Nutrition Information in the Supermarket J. EDWARD RUSSO RICHARD ST AELIN CATHERINE A. NOLAN GARY J. RUSSELL BARBARA L. METCALF* Lists of nutrition information posted in supermarkets were designed to reduce the information-processing costs of comparing altemative foods. In Experiment 1, lists of vitamins and minerals increased nutrition knowledge but had no influence on actual purchases. In Experiment 2, a list of added sugar-a negative component of food-increased the market share of low-sugar breakfast cereals at the expense of high-sugar brands. We conclude that effort-reducing displays are a successful technique for increasing information use, especially for the more highly valued neg- ative nutrients. T he focus of this article is the design of information programs that successfully alter consumers' knowledge and purchase patterns. We present the results of two studies that tested specific features of these in- formation programs in the context of purchasing nu- tritious food. Finally, we discuss the implications of these findings for developing new information programs • J. Edward Russo is Associate Professor, Johnson Graduate School of Management, Cornell University, Ithaca, NY 14853. Richard Staelin is the Edward and Rose Donnell Professor of Business Ad- ministration, Fuqua School of Business, Duke University, Durham, NC 27706. Catherine A. Nolan is a former student for the Master of Science degree in the School of Nutritional Science, University of Washington, Seattle, WA 98195. Gary J. Russell is Assistant Professor, School of Business Administration, University of California, Berkeley, CA 94720. Barbara L. Metcalf is a former research project manager, Graduate School of Business, University of Chicago, Chicago, IL 60637. The research associated with Experiment 2 was submitted in partial fulfillment of the Master of Science degree in nutrition edu- cation at the University of Washington. The authors gratefully ac- knowledge the cooperation of Jewel Food Stores of Chicago-espe- cially Jane Armstrong, Steve Rubow, Frank Shepard, and Mary Anne Vydra-and of Associated Grocers and the Klauser Corporation of Tacoma, Washington. This research benefited from the advice ofsev- eral individuals and organizations, notably Randall G. Chapman, Eric J. Johnson, Joan Karbeck, Bernard Megrey, Leo J. Shapiro, Raymond C. Stokes, Werner Wothke, Policy Research Corporation, and an Oversight Committee of the National Science Foundation. We also thank the U.S. Food and Drug Administration for suspending food labeling regulations to enable the field test of nonstandard in- formation formats. Major support for this research was provided by NSF Grant DAR 76-81806, with additional support from a bequest from C. N. Newman to the Graduate School of Business of the Uni- versity of Chicago, the faculty research fund of the Johnson Graduate School of Management, the Food Marketing Institute, and the Grad- uate School of the University of Washington. This report reflects only the opinions of the authors; its findings are not necessarily en- dorsed by the aforementioned individuals or organizations. 48 in other contexts such as hazardous consumer products and energy-consuming appliances. We begin by developing a conceptual framework to explain how consumers decide to use product infor- mation. The basic premise is that consumers conduct an informal cost-benefit analysis and use information only when the perceived benefits exceed the costs. This premise should hold even if the information is free (as it is for nutrition information) since consumers' effort to process the information is still a cost (Mazis et al. 1981; Russo 1981). A cost-benefit view implies two general strategies for increasing information use: (1) increase the perceived benefits, or (2) decrease the effort costs. The first ap- proach, benefit enhancement, characterizes traditional educational and motivational campaigns. The second, the one used in this study, attempts to reduce the costs associated with gathering and using available infor- mation. This effort reduction approach has been shown to be effective in a similar setting where the information was unit prices rather than nutrition (Russo 1977). INFORMATIONAL FORMATS: A COST-BENEFIT ANALYSIS Benefits When shoppers are asked about the importance of nutrition and of nutrition information, typical respon- ses are quite positive (Dietrich 1980; Marketing Science Institute 1980). For example, among a list of 13 factors in food decisions, vitamins and minerals were men- tioned by 76 percent of shoppers, the second largest number (General Mills 1980). However, these claimed benefits need to be qualified by the meaning of nutrition. © JOURNAL OF CONSUMER RESEARCH • Vol. 13. June 1986
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Nutrition Information in the Supermarket
J. EDWARD RUSSO RICHARD ST AELIN CATHERINE A. NOLAN GARY J. RUSSELL BARBARA L. METCALF*
Lists of nutrition information posted in supermarkets were designed to reduce the information-processing costs of comparing altemative foods. In Experiment 1, lists of vitamins and minerals increased nutrition knowledge but had no influence on actual purchases. In Experiment 2, a list of added sugar-a negative component of food-increased the market share of low-sugar breakfast cereals at the expense of high-sugar brands. We conclude that effort-reducing displays are a successful technique for increasing information use, especially for the more highly valued negative nutrients.
T he focus of this article is the design of information programs that successfully alter consumers'
knowledge and purchase patterns. We present the results of two studies that tested specific features of these information programs in the context of purchasing nutritious food. Finally, we discuss the implications of these findings for developing new information programs
• J. Edward Russo is Associate Professor, Johnson Graduate School of Management, Cornell University, Ithaca, NY 14853. Richard Staelin is the Edward and Rose Donnell Professor of Business Administration, Fuqua School of Business, Duke University, Durham, NC 27706. Catherine A. Nolan is a former student for the Master of Science degree in the School of Nutritional Science, University of Washington, Seattle, WA 98195. Gary J. Russell is Assistant Professor, School of Business Administration, University of California, Berkeley, CA 94720. Barbara L. Metcalf is a former research project manager, Graduate School of Business, University of Chicago, Chicago, IL 60637. The research associated with Experiment 2 was submitted in partial fulfillment of the Master of Science degree in nutrition education at the University of Washington. The authors gratefully acknowledge the cooperation of Jewel Food Stores of Chicago-especially Jane Armstrong, Steve Rubow, Frank Shepard, and Mary Anne Vydra-and of Associated Grocers and the Klauser Corporation of Tacoma, Washington. This research benefited from the advice ofseveral individuals and organizations, notably Randall G. Chapman, Eric J. Johnson, Joan Karbeck, Bernard Megrey, Leo J. Shapiro, Raymond C. Stokes, Werner Wothke, Policy Research Corporation, and an Oversight Committee of the National Science Foundation. We also thank the U.S. Food and Drug Administration for suspending food labeling regulations to enable the field test of nonstandard information formats. Major support for this research was provided by NSF Grant DAR 76-81806, with additional support from a bequest from C. N. Newman to the Graduate School of Business of the University of Chicago, the faculty research fund of the Johnson Graduate School of Management, the Food Marketing Institute, and the Graduate School of the University of Washington. This report reflects only the opinions of the authors; its findings are not necessarily endorsed by the aforementioned individuals or organizations.
48
in other contexts such as hazardous consumer products and energy-consuming appliances.
We begin by developing a conceptual framework to explain how consumers decide to use product information. The basic premise is that consumers conduct an informal cost-benefit analysis and use information only when the perceived benefits exceed the costs. This premise should hold even if the information is free (as it is for nutrition information) since consumers' effort to process the information is still a cost (Mazis et al. 1981; Russo 1981).
A cost-benefit view implies two general strategies for increasing information use: (1) increase the perceived benefits, or (2) decrease the effort costs. The first approach, benefit enhancement, characterizes traditional educational and motivational campaigns. The second, the one used in this study, attempts to reduce the costs associated with gathering and using available information. This effort reduction approach has been shown to be effective in a similar setting where the information was unit prices rather than nutrition (Russo 1977).
INFORMATIONAL FORMATS: A COST-BENEFIT ANALYSIS
Benefits
When shoppers are asked about the importance of nutrition and of nutrition information, typical responses are quite positive (Dietrich 1980; Marketing Science Institute 1980). For example, among a list of 13 factors in food decisions, vitamins and minerals were mentioned by 76 percent of shoppers, the second largest number (General Mills 1980). However, these claimed benefits need to be qualified by the meaning of nutrition.
© JOURNAL OF CONSUMER RESEARCH • Vol. 13. June 1986
NUTRITION INFORMATION
Although nutrition is often defined in terms of positive attributes-the components offood necessary for good health-many consumers are more concerned about "negative" nutrients-the components that most consumers want to avoid or reduce their consumption of. Negative nutrients include calories, cholesterol, sodium, sugar, and various chemical additives. 1
Survey evidence reveals that consumers see both positive and negative nutrients as important, but emphasize the negative (e.g., Heimbach and Stokes 1979, 1982; Marketing Science Institute 1980). Heimbach (1981) reports a survey of the rated importance of 38 food components, of which 29 were positive and nine were negative. All nine negative nutrients ranked in the top 12 in importance (protein, iron, and calcium completed the list). Further, such attitudes seem to translate directly into food choices. Putnam and Weimer (1981) report that roughly two-thirds of surveyed households claimed to have made at least one recent change in food consumption, almost always to avoid a negative nutrient. Of the 10 most frequently mentioned reasons for changes in eating habits, nine were reductions in negative food components.
Although negative nutrients are more highly valued, the positive nutrients are all comparable-a property essential to the exploration of our new information formats. Because almost all adults need roughly the same daily amounts of the eight nutrients on the food label, each can be expressed in a standard unit such as the U.S. Recommended Daily Allowance (U.S. RDA). This measure is comparable across both nutrients and people. As will become clear in the next section, the full range of proposed information formats can only be tested when such comparability exists. Thus, in Experiment 1 we use only positive nutrients-a pragmatic choice dictated by the comparability (and availability) of positive nutrients.
Costs The costs facing shoppers who want to base their
brand choice on nutrition information are not small. Currently they must collect the nutrition information from many different product labels, comprehend it, and then determine how to aggregate the different nutrients to identify the most nutritious brand. If taken at all seriously, this task requires considerable effort. It involves at least three types of effort costs: the collection, comprehension, and computation of information.
Collection Effort. Collection cost is the time and effort needed to acquire the relevant information. For
IThe distinction between positive and negative nutrients is necessarily imperfect. Many of the negative nutrients-e.g., calories and sodium-are essential to good health. Also, there are small groups who need to consume more rather than less of these nutrients-e.g., the seriously underweight (calories) and those who work in desert climates (sodium). In general, however, the positive-negative distinction applies to the majority of supermarket shoppers.
49
the majority of processed, packaged foods that carry nutrition labels, this means moving along the supermarket aisle to examine the various package facings. For some products, the nutrition information must be obtained from the manufacturer or some other source.
Computation Effort. Once the information is gathered, consumers must somehow combine it into an overall evaluation of nutritional quality. This data reduction task is difficult, more so because most supermarket shoppers are under time pressure and without a calculational aid. As a consequence, it is unlikely that consumers would use all attributes to achieve an overall evaluation. Instead, they would probably simplify the task by ignoring some nutrients. Unfortunately, this strategy can work poorly because each nutrient is essential to health, and no one food attribute is a good predictor of the availability of the other attributes.
Comprehension Effort. Comprehension costs depend on the level of shoppers' nutrition knowledge. Empirical evidence indicates that many consumers lack adequate nutrition knowledge, which suggests that they would experience relatively high comprehension costs. For example, in a survey whose details are reported shortly, we assessed the understanding of two basic concepts, nutrition and U.S. RDA. Over one-third of shoppers failed to identify the correct definition of nutrition, choosing instead such alternatives as low in harmful ingredients, high in fiber, and so forth. Similarly, over one-third did not know the U.S. Recommended Daily Allowance referred to the amount and kind of nutrition in the food. Heimbach (1981) reports a similar finding.
Information Formats That Reduce Processing Costs
The cost of collecting nutrition information can be lowered by gathering and posting nutrient values for all brands on a single list. Exhibit 1 shows one possible format, the Matrix, for TV dinners-one of the product categories tested in Experiment 1.
Although greatly reducing collection effort, this Matrix display does little to lower the computation and comprehension costs. To reduce shoppers' computational effort the information can be summarized in a single overall measure of nutritional quality. The summary measure we devised-the Nutrition Quotient (NQ)-is described in the next section. The resulting information display-the Summary format-is shown in Exhibit 2.
Shoppers' imperfect nutrition knowledge and our plan to use an unfamiliar summary rating posed significant comprehension costs. To lower these costs we needed a mechanism that readily communicates the meaning of the summary values to the consumer. The
50 THE JOURNAL OF CONSUMER RESEARCH
EXHIBIT 1
THE MATRIX FORMAT
Choose A Balanced Diet For Better Nutrition
ThiS chart is part of a research project under the direction of Associate Professor J. Edward Russo, Consumer Behavior Laboratory, Graduate School of Busin .... University 01 Chicago. Chicago. Dnnols. 60637. Products are not included on the list if an estimate for each of the nutrients listed to the right was not available from either the manufacturer or the
U.S.DA Handbook: "Nutritive Values of American Foods." Additionally. new items or items not carried by most stores are not included. Inclusion of any product on the list does nol constitute an endorsement. For more information, contact the Consumer Behavior Laboratory.
TV dinners differ in weight. The nutritk>n listed depends on the weight of each dinner.
Percentage of U.S. Recommended Daily Allowance (U.S. RDA)
" ~_~ {iI ~ &~ § f
~....- $
c.. !9 .# I # ~.:r TV DINNERS
Serving Size 1 DI"ner ,,"0 ~ ~§ "" o c..'t:'
Beans And Franks. Banquet Beef Chop Suey. Banquet Beef EnchUada. Swanson Beef Tenderloin. Steak House Beef. Banquet
......... 10.75 591 25 40 ... 12.00 282 20 6
.............. 15.00 570 30 50 ..................... 9.50 920 70 2 .................... 11.00 312 45 4
Beef. Chopped. Banquet .. 11.00 443 30 30 60 60 60
Beef. Chopped. Hungry-Man. Swanson Beef. Hungry-Man. Swanson
.18.00 730 .. 17.00 540
Beef. Swanson .... 11.50 370 Chicken. BBQ. Hungry-Man. Swanson ......... 16.50 760
Chicken. BBQ. Swanson ........... 11.25 Chicken. Crispy. Swanson ..... 10.75 Chicken. Fried. Banquet. . . . .............. 11.00 Chicken. Fried. Hungry-Man. Swanson ..... 15.75 Chicken. Hungry-Man. Swanson ........... 19.00
Chicken. Man-Pleaser. Banquet Chicken. Swanson ... Chicken. Three Course. Swanson
.. 17.00 . ....... 11.50
Chicken. Western Style. Hungry-Man. Swanson Chicken. Western Style. Swanson
.. 15.00 .... 17.75 · .. 11.75
Chopped Sirloin. Steak House FIsh DInner. Banquet Fish N Chips. Swanson Italian, Banquet ..... . Lasagna. Hungry-Man. Swanson
.. 9.50
.. 8.75 ....... 10.25
.11.00
.17.75
Macaroni" Cheese. Swanson .12.50 Meat Loaf. Banquet .. . . . . . . . . . . .. . 11.00 Meat Loaf. Man-Pleaser. Banquet . . . . . . .. .19.00 Meat Loaf. Swanson ................... . .10.75 Mexican Dinner. Banquet ..... 12.00
Noodles And Chicken, Swanson ...... 10.25 Rib Eye. Steak House ......... 9.00 Salisbury Steak. Banquet ............. 11.00 Salisbury Steak. Hungry-Man. Swanson ..... 17.00 Salisbury Steak, Man-Pleaser. Banquet ... 19.00
Salisbury Steak. Swanson Salisbury Steak. Three Course. Swanson Sirloin, Chopped. Swanson
· .11.50 .... 16.00 ... 10.00
....... 9.50 · .12.50
Sirloin. Steak House Spaghetti. Swanson
Turkey. Banquet Turkey. Hungry-Man. Swanson Turkey. Man-Pleaser. Banquet Turkey. Swanson Turkey. Three Course. Swanson
Veal Parmigiano Banquet
......... 11.00 ...... 19.00
............ 19.00 ... 11.50
.... 16.00
Veal Parmigiana. Hungry-Man. Swanson · .11.00
.... 20.50
530 25 650 50 530 40 910 100 730 90
1016 90 570 60 630 50 890 70 460 35
760 382 450 446 740
390 412 916 530 571
390 820 390 870 823
500 490 460 920 410
293 740 620 360 520
421 910
90 30 50 30 40
15 30 60 30 35
20 70 30 70 60
45 35 50 70 20
35 100 60 45 60
30 60
90 45 2 6
110
20 4
100 2 15
90 40 30 10 10
2 50 6
90 20
100 45 160 10 25
10 2 80 10 90
<> 6 80 2 15
80 20 150 60 15
140 10
,,'t:' J ~ oS- ~ ~
15 20 10 868 <> 15 10 30 20 20 10 8 15
10 15 4 10 6
4 2 35 25 30
10 10 4 <> <>
35 20 10 15 20
<> 15 25 20 15
2 30 10 6 15
<> 20 15 30 10
35 40 25 30 30
20 6
8 20 20 10 30
10 30 10 25 20
15 10 20 25 15
20 20 15 35 40
10 10 20 10 15
10 15 10 15 15
8 20 15 20 15
10 20 20 10 20
15 20
15 20 30 20 35
20 15 15 25 20
20 15 10 35 20
20 8 8 15 35
20 15 20 15 10
8 20 10 25 15
15 15 15 20 10
10 15 20 10 15
15 30
15 15 20 45 30
20 45 30 30 60
40 50 35 80 70
60 45 50 40 20
50 20 25 15 25
6 20 30 20 15
15 40 20 50 30
25 30 30 45 15
35 60 45 35 40
20 30
o conllllns less than 2% of the U.S. RDA of the .. nutrients.
NOn;:: The U. s. RON _ he ... ~ prtrnarIy .. odullllnd _ ...... or other ~ woh spodIIl he.Oh considorallon~ this chldren o~r 4. For children under 4, .........,t or mntng. guide should be "UIIId with the help of a physician and/or
die .. ...,.
>:>~ (j ~ C;
15 4
20 4 4
6 6 4 4 8
8 10 30 10 15
45 6 8 10 6
4 6 4 6
25
20 8 15 8
50
4 4 10 10 15
8 8 2 4 8
8 10 15 6 10
20 25
20 15 25 35 30
20 25 20 20 30
15 15 25 30 20
35 15 15 15 15
35 10 10 25 30
8 25 40 20 35
8 30 20 30 30
15 20 20 35 15
15 25 25 10 20
15 30
For Best Nutrition Eat a Variety of Foods, Incladlng Those Not on This Chart.
NUTRITION INFORMATION
EXHIBIT 2
THE SUMMARY FORMAT
Add Stars to A Balanced Diet The star rarmg makt:!s It ~asle' to ~e at a glance whICh foods provKie "bener nutrition" based on rh. NUTRITION QUOTIENT tlmportant. See detailed explana· tion to the righll A food gets **** II Irs NUTRITION QUOT1ENT is above 10; *** If above 4. ** if above 2. and * if above I
For Better Nutrition The NUTRITION QUOTIENT of each product IS a way of measunng lhe
average "nutritional return" you get for each calorie you eat The nutritional return is based on the U S RDAs in a standardized serving for each of the following nutrients protein, vttamin A vitamin C. thiamine. riboOa\lin. niacin. calcium. and Iron. Using the NlTTRITION QUOTIENT formula. developed by J Edw.rd Ru5lO_ • NUTRITION QUOTIENT 01 Ilndlc ..... nutritional return equal to the cak>nes consumed Numbers higher than 1 mean more nutritional return per catone. but numbers lower than 1 mean Ie ..
ThIS chan IS pan of a reseerch p'O)ecl under Ih~ dnecnon of NSt)(lalt' Prof8SS0r J Edward Russo. Consumer 8ehaw)f Laboratory, Graduale School of Busineu. Unlvenily of Chicago. Chicago. illinoIS. 6Ob37 Products are not included on the list if an estimate tor each of the nUln~nt$. listed to the left was not available from either the manufacturer or the
Nutrition Quotient
* * 2.2
* 19
* 1 7
* 1.7
* 1.6
* 1.5
* 1.4
* 1.4
* 1.4
* 1.3
* 1.3
* 1.3
* 1.2
* 1.2
* 1.2
* 1.2
* 1.2
* Ll
* Ll
* 1.0
* 1.0
* 1.0
* 1.0 _9 _9
9 _9 .9 .9 _9
_8 _8 _8 .8 .8
_8 .8 _8 _7 _ 7
_7 _7 _ 7 _7 _7
_7 _6
U S_ D.A. Handbook "Nuhttlv. V.I .... of American Foods'" AddllionaU~_ new items or items not canted by most Slores are not Included Incluskm of any product on the list does not consdtute an endorsement For more informaHon. contact the Consumer Behavior Laboratory
TV DINNERS .f::~~
~~G Serving Size 1 Dinner ~o
<t Turkey. Banquet lUJO Veal Parmlglan. Banquet 11.00 Turkey. Man-Pleaser. Banquet 19_00 Turkey. Swanson _11.50 Chicken. Fried. Banquet 11.00
hallan. Banquet lUJO Sirloin. Chopped. Swanson 10_00 Salisbury Steak. Banquet 11.00 Beef. Banquet 11.00 Macaroni & Cheese. Swanson 12_50
Beef. Chopped. Banquet _11.00 Chicken. BBQ. Hungry-Man. Swanson 16_50 Beef. Swanson .11.50 Fish DInner. Banquet 8_75 Meat Loaf. Banquet 11.00
Meat Loaf. Man-Pleaser. Banquet _ 19.00 Turkey. Three Course. Swanson 16.00 Turkey. Hungry-Man. Swanson _19_00 Chicken. Hungry-Man. Swanson 19_00 Chicken. Man-Pleaser. Banquet 17.00
Chicken. Swanson _ .11.50 Mexican Dinner. Banquet ___ 12_00
Chopped Sirloin. Steak House 9_50 Saillbury Steak. Man-Pleaser. Banquet _. _. _ ... _19.00 ChIcken. Fried. Hungry-Man. Swanson _15_75
........ gna. Hungry-Man. Swanson _17_75 Beef, Hungry-Man. Swanson _17.00 SaIIebwy Steak, Three Course. Swanson .16.00 Beef Enchdacla. Swanson .- _______ . _15.00
ChIcken. Three Course. Swanson _15.00
Beef Chop Suey. Banquet _12.00 FWt 'N ChI .... Swanson _10_25 Beef, Chopped. Hungry-Man. Swanson _18.00 Chlcken.C~.SwahKm .. _____ 10.75 Beana And Franke. Banquet .10_75
Chicken. BBQ. Swanson 11.25 SpagheItt, Swanson 12.50 Chicken. Weam Style. Swanson 11.75 Rib Eye, StIrak House ............ 9.00 Meat l.oIIf, Swanson . _ 10.75
SaJllbury Steak, Hungry Man. Swanson 17_00 Beef Tenderloin, StIrak House .. - 9.50 Sirloin. StIrak House _____ 9_50 SaJIIbury Steak, Swanson ........... 11.50 Veal hrmItIIana. Hungry-Man. Swanson 20.50
ChIckea, Weam Style, Hungry-Man. SwahKm . _ 17.75 NoodIea And Chicken. SwalllOn -- 10.25
NOTE: Tho u.s. ROAo IoIod ... aM prtrnorIj/ 10 oIIuIIt one! chld.l_ 4-For ciIIo*n ....... 4, ....-or n ..... -. or oIhor poopIo wtth IpICIII
51
Fo ..... Nldiftlo • .EM. V ..... of Foode, lad ..... n.o •• Not oa ...... C ....
52
method used in this study was to display star symbols2
beside each summary value as shown in Exhibit 2. Stars were chosen because they are a positive symbol and familiar to most consumers from ratings of movies and restaurants. Consumers could read the accompanying explanation to understand the basis of the star rating.
Since the Summary format not only reduces the collection costs but also reduces the comprehension costs, it was our belief that this format would dominate the Matrix format in market acceptance and use. However, before we present the results of an experiment testing this premise, we will discuss how we constructed a summary measure.
Summary Rating of Nutrition: The Nutrition Quotient
The selection of an overall summary ofthe nutritional value of a food was based on several criteria. First, we limited it to calories and the eight leader nutrients listed on food labels-i.e., protein and the seven vitamins and minerals as shown in Exhibit 1. These are the only nutrients that have values available for nearly all brands. Second, the summary rating had to be objective, not a function of subjective judgments such as which nutrients are most important. Finally, it had to be transparent enough that typical shoppers could easily understand it. The design of such a measure required two decisions: in what units to measure each nutrient, and how to combine the nine individual values into an overall rating.
Unit of Measure. Each of the eight positive nutrients was measured in terms of its nutrient density, defined as the amount of that nutrient contained in a standard number of calories of the food. For each positive nutrient we began with the percent of the U.S. RDA contained in a single serving. For the ninth nutrient-calories-we assumed an average daily allotment of 2300 calories, as recommended by the American Medical Association (Council on Foods and Nutrition 1973). The nutrient density is then defined as the percent of the U.S. RDA of the given nutrient divided by the percent of the daily caloric standard contained in a serving of the food. For example, one serving of frozen peas contains 35 percent of the U.S. RDA of Vitamin C, 2 percent of the U.S. RDA of calcium, and 70 Calories (or 3.04 percent of the daily caloric allowance). Thus the nutrient density of Vitamin C in frozen peas is 35/ 3.04 = 11.50. The density of calcium in frozen peas is 2/3.04 = 0.66. The concept of nutrient density can be
2This system partitioned the NQ continuum into five categories, symbolized by 0 to 4 stars. One star is awarded if the NQ exceeds \, 2 stars if it exceeds 2, 3 stars if above 4, and 4 stars if above 10. To use our summary rating of nutrition, shoppers had only to count the number of stars and realize that more stars mean a more nutritious food.
THE JOURNAL OF CONSUMER RESEARCH
traced to the early 1900s; for a brief history see Wyse et al. (1976).
For any single nutrient the density 1.00 has special meaning. To assure consumption of 100 percent ofthe U.S. RDA of any nutrient, the nutrient density aggregated over all foods eaten in a day (the net nutrient density) should theoretically total at least 1.00 for the average person consuming 2300 calories. The U.S. RDAs are set safely above the minimum requirements; most people who consume a little less than the daily standard of a nutrient are still safe. Nonetheless, as the net density for a nutrient drops further below 1.00, there is an increasing risk of deficiency.
Note that a summary measure cannot easily be constructed for negative nutrients such as cholesterol, sodium, and sugar. The NQ relies on the comparability of a U.S. RDA across people and across nutrients. In contrast, the maximum desirable intake for negative nutrients varies with individual medical condition; no general standard can be set. This is why only the positive nutrients could be used to construct the Summary format.
Combination Rule. To combine the eight individual nutrient densities into an overall (summary) rating, we chose a simple average-i.e., a linear combination giving equal weight to each nutrient. Except for extreme differences in importance weights, a simple unweighted average is nearly as good as statistically optimal weighting of the individual components (Dawes 1979). Equal weighting of each nutrient also avoids controversial SUbjective judgments of differential importance, a feature that has proved valuable in other policy applications (e.g., Hammond and Adelman 1975). Combination policies nearly identical to our NQ have been suggested by Guthrie (1977) and La Chance (1975).
The Limitations of a Summary Measure
Intelligent use of the NQ requires a consumer to understand that it is only a guideline. Its advantage as a simplified measure of total nutrition comes at the expense of precise information about individual nutrients. Thus it is quite possible to select a daily menu with a composite NQ well above 1.00, yet ingest less than 100 percent of the RDA of at least one nutrient. No simple summary measure can ever guarantee a nutritionally balanced diet; only a rule that is based on each nutrient can accomplish this.
On the plus side, the presentation of the NQ to shoppers may improve the nutritional quality of their food purchases by lowering the computational costs and thereby increasing the use of nutrition information during their purchase decisions. We believe that this is the more important standard. The value of the NQ should be judged not by whether it can guarantee an ideal diet, but whether it helps consumers move closer to that ideal.
NUTRITION INFORMATION
The Result of Information Provision: A Hierarchy of Effects
There is some debate over whether the goal of an informational program should be changes in cognition or changes in behavior (Bettman 1975, 1979). In this study, which deals with health related issues, we take the view that the behavioral goal is by far the more important. Nonetheless, even though a change in purchases may be desired as the ultimate effect, prior cognitive effects must be achieved (Mazis et al. 1981). First, shoppers must become aware of the new information. Only then can either greater knowledge of nutrition or a more favorable attitude toward nutrition take place. The latter would presumably be reflected in an increased salience of nutrition during purchase decisions. Only if knowledge or attitude are improved would we expect a change in food purchases. This conceptualization of sequential impact is the familiar hierarchy of effects (Preston 1982) and will be used to test the results of our first experiment.
EXPERIMENT 1 Overview
The main experimental manipulation was the posting of alternative formats of nutrition information in different supermarkets of the same chain. Over a period of time this information had the opportunity to influence shoppers' awareness, knowledge, attitude, and purchases. The effects on awareness, knowledge, and attitude were measured for individual shoppers intercepted in the store during check out. Purchase effects were determined from a complete record of store sales. The remaining aspects of the design involve a special "promotion," two variations in conveying the information, and the time sequence of the information presentation. Each of these design variables will be discussed, and the full design is summarized in Table 1.
Store Selection Jewel Food Stores provided access to a pool of sub
urban supermarkets that were patronized by middleclass shoppers. These stores were screened on store size, hours of operation, and each community's overall socioeconomic status (SES) ranking (based on median income, education, average home value, average rent, and assessed property valuation). Only stores located in the areas that fell in the middle two quartiles of this SES ranking qualified as middle class. Based on these characteristics, we selected 14 closely matched middle-class stores. These 14 were randomly assigned to the 12 experimental and two control treatments.3 In addition,
3The design as described is actually only half ofthe complete study. Also tested was a second set of 14 stores from low SES minority neighborhoods in the city of Chicago. The low SES data will not be reported because they add little to the middle SES results. Generally, the low SES effects were similar but much smaller.
53
Jewel provided weekly records of inventory movement for every brand and size within the product categories tested.
Product Categories The information programs were tested on six product
categories: breakfast cereal, frozen vegetables, canned soup, canned and bottled fruit and vegetable juice, canned and bottled fruit, and frozen TV dinners. These product categories represent a variety of packaged food products. All exhibit a large range of nutrition quotients; we avoided categories with negligible nutritional differences, such as peanut butter. Moreover, we defined product categories broadly rather than narrowly-e.g., all frozen vegetables not just all brands of frozen peas. As a result, at least one item in each product class is purchased by over 50% of all U.S. households.
Source of Nutrition Information For the six product categories, we were able to obtain
complete nutrition information on 94 percent of the 498 nutritionally distinct items that passed through Jewel's delivery records.4 The remaining six percenttypically mixed foods such as Clamato (tomato and clam) juice, frozen mixed vegetables, and "snack packs" of assorted breakfast cereals-was excluded from the study. Thus, the products with posted nutrition information were: breakfast cereal, 94; frozen vegetables, 128; canned soup, 63; bottled juice, 62; canned fruit, 74; and TV dinners, 47.
Information Formats The primary independent variable is the format of
the nutrition information. In addition to the Matrix (Exhibit 1) and Summary (Exhibit 2) formats, two others were tested. The Complete format is a composite of the first two: it provides both the nine individual nutrients and an overall nutrition rating, the NQ, and its star category. Such complete information should enable a consumer to better understand a food's NQ rating by showing the basis for a high or low NQ. The Complete format comes closest to the ideal of "full disclosure" and is probably the most attractive from a public policy perspective.
The fourth format-Intermediate-is a partIal summarization of the Matrix information. The seven vitamins and minerals are aggregated into an average percentage ofU .S. RDA. The Intermediate format contains three nutrition attributes: calories, protein, and an average of the seven vitamins and minerals. It does not contain the NQ.
4Given that this information was "publicly" available, one way of viewing the nutritional lists is that they substantially lower a consumer's search costs.
54
Information delivery mode
Baseline (no information) (Weeks 1-4)
Posting-only (Weeks 5-17)
Posting-plus-promotion (Weeks 18-33)
Posting-extended (Weeks 18-33)
Control (no information)
(Weeks 5-33)
Take-home copy
yes
no
yes
no
yes
no
yes
no
yes
no
yes
no
Matrix
Store 1 PUR, INVW
Store 2 PUR
(N)
(N)
Store 1 PUR, INVW
(N)
(N)
(N)
X
X
(N)
(N)
THE JOURNAL OF CONSUMER RESEARCH
TABLE 1
DESIGN OF EXPERIMENT 1
Information format
Summary Complete
Alphabetic
Stores 3-4 PUR
Store 5 PUR
Store 6 PUR Store 7 PUR,
INVW
Store 8 PUR, INVW
Store 9 PUR, INVW
X
Numeric
Store 10 PUR,INVW
Store 11 PUR
Alphabetic
X Store 9 PUR, INVW
(N) (N)
Numeric
Stores 6,8 Store 11 PUR,INVW PUR,INVW
(N) (N)
Alphabetic
X X
X X
Numeric
Store 7 PUR, Store 10 INVW PUR,INVW
X X
Intermediate
Store 12 PUR,INVW
X
(N)
(N)
Store 12 PUR,INVW
(N)
(N)
(N)
X
X
(N)
(N)
Control
Stores 1, 7-10, 12-14 PUR, INVW Stores 2-6,11 PUR
Stores 13-14 PUR,INVW
NOTE: The symbols are PUR: Purchase record data; INVW: Interview data; (N): Nonsensical combination of factors; X: No data collected.
Each of these four formats actually consists of three components: a heading, followed by text providing justification for the information, and finally, the specific nutrient information. Each component differs slightly across formats. For the Summary format, the headline
says, "Add Stars to a Balanced Diet for Better Nutrition." The text then describes how the NQ measure was obtained, and the U.S. RDA is mentioned but no specific definition is given. For the Matrix format, the headline says, "Choose a Balanced Diet for Better Nu-
NUTRITION INFORMATION
trition." The following text then omits any discussion of the NQ since it is not part of the information displayed. The Complete format employs the same headline and text as Summary, while the Intermediate format uses the heading and text of Matrix.
Other Information Program Variables Numeric versus Alphabetic Ordering. The Summary
format in Exhibit 2 is organized by Nutrition Quotient-i.e., the foods are ordered from the highest NQ to the lowest. Although such an ordering makes it easy to find the highest or lowest NQ foods, some shoppers may want to find the NQ of their favorite brand or of the only five brands everyone in their family will eat. For this purpose an alphabetical list may be easier to use. Also, an alphabetical list appears less prescriptive, more neutral, and may be more readily accepted.
We tested both the numerically and alphabetically ordered versions. It should be noted, however, that since the Matrix and Intermediate formats did not include summary ratings, there was no logical way to numerically order the brands in these two displays. This has two implications. First, given the nonsensical nature of the Matrix-Numeric and Intermediate-Numeric design cells, there were only six applicable format by ordering conditions. Second, since it was inappropriate to use a complete factorial design of the format and ordering variables, these two variables are partially confounded. The independent contributions of each will have to be assessed via statistical analysis rather than directly from cell means.
Take-Home Copies. The availability of a take-home copy of the nutrition information should enhance the effect of the lists. The supermarket aisle is hardly the ideal place to study nutrient lists, especially long ones. Also, other family members might need to be involved. For these reasons we tested whether take-home copies of a list would increase its impact. The presence or absence of take-home copies doubled the number ofpossible experimental treatments to 12. These were the two take-home conditions combined factorially with the six format by ordering conditions. Testing these treatments required 12 different supermarkets. Two control stores brought the total number to 14. These stores and their respective treatment conditions are displayed in the Posting-Only block of Table 1.
Physical Display The nutrition information was printed on 24 X 18
inch posters. For the two product categories with the largest number of items-breakfast cereal and frozen vegetables-two sides of a poster were required to present all of the information in the Matrix and Complete formats. In all other cases, complete information for all items was printed on one side and then duplicated on the other side. In a Take-Home treatment, a small,
55
bright green tag saying "Take Home a Copy" was placed on each poster. These posters were hung on "over-thewire" hangers that protruded from the shelving facing shoppers as they walked down the aisle. The bottom of each poster was approximately five feet above the ground. One, two, or three posters were hung in each product category, depending on the total linear feet of shelving.
The take-home copies were 8.5 X 11 inch replicas of the posters. They hung from the shelving directly below a poster. When only one side was needed to present the nutrition information, the reverse side was used to explain the role of the eight nutrients.5 For the two product classes requiring both sides to list the Matrix and Complete information, the take-home copies were expanded to two double-sided sheets. Two sides were used to duplicate the posted nutrition information, one side had the added explanation, and the fourth side was blank. Each week the condition of the posters was monitored by a member of the project staff. Whenever a poster was missing or damaged it was replaced. Take-home copies were replenished as needed.
Special Promotion: Handout Copies The final design variable was equivalent to a special
promotion designed to increase consumer awareness of the nutrition lists in the store. As with the take-home copies, the special promotion should intensify the effects of the specific nutrition list being tested. This special promotion was accomplished by handing out a takehome copy of the breakfast cereal list to shoppers as they entered a store. (Breakfast cereal was chosen because it had the highest household purchase rate of the six product categories.) This handout was performed by professional interviewers who also attempted to convey the benefits of the information in a 30-second explanation. This special promotion resulted in roughly one-third of the regular shoppers in each store receiving the handout (73 percent of whom also received the 30-second recitation of benefits). It took place during the mid-portion of the 33 weeks of the experiment and helped to partition it into three distinct time phases.
These time phases were as follows. The first four weeks were used to assess initial store differences; no stores received the posters or promotion. During weeks 5-17 the nutrition information was posted without any special promotion, with one treatment per store. This phase of the experiment is referred to as Posting Only. In weeks 18 and 20 the special promotion was conducted in a subset of stores. This intense information delivery (in addition to the posters) is referred to as the
SThe back side consisted of the headline, "The nutrients represented on the charts-why are they so important," followed by text that told what a nutrient does, a warning that it is important to eat a variety of foods each day, and specific information on the values and food sources of each of the eight nutrients.
56
Posting-Plus-Promotion phase of the experiment. This phase ran from weeks 18 through 33 in six treatment stores. Finally, two of the stores used in the PostingOnly phase retained the posters but did not receive the special promotion. These two stores comprised the Posting-Extended condition, which also ran during weeks 18 through 33. In summary, four changes took place at week 18.
t. Six of the 12 treatment stores received a special promotion by handing out take-home copies to shoppers as they entered the store during Weeks 18 and 20. These stores are referred to as Posting-PIus-Promotion.
2. These 6 stores were classified as Take-Home stores since it was nonsensical to continue the No-TakeHome treatment once take-home copies were handed out to customers. As a result, the estimated effects of the Take-Home variable are restricted to the PostingOnly portion of the experiment.
3. Two of the 12 original treatment stores retained posters without the special promotion. These two stores are referred to as Posting Extended.
4. The remaining four stores were dropped from the experiment.
Table 1 shows the identity of the specific stores in each phase.
In-Store Survey of Shoppers As mentioned earlier, there were 12 treatment and
two control stores in the initial Posting-Only phase of the experiment. For each of these stores, Jewel supplied complete sales data for each brand and size for each week. These data were supplemented by an in-store survey designed to assess shoppers' awareness, knowledge, attitude, and purchases, as well as selected demographic characteristics. Because of cost constraints these in-store surveys were restricted to only six treatment stores, selected to enable estimates of the effects of the four information formats, the Numeric versus Alphabetic ordering (for the Complete format), and the presence or absence of the take-home copy (for the Summary format). In addition, both control stores were surveyed, but at one-half the normal frequency, and then the responses were combined. This process of interviewing six treatment stores plus the two control stores was continued during the subsequent PostingPlus-Promotion phase. In addition, the two stores placed in the Posting-Extended condition were surveyed.
Measures Taken Awareness. To determine awareness, shoppers were
asked if they had seen any nutrition posters and, if so, in which product categories they had seen them. They responded yes or no to 12 product categories, the six for which the information was actually posted and six
THE JOURNAL OF CONSUMER RESEARCH
distractor categories. From these responses we derived a measure of "proven" awareness ranging from O-no awareness to 6-maximum awareness.
Knowledge. Three types of nutrition knowledge were assessed: Basic, Advanced, and Comparative. Basic nutrition knowledge was measured from shoppers' responses to two questions asking "which of the [following) statements you think best describes what U.S. Recommended Daily Allowance is" (the amount and kind of nutrition in the food in contrast to all ingredients or unit cost), and "what you think nutritious foods are" (high in vitamins, minerals and protein in contrast to low in calories, high in fiber, or low in harmful ingredients). Advanced knowledge was measured by averaging the responses to three questions: (1) whether too much of some vitamins can be harmful (yes); (2) whether eating a variety of foods is ordinarily sufficient (yes); and (3) whether fortification of seven vitamins and minerals is sufficient (no). Finally, Comparative nutrition knowledge was determined from four pairwise comparisons of overall nutrition-e.g., apple juice versus tomato juice (tomato is superior).6
Attitude. Attitude toward nutrition was assessed by averaging the responses on a five-category bipolar agreement scale to two statements: "Supermarkets should provide nutritional information even if it adds a few cents to prices," and "There's too much talk these days about what's good and what's bad for you when it comes to food." (The direction ofthe latter statement was reversed before its responses were averaged.) In addition, we assessed shoppers' attitudes toward the helpfulness of the store: "In general, do you think that supermarkets like this one are doing a great deal, something, or not too much to help you buy nutritional foods?"7
Purchase. Interviewers also audited a shopper's market basket for all purchases in any of the six product categories tested. Only shoppers who purchased at least one product on the nutrition lists were accepted as qualified experimental subjects. In addition, the marketbasket audit enabled a check of interviewer accuracy as described in the Appendix. Note that all data analyses of actual purchases are based on complete sales records
6None of the information supplied on either side of the take-home copies directly provided the answers to the Basic or Advanced knowledge questions. However, the reverse side did say that it is important to eat a variety of foods every day (a component of Advanced knowledge), and the front side used the term U.S. RDA (a component of Basic knowledge). Information used in the Comparative knowledge measure was directly available from the Summary and Complete formats. We discuss these issues in more depth in the results section.
7 A reviewer pointed out that some shoppers may have interpreted this question differently than was intended. Instead of assessing the helpfulness of the store they were shopping in, the phrase "supermarkets like this one" may have signalled other supermarkets-i.e., not this one.
NUTRITION INFORMATION
provided by the store and not on the market-basket audits.
Demographics. Also recorded were standard sociodemographic characteristics (e.g., age, education, income, race, and sex) and several health-related facts. The latter included any health problems that required avoiding calories or seeking vitamins or minerals and an unobtrusive estimate of body size (very heavy, somewhat heavy, normal range, somewhat thin, and very thin). The order of the questions was: (1) a mixture of attitude, Basic, and Advanced knowledge, (2) Comparative knowledge, (3) awareness, (4) demographics, and (5) the market-basket audit.
Interviewing was done 12 times, twice during the baseline period, five times during weeks 5-17, and five times during weeks 18-33. Within each store shoppers were selected randomly with two qualifications: a minimum of 18 years of age and at least one purchase in the six experimental product categories. Interviewing was completed in the afternoon and early evening of a single day. The day of the week was selected from among Wednesday through Saturday and balanced over treatment. Each administration of the questionnaire yielded about 40 completed interviews per store. The total number of interviews was 3,254. Of these, 892 were untreated shoppers-i.e., in experimental stores during the pretest phase or in the control stores. The data from the 463 shoppers in the Posting-Extended phase were used only to estimate time trends, and were excluded from other analyses.
RESULTS
The analysis begins with the impact of the nutrition information on knowledge and attitude and then examines the actual purchases. It is designed to answer two main questions: did the nutrition information have any impact and, if so, which information formats and other factors were the most effective? Because of a problem with interviewer bias (see the Appendix), we had to exclude all data collected from Store 12. As is clear from the design shown in Table 1, this meant that we were unable to estimate the effects of the Intermediate format. Thus we will compare the Matrix, Summary, ann Complet~ formats only.
Plan of Analysis
Since the statistical analyses for the awareness, lcnowlp,dge, and attitude measures are essentially iden-
cal, we describe the basic plan before presenting any I ~s llt<;. Our original design was not orthogonal but partie, 1 y confounded. Therefore, we used an analysis of COvariance to estimate the influence of the four treatment variables: format, information delivery mode (Posting Only and Posting Plus Promotion), ordering (Alphabetic and Numeric), and take-home copy (Take-
57
Home versus No-Take-Home). We used a linear model comprised of two submodels: (1) a control sub model of demographic characteristics and historical time parameters; and (2) a treatment submodel with eight terms-three formats by two information delivery modes (i.e., six parameters) and single parameters for the ordering and take-home variables. For each of these analyses, data were used from both control and treatment stores covering both the baseline and treatment phases. Conceptually, the control stores were used to control for time effects while the baseline data were used to control for individual store effects.
Instead of reporting estimates of the individual model parameters, we present "average" effects calculated from the parameter estimates (i.e., the regression coefficients and the resulting covariance matrix). Specifically, we substituted in the values for the average consumer surveyed, removed any historical time trends (as determined from the control stores), averaged over format (for the two information delivery modes), averaged over delivery modes (for the three formats), assumed that take-home copies were available, and for the Summary and Complete formats assumed that the order was Numeric. These latter two assumptions implicitly reflect our prior belief that take-home copies and Numeric ordering are the more effective methods of delivery.
Such an approach allows us to test the impact of each of the four independent variables assuming the other three are held constant. However since our design was not orthogonal, each of the independent variables shared some joint variance. Statistically this meant that not all of the explanatory power from the combination of predictors could be assigned to one variable alone. As a result, the statistical significance of each of the independent variables was based on a partial F-test derived from adding it as the last term in the treatment model.
The nonorthogonality of the design also restricts the interpretation of the ordering and take-home copy variables. The Numeric versus Alphabetic difference was nested within the Complete format (see Table 1). Thus, the interpretation of ordering effects is confined to this format; generalization to other formats is speculative. Also, Take-Home versus No-Take-Home is nested within the Summary format and the PostingOnly phase, and its interpretation is similarly restricted.
Differences among the average effects ofthree formats were tested using a Scheffe test (Neter and Wasserman 1974) modified to account for the intercorrelations among the three estimates (since they are computed from the same underlying treatment and control parameters). The largest of the three variances of the pairwise differences (where each pairwise variance included the correlation between the two estimated effects) was used to derive a conservative MSE (mean square error). This conservatism, and that inherent in the Scheffe test, were partially compensated for by setting the criterion
58 THE JOURNAL OF CONSUMER RESEARCH
TABLE 2
CHANGES IN KNOWLEDGE AND ATTITUDE
Nutrition knowledge Attitudes
Treatment
Overall effect
Posting only Posting-plus-promotion
Matrix Complete Summary
Numeric ordering Take-home copy
Untreated (initial level)
• p < 0.05, two-tailed.
Basic Advanced
.OS* .OS*
.20*· .04· -.04b .12*b
.12*· .13*·
.05· .10*·
.07*· .00b
-.07 .OS* .15* -.OS
.S4 .S1
Store Comparative Nutrition helpfulness
.OS* .17* .05*
.04· .17*· .03*·
.OS*· .1S*· .07*·
.OOS .43*· .15*·
.14*b .17*b -.OS*b
.05· -.10'" .Osc
.1S* .SO* -.25*
.1S* -.1S* -.07
.44 .S2 .57
NOTE: Tests lor inlormation delivery mode and lormat effects on knowledge are one-tailed, since only positive effects should have occurred. (The one observed negative effect would not have been significant il a two-tailed test had been performed.) All othar tests are two-teiled. In each column values with different superscripts are stetistically different from othar levels ollhe same lactor at p < 0.10.
for statistical significance at p < 0.10 rather than p < 0.05. The difference between the two information delivery modes (Posting-Only and Posting-PIus-Promotion) was tested by a t-test that also reflected the intercorrelation of the two estimates. For format and delivery mode, reliably different levels are indicated by different superscripts in the appropriate table. These superscripts are set equal (and no tests performed) if the conservative partial F-tests described earlier yield a nonsignificant effect for that variable as a whole.
Changes in Nutrition Knowledge Three types of nutrition knowledge were assessed:
(1) Basic, (2) Advanced, and (3) Comparative. All three measures can be interpreted directly as the proportion of correct responses to the pertinent questions of the in-store survey. For instance, 1.00 represents a correct answer to every question comprising the measure, and all the reported change scores represent increases (or decreases) in the proportion of correct responses. The tests for changes in knowledge were based only on the subset of shoppers who scored positively on the awareness test-i.e., who were aware of at least one of the six nutrition posters. There should be (and was) little knowledge change for the unaware shoppers. The restriction to aware shoppers means that the following analyses of knowledge changes report the effectiveness of formats conditioned on awareness, specifically on consumers' awareness of at least one nutrition poster. The changes in knowledge, including statistical tests, are reported in Table 2.
Overall Effect. The overall effect of the information treatment was significant for all three types of knowledge. As a standard against which to judge the magnitude of the estimated changes, Table 2 displays the absolute levels observed for the untreated shoppers. Over all treatment conditions nutrition knowledge was in-
creased 22, 21, and 11 percent of the maximum possible for Basic, Advanced, and Comparative, respectively.
Posting-Only versus Posting-Plus-Promotion. The effect of information delivery mode on Advanced and Comparative knowledge generally conforms to our expectations, namely the superiority of Posting-Plus-Promotion over Posting-Only (significant for Advanced, but not for Comparative knowledge). We have no explanation for the contrary result for Basic knowledge.
Format. Before examining which format was most effective in increasing nutrition knowledge, let us review what was expected. Since none of the three formats provided direct answers to the Basic or Advanced knowledge questions, any format differences should be associated with ease of comprehension and use. 8 Then the effort reduction hypothesis predicts that the Summary format should be the most effective, possibly joined by Complete, while Matrix should be the least effective. This pattern should hold especially for Comparative knowledge where shoppers were asked to compare different foods within a product category-e.g., apple and tomato juice. Although the Matrix format gathers the individual nutrients into one place, it does not fully facilitate such comparisons. In contrast, the Summary and Complete formats make interbrand
80ne of the Advanced knowledge questions asked whether it is difficult to get enough nutrition just by eating a variety of foods. A reviewer pointed out that the Summary and Complete headline used the phrase "Add Stars to" (versus the Matrix phrase "Choose") a balanced diet. It is possible for the former phrase to leave the impression that it is necessary to augment (add to) one's intake to get a balanced diet. Such an interpretation would imply that the Matrix format would score higher on Advanced knowledge. The same reviewer pointed out that the Complete and Matrix formats equate the term U.S. Recommended Daily Allowance (used in the Basic knowledge question) to the abbreviation U.S. RDA while the Summary format does not. Such a difference might cause the Matrix and Complete formats to score higher on Basic knowledge.
NUTRITION INFORMATION
comparisons quite simple, since the NQ summarizes over all nutrients.
The data in Table 2 confirm our expectations only for Comparative knowledge, where both the Summary and Complete formats are superior to Matrix (which shows no effect whatsoever). Note, however, that only the superiority of Complete is statistically significant. For Basic and Advanced knowledge, the effort reduction hypothesis was not supported. Matrix is superior to both Summary and Complete, though not significantly so except for Advanced knowledge and the Summary format. We discuss the possible reasons for this result shortly.
Numeric Ordering. For Comparative knowledge the Numeric ordering should prove superior to Alphabetic because the former facilitates the essential interbrand comparisons. In accord with this prediction, the observed effect (0.16) is both large and significant. No differential effects on Basic and Advanced knowledge were expected. For Advanced knowledge, however, the Numeric ordering did prove reliably superior.
Take-Home Copy. As might be expected, the presence of a take-home copy aided Comparative knowledge. When a reduced copy of the posted list could be taken and read at home, Comparative knowledge increased more than when such copies were not available. The size of the effect (0.18) suggests that the take-home copy played a major role in the increase in Comparative knowledge. Moreover, it implies that shoppers learned which foods provide superior nutrition less by reading the nutrition posters in the store than by taking a copy away with them. For Basic and Advanced knowledge, we presumed that take-home copies would be helpful but not essential (see Footnote 6). Only the former showed a significant effect.
Conclusion. Making nutrition information more readily available increased all three types of nutrition knowledge, although the effects differed by type of knowledge and format. Contrary to the effort reduction hypothesis, the Matrix format was not inferior to the Complete or Summary formats. In fact it scored highest on these measures (although the difference was not always statistically significant). Likewise there was no consistent effect of Numeric ordering or Take-Home Copy. For Comparative knowledge, just the opposite was found. The Matrix format did not provide the needed information and failed to produce any knowledge increase, while significant effects were found for Complete, Numeric ordering, and Take-Home Copy. In each case these treatments enhanced the acquisition of the relevant information needed to make comparisons between different alternatives in a product class.
Changes in Attitudes Two attitudes, Nutrition Attitude and Store Help
fulness, were assessed. Both attitude measures ranged
59
between 0 and 1, with 0 most negative, 1 most positive, and 0.5 signifying indifference. All attitude scores reported in Table 2 are changes on this scale.
Nutrition Attitude. The overall effect of the nutrition information on Nutrition Attitude was large and positive, showing an increase of 0.17 above a base level of 0.62. There was only a negligible difference between Posting-Only and Posting-Plus-Promotion.
However, the format differences are striking. The Matrix display is clearly the most effective at generating favorable attitudes toward nutrition. Just as noteworthy is the significant negative impact of the Summary format. Those shoppers exposed to the Summary format exhibited a more negative attitude toward nutrition with a decrease of 0.10 on the 0 to 1 scale. We discuss this result shortly.
Both ordering and take-home copy caused significant attitude changes. A Numeric ordering was much more effective than an Alphabetic one. Apparently consumers found it more useful when viewing the Complete posters. (Recall that the contrast between Alphabetic and Numeric ordering was confined to the Complete format.) The significant negative effect of a take-home copy is part of the story of the Summary format, to which we now turn.
The Rejection o/the Summary Format. We believe that the negative change in attitude generated by the Summary format is the key to why it was so ineffective. This format is being rejected. The net impact is not neutral, as if the information were useless, but negative, as if shoppers don't want nutrition information presented this way. This rejection hypothesis is consistent with the relatively small effects on knowledge, and it may also explain the intermediate position of the Complete format. Adding the NQ measure to a Matrix poster to produce the Complete format reduces the Matrix's efficacy. The rejection interpretation is also consistent with the negative impact of a take-home copy on Nutrition Attitude. Recall that this factor was measured only for the Summary format (see Table 1). Consequently, its negative impact reflects the fact that offering more ofthe Summary format via take-home copies only makes things worse. If such an interpretation is correct, the negative effect of Take-Home Copy should not generalize to formats that generated positive attitude changes. Thus, the effect of a take-home copy on the Matrix and Complete formats would presumably have been positive. It is our belief that the observed negative Take-Home effect is derived from and is confined to the Summary format.9 We consider further what might
9We considered the possibility that at least part of the rejection of the Summary might have been a reaction to the handout manipulation. That is, shoppers' negative attitude to the Summary format may have been compounded by how they received this informationactively prof erred by interviewers in the Posting-Plus-Handout presentation mode. For the Summary format alone the attitude changes
60
have caused the rejection of Summary in the General Discussion.
Store Helpfulness. Does the impact of nutrition information reflect directly on the supermarket chain that provides it? Initially we expected either a null or positive effect, but the observed decreases in Nutrition Attitude may have spread to shoppers' attitude toward the store. The findings are reported in Table 2.
The overall impact is significantly positive, with a larger effect for Posting-Plus-Promotion than PostingOnly (but not significantly so). Again there are differential effects for format, with Matrix superior to both Complete and Summary. There are also two unexpected effects, the negative impact of the Complete format and of Numeric ordering. We can offer no plausible explanation for either, but note that the two estimates are highly correlated due to our design. The most important result, however, is simply that the nutrition information generated significant customer good will. It appears that the store gets credit for providing the nutrition information, especially in the most attractive format, Matrix.
The Role of Awareness
Earlier we described a hierarchy of effects with changes in knowledge and attitude preceded by awareness (and followed by purchases). The preceding analyses (reported in Table 2) were conditional on the fact that shoppers were aware of at least one poster. This raises the question of whether the Matrix format was also best at generating awareness or whether the differential effects on knowledge and attitude occurred only after holding awareness constant.
Direct Observation of Eye Fixations. The first step in achieving awareness is to insure that shoppers look at the nutrition posters hanging in the aisles. To assess actual looking rates, we videotaped shoppers for 16 hours as they passed by a single nutrition chart (for canned fruit) posted in an additional matched store. Four treatments (Matrix, Summary, Complete, and Intermediate), four days of the week (Tuesday through Friday), and four times of the day (1-2 p.m., 3-4 p.m., 4:30-5:30 p.m., and 6-7 p.m.) were combined in a Latin square design. The video camera was mounted above the shelving, unnoticed by shoppers as they walked down the aisle.
The proportion of shoppers fixating the poster is reported in Table 3 for the three formats with usable data from the in-store interviews. A chi-square test for differences among the formats yields only a marginally
for the Posting-Only and Posting-Plus-Handout conditions were -0.13 and -0.07, respectively. The difference between these values is neither significant nor in the predicted direction. Thus, it would seem that the negative effect of the Summary format cannot be attributed to how shoppers received the information, but to the format itself.
THE JOURNAL OF CONSUMER RESEARCH
TABLE 3
EYE FIXATIONS ON THE NUTRITION POSTERS
Probability of reading
Proportion Proportion given a Format fixating reading- fixation
Matrix .068 (176)b .023 .33 Complete .021 (191) .005 .25 Summary .045 (155) .013 .29 Mean .045 .014 .29
• Reading requires a minimum fixation duration of 1 second. • The numbers In parenlhesas are sample sizes, I.e., the number of shoppers who passed
down the aisle where the poster hung.
significant effect: chi-square (2) = 4.86, (p = 0.09). Nonetheless, the Matrix poster attracted 50 percent more eye fixations than the Summary format, in general agreement with the results for knowledge and attitude changes. It is also worth noting that the absolute eye fixation rates are not high; they comprise less than 5 percent of all shoppers passing down the aisle and only eight percent of shoppers interested in the product category, as indicated by their looking at the shelves with canned fruit for at least one second. Also, fewer than two percent of shoppers (three percent of interested shoppers) actually read the poster. (Reading is defined as fixating for more than one second.) These rates should be neither surprising nor discouraging. The nutrition posters-black and white charts of names and numbers-are hardly eye-catching by the standards of point-of-purchase advertising. However, over many weeks and exposure opportunities the cumulative awareness level should exceed the few percent observed here.
Surveyed Awareness The in-store interviews provided a measure of cu
mulative awareness that ranged from 0 (indicating no awareness) to 6 (the maximum number of posters that could be seen). This measure was partitioned to capture both breadth and depth of awareness. The proportion of shoppers aware of at least one poster is a breadth measure. The number of posters correctly identified, given some (nonzero) awareness, is a measure of the depth or intensity of awareness. The distinction between breadth and depth is analogous to trial and repeat purchase. Awareness of a single poster may represent trial in the passive sense of an incidental encounter with the information. This is especially likely in the Posting-PlusPromotion phase of the experiment. In contrast, awareness of more posters indicates an increasingly deliberate, active seeking of the information. Both awareness measures were analyzed by the plan described earlier. We report here only the main findings (see Russo et al. 1985 for more detail).
NUTRITION INFORMATION
In terms of breadth of awareness-i.e., generating any awareness at all-the Matrix and Complete formats were superior. The respective proportions of shoppers aware of at least one poster were 0.39 and 0.42, both of which are significantly above the 0.19 proportion for Summary. Turning to the depth of awareness-i.e., the number of posters shoppers were aware of (for those who were aware of at least one)-we found the following: 5.3 for Matrix, 3.3 for Complete, and 2:5 .for Su~mary. All three pairwise differences are statIstIcally SIgnificant. This latter pattern of results clearly foreshadows the knowledge and attitude changes (except for Comparative knowledge), with Matrix superior to Summary, and Complete typically intermedi~te. Thus, not only do we find the Matrix format supenor to t~e Summary in generating some awareness, but once thIS awareness is generated, shoppers tend to seek out these posters. We believe it is the latter effect that leads to the differential changes in Basic and Advanced knowledge reported in Table 2. The only exceptio~ is Comparative knowledge. As noted above, the Matnx format is essentially useless in communicating comparisons among foods in the same product category.
Time Course of Effects
Were the effects of the nutrition information mainly transitory, drawing initial attention but declining as the study wore on, or were they slowly but steadily increasing throughout, suggesting greater long-term effec~s than observed here? The time course can be most dIrectly measured in terms of awareness, which in turn drives the changes in knowledge and attitude. We analyzed both the proportion of aware shoppers and the number of posters they were aware of. Because the resu~ts for the latter variable duplicate those for the proportIOn of aware shoppers in all important respects, we report only the former.
To assess the time course of awareness, we added linear and quadratic time trends for both the PostingOnly and Posting-Plus-Promotion phases to the model described earlier. We also added the data from the two Posting-Extended stores, which allowed us to estimate the time effects of the Posting-Only delivery mode over the full 29 weeks. In this expanded model the different formats were still represented by single parameters, which now signify different level shifts of a common time trend. In a second model, we estimated time trends individually for each of the three formats using a total of 12 time parameters. The improvement in explained variance due to this second model was at chance level; the eight additional parameters had a partial F < 1.00 (p > 0.50). Thus, we report only the results of the first model where the time trend is estimated separately for Posting-Only and Posting-Plus-Promotion, but is assumed to be common over the three formats. These estimated trends are displayed in Figure A.
The Posting-Only time trend is not significantly dif-
FIGURE A
TIME COURSE OF NUTRITION INFORMATION AWARENESS
Proportion of shoppers aware of nutrition information
.6 Posting
.5
B Complete o Matrix * Summary
Post-plus-promotion t Complete
.4 • Matrix * Summary
.3
B 0
B 0
0 * .2
~ B .1 *
t
~ * t * t t * t B • * * B B
1t ~ * * B *
B
0
61
20 40 60 80 IT 100 IT 120 140 160 180
Handout Time (days)
ferent from a horizontal line. The Posting-PIus-Promotion trend is significantly nonhorizontal; for the linear term, t = 2.06 (p < 0.05), and for the quadratic term, t = 1.87 (p = 0.07). Both are two-tailed tests. A sense of the variability about these estimates can be gathered from the raw proportions, which are also plotted in Figure A. These observed values are adjusted only for main effects of format within delivery mode. (They differ somewhat from the estimates reported earlier because the present values are not adjusted by the control model and because the earlier ones did not include the Posting-Extended data for the Summary and Complete formats.)
The story told by the time trends seems straightforward. For the Posting-Only phase, a substantial portion of the maximum breadth of awareness is immediate. After one week the proportion of aware shoppers averaged over formats is 16 percent; it reaches a maximum of 20 percent 16 weeks later. Both the rise and decay of awareness are slow and not significantly different from zero.
For the Posting-Plus-Promotion delivery mode, the time trend is quite different. The proportion of aware shoppers averaged over formats increases rapidly from an estimated 33 percent one week after the first handout to a maximum of 54 percent after seven weeks, and then decreases rapidly. The handout is successful in broadening awareness, but like many promotions, without repetition its impact decays.
Who Was Affected by the Nutrition Information?
Were the shoppers who were most influenced by the nutrition posters those who most needed nutrition
62
knowledge, such as people with diet-related health problems? Or do we find the familiar result of the rich getting richer, or in this case, the educated getting more educated?
We estimated the increase in nutrition knowledge for those shoppers who reported that they needed to eat fewer calories or more nutrients (about nine percent of the sample) relative to those who did not. This analysis is described in full in Russo et al. 1985. The differences between the target and normal groups were: 0.10 for Basic knowledge, 0.01 for Advanced knowledge, and 0.03 for Comparative knowledge. In all three cases, those who needed the knowledge actually learned more. Similar, though weaker, effects were found when shoppers rated Very Heavy (nine percent of the sample) or Very Thin (three percent) were compared to those in the normal range of body size. And even larger, equally consistent effects were found for the two attitudes.
In contrast, there were smaller, often negligible, effects for traditional demographic predictors like education and income. Though shoppers high on these two characteristics started out higher on all three types of nutrition knowledge, the nutrition posters did not increase their knowledge more than they did for shoppers lower in education or income. The essence of this analysis is that the impact ofthe nutrition information was greatest on those people who most needed it.
Changes in the Food Purchased
The ultimate stage in the hierarchy of effects is a change in purchases. The nutrition information created both awareness and consequent improvements in knowledge and attitude. Were these cognitive changes powerful enough to overcome purchase factors like taste, price, and habit?
To test for purchase changes, we used the complete sales records and calculated the average of several nutrition characteristics over all brands and sizes within a product category for each store week. The most important of these was the NQ. However, we also examined calories, average nutrition (mean percent U.S. RDA of the eight nutrients), and each of the eight nutrients separately.
To test for effects on these nutrition characteristics, we pooled the results over all six product categories by normalizing each category by its mean and standard deviation. The resulting values can be interpreted as standard deviations above or below the grand mean.
Multiple linear regression analyses were performed on the normalized store-week means. The control portion of the model contained terms for historical time (determined from control stores) and store base rates (determined from the four weeks of baseline data). Two versions of the treatment submodel were used: a single parameter to assess an overall effect of treatment versus control and a full treatment model (10 parameters) that
THE JOURNAL OF CONSUMER RESEARCH
included the four formats crossed with the two information-delivery modes and single terms for Numeric (versus Alphabetic) ordering and for the presence of a take-home copy.
Based on the single parameter treatment model, the overall effect of the treatment on NQ-the primary measure of nutritional quality-was not significant. Indeed, the effect itself, -0.029 (i.e., 0.029 standard deviations below the mean NQ), was not even positive. The full treatment model revealed no significant differences among the treatments. The same tests were performed separately on the NQ of each product category. Again, no significant differences were found.
The identical analysis was performed on the other nutrition criteria, with the same results. No significant effects were found for calories, average nutrition, or any of the eight individual nutrients. We specifically examined combinations of product categories and nutrients that seemed highly related, such as fruit juice and Vitamin C. Again no significant effects of the information treatment were found.
The full nullity of the result is possibly best conveyed by examining those individual products with the highest NQs. For five of the six product categories, there were a few products clearly rated as superior in NQ. For example, among TV dinners, three of the top four are turkey dinners (see Exhibit 2). For each product category the combined sales of this top group were tested for an effect of treatment using the single-parameter model described earlier. In none of the five tests was there a significant effect of treatment. The mean percentage changes in sales attributed to the treatment in these analyses were: breakfast cereal (Total and Product 19), - 3.2 percent; frozen vegetables (broccoli and spinach without sauces), -1.8 percent; canned soup (vegetable, either plain or with beef, chicken, or turkey), +0.9 percent; canned and bottled juice (tomato juice), +0.1 percent; TV dinners (turkey), + 1.1 percent. Over these five product categories the net effect (-0.6 percent) was not even positive. Given no effect even for the most promising products, the conclusion of no purchase impact of the nutrition information seems inescapable.
DISCUSSION
The new effort-reducing formats succeeded in improving both nutrition knowledge and attitude, but not purchases. In this section we discuss possible explanations for this null result, which will then lead to a second experiment. A consideration of why Summary was the least effective format, contrary to the effort-reduction hypothesis, is postponed until the general discussion.
Why Was There No Change in Purchases?
We begin by ruling out several possible explanations. First, the absence of a conventional promotional cam-
NUTRITION INFORMATION
paign can be dismissed. The levels of awareness seem entirely adequate (see Figure A). Second, note that the absence of purchase changes held across all treatment conditions, even those most successful in changing knowledge and attitude. This uniformity implies that the explanation cannot rely on the particular format, but depends on more general factors. Finally, consider the possibility that there may not have been sufficient effort reduction, especially since the most effort-reducing format-Sum mary-turned out to be the least effective. This explanation is belied by the ability of all formats, including Summary, to change knowledge and attitude. These cognitive changes show that the strategy of effort reduction was successful. The question is why its success stopped at cognition and did not affect purchases.
We believe that the absence of purchase changes in the present study was caused by the low perceived benefits of positive nutrients like vitamins and minerals relative to the high cost of changing purchases. The effort-reducing formats improved knowledge and attitude because they lowered the information-processing costs of such cognitive changes to only a few minutes of time. However, purchase changes involve the additional costs of overcoming eating habits and taste preferences. We believe that these costs were high enough relative to the perceived benefits of improving nutritional intake to preclude purchasing changes.
The preceding reasoning is incomplete in one important respect. Given that the information-processing costs for a purchase change were lowered, the total cost (processing plus overcoming habits and tastes) must also have been lowered. Consequently, there should have been some shopper segment for whom costs and benefits initially had been roughly equal but for whom the reduced costs are now less than the benefits. Thus, we should have found some small change in purchases, instead of the zero (actually, slightly negative) effect that was observed. Further, there is a plausible candidate for this shopper segment. The in-store interviews identified dietarily at-risk shoppers who exhibited major changes in knowledge and attitude. Given these shoppers' recognition of their own at-risk status and their large cognitive changes, how it is possible that they exhibited absolutely no change in food purchased?
We believe that the answer lies in the use of a daily multivitamin pill and other food supplements. That is, any consumers who believe that their diets leave them susceptible to a nutrient deficiency have available to them a much less painful path to sufficiency than changing their food consumption patterns. They can take a multivitamin or similar food supplement. 1O According to a recent FDA national survey, 40 percent of
IOThis suggests that the place to look for purchasing changes was not the posted product categories but sales of multivitamins. Unfortunately, we did not monitor the sales of this product category.
63
adults (excluding pregnant and lactating women) consume at last one vitamin/mineral supplement. 11 In the western states, where food supplement usage is believed to be higher, Schutz et al. (1982) found that 67 percent of adults use food supplements. Thus, even those shoppers who perceive positive nutrients as important need not use nutrition information about available products to solve their problem. Stated differently, the technology of food supplements has broken the hierarchical link between knowledge and purchase. For any shopper with a sufficiently high level of perceived benefits for the positive nutrients, food technology has provided an easier way to attain those benefits than changing what is eaten.
Conclusion
We believe that the failure to influence food purchases is due to the low perceived benefits of positive nutrients coupled with the ease of taking a multivitamin supplement to compensate for any dietary deficiencies. This suggests switching to negative nutrients where the survey evidence cited earlier indicates higher perceived benefits. Furthermore, no food supplement can reduce the intake of cholesterol, sodium, sugar, and so on. Thus, though food technology may help solve the problem of excess consumption by creating foods low in negative nutrients like egg substitutes (without cholesterol) and light beers (Molitor 1981), a change in purchase is still required. Consequently, providing information about negative nutrients may provide a better test of the effort-reduction hypothesis.
EXPERIMENT 2
The switch from positive to negative nutrients limits the possible formats. The absence of a daily standard for each negative nutrient prohibits any aggregation across nutrients, eliminating the Summary, Complete, and Intermediate formats. A further problem is the difficulty of obtaining the levels of negative nutrients for all the foods in a product category. Little of this information is on the food label, and manufacturers are not eager to publicize the negative effects of their products. Two exceptions are sodium and sugar, which are either on the label or available from manufacturers for many product categories.
For these reasons we decided to test only the simplest list format, displaying a single negative nutrient for one product category-the added sugar in breakfast cereals. There is considerable evidence that most shoppers consider sugar to be negative, are trying to reduce sugar consumption, and do select alternative products with less sugar (General Mills 1980; Heimbach 1981).
IIRaymond C. Stokes, personal communication, 20 March 1984.
64
Design
Two matched supermarkets were used in a single pretest-test-posttest design with a control. The durations of the three phases were: pretest, 24 weeks; test, eight weeks; and posttest, eight weeks. Given a cereal purchase cycle of approximately two weeks, the test and posttest durations should assess the effect of repeat as well as initial purchase.
There was only one information manipulation, the provision of a nutrition list in a single format. The format used was a list of the added sugar (sucrose) for all brands of breakfast cereal, and the delivery mode was Posting-Only with some in-store promotion as will be described shortly. In addition, take-home copies were provided in the one experimental store, and both Alphabetic and Numeric ordering of the list were used simultaneously. Thus, the research question is whether shoppers' purchases will be influenced by a posted list of the added sugar in breakfast cereals (without any special handouts, but with take-home copies and both Alphabetic and Numeric orderings shown).
Sales records were obtained from checkout scanners. Two dependent measures were constructed from these data: the average sugar density (grams per one ounce serving) over all cereals sold, and the market shares of the high- and low-sugar cereals. If the information program is successful in altering purchases, the mean sugar density should decline and the market share of the low (high) sugar cereals should increase (decrease). No shopper interviews were conducted, so there are no assessments of awareness, knowledge, or attitude.
Sugar Information
Cereals were divided into four categories by sugar level, as defined by the number of teaspoons (about four grams) of added sugar per one ounce serving. 12
The teaspoon measure was judged to be familiar and easy to understand for almost all shoppers. The total number of brands common to both participating stores was 82, partitioned by sugar level (in teaspoons per serving) as follows: 26 at 0-1 teaspoon, 21 at 1-2 teaspoons, 19 at 2-3 teaspoons, and 16 at 3-4 teaspoons. 13
No brand added more than 4 teaspoons of sucrose per ounce of cereal.
12Unfortunately some cereal manufacturers included the dried fruit in the total added sucrose content, and others listed separately the grams of sucrose the dried fruit added. For this study all dried fruit was treated as added sugar. A few cereals listed no forms of added sugar, but listed minor amounts of sucrose. This represents the natural sugars within the cereal grain. Because the educational campaign was on added sugar, these cereals were listed as containing no added sugar per serving.
13Besides the 82 brands for which information was posted, we were unable to obtain from the manufacturer the added sugar content for two brands of granola breakfast cereal.
THE JOURNAL OF CONSUMER RESEARCH
Physical Display One four-panel numerically ordered list and one
three-panel alphabetically ordered list were hung parallel to the cereal aisle one foot above the top shelf. Although this position is well above eye level, the print could be read easily by any shopper standing in the aisle. The numeric panels were each 22 X 13 inches; the alphabetic, 21 X 17 inches. All panels were mounted on red poster board. The four numeric panels corresponded to the four sugar level categories.
Take-home copies were replicas of the alphabetical list reduced to 12 X 10 inches. Additional information about why and how to eat less sugar was provided on the back. The take-home copies were available in 15 recipe card holders mounted on the shelf facing of the cereal aisle. The holders were refilled weekly.
In-Store Promotion
Nine small signs encouraged shoppers to go to the cereal aisle to find out more about sugar and cereal. These signs were attached to the shelf facing in the cereal aisle and where related products like milk and frozen juices were located. In the cereal aisle, two fluorescent green arrows (2 X 3 feet) drew shoppers' attention to the nutrition posters. These arrows were mounted about nine feet above the center of the aisle.
RESULTS The effect of the sugar posters on purchase behavior
can be measured by the average sugar density (in grams per one ounce serving) over all cereals purchased. The differences in sugar density between the Experimental and Control stores are shown in Figure B. Two things are immediately clear from this plot. First, the effect of the posted information occurred quickly and was relatively stable. The actual average decrease in this difference between pretest and test levels was -0.37-i.e., the posters in the Experimental store resulted in a 0.37 gram decrease in sugar per ounce of cereal purchased. This difference was significant at the 0.05 level using a one-way analysis of variance and planned comparisons. (Details of these and other analyses can be found in Nolan 1983.) Second, the statistically significant effects disappeared entirely and almost immediately when the posters were removed. The average difference between pretest and posttest differences was 0.01.
Similar findings were found when market share differences were calculated for the four sugar level categories. Table 4 shows these results. The left data column contains the mean market share of each category, estimated from the pretest period and averaged over both stores. Next the differences between ExperimentalControl market shares are shown for each phase. But most important are the test-pretest and posttest-test differences. The test-pretest value estimates the effect of
NUTRITION INFORMATION
FIGURE B
DIFFERENCE IN AVERAGE SUGAR LEVEL OVER ALL BREAKFAST CEREALS PURCHASED IN EXPERIMENTAL
Sugar density (gm/l oz. serving)
.2
-.2
-.4
AND CONTROL STORES
Pretest Test Posttest
-.6~~~~~~~~~~~~~~~~~~~ 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
Weeks
posting the sugar lists. The market share of the category lowest in sugar increased 2.7 percent, while that highest in sugar decreased 2.3 percent. Consistent but smaller changes occurred for the two intermediate categories. All these effects disappeared after the posters were removed. The same ANOV A and planned comparisons were computed, with the latter shown in Table 4. All four differences for the two extreme sugar levels are statistically significant (p < 0.05). None of the four differences for the two intermediate categories is significant.
Conclusion Experiment 2 was designed to test a single effort-re
ducing information format for a negative nutrient. The results indicate an immediate and powerful effect on purchases. The market share of the low-sugar cereals increased 2.7 percent from a base of 43.4 percent. This is a large sales increase by food retailing standards. To achieve such an increase through manufacturer promotion, such as advertisements or coupons would typically require large expenditures, longer time periods, or both. Possibly more indicative of the impact of the posted sugar lists is the decrease in the market share of the high-sugar cereals, which dropped 2.3 percent from a base of 18.2 percent. Collectively, these brands lost 13 percent of their sales. We believe that these effects are directly attributable to a reduction in the effort needed to compare brands of breakfast cereal on added sugar.
GENERAL DISCUSSION The combined results of our two experiments raise
several issues. We address in turn: the contrast between the effort-reduction technique used here and traditional promotional methods; why the summary rating, which maximally reduced effort, was the least successful for-
65
mat; and implications for future information formats, including formats for other kinds of product information.
Effort Reduction versus Traditional Promotion
Our information-provision strategy was based on decreasing processing effort instead of trying to enhance the perceived benefits of nutrition. What has been the relative effectiveness of each approach? In the years since Experiment 1 was performed (in 1979), several field studies have been conducted that were designed to improve the nutritional quality of the food people buy. They are summarized in Table 5.
First, note that studies trying to enhance perceived benefits have uniformly failed to change purchases. For instance, the National Heart, Lung and Blood Institute (1983) had the full cooperation of Giant Foods in Washington, D.C. (e.g., 90 experimental stores), a radio and newspaper promotional campaign, professional information materials, and a full year to observe effects. Yet there was no impact on shoppers' purchases. It is especially important to note that a media campaign did not succeed-i.e., the full range of promotional methods was used without an effect. 14
In contrast, those studies that reduced effort were somewhat more successful in changing purchases. Pettersen (1979) found a statistically significant but unsystematic change. Further, her results may have been influenced by the unit price attribute of the information display, since brands were listed in order of unit price, not nutrition. Muller (1984) found a statistically significant change in sales for a matrix format of nutrients. Because his study ran for only two weeks and observed a significant effect only for the first week, a longer replication is needed. Nonetheless, a reanalysis of his data suggests that his effect is reliable. Muller posted an information matrix that included both positive and negative nutrients in two orderings, either all the negative nutrients first or all last. 15 If Muller's results confirm our own, a reanalysis of his original data (Muller 1982) should show a larger purchase change when the negative nutrients are highlighted.
Of the five product categories tested, two (mayonnaise and ketchup) contained no interbrand differences in nutrition, showed no purchase effects, and will be excluded. For the remaining three product categories
14Changing dietary patterns through the kind of benefit enhancement that promotion conveys may take years. Certainly advertisers and manufacturers take this view when they promote products, even superior ones. Also, the execution of the promotional campaign may have been faulty. For an analysis of what it takes to create an effective media information campaign, see Atkin (1981).
15Muller also used four numbers of nutrients: I, 2, 4, or 8. This meant that often only negative or only positive nutrients were posted. For example, when only one nutrient was posted, it was necessarily a negative one in the ordering with all negatives first.
66 THE JOURNAL OF CONSUMER RESEARCH
TABLE 4
CHANGES IN MARKET SHARE OF BREAKFAST CEREALS CATEGORIZED BY AMOUNT OF ADDED SUGAR
Experimental-control difference in market Difference in the experimental-control share differences
Sugar level Initial market category" shareb Pretest Test Posttest Test-Pretest" Posttest-Test"
1 (Lowest) .432 -.012 .015 -.011 .027d -.026d
2 .134 .019 .028 .021 .009 -.007 3 .251 .014 .001 .004 -.012 .002 4 (Highest) .182 -.021 -.045 -.014 -.023d .031 d
• category numbers Indicate the maximum amount of added sugar in teaspoons per 1 oz. serving. Thus. all breakfast osreals in Category 1 contain between 0 and 1 teaspoon of added sugar; those in Category 4 contain between 3 and 4 teaspoons.
• Initial market shares are computed from the pretest period only . • The stetistlcal teats are planned comparisons besed on one-way ANOVAs of the weakly dillerenos scores performed separately on each of the four categories . • p < O.OS.
TABLE 5
FIELD STUDIES OF DIET-RELATED INFORMATION
Results
Strategy for Information Food Accompanying Duration Purchase Knowledge Citation change provided category promotion (weeks) change change
Soriano and Dozier increase heart health dairy none 17 no yes (1978) perceived messages
benefits
Pettersen (1979) reduce summary rating, breakfast cereal none 8 yes· nla processing sugar, and unit effort price
Jeffrey et al. (1981) increase heart health dairy none 26 no no perceived messages benefits
Olson et al. (1982) increase complex high carbohydrate radio and 16 nob nob perceived carbohydrates foods newspaper benefits
National HLBI increase heart health fats and oils, radio and 52 no yes (1983) perceived messages dairy, and newspaper
benefits meats
Muller (1984) reduce matrix of relevant various packaged none 2 yesC nla processing nutrients foods effort
Levy et al. (1985) reduce shelf tags for 3 20 varied food radio and TV 104 yes nla processing negative nutrients categories effort
Experiment 1 reduce matrix and summary various packaged none 29 no yes processing of positive foods effort nutrients
Experiment 2 reduce list of sugar added breakfast cereal none 8 yes nla processing effort
• A purchase change occumsd, but it was not systematic and may be explained by unit prices rather than nutrition information.
• Claimed changes In both purchases and knowledge are disabled by the experimentel design, which had no historical control and relied on self-selected shoppers as intervieWees .
• A significant effect was observed only for the first of two weaks. See discussion In !9xt.
NUTRITION INFORMATION
(bran cereal, 7 brands; canned cream of mushroom soup, 2 brands; and macaroni and cheese dinner, 3 brands), a comparison of the distributions of market shares indicates that: (1) both the positive and negative content/orders altered purchases toward more nutritious foods, and (2) those information matrices that emphasized the negative nutrients were twice as effective in causing the purchase of more nutritious foods. This reanalysis should not be considered definitive since the original data were not used but rather the ridit statistics reported in Tables 34, 37, and 38 of Muller (1982). Nonetheless, the results are consistent with our finding that effort-reducing formats like matrices can change food purchases, at least for negative nutrients.
Overall, the studies catalogued in Table 5 support the effort-reduction strategy. This is not to say that benefit enhancement will not work. However, changing purchase patterns through promotion is expensive and time consuming. Those concerned with nutrition policy may get larger results from their limited resources by adopting effort reduction over promotional strategies.
Summary Ratings versus Full Matrices
The effort-reduction hypothesis clearly favored the Summary over the Matrix format. Why then was Summary almost uniformly inferior in shoppers' eyes? We believe that the answer lies in a lack of credibility and acceptance. The NQ measure was new, unfamiliar to shoppers, and not supported by an official endorsement or a promotional/educational campaign. Further, even if credible to some shoppers, it may have been rejected as too prescriptive, ignoring all individual differences in its judgment of what is good or bad to eat. Shoppers may want to apply their individual values-e.g., some may prefer more emphasis on Vitamin C while others may be more concerned about iron or calories. This resistance to prescription has been reported informally elsewhere (Hammonds 1978).
One last possible explanation should be acknowledged. The real value of the Matrix may have been calories-the one negative nutrient that the Summary format did not display. The intermediate effects of the Complete format disable this hypothesis. If calories were all that mattered to shoppers, the Complete format should have been as effective as the Matrix. Its intermediate effectiveness argues for a negative impact of adding the NQ measure to the full matrix, consistent with a lack of credibility or acceptance of this summary rating.
The rejection of the NQ carries a lesson for other information provision programs. It may be much more difficult to achieve consumer acceptance of a summary measure than of individual attributes, especially objective ones. Thus a promotional campaign, preferably with a credible endorsement, may have to accompany any summary rating (and, possibly, any subjectively judged attributes). This is not to suggest that summary
67
ratings should be avoided. To the contrary, their large effort-reduction benefits make summary ratings enormously attractive. And where credibility has been established they are also very successful. Star ratings of movies, restaurants, and hotels are widely used, and Consumer Reports groups and ranks a wide range of products on overall value. A recent survey of German consumers assessed the use of the ratings of overall quality and of individual attributes published in the German equivalent of Consumer Reports (Silberer 1985). Of recent purchasers of durable goods who used the magazine's product information, 67 percent reported using the summary rating while only 28 percent used the individual product attributes. For nondurables 54 percent used the summary rating and 11 percent used the individual attributes. Furthermore, for both durables and nondurables, about 38 percent reported using only the summary rating in guiding their purchase (and roughly 10 percent reported using only the individual attributes). The dominant use of the summary rating coupled with the acceptance of star ratings of movies and restaurants suggests that, given credibility, a summary rating will be more effective than a matrix of information. These data also suggest that prescriptiveness is at most a minor problem, one that can be overcome by credibility of the summary ratings. In sum, the lesson learned from the failure of the NQ measure is that any summary rating must be accepted as credible before its effort-reduction benefits can be realized.
The Future of Information Provision Programs
The effort-reduction technique can provide the rationale for a wide range of information provision programs. Certainly the list format works (Experiment 2; Russo 1977), and the evidence suggests that the matrix format is worth pursuing (Experiment 1; Muller 1984). We believe that summary ratings will also succeed, but only in conjunction with a program to establish their credibility and acceptance. These formats are hardly limited to nutrition information but can be used whenever products or services can be usefully represented by performance attributes, a credible summary rating, or both.
Who should initiate such programs? We believe that one possibility is retailers. Experiment 1 shows that the nutrition posters can significantly increase customer good will. Presumably, more sought-after information-such as negative nutrients or potential hazardscan produce greater effects, eventually leading to an increased customer base. Besides having a direct incentive, retailers are in the best position to include many different brands in a unified information format and to display the result at the point of purchase. We hope that supermarket chains will develop displays and posting arrangements for the cost-effective presentation of nutrition information. Other opportunities for super-
68
markets include unit prices (Russo 1977) and hazard information for household products like bleach (Bettman, Payne, and Staelin 1986). Further, the possibility of effort-reducing displays extends to other multibrand retailers such as home appliance stores, video specialists, and others.
New technologies will provide further opportunities to translate the effort-reduction technique into effective information programs. Personal computers, sophisticated display devices, and telecommunications technology (e.g., Talarzyk and Urbany 1983) will move information provision from a passive to an active mode. The posters tested in our research are entirely passive. In contrast, an in-store video terminal for displaying the same nutrition information could allow some active shopper participation-e.g., by selecting product categories of interest, an alphabetic versus numeric ordering, a single attribute like calories, and so forth. As artificial intelligence is incorporated into information programs via question-answering capabilities and expert systems (Duda and Shortliffe 1983), the shopper can assume an increasingly active role in selecting personally relevant product information.
The change from passive to interactive information technologies will not, however, change the value of effort reduction. We believe that consumers will always base the decision to use product information on a cost-benefit analysis, where information-processing effort is a major cost. Decreasing processing effort will always be an important tactic for making information provision programs more successful. We see effort reduction as changing the information environment to adapt to people. In contrast, benefit enhancement exhorts people to greater efforts to seek information from a relatively inhospitable environment. The results of our studies and those in Table 5 suggest that it may be more effective to change the shopping environment to adapt to people than to change people to adapt to an effortful environment.
APPENDIX
Interviewers were partially confounded with store and thus with treatment. We decided not to rotate interviewers across stores to save the time and cost of interviewers' travel to the 14 widely scattered suburban stores and because of assurances from our highly regarded interviewing firm that their experienced interviewers were completely reliable. Nonetheless, after the data had been collected we checked for any interviewer bias. We could devise no single valid criterion for identifying bias. Instead, we adopted a strategy of assessing each of the interviewers on multiple criteria and drawing the conclusion of bias only if a pattern emerged. Five criteria were used. The first three were based on comparisons of the market baskets of shoppers (as recorded by each interviewer) with the warehouse withdrawal data for the same stores and the same weeks. This comparison
THE JOURNAL OF CONSUMER RESEARCH
was made on NQ, calories, and average nutrition (the mean percent U.S. RDA of the eight nutrients). We also compared the education and income levels of each interviewer's respondents with the education and income of all other interviewers who worked at the same store. Significant differences in both education and income were treated as an indication of a bias. Finally, we examined the records of the survey firm and our own records for evidence of irregularities and corrective reprimands.
Because of the likelihood of false alarms, we discarded an interviewer's data only if bias was detected by more than two of the five criteria. In recording the contents of the market basket a bias might have been detected only because the interviewer was careless. For example, there may have been a tendency to recognize brands with unique packaging or the ones she herself usually bought. Although this is a bias, it is unrelated to the treatment. Similarly, an education or income bias could only be measured relative to other interviewers. If there were just two interviewers in a store and significant differences in both education and income are observed, it is still not clear which of the two interviewers is biased.
With one exception, no interviewer failed more than two criteria, and even in these cases no systematic pattern of bias was apparent. The one exception failed all five criteria. All data collected by this interviewer were excluded from all analyses. Unfortunately, this one biased interviewer was responsible for 8 of the 10 survey administrations during the experimental phases in Store 12, the only store that tested the Intermediate format. Too little data remained to permit estimation of this treatment. Thus, the Intermediate format had to be excluded from all analyses of the interview data (but not the purchase data, which were based on warehouse withdrawal records). In addition, one other interviewer initially misunderstood the interview protocol and the data from her first week were dropped.
We examined the consequences of two other alternatives: (1) dropping all those interviewers who gave any indication of bias, and (2) removing the bias statistically. Dropping interviewers with any indication of bias would have prohibited estimation of over half of the treatments. Although not comfortable with even marginal indications of bias, we judged the potential threat to validity not so great as to justify dropping so many treatment conditions. The second alternative was to estimate interviewer effects as part of the treatment model. Because interviewers were assigned only to stores close to their homes (to minimize travel costs), the interviewers were highly correlated with stores and, therefore, treatments. This correlation prevented any reliable estimation of both interviewer and treatment terms. Thus, a post hoc statistical solution to the problem of bias was not possible.
This brief discussion of potential interview bias is a good illustration of why researchers, especially in costly unrepeatable field experiments, should never accept as-
NUTRITION INFORMATION
surances that something will not be a problem. In hindsight, we should have rotated interviewers over treatments.
[Received April 1984. Revised September 1985.]
REFERENCES Atkin, Charles K. (1981), "Mass Media Information Cam
paign Effectiveness," in Public Communication Campaigns, eds. Ronald E. Rice and William J. Paisley, Beverly Hills, CA: Sage, 265-279.
Bettman, James R. (1975), "Issues in Designing Consumer Information Environments," Journal of Consumer Research, 2 (December), 169-177.
--- (1979), An Information Processing Theory of Consumer Choice, Reading, MA: Addison-Wesley.
---, John W. Payne, and Richard Staelin (1986), "Guidelines for Designing an Effective Labeling System: Cognitive Considerations in Presenting Risk Information," in Hazardous Product Labeling and Precautionary Behavior, eds. Kip Viscusi and Wes Magat, Cambridge, MA: Harvard University Press, (forthcoming).
Council on Foods and Nutrition (1973), "Improvement of the Nutritive Quality of Foods: General Policies," Journal of the American Medical Association, 225 (27 August), 1116-1118.
Dawes, Robyn M. (1979), "The Robust Beauty of Improper Linear Models," American Psychologist, 34 (July), 95-106.
Dietrich, Robert (1980), "Shoppers at the Turning Point," Progressive Grocer, 59 (7), 65-67.
Duda, Richard O. and Edward H. Shortliffe (1983), "Expert Systems Research," Science, 220 (15 April), 261-268.
General Mills (1980), "A Summary Report on U.S. Consumers' Knowledge Attitudes and Practices About Nutrition," prepared by Maracom Research Corp., published by General Mills, Inc., P.O. Box 1113, Minneapolis, MN, 55440.
Guthrie, Helen A. (1977), "Concept of a Nutritious Food," Journal of the American Dietetic Association, 71 (July), 14-19.
Hammond, Kenneth R. and Leonard Adelman (1975), "Science, Values, and Human Judgment," Science, 194,389-396.
Hammonds, Tim (1978), "Nutrition and the American Food System: Part II," paper presented at the Food Marketing Institute/Family Circle/Community Nutrition Institute National Nutrition Conference, Washington, D.C., October 22.
Heimbach, James T. (1981), "Defining the Problem: The Scope of Consumer Concern with Food Labeling," in Advances in Consumer Research, Vol. 8, ed. Kent B. Monroe, Ann Arbor, MI: Association for Consumer Research,474-476.
-- and Raymond C. Stokes (1979), "FDA 1978 Consumer Food Labeling Survey," Food and Drug Administration, Department of Health and Human Services, Washington, D.C.
--- and Raymond C. Stokes (1982), "Nutrition Labeling and Public Health: Survey of American Institute ofNu-
69
trition Members, Food Industry, and Consumers," American Journal of Clinical Nutrition, 36,700-708.
Jeffrey, Robert W., Phyllis L. Pirie, Barbara S. Rosenthal, Wendy M. Gerber, and David S. Murray (1982), "Nutrition Education in Supermarkets: An Unsuccessful Attempt to Influence Knowledge and Product Sales," Journal of Behavioral Medicine, 5 (2), 189-200.
La Chance, Paul A. (1975), "Critique ofIndex of Food Quality (lFQ)," Journal of Nutrition Education, 7 (Oct.-Dec.), 136.
Levy, Alan S., Odonna Mathews, Marilyn Stephenson, Janet E. Tenney, and Raymond E. Schucker (1985), "The Impact of a Nutrition Information Program on Food Purchases," Journal of Public Policy and Marketing, 4, 1-13.
Marketing Science Institute (1980), "Determinants of Food Usage Behavior: A Market Segmentation Approach," Cambridge, MA: Marketing Science Institute.
Mazis, Michael B., Richard Staelin, Howard Beales, and Steven Salop (1981), "A Framework for Evaluating Consumer Information Regulation," Journal of Marketing, 45 (1), 11-21.
Molitor, Graham T.T. (1981), "Look to Europe's Policy Vanguard," Food Development, September, 15-18,20, 24.
Muller, Thomas E. (1982), "The Impact of Consumer Information on Brand Sales: A Field Experiment with Point of Purchase Nutritional Information Load," unpublished dissertation, Faculty of Commerce, University of British Columbia, Vancouver V6T IY8.
--- (1984), "Buyer Response to Variations in Product Information Load," Journal of Applied Psychology, 69 (2), 300-306.
National Heart, Lung, and Blood Institute (1983), "Foods for Health: Report ofthe Pilot Program," National Institute of Health Publication No. 83-2036, Bethesda, MD 20982.
Neter, John and William Wasserman (1974), Applied Linear Statistical Models, Homewood, IL: Richard D. Irwin.
Nolan, Catherine A. (1983), "An Evaluation of a Point of Purchase Nutrition Education Program," unpublished thesis, University of Washington, Seattle, WA 98195.
Olson, Christine, Carole A. Bisogni, and Patricia F. Thonney (1982), "Evaluation of a Supermarket Nutrition Education Program," Journal of Nutrition Education, 14 (4), 141-145.
Pettersen, Magnhild A. (1979), "Cereal Buying Guide as an Aid to Consumer Choice of Ready-to-Eat Cereals: An Experimental Study," unpublished thesis, University of Maryland, College Park, MD 20742.
Preston, Ivan L. (1982), "The Association Model of the Advertising Communication Process," Journal of Advertising, II (2), 3-15.
Putnam, Judy Jones and Jon Weimer (1981), "Household Diet Changes Linked to Nutrition Concerns," staff paper ofthe Economics and Statistics Service, U.S. Department of Agriculture, Washington, D.C. 20250.
Russo, J. Edward (1977), "The Value of Unit Price Information," Journal of Marketing Research 14 (May), 192-201.
--- (1981), "The Decision to Use Product Information at the Point of Purchase," in Theory in Retailing: Traditional and Nontraditional Sources, eds. Ron Stampfl and Elizabeth Hirschman, Chicago: American Marketing Association, 155-167.
70
---, Richard Staelin, Gary J. Russell, and Barbara L. Metcalf (1985), "Nutrition Information in the Supermarket," Report No. 85-100, Cambridge, MA: Marketing Science Institute.
Schutz, Howard G., Marsha Read, Robert Bendel, Vijay S. Bhalla, Inez Harrill, J. Edgar Monagle, Edward T. Sheehan, and Bluebell R. Standal (1982), "Food Supplement Usage in Seven Western States," American Journal of Clinical Nutrition, 36 (November), 897-901.
Silberer, Ganter (1985), "The Impact of Comparative Product Testing upon Consumers' Selected Findings of a Research Project," Journal of Consumer Policy, 8 (I), 1-28.
THE JOURNAL OF CONSUMER RESEARCH
Soriano, Esteban and David M. Dozier (1978), "Selling Nutrition and Heart-Healthy Diet Behavior at the Point-ofPurchase," Journal of Applied Nutrition, 30,56-65.
Talarzyk, W. Wayne and Joel E. Urbany (1983), "Videotex and Retailing-Overview and Implications for Key Retail Groups," Working Paper Series 83-44, College of Administrative Science, Ohio State University, Columbus, OH 43210.
Wyse, Bonita, Ann W. Sorenson, Arthur J. Wittwer, and R. Gaurth Hansen (1976), "Nutritional Quality Index Identifies Consumer Nutrient Needs," Food Technology, January, 20-40.
Related Documents
