the sugar index of the breakfast cereal consumption in the ACNielsen data, 1999.

P 6
e
r 5
c
e 4
n
t
3

0
1 5 9 13 17 21 25 29 33 37 41 45 49 53

Sugar I ndex

Figure 3. Distribution of the fiber index of the breakfast cereal consumption in the ACNielsen data, 1999.

16
 30

25

P 20
e
r
c 15
e
n
t
10

0
0 4. 5 9 13. 5 18 22. 5 27 31. 5 36 40. 5 45

Fi ber I ndex

17
Table1. Nutrition Index for breakfast cereals calculated by Consumer Union and reported in Consumer
Report, October 1986, and their nutrition information per 100 grams as reported in USDA food nutrition
data.4

Protein, Fat, Sugar, Fiber, Sodium,

Brand Index
gm gm gm gm mg
All-Bran (Kellogg) 60 12.5 3.32 19 32 258
Apple Jacks (Kellogg) 33 3 2 49 3.2 475
Bran Buds (Kellogg) 51 7 2.15 27 43 676
Cap'n Crunch (Quaker Oats) 22 4.35 5.81 43.57 2.5 749
Cap'n Crunch's Crunch Berry (Quaker Oats) 22 4.45 5.65 44.52 2.5 699
Cap'n Crunch's Peanut Butter Crunch 20 7.05 9.15 33.19 2.9 742
Cheerios (General Mills) 51 11 6 4 9 910
Cocoa Krispies (Kellogg) 29 3.4 3.2 45 3.2 613
Cookie-Crisp, Chocolate Chip (Ralston Purina) 27 4 3 42 1.5 594
Corn Chex (Ralston Purina) 40 7 0.9 10.77 2 959
Kellogg's Corn Flakes 44 7 0.8 7 3.5 725
Crackin'Oat Bran (Kellogg) 36 8.3 14.6 31 11.7 286
Kellogg's Bran Flakes 49 10 2 17 17.5 715
Froot Loops (Kellogg) 31 3.4 4.1 47 3.1 471
Frosted Mini-Wheats (Kellogg) 62 9.8 1.6 19.6 10 10
Frosted Krispies (Kellogg) 27 3.4 0.8 40 0.3 726
Golden Grahams (General Mills) 27 5 3.5 35 3 895
Honey Nut Cheerios (General Mills) 27 9 4 35 6 898
King Vitamin (Quaker Oats) 31 6.45 3.56 20.2 3.8 838
Kix (General Mills) 40 6 2 11 3 891
Life (Quaker Oats) 49 9.92 4.38 19.48 6.6 513
Lucky Charms (General Mills) 29 7 3.8 43 5 678
Product 19 (Kellogg) 42 7.7 1.4 13.3 3.3 690
Kellogg's Raisin Ban 38 8.5 2.5 32 11.9 593
Rice Chex (Ralston Purina) 42 6 1 8 1 941
Rice Krispies (Kellogg) 42 6.2 1.3 9.1 0.4 966
Quaker Puffed Rice 67 7 0.9 0 1.4 5
Special K (Kellogg) 53 22.5 1.55 12.9 2.4 721
Corn Pops (Kellogg) 36 3.7 0.73 45.2 0.8 386
Frosted Flakes (Kellogg) 29 3.3 0.52 38 3.2 479
Honey Smacks (Kellogg) 40 6.4 1.85 56 3.7 186
Total (General Mills) 49 8 2.5 17 8 639
Trix (General Mills) 27 3 3.8 44 3 647
Wheat Chex 56 10 2 10 11 891
Wheaties (General Mills) 47 10 3.2 14 10 725
Quaker Puffed Wheat 78 16.26 2.15 1.36 9.4 5
Cinnamon Life (Quaker Oats) 49 8.91 4 26.59 6.2 478
Oh's Honey Graham (Quaker Oats) 18 3.97 7.41 45.07 2.1 600
Sun Country Granola with Raisins 51 11.77 18.02 20.42 5.2 33
Fiber One (General Mills) 67 8 2.7 0 48 429
All-Bran with Extra Fiber (Kellogg) 67 11.3 3.5 0.37 50 475

4
Consumer Reports Magazine provides nutrition information of breakfast cereals with corresponding indices. But, nutrition
information was provided only for some of the breakfast cereals for which ratings were reported. Moreover, when level of a key
component is small, the Magazine reports “trace”, which cannot be used to recover weights assigned to each key nutrient
component in the index. For these two reasons, we use USDA food nutrition data as a source of the breakfast cereal nutrition
content.
18
 Protein, Fat, Sugar, Fiber, Sodium,
Brand Index
gm gm gm gm mg
Crispix (Kellogg) 44 6.8 0.8 10.3 0.5 724
Raisin Nut Bran (General Mills) 42 9.39 8 29 9.2 455
FrankenBerry (General Mills) 24 3 2.6 47 0.9 711
Country Corn Flakes (General Mills) 44 6 1.3 8 1.7 877
Count Chocula (General Mills) 27 4 3.6 47 1.8 584
Cocoa Puff (General Mills) 24 4 3.2 47 2.3 571
Cinnamon Toast Crunch (General Mills) 22 5 11 34 4 687
Kaboom (General Mills) 31 9 3.7 20 6 950
Post Natural Bran Flakes 56 9.4 2.2 18.9 17.6 732
Cocoa Pebbles (Post) 27 3.5 4.2 44 1.6 541
Fruity Pebbles (Post) 27 3.6 3.9 44 0.7 584
Alpha-Bits (Post) 29 8.5 4.1 39 4.1 661
Fruit & Fibre (Post) 44 7.1 5.6 29.8 9.7 509
Super Golden Crisp (Post) 42 5.5 1.4 53.9 0 150
Grape-Nuts (Post) 56 10.8 1.9 12 8.7 610
Grape-Nuts Flakes (Post) 49 10 2.9 17.6 8.8 482
Honey-Comb (Post) 29 5.2 2.1 38.3 2.5 743
Post Natural Raisin Bran 40 7.9 1.8 33.4 13.1 611
Post Toasties Corn Flakes 44 6.7 0.1 6.5 4.5 949
Nabisco Shredded Wheat 71 10.4 1.2 0.8 11.5 7
Nabisco Shredded Wheat'N Bran 73 12.5 1.4 1 13.4 5
Nabisco Shredded Wheat Spoon Size 73 10.3 1.1 0.9 11.4 7
Post 100% Bran 58 12.7 2.1 24.4 28.6 417
Bran Chex (Ralston Purina) 51 7 2.5 22 13 657
Uncle Sam (US Mills) 69 15.98 11.6 1.56 20.3 206
Familia Genuine Swiss Muesli (Biofamilia) 56 9.5 6.3 26.2 8.5 50

19
Table 2. Weights for each component in the CU index.
Dependent variable is Consumer Report Magazine Index.
Standard
Variable Coefficient t Value Pr > |t|
Error
Intercept 63.21 2.60 24.28 <.0001
Protein 0.88 0.18 4.89 <.0001
Total lipid (fat) -0.93 0.15 -6.29 <.0001
Sugars -0.49 0.04 -12.69 <.0001
Fiber 0.26 0.05 5.41 <.0001
Sodium -0.02 0.00 -13.92 <.0001

Number of observations 67
Adj R – square 0.94
F statistics 211.41

Table 3. Purchased breakfast cereals Healthiness indices5

Healthiness Index Range Mean Std. Dev Minimum Maximum
CU 1 to 100 44.76 9.01 22.55 75.44
Sugar 1 to 100 23.58 9.87 0 53.90
Fiber 1 to 100 6.52 3.06 0 48.00

5
One observation in the database is purchase of buckwheat groats which contains a lot of protein and almost no fat, sugar and
sodium. The index for this purchase is greater than 100. We removed this purchase from the further analysis.
20
Table 4. Variable definitions, their summary statistics and expected effects on the independent variable.
Sample mean Expected Expected
Expected
and effect on effect on
Variable Definition effect on
standard sugar fiber
CU index
deviation index index
25.46
Income Income per household member per year, $1000 ? ? ?
(18.16)
1 if average of education levels of male and female
0.34
Education heads of households is at least completed college + - +
(0.47)
degree, 0 otherwise
1 if average age of male and female heads of household 0.52
Old + - +
is 50 or older, 0 otherwise (0.5)
1 if average age of male and female heads of household 0.38
Middle ? ? ?
is between 35 to 49, 0 otherwise (0.49)
1 if average age of male and female heads of household 0.10
Young - + -
is under 35, 0 otherwise (0.30)
0.1
Babe 1 if children under 6 years old are present, 0 otherwise ? ? ?
(0.3)
0.16
Youth 1 if children 6-12 years old are present, 0 otherwise - + -
(0.37)
0.15
Teen 1 if children 13-17 years old are present, 0 otherwise - + -
(0.36)
0.76
Male head 1 if male head present, 0 otherwise ? ? ?
(0.42)
0.91
Female head 1 if female head is present, 0 otherwise + - +
(0.29)
Female head not 1 if female head is present and do not work, 0 0.30
+ - +
working otherwise (0.46)
Male head not 0.18
1 if male head is present and do not work, 0 otherwise ? ? ?
working (0.38)
Ratio of number of private label purchases to total 0.18
Private label + - +
number of purchases6 (0.10)
Ratio of number of purchases on special prices to total 0.28
Special price - + -
number of purchases6 (0.21)
Weighted by volume average price of least healthy 0.63
Unhealthy price + - +
cereals in each market (0.03)
Weighted by volume average price of most healthy 0.41
Healthy price - + -
cereals in each market (0.03)
0.1
Black 1 if black race, 0 otherwise ? ? ?
(0.3)
0.01
Oriental 1 if oriental race, 0 otherwise ? ? ?
(0.11)
0.06
Hispanic 1 if Hispanic origin, 0 otherwise ? ? ?
(0.24)
0.34
South US region 1 if household is in South US region ? ? ?
(0.47)
0.20
West US region 1 if household is in West US region ? ? ?
(0.40)
0.20
East US region 1 if household is in East US region ? ? ?
(0.40)
0.25
Central US region 1 if household is in Central US region ? ? ?
(0.43)

6
In these ratios, the denominator is total number of household’s purchases in all food categories, not only breakfast cereals, and
the nominator is number of private label purchases (or purchases on special prices) in all food categories also.

21
Table 5. Results of the regression analysis. 7
Variable CU index Sugar index Fiber index
Intercept 40.31*** 28.89*** 4.01***
(2.45) (2.53) (0.87)
Income 0.03*** -0.03*** 0.02***
(0.01) (0.01) (0.00)
Education 0.81*** -0.87*** 0.36***
(0.25) (0.27) (0.09)
Old 5.06*** -6.12*** 1.2***
(0.41) (0.44) (0.14)
Middle 2.19*** -2.58*** 0.31**
(0.39) (0.42) (0.13)
Babe -0.81** 0.48 -0.23*
(0.39) (0.43) (0.14)
Youth -3.22*** 3.47*** -0.68***
(0.33) (0.36) (0.11)
Teen -2.82*** 3.22*** -0.53***
(0.33) (0.36) (0.12)
Male head -0.70** 0.50 -0.12
(0.29) (0.31) (0.10)
Female head 0.45 -0.45 0.05
(0.42) (0.45) (0.15)
Female head not working 1.10*** -1.26*** 0.28
(0.26) (0.28) (0.09)
Male head not working 1.38*** -1.47*** 0.44
(0.33) (0.36) (0.12)
Private Label 4.05*** -3.30*** 0.56
(1.13) (1.24) (0.40)
Special price -3.26*** 3.14*** -0.61***
(0.52) (0.57) (0.18)
Unhealthy price 8.69** -6.97* 2.41*
(3.69) (3.63) (1.32)
Healthy price -13.54*** 6.88** -0.27
(3.70) (2.87) (1.47)
South 0.62** -0.04 0.19*
(0.32) (0.35) (0.11)
West 1.25*** -0.67* 0.17
(0.36) (0.39) (0.12)
East 0.85** -0.40 0.13
(0.36) (0.37) (0.13)
Black 1.41*** -0.48 -0.89***
(0.35) (0.38) (0.12)
Oriental 0.41 -1.46 -0.62**
(0.89) (0.97) (0.31)
Hispanic -0.73 0.70 -0.17
(0.44) (0.48) (0.15)

Number of observations 5892 5892 58848

R-square 0.16 0.16 0.12
LM test statistic for the error
0.41 1.67 0.85
component

7
We tested for and did not find any multicollinearity problem in three models.
8
In this regression smaller number of observations is used because in one market, represented only by 8 households, none of the
households bought breakfast cereals that are least healthy according to our definition based on fiber index.
22