CONSUMER PERCEPTIONS, PREFERENCES, AND WILLINGNESS TO PAY FOR
SUSTAINABLY LABELED FOOD PRODUCTS
By
Hillary M. Sackett

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Agricultural, Food, and Resource Economics - Doctor of Philosophy
2013

ABSTRACT
CONSUMER PERCEPTIONS, PREFERENCES, AND WILLINGNESS TO PAY FOR
SUSTAINABLY LABELED FOOD PRODUCTS
By
Hillary M. Sackett
“Sustainably produced” food labels have rapidly grown in popularity over the past decade.
However, because there is no government agency overseeing certification of these production
methods, consumers may be generally confused about the production attributes that may
or may not be present in a sustainable food system. Moreover, the ability of a firm to differentiate their product depends critically on an accurate understanding of the perceptions
consumers hold about the implications of a credence labeling claim. Building upon existing
work evaluating other food attribute labels and the impact of consumer inferences, this work
begins to address gaps in the literature regarding food products with sustainable production
claims.
To achieve these goals, a comprehensive set of information was gathered from consumers,
through two complementary activities: a nation-wide survey on food consumption habits including simulated shopping scenarios designed to reveal tradeoffs perceived by the consumer
and a valuation field experiment designed to capture the same information in an in-person,
non-hypothetical setting. The juxtaposition of the hypothetical and non-hypothetical results enhances the reliability and appropriateness of the economic analysis and marketing
recommendations. Collectively, these two approaches yield a more detailed understanding
of consumer behavior.

The first essay initiates the process of examining consumer inferences and valuations of
food products making “sustainably produced” claims utilizing a Best-Worst scaling framework to identify what consumers believe “sustainably produced” labels mean and their preferences for each of the sustainable farming practices considered. The results of which suggest
that consumers perceive farm size and local production as highly important elements of sustainable agriculture. Additionally, consumer preferences over economic attributes such as
consumer food prices and financial stability of farmers exhibit high heterogeneity, indicating
segmentation in the sample and potential for targeted marketing management.
The second essay analyzes data from a hypothetical choice experiment to better understand consumer purchasing behavior when faced with competing food production attributes
such as “sustainable”, “organic” and “local”. The primary goal in this piece is to estimate preferences for “sustainably produced” food products and determine how they may be
affected by varying degrees of information about sustainable agricultural systems. Additionally, the willingness to pay measurements estimated in this essay provide insight into the
trade-offs perceived between current eco-labeling schemes, and the potential for differentiating “sustainably produced” products from their “organic” and “local” counterparts.
The third and final essay compares the results of the experimental auction and the hypothetical choice experiment to further examine the influence of sustainable, organic, and local
production labels on food choice and to estimate consumer willingness to pay for sustainably produced apples and steak. The findings suggest a preference ordering in experimental
bids, differing from that found in the hypothetical choice experiment. However, upon deeper
investigation, the estimates do not provide sufficient evidence of hypothetical bias in survey
responses. Furthermore, the data indicate statistically insignificant differences in willingness
to pay for the sustainable, organic, and local products, implying a weak level of differentiation between labels. Overall, this work supports other experimental findings in the literature
of positive price premiums on these credence attributes.

ACKNOWLEDGEMENTS
I would like to thank my committee, namely Dr. Robert Shupp, Dr. Glynn Tonsor, Dr.
Phil Howard, and Dr. Robert Brent Ross for their guidance in this research. The entire
committee provided feedback and additional insight that resulted in higher quality output.
I am especially appreciative of Dr. Shupp’s contributions to this work and for taking on the
role of teacher by introducing me to the field of experimental economics, the role of mentor
through his support and encouragement as my major professor in my academic program,
and for giving me trust and freedom in exploring new ideas and the generous funding to do
so. Special thanks to Dr. Tonsor for continuing to provide guidance even after relocating to
Kansas State University, and for challenging me to reach my full potential by never giving
easy answers. Appreciation goes to Dr. Howard for always offering alternative viewpoints
and for inspiring me to continue working in the area of sustainable food systems. Warm
gratitude to Dr. Mariah Ehmke for providing me with valuable advice on life and career,
and for setting an example that I aspire to.
I have made many wonderful connections within the food community at Michigan State
University and in the Greater Lansing area during my graduate study. Specifically, I thank
Julie Cotton for sharing her enthusiasm for local food systems and for the amazing work
she is doing to expand opportunities for students interested in sustainable agriculture at
MSU. I also thank all of the beautiful souls at MSU’s Student Organic Farm, especially the
2010 Farmer Training Program cohort, for shared meals and showing me first hand what
farm-to-table is really all about. A huge outpouring of love to Megan Sheahan, Rachel
Ford, Miriam and Juha Sohlberg, and Ty Lawson for their delicious food, adventures, and
friendship, without which I never would have made it through.
Finally, I would like to thank my family, both Sackett and Brian/Kleinhardt, for their
unending support and love. A very special thanks to my wife, Rachel, for her constant
companionship, for keeping me nourished in every sense of the word, for helping me keep
things in perspective, and for working towards our shared dreams.
iv

TABLE OF CONTENTS
LIST OF TABLES

vii

KEY TO ABBREVIATIONS

ix

Chapter 1: General Introduction
Overview . . . . . . . . . . . . . . . . . . .
Objectives . . . . . . . . . . . . . . . . . . .
Literature Review . . . . . . . . . . . . . . .
Preliminary Findings . . . . . . . . . . . . .
Likert Scale Sustainability Perceptions
Locally Grown Perceptions . . . . . . .
Social Issues Importance . . . . . . . .
Organization of Dissertation . . . . . . . . .

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

1
1
2
2
3
3
5
5
5

of Sustainable Farming Practices
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.

7
7
9
9
11
13
16
16
19
20
21
22
25
28

APPENDICES
Marginal Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Apple Survey Attribute Definitions . . . . . . . . . . . . . . . . . . . . . . . . . .
Beef Survey Attribute Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . .

31
32
41
42

REFERENCES

43

Chapter 3: Consumer Preferences for Sustainably Produced Steak and Apples
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Discrete Choice Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . .
Alternative Specific Design . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Information Treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

46
46
48
48
49
51
53

Chapter 2: Consumer Perceptions
Introduction . . . . . . . . . . . . .
Research Methodology . . . . . . .
Best-Worst Scaling . . . . . .
Sustainability Attributes . . .
Data Collection . . . . . . . . . . .
Survey Question Example . .
Analysis . . . . . . . . . . . . . . .
Results and Discussion . . . . . . .
Wide-Form Data Example . .
Long-Form Data Example . .
Conditional Logit . . . . . . .
Latent Class Model . . . . . .
Conclusions . . . . . . . . . . . . .

v

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.

Survey Question Example .
Analysis . . . . . . . . . . . . . .
Multinomial Logit . . . . . .
Conditional Logit . . . . . .
Random Parameters Logit .
Results and Discussion . . . . . .
Information Treatment Tests
Pooled Model Results . . . .
WTP comparison . . . . . .
Conclusions . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.

54
54
57
57
58
59
59
62
66
69

APPENDICES
USDA Information Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Food Alliance Information Treatment . . . . . . . . . . . . . . . . . . . . . . . . .
Choice Experiment Instructions, Definitions, and Cheap Talk Script . . . . . . . .

70
71
73
76

REFERENCES

78

Chapter 4: Willingness to Pay for Sustainably Labeled Steak and Apples:
An Application in Experimental Auctions
81
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Experimental Auctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Experimental Design and Implementation . . . . . . . . . . . . . . . . . . . 83
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Pooled Sample Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Raw Data Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Hypothesis Testing of Equal WTP Across Samples . . . . . . . . . . . . . . . 97
Tobit Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
APPENDIX
104
Survey Instrument . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
REFERENCES

137

Chapter 5: Summary

140

vi

LIST OF TABLES
Table 1

Likert Scale Sustainability Perceptions . . . . . . . . . . . . . . . . . .

4

Table 2

Locally Grown Perceptions . . . . . . . . . . . . . . . . . . . . . . . .

5

Table 3

Social Issues Importance . . . . . . . . . . . . . . . . . . . . . . . . . .

5

Table 4

Demographic Variables and Summary Statistics of Survey Participants

14

Table 5

Apple and Beef Survey Attributes . . . . . . . . . . . . . . . . . . . .

15

Table 6

Best-Worst Survey Question Example . . . . . . . . . . . . . . . . . .

16

Table 7

Wide-Form Data Example . . . . . . . . . . . . . . . . . . . . . . . . .

20

Table 8

Long-Form Data Example . . . . . . . . . . . . . . . . . . . . . . . . .

21

Table 9

Relative Importance of Sustainable Attributes: Logit Estimates . . . .

24

Table 10

LCA Estimates: Apple Attribute as “Most Important” by Cluster . . .

26

Table 11

LCA Estimates: Beef Attribute as “Most Important” by Cluster . . .

27

Table 12

Apple Attribute Importance Measures using Best-Worst Scaling . . . .

33

Table 13

Beef Attribute Importance Measures using Best-Worst Scaling . . . . .

35

Table 14

Apple Production Attribute Variance-Covariance Matrix . . . . . . . .

38

Table 15

Apple Production Attribute Correlation Coefficients . . . . . . . . . .

38

Table 16

Beef Production Attribute Variance-Covariance Matrix . . . . . . . . .

39

Table 17

Beef Production Attribute Correlation Coefficients . . . . . . . . . . .

39

Table 18

Attributes and Levels Used in the Choice Experiment . . . . . . . . .

50

Table 19

Demographic Variables and Summary Statistics of Survey Participants

52

Table 20

Choice Experiment Survey Question Example . . . . . . . . . . . . . .

54

Table 21

H0 : Pooling Across Information Treatments: Apple Survey . . . . . .

59

Table 22

H0 : Pooling Across Information Treatments: Beef Survey . . . . . . .

59

Table 23

Information Treatment Comparison: Apple Survey . . . . . . . . . . .

61

Table 24

Information Treatment Comparison: Beef Survey . . . . . . . . . . . .

61

Table 25

Discrete Response Profile: Apple Survey . . . . . . . . . . . . . . . . .

63

Table 26

MNL, CL, and RPL Parameter Estimates: Apple Choice Experiment .

63

vii

Table 27

WTP Estimates for Apple Labeling Scheme Attributes . . . . . . . . .

63

Table 28

Discrete Response Profile: Steak Survey . . . . . . . . . . . . . . . . .

65

Table 29

MNL, CL, and RPL Parameter Estimates: Steak Choice Experiment .

65

Table 30

WTP Estimates for Steak Labeling Scheme Attributes . . . . . . . . .

65

Table 31

Hypothesis Test of Equal WTP Across Credence Labels: Apple Survey

68

Table 32

Hypothesis Test of Equal WTP Across Credence Labels: Steak Survey

68

Table 33

Summary Statistics of Survey and Experiment Samples . . . . . . . .

87

Table 34

Pooled Sample Comparison: RPL Estimates in Apple Survey . . . . .

92

Table 35

Hypothesis Testing of Pooling Across Participant Sample: Apple Survey 92

Table 36

Pooled Sample Comparison: RPL Estimates in Beef Survey . . . . . .

93

Table 37

Hypothesis Testing of Pooling Across Participant Sample: Beef Survey

93

Table 38

WTP Estimates and Associated Price Premiums for Apple Attributes

94

Table 39

WTP Estimates and Associated Price Premiums for Steak Attributes .

94

Table 40

Comparison of Hypothetical and Non-hypothetical WTP: Apple Survey 97

Table 41

Comparison of Hypothetical and Non-hypothetical WTP: Steak Survey

Table 42

Tobit Regression Results for Apple Bids . . . . . . . . . . . . . . . . . 100

Table 43

Tobit Regression Results for Steak Bids . . . . . . . . . . . . . . . . . 100

viii

97

KEY TO ABBREVIATIONS
GC = Ground Cover and Area Management Practices are Employed
FN = Fertilizer and Nutrient Materials are used minimally
CP = Little to No Chemical Pesticides are Used for Pest Management
CH = Little to No Chemical Herbicides are Used for Weed Management
PM = Pollinator Management is Employed
OP = Other Pests are controlled using preventative measures, and habitat controls
SA = Prohibited use of sub-therapeutic antibiotics
GH = Prohibited use of growth hormones
GM = Prohibited use of genetically modified livestock
AS = Animal Health and Safety are Protected
PF = Feed is Pasture Based and Waste Management Systems Employed
FS = Farm Size is small and Corporate Involvement is limited
PL = Production, distribution, and sale is done locally
CFP = Consumer food prices are affordable
FF = Farmers are financially stable

ix

Chapter 1: General Introduction
Overview
Food produced using sustainable production practices is receiving increasing attention as the
public and private sectors become progressively more invested in “sustainability” and “green
marketing”. There are currently over 200 different seals and logos representing ecological,
ethical, or sustainability attributes in the global food industry (Organic Monitor). Alternative agricultural production systems generate value-added food attributes that provide
consumers with varying degrees of utility, stemming from their environmental, economic,
and social impacts. If farms are to remain viable and contribute to food and farm system
sustainability, they may need to find and exploit these kinds of high-valued niche markets.
Moreover, a firm’s ability to differentiate their product depends critically on an accurate
understanding of how such an alternative credence labeling scheme is perceived by the consumer. Building on existing work evaluating food attribute labels, this research examines
the impact of consumer preferences on purchases of food products making “sustainably produced” claims.
As sustainable food claims gain market momentum, questions naturally arise regarding which attributes of sustainability are most important to consumers, how perceptions of
sustainability may affect purchasing decisions, if consumer segments can be identified that
systematically behave differently in response to these labeling schemes, and if purchasing
decisions over specific sustainability attributes are stable and consistent in hypothetical and
non-hypothetical settings. The limited existing literature on sustainable food labels has
spent little effort identifying which attributes of sustainability consumers believe are important components of agricultural production systems. Additionally, because there is no government agency overseeing certification of these production-level claims, evidence suggests
that consumers may be confused about the methods underlying the “sustainably produced”
label.

1

Objectives
The primary objectives of this research are as follows: first to identify what consumers
believe “sustainably produced” means by determining the aggregate importance and share
of preference for a variety of sustainable farming practices considered. Second, we aim to
estimate willingness to pay for sustainably labeled foods and compare estimates across hypothetical and non-hypothetical preference elicitation. Additionally, we seek to determine the
impact of information on sustainability, from different sources, on WTP. Lastly, we develop
a set of recommendations, based on the results of these objectives, designed to guide meat
and produce industries in the development and implementation of a “sustainably produced”
marketing plan.
To achieve these goals, we gathered a comprehensive set of information from consumers,
through two complementary activities: a nation-wide survey on food consumption habits, including simulated shopping scenarios designed to reveal tradeoffs perceived by the consumer
and a valuation field experiment designed to capture the same information in an in-person,
non-hypothetical setting. The juxtaposition of the hypothetical and non-hypothetical results enhances the reliability and appropriateness of the economic analysis and marketing
recommendations. Collectively, these two approaches yield a more detailed understanding
of consumer behavior.

Literature Review
Surprisingly little economic research has focused on sustainability in the context of agricultural production practices. However, some food attributes that consumers may associate
with sustainable production, such as local, organic, natural, GMO-free, hormone-free, and
environmentally friendly have a considerable research base. Generally, the literature has
found that credence attributes have a positive impact on consumer valuations and lead to
higher WTP, despite the inability to directly identify the attribute either before purchase or
after consumption. “Sustainably produced” is an attribute with no absolute definition, leav2

ing sustainable food labels much more open to consumer perceptions and inferences that go
beyond the information provided on packaging or marketing labels. Understanding the perceptions consumers hold regarding what a “sustainably produced” labeling scheme implies
is crucial to determining if the product will succeed in the market.

Preliminary Findings
Likert Scale Sustainability Perceptions
To collect consumer data used for this analysis, a national web-based survey of 1002 households was disseminated in the summer and fall of 2010. Consumer respondents were recruited
by Decipher, a marketing research and survey programming company, and roughly adhere to
the expected demographic profile of the United States population. The following Likert scale
questions were posed to establish general perceptions of sustainability prior to completion
of the market choice-oriented questions.

3

Table 1: Likert Scale Sustainability Perceptions
Likert Scale* Ranking Share
Survey Question
1
2
3
4
5
6
“I rarely think about the sustainability of 0.06 0.07 0.12 0.22 0.18 0.19
production practices when making food purchases”
“Farmers face lower levels of productivity 0.04 0.09 0.15 0.39 0.18 0.11
when using sustainable production practices”
“Farmers could sustainably produce food 0.05 0.09 0.16 0.33 0.20 0.10
without much additional monetary expense”
“Farmers currently participate in sound sus- 0.02 0.05 0.11 0.39 0.24 0.13
tainable food production practices”
“Organic products are from farms currently 0.03 0.04 0.09 0.31 0.21 0.21
participating in sustainable practices”
“Locally grown products are from farms cur- 0.02 0.04 0.08 0.38 0.24 0.16
rently participating in sustainable practices”
“Smaller farms are more likely to currently 0.02 0.04 0.08 0.31 0.25 0.22
participate in sustainable practices”
“Corporate farms are more likely to cur- 0.07 0.11 0.17 0.33 0.16 0.11
rently participate in sustainable practices”
“All products labeled as sustainably pro- 0.03 0.04 0.13 0.38 0.21 0.15
duced are from farms currently participating
in sustainable practices”
*Scale is measured according to 1 = Strongly Disagree and 7 = Strongly Agree

4

7
0.16

0.05

0.07
0.07
0.10
0.09
0.10
0.06
0.07

Locally Grown Perceptions
When asked, “ Which best describes the proximity from your home you consider locally
grown products to come from?”, the survey participants responded as follows:
Table 2: Locally Grown Perceptions
Distance
10 miles
50 miles
100 miles
500 miles
Within My State
Other

Choice Share
0.17
0.30
0.20
0.07
0.24
0.02

Social Issues Importance
Finally, when asked to rank the following social issues in order of their perceived importance,
the participants responded:
Table 3: Social Issues Importance
Social Issue
Mean Ranking
Food Safety
3.13
Human Poverty
3.88
Health Care
3.99
Consumer Food Prices
4.11
Environment
4.5
Animal Welfare
5.63
Sustainable Agriculture
5.45
Financial Stability of Farmers
5.3

Share of #1 Ranking
0.26
0.20
0.18
0.14
0.11
0.05
0.03
0.02

Organization of Dissertation
This dissertation proceeds as follows, beginning with an examination of consumer perceptions
of select sustainable farming practices in Chapter 2. The second chapter begins to address
gaps in the literature regarding food products with sustainable production claims through
the implementation of a best-worst scaling framework. The goal of the best-worst analysis
5

is to identify what consumers believe “sustainably produced” should mean in the context
of food labels. The primary purpose of Chapter 2 is to introduce an economic application
of best-worst scaling for measuring importance of production attributes in food systems, to
identify which environmental, economic, and social indicators of sustainability currently used
by third-party agricultural certifiers as farm-level evaluation criteria consumers perceive as
important systematic components of food production, and to assess consumer heterogeneity
for insight into potential marketing strengths for third-party certified “sustainable” products.
The dissertation continues to unpack preferences over sustainable food labels utilizing a
hypothetical choice experiment in Chapter 3. The third chapter seeks to estimate preferences for “sustainably produced” food products in a discrete choice modeling framework and
to determine how preferences may be affected by varying degrees of information about sustainable agricultural systems from different sources. Using comparative measures, Chapter
3 begins to shed light on the degree of perceived market differentiation between products
labeled as “sustainable” and their “organic” and “local” counterparts.
Building further on the analysis from Chapter 3, the fourth chapter in this volume examines the hypothetical and non-hypothetical stated preferences of a smaller sub-sample of
consumers that participated in experimental auctions in a laboratory setting. Comparison
of the hypothetical and non-hypothetical WTP measures is used to determine the extent of
hypothetical bias in the sample, and to induce a relationship between hypothetical choices
and bidding behavior. Together these two measures, analyzed in Chapter 4, provide a richer
description of consumer preferences and is believed to be more representative of actual market behavior. In Chapter 5 general conclusions are drawn and a synthesis of the results leads
to a set of recommendations for producers and marketers of sustainably produced foods.

6

Chapter 2: Consumer Perceptions of Sustainable Farming Practices
Introduction
Food produced using sustainable production practices is receiving increasing attention in
both public and private arenas as a greater number of food products are being marketed and
labeled using “sustainable” or “sustainably produced” certification schemes for differentiation. As sustainably produced food gains market momentum, questions arise regarding what
consumers perceive when faced with “sustainably produced” labels. Specifically, consumers
want to know what sustainable claims imply about the environmental, economic, and social
factors of production and farmers want to know what consumers are willing to pay for this
value-added attribute before either party invests heavily in sustainable certification.
Consumer desired attributes of food system sustainability have recently been discussed
in more detail in the literature. For example, a framework for evaluating consumer priorities
with regard to sustainable foods was built by Clonan et al (2010) based on seven guiding
principles of sustainability put forth by Sustain, an alliance for better food and farming. The
authors used a five point Likert scale, embedded in a structured questionnaire, to explore
attitudes towards sustainability components such as fair trade, organic, local, and animal
welfare. This study found that consumers responded positively toward environmental responsibility metrics related to how their food was produced. Similarly, using a discrete
choice modeling method for evaluating consumer attitudes towards sustainability claims on
food products, Saunders et al (2010) also focus on the results of a Likert scale rating of sustainability attributes, in the context of carbon emissions and other contributions to global
climate change. However, the limited literature on sustainable food labels has spent very
little effort focused on identifying attributes of sustainability that consumers believe are or
should be important components of a “sustainably produced” labeling scheme.
Batte(2010) reviews several studies that identify consumer-driven changes in food marketing channels as related to sustainable food claims. In his review, Batte identifies three

7

studies that support the importance of consumer demand in food product differentiation
schemes. The first, a study by Onozaka et. al. (2010) finds evidence from a conjoint choice
experiment of significant heterogeneity in valuing various food differentiation claims among
shoppers in different marketing venues and notes that the consistent significance of selfperceived efficacy in their psychographic model suggests that consumers that believe they
have a role in improving sustainability tend to value sustainable product claims more highly.
In another study, Onozaka and McFadden (2011) explore the increasing use of sustainable
food labels by analyzing the interactive effects of sustainable production claims, finding that
locally grown differentiation is most highly valued. The authors suggest that consumer preferences for local food have been found to go beyond basic quality characteristics and are
increasingly significantly related to sustaining the local economy by supporting local farmers
and conserving local farmland. Batte concludes that further research is needed to identify
how consumer demand for sustainably produced food is affected by the perceived importance of the environmental, economic, and social attributes used in differentiation through
certification and labeling.
To address this gap in the literature, we utilize a best-worst scaling framework to examine what consumers believe “sustainably produced” - in the context of food labels - means
or should mean. Best-worst choice modeling is a relatively new technique for analyzing
consumer preferences or beliefs. This technique has been applied in a variety of settings
including public health research by Flynn et all (2007), international agribusiness marketing
by Umberger et al (2010), and by Scarpa et al (2010) to estimate desired tourism benefits. In an agricultural and food systems context, best-worst choice design and analysis have
been previously demonstrated using empirical wine market examples by Casini, Corsi and
Goodman (2009), Cohen (2009), and Mueller and Rungie (2009). More closely related to
our work, Lusk and Parker (2009) use best-worst scaling techniques to evaluate the methods consumers prefer for producers to use to improve the fat content in ground beef. Most
closely related to the current study, Lusk and Briggeman (2009) determine relative consumer

8

attitudes towards value-added food attributes, such as safety, nutrition, taste, and price using the best-worst framework. Here we examine consumer attitudes towards value-added
sustainable production attributes in a similar fashion.
The primary purpose of this paper is threefold. First, to introduce an economic application of best-worst scaling for measuring the importance of production attributes in consumer
decision making in the context of food systems. Additionally, we seek to identify which environmental, economic, and social indicators of sustainability, currently utilized by third
party agricultural certifiers as farm/ranch-level evaluation criteria, consumers perceive as
important systematic components of food production. Finally, we aim to assess consumer
heterogeneity in perceived importance of sustainable farming practices when facing “sustainably produced” food purchasing decisions for insight into potential marketing strengths for
third party certifiers.

Research Methodology
Best-Worst Scaling
Marketing surveys that measure attribute importance most often utilize a Likert Scale ranking approach. However, this method has several known weaknesses. First, scaled rating
systems do not force respondents to make trade-offs between attributes. Additionally, Likert
Scale ranked data lack a natural interpretation outside of the survey context. To address
these issues, we implemented a best-worst design to investigate preferences for and perceptions of alternative sustainable farming practices. The survey instrument used to collect
consumer data was designed to simplify the choice task for respondents.
Best-Worst analysis requires the survey respondent to simultaneously choose the most
and least important attributes out of a set of competing options. This method is commonly
referred to as “maximum difference scaling”, as the attributes chosen should maximize the
difference in utility realized, for the respondent, on an underlying scale of preference. The
measured level of importance from the best-worst data analysis is applied to a standardized
9

ratio scale that determines, with more certainty, the percentage difference in importance
across attributes. The theoretical foundation for this analysis is provided by Marley and
Louviere (2005) in their development of probabilistic models for analyzing best-worst choice
tasks.
Best-worst scaling, as originally devised by Flynn and Louviere (1992), is capable of addressing relative utility impacts across attributes that traditional discrete choice questions
cannot. To observe trade-off behavior, the specification of attributes from a choice set of
competing alternatives is repeated over a number of variable choice subsets. In this way,
best-worst tasks provide more information than single choice designs, while forcing respondents to consider the extremes of their utility space. Ideally, the stated preference outlined in
each best-worst scenario should approximate observed consumer behavior in retail markets.
Each choice set was created using a 210 main effects orthogonal experimental design that
was balanced with ten attributes, each exhibiting two levels. The specific attributes chosen
for inclusion are motivated in the following subsection and further outlined in the Data Collection section. The orthogonal experimental design yielded twelve alternative choice sets,
broken into two blocks and randomized across participants. Each block contained two questions with five alternatives, three questions with six alternatives, and one question with all
ten alternatives. The presence or absence of each sustainable farming attribute was independent across choice subsets, allowing identification of relative preferences onto a ratio scale.
The additional utility or dis-utility from moving between attribute levels will be estimated
using a logit model framework.

10

Sustainability Attributes
The USDA National Institute of Food and Agriculture provides limited information on the
purported sustainability of particular farming and ranching practices. At the national level,
sustainable agriculture was first addressed by Congress in the 1990 Farm Bill (USDA NIFA)
which states that under the law,
“The term sustainable agriculture means an integrated system of plant and animal production practices having a site-specific application that will, over the long term:
• satisfy human food and fiber needs
• enhance environmental quality and the natural resource base upon which the agricultural economy depends
• make the most efficient use of nonrenewable resources and on-farm resources and integrate, where appropriate, natural biological cycles and controls
• sustain the economic viability of farm operations
• enhance the quality of life for farmers and society as a whole.”
However, the USDA also warns that, “guidelines about what specific practices meet longterm environmental, economic, and social goals and constitute sustainable agriculture is still
under debate”. Therefore, when developing our consumer survey, for the purposes of determining perceptions of the practical farm-level components of a sustainable production
system, we attempted to best follow the advice provided by the USDA by including environmental, economic, and social attributes of sustainability that address the final two points
above.
The USDA provides website url links to a handful of groups that have attempted to develop standards and provide certification services based on their own guidelines. The first of
the groups indicated on the USDA website is Food Alliance (www.f oodalliance.org). Food
Alliance is a non-profit organization that has developed sustainable agricultural practice
11

standards and operates a voluntary certification program based on those standards. Food
Alliance is the most comprehensive certification program for sustainable food production
in North America, employing independent third-party inspectors to audit Food Alliance
certified businesses to determine whether they meet program standards and criteria. Food
Alliance outlines all crop and livestock specific certification criteria and provides the following
general guidelines for all certified production:
• Provide safe and fair working conditions
• Ensure the health and humane treatment of animals
• No use of hormones or sub-therapeutic antibiotics
• No genetically modified crops or livestock
• Reduce pesticide use and toxicity
• Protect soil and water quality
• Protect and enhance wildlife habitat
• Continuously improve management practices
For the purposes of this study, we adhered as closely as possible to the Food Alliance
Whole Farm/Ranch Evaluation Criteria for crop and livestock specific guidelines when choosing the sustainability attributes that would be included in the survey. All crop and livestock
specific Food Alliance Evaluation Criteria is publicly available online and from the authors
by request. Because there is currently no government sponsored certification of sustainable
food production, all “sustainably produced” labels found on food products in the United
States are certified by a private third-party such as Food Alliance, or by the farm of origin
itself. We are interested in identifying which certification-guided environmental, economic,
and social attributes consumers perceive as important indicators of sustainability when faced
with these kinds of purchasing decisions.
12

Other reputable, broadly defined, sustainable agriculture certifying organizations include
Rainforest Alliance and Global G.A.P. The Rainforest Alliance certification standards are
based on the same Sustainable Agriculture Standards (SAS) as Food Alliance. However, the
Food Alliance certification program was chosen as the guiding standard because it had more
detailed crop and livestock specific criteria publicly available at the time the survey was developed. Food Alliance standards took preference over Global G.A.P guidelines because of
its national versus international focus and because Global G.A.P functions to oversee other
accredited certification bodies for a number of different labels, potentially creating confusion.
There are many farming practices with sustainable characteristics that could have been
included for the purposes of this study. The set of sustainable farming attributes were chosen
for inclusion after study of the USDA’s general sustainability guidelines and Food Alliance’s
third party sustainable certification standards. The set of attributes used in this study offer
insight into possible perceived dimensions of sustainability and could be expanded on in
future work. It is worth note that due to the nature of the variable choice sets, the analysis
that follows is conditional on the set of evaluated attributes/farming practices and should
not be interpreted to make judgements on sustainability attributes outside of this context.

Data Collection
This is the first known study to use best-worst scaling to measure consumer perceptions
of production attribute importance in the context of food system sustainability. We chose
best-worst scaling because we are especially interested in determining relative importance
of sustainable agricultural certification criteria currently used in food markets. To collect
consumer data used for this analysis, a national web-based survey of 1002 households was
disseminated in the summer and fall of 2010. Consumer respondents were recruited by
Decipher, a marketing research and survey programming company. A summary of population
demographic statistics can be found in the following table.

13

Table 4: Demographic Variables and Summary Statistics of Survey Participants
Variable
Definition
Apple
Beef
Gender
1= Male; 2 = Female
1.476
1.516
Total Participants
500
502
Age
Average Age in Years
51.48
50.92
Adults
Number of Adults in Household
2.062
2.048
Children
Number of Children in Household 0.48
0.51
Meals
Number of Meals/ week with 6.64
3.19
Product
Shop
% of Total Shopping at Location
Grocery Store
81.97
81.16
Health Food Store
8.49
9.20
Food Co-op
2.05
1.90
Convenience Store
3.0
4.42
Farmers Market
3.72
5.24
Butcher
4.38
2.92
Assistance 1 = On Food Assistance; 2 =
Otherwise
1 = On Food Assistance
13.2 %
11.95 %
2 = Not on Food Assistance
86.8 %
88.05 %
Education Highest Level Completed
1 = Did not graduate from high 2.2 %
2.59 %
school
2 = Graduated from high school, 17.8 %
18.12 %
no college
3 = Attended college, no degree 28.8 %
33.67 %
earned
4 = Attended college, associates 15.4%
12.55 %
or trade degree earned
5 = Attended college, Bachelors 24.6 %
22.52 %
degree earned
6 = Graduate or advanced degree 11.2 %
10.56 %
earned
Household Range of Pre-tax Income
Income
1 = Less than $20,000
20.8%
19.12 %
2 = $20,000 - $39,999
26.4%
28.88 %
3 = $40,000 - $59,999
18.4 %
21.51 %
4 = $60,000 - $79,999
17.6 %
11.95 %
5 = $80,000 - $99,999
8.2 %
7.57 %
6 = $100,000 - $119,000
3.8 %
4.98 %
7 = $120,000 - $139,999
1.8%
2.39 %
8 = $140,000 - $159,999
1.4 %
1.20 %
9 = More than $160,000
1.6%
2.39 %
14

Two versions of the survey were developed; one for apples and one for beef. The following
table lists the sustainable farming practices that were chosen, in accordance with the crop
(apple) and livestock (beef) specific Food Alliance Evaluation Criteria, as the best-worst
attributes that the survey respondents would choose between for each food product.

Table 5: Apple and Beef Survey Attributes
Apple Survey Attributes
Beef Survey Attributes
Ground Cover Management
Prohibited Use of Antibiotics
Limited Fertilizer Use
Prohibited Use of Growth Hormones
Limited Pesticide Use
Prohibited Use of GM Livestock
Limited Herbicide Use
Animal Health and Safety
Pollinator Management
Pastured Feed and Waste Management
Preventative Pest Control
Preventative Pest Control
Farm Size
Farm Size
Geographic Level of Production
Geographic Level of Production
Consumer Food Prices
Consumer Food Prices
Financial Stability of Farmers
Financial Stability of Farmers

Many of these practices fall under current organic certification guidelines endorsed by the
USDA, and all aforementioned attributes are components of sustainable farm certification
by Food Alliance. The included attributes span the three-pronged sustainability framework
suggested by Callens and Tyteca (1999) and are supported by the USDA National Institute
of Food and Agriculture, using environmental, economic, and social metrics for evaluation.
The choice sets for this analysis are developed around ten attributes, each with two
levels indicating the presence or absence of a given farming practice. Consumers are shown
a set of attributes and asked to indicate which is most important (best) and which is least
important (worst) over several repeated choices where the set of attributes varies across
questions. In theory, each respondent undertakes the task of identifying every possible pair
of attributes, calculating the difference in utility between each attribute pair, and choosing
the pair that maximizes this difference. This method is an extension of Thurstone’s (1927)
paired comparison method, which has been utilized frequently in psychological research.
In maximum-difference scaling, the distances between attributes are modeled as pair-wise
15

utilities and estimated in relation to a single attribute level rather than to an entire scenario.
All apple survey participants were faced with the following example scenario, with similar
extension to the beef survey.

Survey Question Example
Which one of the following aspects of apple farming do you believe is the most and least
important in a sustainable apple production system? Please check only one in each column.

Table 6: Best-Worst Survey Question Example
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Little to No Chemical Pesticides are Used for Pest
Management
Pollinator Management is Employed
Other Pests are controlled using preventative measures, and habitat controls
Production, distribution, and sale is done locally
Consumer food prices are affordable

The main effects orthogonal experimental design ensured that each potential best-worst
pair appeared exactly twice in each blocked section of six choice sets and that each attribute
was seen by the respondent exactly four times in this portion of the survey. Definitions were
provided for six of the production attributes, excluding the economic attributes, which were
left to the interpretation of each respondent. Full definitions can be found in the Appendix
to this paper.

Analysis
Two primary approaches are used to analyze best-worst ranked data. “Paired” models are
implemented to make inferences about the latent utility scale, while “marginal” models
aggregate over all pairs that include a given attribute level, to model choice frequencies.
16

Both methods have the same measurement properties and can be analyzed at the respondent
or sample level, yielding similar results. For brevity, the simpler marginal (count-based)
specification and analysis is omitted in this text and is available in the Appendix to this
chapter.
Consider a choice set like the one outlined above, with J = 6 attributes, and therefore
J(J − 1) = 30 possible best-worst combinations that could be chosen by the respondent.
The particular pair of attributes chosen by the consumer represents one choice out of all
J(J − 1) = 30 possible pairs, in a given choice set, that maximizes the difference in their
perceived importance. Let λj formally represent the location of attribute j on the underlying
scale of importance. Therefore the latent, unobservable level of importance for individual i
is given by
Iij = λj + ij

(1)

where ij is the random error. Thus, the probability that the consumer chooses item j as
most important and item k as least important is equal to the probability that the difference
between Iij and Iik is greater than all other J(J − 1) − 1 possible differences in the choice
set. If ij is distributed i.i.d then λj can be estimated for attribute j in a conditional logit
framework. The equation to be estimated is as follows:

choicei = α + λ1 L1 + λ2 L2 + ... + λJ LJ + i

(2)

where each Lj corresponds to one of the J attributes included in a given choice set. Each
attribute is an alternative specific regressor, taking a value in −1, 0, 1.

Using the conditional logit parameter estimates, λj , the probability that attribute j is
chosen as most important (best) and attribute k is chosen as least important (worst) is given
by:

17

P rob(j = best ∩ k = worst) =

λ −λ
e j k
λs −λt − J
J
J
s=1 t=1 e

(3)

The parameter estimates represent the share of importance for the given attribute, relative to the attribute ranked least important, normalized to zero. To return results consistent
with standardized ratio scaling techniques, the share of importance for each attribute, equal
to the forecasted probability of being chosen as most important, takes the following form:

sharej =

λ
e j
J eλi
i=1

(4)

The preference shares must sum to one across all ten sustainable farming attributes. The
above equation should be interpreted as the importance of attribute j on a ratio scale. The
share of preference for a given attribute reflects both the true importance of the attribute
and the relative uncertainty in the importance consumers place on the attribute. However,
this specification neglects consumer heterogeneity, as the estimation requires the assumption
that all individuals in the sample must place the same level of importance on each value,
once they are ranked.
To more deeply explore consumer heterogeneity, a latent class cluster analysis is also
employed. This clustering technique assumes that individuals belong to one of a predetermined number of latent classes. The Bayesian Information Criterion (BIC) determines the
optimal number of latent clusters. This method involves minimizing within-cluster variance
and maximizing across-cluster variance. The probability of membership in a given latent
class is estimated using the logit model parameters. The latent class cluster analysis uses
the individual best-worst scores as the dependent variables in this model. The covariation
across individual best-worst scores measures unobserved utility gains and losses associated
with the inclusion or exclusion of the given attribute for each respondent.

18

Results and Discussion
The raw data contains 1002 observations (rows of data), one for each respondent. Within
each respondent’s observation, two variables for each question corresponds to the respondent’s choice of most important production attribute and choice of least important production attribute, leading to 12 variables for each respondent because each survey contains six
best-worst questions. This type of data is commonly referred to as being in “wide-form”.
Using SAS, the wide-form data was transformed into what is commonly referred to as “longform” data, with one observation providing data for just one alternative for each individual.
Specifically, if a given best-worst question offers J attributes to choose between, then the
long-form data has J(J − 1) observations for that question, as alternatives are now thought
of as potential best-worst pairings among the J potential attributes. Each observation for
a given question includes J explanatory variables, one for each of the production attributes
included in a choice set. In a given observation, all but two of the production attributes
take a value of zero. Of the two remaining attributes in this observation, one production
attribute takes a value of 1 referring to the “most important” choice, and the other takes a
value of −1 referring to the “least important” choice. Thus, the J(J − 1) observations each
refer to a possible pair of production attributes that could be chosen by the respondent as
best and worst simultaneously. Over these J(J − 1) observations, the variable “choice” takes
a value of zero for all but one observation. The remaining observation refers to the actual
best-worst pair chosen by the respondent for that question and the variable “choice” takes
a value of 1.

19

Wide-Form Data Example

Table 7: Wide-Form Data Example
ID Q15 Options Q15 Best Q15 Worst
1
5
1
2
2
5
3
2
3
5
5
4
4
5
3
1
In this example, there are four survey respondents referred to by their ID numbers ranging
from 1 − 4. “Q15 Options” refers to the number of attributes that are contained in choice
set (question) 15 and is the same for all respondents. “Q15 Best” refers to which of the
five attributes the respondent chose as “most important” in choice set (question) 15 and
similarly, “Q15 Worst” refers to which of the five attributes the respondent chose as “least
important” in choice set (question) 15. Here, respondent 3 has answered question 15 by
choosing attribute 5 as “most important” and simultaneously choosing attribute 4 as “least
important”.

20

Long-Form Data Example

ID
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3
3

Table 8: Long-Form Data Example
Choice Q15-1 Q15-2 Q15-3 Q15-4
0
-1
1
0
0
0
-1
0
1
0
0
-1
0
0
1
0
-1
0
0
0
0
0
-1
1
0
0
0
-1
0
1
0
0
-1
0
0
0
1
-1
0
0
0
0
0
-1
1
0
0
0
-1
0
0
1
0
-1
0
0
0
1
-1
0
1
0
0
0
-1
0
1
0
0
-1
0
0
1
0
-1
0
0
0
1
-1
0
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1

Q15-5
0
0
0
1
0
0
1
0
0
1
0
0
1
0
0
0
-1
-1
-1
-1

In the table above, a portion of the wide-form example has been transformed into longform. As noted above, respondent 3 is presented with five sustainable farming production
attributes in question 15. As such, there are twenty (5 ∗ 4) possible best-worst pairs that this
respondent could choose, corresponding to the twenty observations (rows of data). In each
row, each of the five production attributes, labeled Q15-1 through Q15-5 for simplicity, take
a value in {−1, 0, 1}. For example, the first row in this table refers to the possibility that in
Q15, attribute 1 is chosen as “least important” and attribute 2 is simultaneously chosen as
“most important”. The second row in this table refers to the possibility that in Q15, attribute
1 is chosen as least important and attribute 3 is simultaneously chosen as “most important”,
and so on. We see from the wide-form data above that respondent 3 answered question 15
21

by choosing attribute 5 as “most important” and attribute 4 as “least important”. In the
long-form data this choice is reflected in the variable “choice” taking a value of 1 for the
observation corresponding to that specific best-worst pair and zero everywhere else.

Conditional Logit
The long-form data is used in a logit-model analysis here. The summary statistics from
a conditional logit model analysis can be found in the following table for apple production
and beef production, respectively. The regression parameters are transformed into preference
shares, which offer more intuitive interpretation. The preferences shares sum to one, and each
represents the proportional share of importance for the given attribute relative to consumer
food prices, which is normalized to zero.
Half of the respondent sample was randomly assigned to complete the survey on apple
production practices. The results in Table 3 indicate that the attribute corresponding to
“farm size is small and corporate involvement is limited” has the highest preference share. As
noted earlier, the preference share parameters report the importance of a given sustainable
farming production attribute on a ratio scale, reflecting both the true importance of the
attribute as well as relative uncertainty conveyed as the probability that an attribute is picked
as more important than any other. Following farm size, pollinator management received the
second highest preference share. Worth noting is the moderately high preference share on
the attribute corresponding to “production, distribution, and sale is done locally”, with just
under half the importance relative to farm size. “Consumer food prices are affordable” was
chosen as least important more often in every scenario than any of the remaining attributes
and was therefore set as the base category.
These results should also be interpreted in relative magnitude. For example, pollinator
management is inferred to be roughly half as important in a sustainable agricultural system as
farm size to this sample of consumers. In fact, in a sustainable agricultural system, farm size
is found to be just more than two times more important than local production, distribution

22

and sale. The use of off-farm chemical inputs such as fertilizers, herbicides, and pesticides
all ranked between three and four times less important to a sustainable agricultural system
than farm size and only marginally more important than consumer food prices. Overall, this
sample of consumers indicated that the four economic attributes of sustainability included
in this study were located near the boundaries of their utility spectrum.

23

Table 9: Relative Importance of Sustainable Attributes: Logit Estimates
Apple
At- λ
Preference
Beef
At- λ
Preference
tribute
Share
tribute
Share
Farm Size
1.77*
0.28
Farm Size
1.41*
0.28
(0.05)
(0.05)
Pollinator
1.25*
0.15
Local
1.09*
0.17
Management
(0.06)
(0.06)
Ground
1.22*
0.13
Preventative 0.80*
0.12
Cover
Pest Management
(0.06)
(0.06)
Local
1.08*
0.12
Pasture
0.47*
0.09
Based Feed
(0.06)
(0.06)
Preventative 0.90*
0.09
Prohibited
0.43*
0.08
Pest ManAntibiotics
agement
(0.05)
(0.06)
Minimal Fer- 0.78*
0.08
Farmer
0.34*
0.08
tilizer
Financial
Stability
(0.06)
(0.06)
Limited Her- 0.56*
0.06
Prohibited
0.23*
0.07
bicides
GMO
(0.05)
(0.06)
Farmer
0.29*
0.04
Prohibited
-0.08*
0.05
Financial
Growth
Stability
Hormones
(0.05)
(0.05)
Limited Pes- 0.14*
0.03
Animal
-0.10*
0.04
ticides
Health and
Safety
(0.05)
(0.05)
Affordable
DROPPED
0.02
Affordable
DROPPED
0.02
Food Prices
Food Prices
(*) and (**) significant at 0.01 and 0.05 levels, respectively
standard errors are reported in parentheses

24

The remaining half of the respondents were randomly assigned to complete the survey
on beef production practices. The data in Table 5 reveals that, as in the apple survey,
the attribute corresponding to “farm size is small and corporate involvement is limited”
has the highest preference share. The attribute with the second highest preference share
is “production, distribution, and sale is done locally”, suggesting that the locality of meat
production is more important in a sustainable agricultural system to consumers than the
locality of apple production relative to the other included attributes. This difference could
be attributable to the fact that beef production is less place-specific than apple production
due to the more specific climate needs for apple growth. The ratio scaled values indicate
that locality is between one half and three quarters as important as farm size. The attribute
corresponding to “animal health and safety is protected” had the lowest preference share
other than affordable consumer food prices. The use of growth hormones also had a relatively
low preference share.
When standardized to the ratio scale, farm size is found to be seven times more important
in a sustainable beef production system to consumers than animal health and safety. Yielding
the same results as the apple version of the survey, the respondents of the beef survey ranked
financial well-being of farmers and affordable consumer food prices in the bottom third on
the underlying scale of importance.

Latent Class Model
To better qualify distinct consumer segments, this data was also analyzed under a latent
class clustering framework. The ten best-worst attributes formed the dependent variables
over the underlying probability distribution of latent class inclusion. Consumer clusters with
similar perceptions of the ten sustainable attributes are located close to one another in ndimensional utility space. Identifying groups of consumers with similar perceptions is useful
for advertising, pricing, and product development.
The optimal number of latent classes is determined by minimizing the Bayesian Informa-

25

tion Criterion. This resulted in four latent classes of apple consumers and five latent classes
of beef consumers. The following tables outline the latent class structures. In each column,
the parameter reported identifies the item response probability corresponding to participant
endorsement of each attribute as “most important”.
Table 10: LCA Estimates: Apple Attribute as “Most Important” by Cluster
Attribute
“Localvores” “Small Farmers” “Price
“Confused/
Savvy
IndifferShoppers” ent”
(14%)
(30%)
(10%)
(46%)
Ground Cover
0.0
0.1852
0.0
0.0976
Fertilizer
0.0
0.1852
0.0
0.0976
Pesticides
0.0
0.1852
0.0
0.0976
Herbicides
0.3077
0.0
0.0
0.1219
Pollinators
0.0
0.1481
0.0
0.0976
Pest Management
0.3846
0.0
0.0
0.1220
Farm Size
0.0
0.1481
0.0
0.1219
Local
0.3077
0.0
0.0
0.1220
Food Prices
0.0
0.0
1.00
0.0
Farmer Stability
0.0
0.1481
0.0
0.1220
For apples, the first cluster, representing 14% of the sample, is made up of consumers that
believe local production is a highly important aspect of sustainable production. This group
also strongly values preventative pest management and limited use of chemical herbicides.
It is possible that the use of chemicals in production is undesirable for this cluster due to
potential effects on local ecosystems and water resources.
The second cluster, representing 30% of the sample, perceives limited corporate involvement to be important in a sustainable production system, as indicated by their preference for
small farm size and financial stability of farmers relative to other clusters. This group also
emphasizes the importance of other farming attributes that are in line with current land-use
organic standards, such as the use of ground cover and the limited use of chemical inputs.
This group likely places a higher value on organically labeled products, but would like to see
the farms less consolidated in a sustainable system.
The third cluster, representing 10% of the sample, is motivated primarily by consumer
26

food prices. The fourth cluster, representing the largest portion of the sample at 46%, is made
up of consumers that are either indicating indifference across these particular attributes, or
may be confused about how these attributes relate to one another in a systematic way. This
is evidenced by the roughly equal ranking of all attributes in importance to a sustainable
production system. From this latent class assessment we find that 56% of the evaluated apple
consumer sample is either price driven or largely indifferent to varied aspects of sustainability.
Table 11: LCA Estimates: Beef Attribute as “Most
Attribute
“Animal
“Nutrition
“Price Savvy
Rights Ac- Buffs”
Shoppers”
tivists”
(15%)
(24%)
(10%)
Antibiotics
0.0
0.1818
0.0
Growth Hor- 0.0
0.2273
0.0
mone
GMO
0.0
0.0
0.0
Animal
0.3077
0.0
0.0
Safety
Pest Mngmt 0.0
0.0
0.0
Pastured
0.3846
0.2273
0.0
Feed
Local
0.3077
0.0
0.0
Food Prices 0.0
0.0
1.0
Farm Size
0.0
0.1818
0.0
Farmer Sta- 0.0
0.1818
0.0
bility

Important” by Cluster
“Confused/
“Say No to
Indifferent”
GMOs”
(41%)
0.1351
0.1081

(10%)
0.0
0.0

0.0
0.1351

1.0
0.0

0.1081
0.1081

0.0
0.0

0.1351
0.0
0.1351
0.1351

0.0
0.0
0.0
0.0

For beef, the first cluster, representing 15% of the sample, values the ethical treatment of
animals in meat production, evidenced by the high propensity for choosing the preservation of
animal health and safety, and pasture based feed as most important in a sustainable system.
Additionally, this cluster perceives local production and limited corporate involvement as
desired contributions to the sustainable system.
The second cluster, comprising 24% of the sample, is primarily concerned with the safety
and nutritional aspects of their food decisions. This cluster most often chose the prohibition
of sub-therapeutic antibiotics and bovine growth hormone along with pastured feed as the
27

most important attribute of sustainable production systems. Additionally, this cluster of
consumers values small farm size and the financial stability of farmers. Since small farms
may be inclined to have a more direct connection with their consumer base, better food
safety and traceability standards might be perceived as attainable.
The third and fourth clusters share the same characteristics as the “price savvy shoppers”
and “sustainably indifferent” consumers from the apple survey, at 10% and 41%, respectively,
together comprising 51% of the beef sample. However, in the beef survey a distinct fifth
cluster emerges, comprised of consumers that chose the prohibition of genetically modified
livestock as most important in every choice set in which it appeared. While this segment
of the market is small, it may offer significant opportunities for producers that do not use
genetically modified products and for marketing firms that could emphasize this attribute
in a labeling scheme.

Conclusions
Best-worst analysis was applied in this research to investigate the degree of importance consumers give to ten sustainable farming production attributes and in particular was used to
determine behavioral differences across clustered subgroups of the population sample. The
advantages of this methodology compared to more traditional stated preference analysis is
evident in its higher discriminatory power for measuring trade-off decisions and in its wider
applicability and interpretation outside of the survey context. While avoiding common rating
bias, best-worst analysis results can be used in cross-national and cross-regional comparison
studies on diverse populations and their judgment of similar attributes. This study gives
credence to the strength of the best-worst method in yielding clear and simple interpretations. The simplicity of this analysis can be applied by marketing managers to gain insight
into the evaluation behavior of different consumer segments for targeted food labeling.
The information gathered here from consumer data on perceptions of sustainable farming
practices holds large potential for marketing managers. The unique best-worst framework

28

provides greater insight into determinants of market behavior than more commonly used
Likert scale ranking approaches. In both beef and apple surveys, consumers indicated a
strong perceptive correlation between sustainability and the size and locality of the farm of
origin. This analysis suggests, similar to the findings of Onozaka and McFadden (2011), that
consumers perceive quality difference for locally grown and distributed products. Supporting
studies, such as that of Bond, Thilmany and Keeling-Bond (2008) give evidence that preferences for local food products are significantly related to factors affecting farmer viability,
sustaining local farm land, and contributing to smaller, local economies. Our work supports
these findings that scale and geographic range factor heavily into consumer perceptions of
sustainably labeled food products.
Increasing attention drawn towards smaller, local farms has been a response to the widening awareness of global food system business conduct. Distrust has been growing for imported
foods, especially meat products, with recent publicity on country of origin labeling requirements and other high profile food contamination cases. Sustainability claims on food targets
many dimensions of consumer utility from quality and safety concerns to more intrinsic valuation connected to underlying food values such as fairness and environmental impact. In
effect, some sustainability claims may be seen by consumers as substitutable, while others
complementary; another point emphasized by Onozaka and McFadden (2011). The value of
sustainable certification may only contribute marginally to the localness of a food product,
while in other situations it may enhance commitment to a more well-rounded sustainable
farming view point.
While consumers are generally familiar with organic standards outlining land-use and
environmental impact variables of production, this study indicates that environmental indicators of sustainability are less important to consumers than economic dimensions. Based on
our initial results, size, scale, and geographic scope capture the attributes of sustainability
that are most important to consumers. Therefore, differentiating food claims on the level of
locality provides a marketing avenue worth exploring further, as supported by the growing

29

literature on the local foods movement.
Our study also provides supporting evidence that the term “sustainable” could be causing more confusion than it is adding value in credence labeling schemes. The latent class
assessments on both beef and apple consumer samples indicated that 50 - 55% of the evaluated population is highly price-driven or significantly indifferent to varied sustainability
attributes. These results could be interpreted to suggest confusion or indifference, but they
could also imply that consumers view sustainability as a bundle of attributes and that each
alone is not sufficiently important to determine the sustainable nature of an agricultural
system. Therefore, this study may support claims that only an integrated system of some or
all of these attributes can be considered sustainable, as perceived by consumers. Given the
associated heterogeneous economic welfare impacts that would come as a result of acrossthe-board market adjustments, such as government endorsed sustainability labeling, or bans
on select farming practices, this is an important result.

30

APPENDICES

31

Marginal Analysis
Two primary approaches are used to analyze best-worst ranked data. “Paired” models are
implemented to make inferences about the latent utility scale, while “marginal” models aggregate over all pairs that include a given attribute level, to model choice frequencies. Both
methods have the same measurement properties and can be analyzed at the respondent or
sample level, yielding similar results. For brevity, the simpler marginal (count-based) analysis specification is omitted in the published text.
For marginal (count-based) data analysis, individual best-worst scores are calculated
for each attribute as the summation of the number of times each respondent indicated the
attribute as most important less the summation of the number of times each respondent
indicated the attribute as least important. The larger the best-worst score, the higher the
attribute is ranked on the underlying scale of importance. Individual best-worst scores, for
each attribute, are aggregated across the sample. The square root of the aggregate frequency
of “chosen as best” divided by the aggregate frequency of “chosen as worst” is used to standardize scores to a ratio scale. This simple calculation is merely an empirical model-free
estimate of the random utility model parameters. The highest

BEST
W ORST is scaled to 100

accordingly, and all other attributes scaled relative to this attribute.
The ratio scaled value for each attribute can be interpreted as the probability that the
average respondent believes the given attribute is most important relative to the remaining
J − 1 attributes. The count-based, standardized ratio scale ranks the attributes in the same
order as all other preliminary measures of attribute importance. In fact, the natural log of
this quantity is a good estimate of the conditional logit model parameters ran on the same
data. We should note here that the ratio scaled probabilities do not add to one because each
choice scenario only contains a subset of the total choice set of ten attributes. Therefore,
the probability of being chosen as most important is dependent on the inclusion of the other
attributes in the choice scenario.

32

Table 12: Apple Attribute Importance Measures using Best-Worst
Attribute
Most Least Agg
SQRT Std
Mean
B-W B/W Ratio B-W
Farm Size is small and 759
120
639
2.51
100
0.64
Corporate Involvement is
limited
Pollinator Management is 388
136
252
1.69
67.16 0.23
Employed
Ground Cover and Area 370
138
232
1.64
65.11 0.23
Management Practices
are Employed
Production, distribution, 392
229
163
1.31
52.02 0.16
and sale is done locally
Other Pests are con- 247
193
54
1.13
44.98 0.05
trolled using preventative measures, and habitat controls
Fertilizer and Nutrient 196
205
-9
0.98
38.88 -0.01
Materials are used minimally
Little to No Chemical 223
341
-118
0.81
32.15 -0.12
Herbicides are Used for
Weed Management
Farmers are financially 135
438
-303
0.56
22.07 -0.3
stable
Little to No Chemical 162
531
-369
0.55
21.96 -0.37
Pesticides are Used for
Pest Management
Consumer food prices are 140
681
-541
0.45
18.03 -0.54
affordable

Scaling
SD
Ind
B-W Ratio
1.61
100

0.87

35.94

0.91

35.94

1.07

25

0.77

7.81

0.79

-1.56

0.93

18.75

1.09

46.88
57.81

1.13

1.32

84.38

Half of the respondent sample was randomly assigned to complete the survey on apple
production practices. The results indicate that the attribute corresponding to “farm size
is small and corporate involvement is limited” has the highest best-worst score, as well as
being chosen as “most important” with the highest frequency. The ratio scaled parameter on
the farm size attribute infers that any given respondent would choose it as most important
with a probability of 0.64. Following farm size, pollinator management was ranked with a
probability of choice of 0.23. Worth noting is the relatively high best-worst score on the

33

attribute corresponding to “production, distribution, and sale is done locally”, with about
one fourth of the importance relative to farm size. “Consumer food prices are affordable”
received the lowest best-worst score and was chosen as least important more often in every
scenario than any of the remaining attributes.
These results should also be interpreted in relative magnitude. For example, pollinator
management is inferred to be roughly one third as important in a sustainable agricultural
system as farm size to this sample of consumers. In fact, in a sustainable agricultural system, farm size is found to be four times more important than local production, distribution
and sale, and is five times more important than affordable consumer food prices. The use
of off-farm chemical inputs such as fertilizers, herbicides, and pesticides all ranked between
three and four times less important to a sustainable agricultural system than farm size and
only marginally more important than consumer food prices. Overall, this sample of consumers indicated that the four economic attributes of sustainability included in this study
were located on the boundaries of their utility spectrum.

34

Table 13: Beef Attribute Importance Measures using Best-Worst Scaling
Attribute
Most Least Agg
SQRT Std
Mean SD
B-W B/W Ratio B-W B-W
Farm Size is small and 729
117
612
2.5
100
0.61
1.53
Corporate Involvement is
limited
Production, distribution, 519
119
400
2.09
83.66 0.4
1.06
and sale is done locally
Pests are controlled using 351
153
198
1.51
60.68 0.2
0.95
preventative measures,
cultural and nutritional
controls
Feed is Pasture Based 221
208
13
1.03
41.29 -0.01 0.74
and Waste Management
Systems Employed
Prohibited use of sub- 216
227
-11
0.98
39.08 -0.01 0.84
therapeutic antibiotics
Farmers are financially 223
304
-81
0.86
34.31 -0.08 1.05
stable
Prohibited use of geneti- 238
374
-136
0.8
31.96 -0.14 1.08
cally modified livestock
Consumer food prices are 194
517
-323
0.61
24.54 -0.32 1.27
affordable
Prohibited use of growth 181
506
-325
0.6
23.96 -0.32 1.14
hormones
Animal
Health
and 128
475
-347
0.52
20.8
-0.35 1.09
Safety are Protected

Ind
Ratio
100

65.57
32.79

-1.64

-1.64
13.11
22.95
52.46
52.46
57.38

The remaining half of the respondents were randomly assigned to complete the survey on
beef production practices. The data reveals that the attribute corresponding to “farm size is
small and corporate involvement is limited” has the highest best-worst score and the highest
frequency of choice as “most important” across all respondents. The attribute with the second highest best-worst score is “production, distribution, and sale is done locally”, suggesting
that the locality of meat production is more important in a sustainable agricultural system
to consumers than the locality of apple production relative to the other included attributes.
The ratio scaled values indicate that locality is roughly three quarters as important as farm
size. The attribute corresponding to “animal health and safety is protected” had the lowest

35

best-worst score and was chosen as least important more times in every scenario than any of
the remaining nine attributes. Affordable consumer food prices and prohibited use of growth
hormones also had relatively low best-worst scores.
When standardized to the ratio scale, farm size is found to be more than five times more
important in a sustainable agricultural system to consumers than animal health and safety.
Yielding the same results as the apple version of the survey, the respondents of the beef
survey ranked financial well-being of farmers and affordable consumer food prices in the bottom third on the underlying scale of importance. The use of sub-therapeutic antibiotics and
pasture based feed and waste management systems both exhibit a near zero mean, indicating
indifference among consumers.
The aggregated best-worst scores do not convey information about the heterogeneity that
may be present across sample consumers. The larger the range of heterogeneity across consumers, the more the market will respond to targeted channels of communication through
labeling schemes. The standard deviation of the best-worst score for each attribute measures the extent of variation amongst consumers. A higher standard deviation indicates a
wider variety of relative importance for a given attribute. Conversely, a smaller standard
deviation is indicative of general agreements across consumers on relative importance of a
given attribute.
The standard deviation is bounded by [-4,4] because each attribute could be chosen as
most or least important a maximum of four times in each version of the survey. For the
purposes of this study, a standard deviation above one can be interpreted as high heterogeneity across consumers. Attributes at both ends of the spectrum of relative importance
exhibit varying degrees of heterogeneity. Attributes with a high best-worst score and high
standard deviation are likely to be very important to a select subset of consumers. Additionally, attributes with a low best-worst score and high standard deviation have potential
in niche markets, since the attribute appeals to a much smaller consumer segment.
Distinct drivers of heterogeneity are important to identify which attributes are related or

36

jointly important for the same consumer segment. The variance-covariance matrix outlines
the pairs of attributes that vary simultaneously. For example, attributes that are highly covariable will both exhibit a high best-worst score within the same consumer group. Similarly,
attributes that are negatively correlated will likewise drive the same segment of consumers
but to opposite conclusions on importance. For this reason, correlation coefficients are often
easier for interpretation, due to their natural bounding in [-1,1]. The basis for the cluster analysis used in the body of this paper comes from attributes that tend to be tracked
together over consumers. Therefore, greater numbers of statistically significant correlation
coefficients imply a more structured market.
The highly negative correlation between farm size and consumer food prices indicates that
these attributes are likely to move in opposite directions of importance within a given group
of consumers. Another interesting result is the highly negative correlation between farm size
and chemical pesticide use. Additionally, the attribute corresponding to local production,
distribution and sale is significantly negatively correlated with both chemical pesticide and
herbicide use. Together, these observations suggest that there may be two distinct consumer
segments in this market; one that views traditional organic certification standards, outlining
chemical use and land management, as important components of sustainability, and one that
believes supporting local economic enterprise and small business is more important to the
sustainable model of production. We should note that these two beliefs do not need to be in
conflict, but they do offer insight into the perception of sustainability as it relates to more
well known organic food standards and how the systems may or may not compliment one
another.

37

GC
FN
CP
CH
PM
OP
FS
PL
CFP
FF

Table
GC
0.83
-0.05
-0.25
-0.12
0.12
0.09
-0.16
-0.03
-0.23
-0.15

14: Apple Production Attribute Variance-Covariance Matrix
FN
CP
CH
PM
OP
FS
PL
CFP

FF

0.63
0.04
0.07
-0.01
-0.03
-0.26
-0.13
-0.13
-0.09

1.19

GC
FN
CP
CH
PM
OP
FS
PL
CFP
FF

Table 15: Apple Production Attribute Correlation
GC
FN
CP
CH
PM
OP
FS
1
-0.06 1
-0.24 0.04
1
-0.15 0.1
0.45
1
0.15
-0.01 -0.22 -0.15 1
0.12
-0.05 0.1
-0.03 0.03
1
-0.11 -0.2
-0.31 -0.22 -0.12 -0.22 1
-0.04 -0.15 -0.28 -0.24 -0.09 -0.2
-0.01
-0.2
-0.13 -0.15 -0.18 -0.28 -0.11 -0.24
-0.15 -0.12 -0.19 -0.28 -0.07 -0.18 -0.17

1.28
0.46
-0.22
0.08
-0.54
-0.34
-0.22
-0.24

0.86
-0.12
-0.02
-0.32
-0.25
-0.22
-0.28

0.75
0.02
-0.18
-0.08
-0.33
-0.07

38

0.59
-0.27
-0.16
-0.11
-0.15

2.58
-0.03
-0.51
-0.3

1.15
-0.09
-0.06

1.76
0.14

Coefficients
PL
CFP

1
-0.06
-0.05

1
0.1

FF

1

SA
GH
GM
AS
PF
OP
FS
PL
CFP
FF

Table 16: Beef Production Attribute
SA
GH
GM
AS
PF
0.71
0.18
1.31
-0.09 0.28
1.16
-0.07 -0.1
0.02
1.19
-0.01 -0.04 -0.08 0.04
0.54
-0.05 -0.14 -0.09 -0.06 0.04
-0.27 -0.65 -0.35 -0.37 -0.13
-0.05 -0.26 -0.27 -0.24 -0.15
-0.15 -0.29 -0.36 -0.25 -0.14
-0.19 -0.27 -0.22 -0.12 -0.14

Variance-Covariance Matrix
OP
FS
PL
CFP

FF

0.9
-0.22
-0.07
-0.11
-0.18

1.6
0.13

1.1

SA
GH
GM
AS
PF
OP
FS
PL
CFP
FF

Table 17: Beef Production
SA
GH
GM
AS
1
0.19
1
-0.09 0.22
1
-0.09 -0.09 0.02
1
-0.02 -0.05 -0.1
0.06
-0.05 -0.13 -0.09 -0.05
-0.21 -0.37 -0.21 -0.22
-0.05 -0.21 -0.24 -0.2
-0.14 -0.19 -0.26 -0.18
-0.21 -0.21 -0.19 -0.09

Attribute Correlation Coefficients
PF
OP
FS
PL
CFP

FF

1
0.06
-0.12
-0.19
-0.15
-0.18

1

39

1
-0.15
-0.07
-0.09
-0.18

2.33
0.05
-0.31
-0.08

1
0.03
-0.16
-0.05

1.12
-0.12
-0.01

1
-0.1
-0.02

1
0.09

The beef survey results highlight the high heterogeneity across consumers in their importance ranking of the use of growth hormones, genetically modified livestock, and the
protection of animal health and safety. It is interesting to note that these three attributes
also exhibited highly negative correlation with farm size. These three attributes hinge critically on food safety and human health and nutrition dimensions of consumer utility and
may be motivated by intrinsic food values deeply connected to eco-responsibility movements
and the emergence of socially alternative food markets. These results support the findings
of Umberger, McFadden and Smith (2009) examining consumer valuation of hormone and
GM free claims, leading us to the same conclusion that social dimensions of food values have
the potential for creating distinct segments in the consumer market.

40

Apple Survey Attribute Definitions
In the next section you will be asked to choose which aspects of sustainable apple farming
are most and least important to you. Please take the time to read the following definitions
as related to sustainable production practices to better help you in your responses.

Ground Cover and Area Management Practices are Employed: Adjacent areas are planted
with hedgerows, windbreaks, or other low-maintenance plantings to encourage specific beneficial organisms. Within tree rows, ground cover or mulch are selected and maintained to
improve soil microbial activity, organic matter levels and nutrient cycling.
Fertilizer and Nutrient Materials are Used Minimally: Soil quality, including organic
matter content, is established at planting and maintained at an optimum level to minimize
commercial fertilizer needs.
Little to No Chemical Pesticides are Used for Pest Management: Chemical pesticides are
not used. Alternative strategies are employed, including biopesticides, mating disruption,
trap out and/or augmentation with beneficial organisms.
Little to No Chemical Herbicides are Used for Weed Management: Soil quality and ground
cover in the orchard and adjoining areas are planned and managed to prevent weeds and
weed seed immigration into the orchard. Cultural, mechanical or biological methods are
used to control weeds.
Pollinator Management is Employed: Bees are not placed in the orchard until blossoms
are open. Pesticides hazardous to bees are not used, or only if needed in an emergency.
Other Pests are Controlled Using Preventative Measures, and Habitat Controls: Habitat
is modified around orchards to reduce nesting and perching sites for pest birds.

41

Beef Survey Attribute Definitions
In the next section you will be asked to choose which aspects of sustainable cattle farming
are most and least important to you. Please take the time to read the following definitions
as related to sustainable production practices to better help you in your responses.

Prohibited Use of Sub-therapeutic Antibiotics : Animals may only be treated with antibiotics when necessary for treatment of illnesses, provided they are not slaughtered within
45 days of last treatment.
Prohibited Use of Growth Hormones: The use of hormone treatments, including implants,
to enhance growth is not permitted.
Prohibited Use of Genetically Modified Livestock: Animals produced through embryo
transfer and those whose genetic material has been altered are not permitted.
Animal Health and Safety are Protected: Animal nutrition on the farm results in superior
health as related to breeding success, weight gain, and freedom from illness. Policies are in
effect for low-stress handling, preventative health measures, and regular maintenance and
repair of facilities so as to prevent injury.
Feed is Pasture Based and Waste Management Systems Employed: Cattle receive majority of nutritional intake through grazing activity and animal movement is directed based
on cattleâĂŹs natural action and reaction to the situation. Manure resources are used to
close the nutrient cycle on the farm, but only to the extent that overall nutrient levels are
adequate and not excessive. Excess manure, if any, is put to good use off farm.
Pests are Controlled Using Preventative Measures, Cultural and Nutritional Controls:
Preventative measures and/or cultural controls such as movement of cattle, sanitation, and
composting are used to reduce or eliminate the need for insecticides and miticides. Animals
are free to choose and move to habitats that are most comfortable such as shady areas, windy
spots

42

REFERENCES

43

REFERENCES
Batte, M.T. 2010. “Consumer-driven changes in food marketing channels: organics and sustainable food systems in the United States: Discussion.”American Journal of Agricultural
Economics 93 (2): 604-605.
Bond, C.A., D. Thilmany, and J. Kelling-Bond. 2008. “Understanding consumer interest in product and process based attributes for fresh produce.” Agribusiness 24: 231-252.
Callens, I. and D. Tyteca. 1999. “Towards indicators of sustainable development for firms:
A productive efficiency perspective.” Ecological Economics 28: 41-53.
Casini, L. and A. M. Corsi. 2009. “Consumer preferences of wine in Italy applying bestworst scaling.” International Journal of Wine Business Research 21(1): 64-78.
Clonan, A., Holdsworth, M., Swift, J., and Wilson, P., 2010. “UK consumers priorities
for sustainable food purchases.” Available at http://purl.umn.edu/91948.
Cohen, E. 2009. “Applying best-worst scaling to wine marketing.” International Journal
of Wine Business Research 21: 8-23.
Flynn, A., and J.J. Louviere. 1992. “Determining the appropriate response to evidence
of public concern: the case of food safety.” Journal of Public Policy and Marketing 11:12-25.
Flynn, T.N., J.J. Louviere, T.J. Peters, and J. Coast. 2007. “Best-worst scaling: What
it can do for health care research and how to do it.” Journal of Health Economics 26:171-189.
Lusk, J.L. and B. Briggeman. 2009. “Food Values.” American Journal of Agricultural
Economics 91: 1-13.
Lusk, J. and N. Parker. 2009. “Consumer preferences for amount and type of fat in ground
beef.” Journal of Agricultural and Applied Economics 41(1):75-90.
Marley, A.A.J., and J.J. Louviere. 2005. “Some probabilistic models of best, worst, and
best-worst choices.” Journal of Mathematical Psychology 49: 464-480.
Mueller, S., and C.Rungie. 2009. “Is there more information in best-worst choice data?
Using attitude heterogeneity structure to identify consumer segments.” International Journal of Wine Business Research 21: 21-40.
Onozaka, Y. and D. Thilmany McFadden, 2010. “Defining sustainable food market segments: do motivations and values vary by shopping locale?” American Journal of Agricultural Economics 93(2): 583-589.

44

Onozaka, Y. and D. Thilmany McFadden. 2011. “Does local labeled complement or compete
with other sustainable labels? A conjoint analysis of direct and joint values for fresh produce
claims.” American Journal of Agricultural Economics 93 (3): 693-706.
Saunders, C., M. Guenther, W. Kaye-Blake, S. Miller, and P. Tait. 2010. “Consumer
attitudes towards sustainability attributes on food labels.” New Zealand Agricultural and
Resource Economics Society Inc. Available at http://purl.umn.edu/96944.
Scarpa, R., S. Notaro, J. Louviere, and R. Raffaelli. “Exploring Scale Effects of Best/Worst
Rank Ordered Choice Data to Estimate Benefits of Tourism in Alpine Grazing Commons.”
American Journal of Agricultural Economics 93(3): 813-828.
L L.Thurstone . ”The Method of Paired Comparisons for Social Values, “Journal of Abnormal and Social Psychology, 21, (1927): 384-400.
Umberger, Wendy J., Stringer, Randy and Mueller, Simone C., 2010. “Using Best-Worst
Scaling to Determine Market Channel Choice by Small Farmers in Indonesia” 2010 Annual
Meeting, July 25-27, 2010, Denver, Colorado 90853, Agricultural and Applied Economics
Association.
USDA NIFA. http://www.nal.usda.gov/afsic/pubs/terms/srb9902.shtml#toc2

45

Chapter 3: Consumer Preferences for Sustainably Produced Steak and Apples
Introduction
As suggested by our results in Chapter 2, sustainability, as a concept, remains elusive to
many. The word “sustainable” has been attributed to everything from the foods we eat, to
the structures we build, to the way we raise our families and run our businesses. While there
is certainly a breadth of literature on sustainability, exploring how it is measured and managed, there are surprisingly few studies examining sustainability attributes in the context of
food production. Two studies, to our knowledge, address consumer attitudes and preferences
towards sustainability attributes of food. First, Saunders et al (2010) use a discrete choice
model to investigate consumer decisions to purchase sustainably labeled foods, after displaying information about reduced carbon emissions. Alternatively, Clonan et al (2010) assess
attitudes towards sustainably produced food products through a questionnaire about previous purchases and a Likert scale ranking of preferences. Both studies were used to inform
the choice of sustainable farming attributes addressed in our survey. This work addresses a
gap in the literature by evaluating preferences for sustainably labeled food products within
the context of a choice experiment with varying degrees of information on sustainable agricultural systems.
Alternative agricultural production systems generate value-added food attributes that
are used for product differentiation and provide consumers with a means by which to control their ecological footprint through their consumption choices. “Sustainably produced”
food labels have rapidly grown in popularity over the past decade (Batte 2011). However,
because there is no government agency overseeing certification of these production methods,
evidence suggests that consumers are generally confused about the production attributes
that may or may not be present in a sustainable food system. This paper analyzes data
from a hypothetical choice experiment to better understand consumer purchasing behavior
when faced with competing food production attributes such as “organic” and “local”.

46

Choice experiments have a long history in the measurement of passive use values in environmental economics. Specifically, passive use value refers to the economic value derived
from a change in environmental quality that may not be directly observable in market behavior (Adamowicz 1998). That is, changes in food production attributes create an additional
dimension of consumer utility, which may be traded for other quality indicators, such as
price and certification. In real markets, consumers are faced with consumption choices over
bundles of attributes that can be modeled in a stated preference framework.
Marketing, transportation, and psychology literature led the field for some time in the
adaptation of choice experiments. The method arose from conjoint analysis, but differs in
the choice task to be completed. Unlike conjoint analysis, which utilizes scaled ranking or
rating systems, choice experiments more closely mirror the attribute bundles of competing
alternatives that are found in real markets. Choice experiments are compatible with random
utility theory and are thus useful for determining the share of preference a given attribute has
in a particular market. Therefore, hypothetical choice experiments provide a richer description of the attribute trade-offs that consumers are willing to make, than more traditionally
used contingent valuation methods.
In the past, choice experiments have been used successfully by economists to measure
the effect of environmental improvements and the value of quality differentiation (Boxall
and Adamowitz (1998)). Other economists have also employed choice experiments to value
food attributes such as local, organic and natural as well as more intrinsic values in the
realm of food nutrition and safety such as traceability, animal welfare and genetic modification. These studies generally show that, on average, consumers are willing to pay positive
price premiums for food produced outside of the conventional agricultural model. For example, Lusk, Norwood and Pruitt (2006) find positive price premiums associated with a ban on
sub-therapeutic antibiotics in the pork production industry and Liljenstolpe(2010) finds that
consumers indicate preference for food safety and animal welfare dimensions of value-added
pork attributes. Additionally, Onazaka et al (2008) find supporting evidence of higher price

47

premiums on organically certified and locally grown produce. Most closely related to this
work, Clonan et al (2010) find that stated purchasing behavior demonstrates that free range
and local products have higher market preference over other sustainability attributes.
We seek to estimate preferences for “sustainably produced” food products in a discrete
choice modeling framework and to determine how preferences may be affected by varying
degrees of information about sustainable agricultural systems. If there is a prevalence of
confusion about the attributes implied by sustainable labeling schemes, it is important to
know if providing information about specific sustainable production attributes can lessen
this uncertainty. Moreover, if information is successful in decreasing uncertainty over choice,
it could be used towards better market differentiation between sustainable and organic principles. Additionally, the willingness to pay measures estimated in this paper provide insight
into the value trade-offs perceived between current eco-labeling schemes, and the potential
for differentiating “sustainably produced” products from their “organic” and “local” counterparts.

Research Methodology
Discrete Choice Experiments
In a discrete choice modeling framework, all respondents are assumed to be utility maximizers, facing a choice among competing alternatives that return different levels of utility.
The analyst cannot directly observe respondent utility, but can observe attributes about
the competing alternatives. Specifically, define a random utility function (Uij ) as the ith
consumer’s utility of choosing option j. Then,

Uij = Vij + ij

(5)

where Vij is the deterministic component and ij is the stochastic error (Adamowicz et al
(1998)). Thus the probability that consumer i chooses alternative j is given by,

48

P rob(Vij + ij ≥ Vik + ik ) ; ∀k ∈ Ci

(6)

where Ci is the choice set of alternative bundles faced by consumer i. This choice experiment was constructed from alternative bundles of three attributes (label, certification, and
price) with varying levels.

Alternative Specific Design
An alternative specific design was utilized to capture labeling scheme tradeoffs. In a more
commonly used general design, the respondent is faced with a choice of the status quo
against one or more competing alternatives in which all attributes are allowed to vary across
all levels. Whereas, in an alternative specific design, each choice scenario presents each of
the level alternatives of a given attribute (Label) while allowing the remaining attributes to
vary across all levels. The respondents of our survey were faced with a number of choice
scenarios, and asked to choose between the product labeled sustainable, the product labeled
organic, the product labeled local, and the typical unlabeled product. The first three labeled
products varied in certification and all products varied in price in each choice scenario.
Additionally, each choice scenario presented the option of not buying any of the products
presented. The inclusion of opt-out more realistically mirrors true market opportunities. By
allowing respondents to opt-out of each choice scenario, we remove the assumption of market
participation which will reduce potentially “forced” choices.
The label attribute took four levels; sustainable, organic, local, or typical. Each label
was defined to describe the type of agricultural system within which the food was produced.
The certification attribute took one of three levels; USDA, Private Third Party or Self.
The “typical” product was reported to have no certification. All certifications referred to
verification of all processes used in production and claims made by an accompanying labeling
scheme. Private third party certification was verified by an independent entity unrelated to
the farm of origin or retailer of the product. “Self” certification referred to labeling claims
49

made by the farmer producing the food. The price attribute took one of three discrete levels:
(0.99/lb, 1.49/lb, or 1.99/lb) for apples and (5.99/lb, 8.99/lb, or 11.99/lb) for ribeye steak,
as determined by the market at the time of the survey design. The following table outlines
the choice experiment attributes and levels used in our design. For full descriptions and
definitions please refer to the Appendix.
Table 18: Attributes and Levels Used in the Choice Experiment
Attribute
Levels
Label

Sustainable
Organic
Local
Typical

Certification

USDA
Private Third Party
Self

Product Specific Prices
Price (Apple)

0.99/lb
1.49/lb
1.99/lb

Price (Steak)

5.99/lb
8.99/lb
11.99/lb

50

Data Collection
The data used for this analysis were gathered in our 2010 national web-based survey of
1002 households, described in Chapter 2. Two survey versions were developed; one for
apples and one for steak (beef). All respondents in the sample were randomly assigned
to complete only one product version of the survey. The choice experiment portion of the
survey was preceded by questions about perceived importance of varying sustainable farming
attributes using a best-worst framework, analyzed in the previous Chapter, as well as several
Likert scale ranking tasks assessing previous knowledge of agricultural production system
characteristics and food consumption history. This combination of respondent information
provides our analysis with a comprehensive assessment of the perceptions and preferences
related to sustainability attributes. Immediately prior to the hypothetical choice experiment
each respondent was provided with one of two information treatments about sustainable
agricultural practices. A summary of survey participant demographic statistics is repeated
here for reference.

51

Table 19: Demographic Variables and Summary Statistics of Survey Participants
Variable
Variable Description
Apple
Steak
Gender
1= Male; 2 = Female
1.476
1.516
Total Participants
500
502
Age
Average Age in Years
51.48
50.92
Adults
Number of Adults in Household
2.062
2.048
Children
Number of Children in Household 0.48
0.51
Meals
Number of Meals/ week with 6.64
3.19
Product
Shop
% of Total Shopping at Location
Grocery Store
81.97
81.16
Health Food Store
8.49
9.20
Food Co-op
2.05
1.90
Convenience Store
3.0
4.42
Farmers Market
3.72
5.24
Butcher
4.38
2.92
Assistance 1 = On Food Assistance; 2 =
Otherwise
1 = On Food Assistance
13.2 %
11.95 %
2 = Not on Food Assistance
86.8 %
88.05 %
Education Highest Level Completed
1 = Did not graduate from high 2.2 %
2.59 %
school
2 = Graduated from high school, 17.8 %
18.12 %
no college
3 = Attended college, no degree 28.8 %
33.67 %
earned
4 = Attended college, associates 15.4%
12.55 %
or trade degree earned
5 = Attended college, Bachelors 24.6 %
22.52 %
degree earned
6 = Graduate or advanced degree 11.2 %
10.56 %
earned
Household Range of Pre-tax Income
Income
1 = Less than $20,000
20.8%
19.12 %
2 = $20,000 - $39,999
26.4%
28.88 %
3 = $40,000 - $59,999
18.4 %
21.51 %
4 = $60,000 - $79,999
17.6 %
11.95 %
5 = $80,000 - $99,999
8.2 %
7.57 %
6 = $100,000 - $119,000
3.8 %
4.98 %
7 = $120,000 - $139,999
1.8%
2.39 %
8 = $140,000 - $159,999
1.4 %
1.20 %
9 = More than $160,000
1.6%
2.39 %
52

Information Treatments
Half of the respondents were randomly assigned to receive general information about sustainable agriculture from the USDA website. This information treatment outlined general
principles of sustainability such as, “resource conserving”, “socially supportive”, and “economically viable”. Alternatively, the remaining half of respondents received information
about sustainable agricultural practices that are components of a sustainable certification
scheme from the aforementioned third party certifier, Food Alliance. This information treatment provided eight detailed standards of sustainable agriculture such as energy conservation
and waste recycling, reduced use of chemical inputs, and fair and ethical treatment of workers and livestock. Please see the Appendix for full information treatments.
Following the information treatment, a brief cheap talk script was included to mitigate
the problems associated with hypothetical bias. The effectiveness of cheap talk scripts has
been repeatedly confirmed in the literature. Notably, Aadland and Caplan (2006) suggest
neutral scripts that avoid assumptions about positive bias, as different subsamples may react
differently dependent on factors such as market familiarity. Tonsor and Shupp (2011) utilize
a large national survey and split-sample experimental design, finding that cheap talk scripts
may not only influence the level of willingness to pay estimated for representative consumers,
but also, in general, produce more reliable estimates.
To maintain orthogonality and independence across our choice experiment a main effects orthogonal experimental design was employed. In a main effects orthogonal design,
a subset of the full factorial design is selected such that all linearly additive utility terms
are identifiable. Specifically, the ORTHOPLAN procedure in SPSS Conjoint identified an
orthogonal design based on two attributes, each with three levels, yielding a choice set of
eighteen alternatives. The design utilized in this study is balanced, that is, each level of
the non-alternative specific attributes (price and certification) occurs with equal frequency
across the entire choice set. Therefore, each attribute has equivalent statistical power in
explaining preference. For the sake of brevity, and to lower the complexity of the overall
53

task, the respondents were randomly assigned to answer one of two blocks of nine questions.
The order of the alternatives was randomized for participants to mitigate any ordering bias.
An example of a question faced by all apple survey respondents follows:

Survey Question Example
Which one of the following apple displays listed below would you choose to purchase from?

Table 20: Choice Experiment Survey Question Example
Sustainable
Organic
Local
Typical I choose not to
purchase any of
these options.
Certification Private 3rd Party
USDA
Self
Price
$1.49 /lb
$1.99 /lb $0.99 /lb $1.49 /lb
2
2
2
2
2
Label

Analysis
The choice experiment structure allowed each respondent to select from four versions of
the food product: (1) sustainably labeled, (2) organically labeled, (3) locally labeled, (4)
unlabeled, or to opt-out of the purchase. This data is analyzed in a random utility framework.
In each scenario, the choice of an alternative represents a discrete choice from a set of
competing alternatives. Each alternative in a given choice scenario can be represented with
a utility function that contains a deterministic and a stochastic component. An individual
will choose a given alternative only if the utility representing that alternative is greater than
all other options in the choice scenario. If the stochastic component ij is distributed i.i.d.
extreme value then Adamowicz et al (1998), Boxall and Adamowicz (2002) and Lusk et al
(2003) have shown that the probability of consumer i choosing alternative j is equal to,

P rob(j) =

sV
e ij
sV
k∈C e ik
54

(7)

where s is a scale parameter. In a single sample the scale parameter cannot be directly
identified and is thus assumed to be 1, according to Lusk et al (2003). However, if data
is collected from more than one sample, the relative scale parameter can be calculated and
accounts for the difference in the variation of unobserved effects on error variance heterogeneity. To test whether data from the two information treatments (USDA versus Food
Alliance) can be pooled, we estimate the relative scale parameter, across the two data sets
to control for potential differences in variance. A likelihood ratio test of parameter equality, after controlling for scale, is undertaken to determine if pooling is appropriate for the
following parametric analysis. The parameters contributing to the probability of choice are
then first estimated in a multinomial logit model (MNL).
For choice problems involving three or more alternatives, the multinomial technique is
most often employed. In the multinomial framework, the data on a dependent variable can
fall into one of several mutually exclusive categories. However, analysis is not necessarily
straightforward, as there are many different models for the probabilities of the multinomial
distribution. Model selection depends on whether or not some of the individual specific
regressors vary across alternatives and whether the categories are ordered or unordered
(Cameron and Trivedi 2009). For the purposes of this analysis, we employ and compare the
fit of three models: the multinomial logit, the conditional logit, and the random parameters
logit.
In each of these models, the left hand side variable represents choice and is coded as
a dummy variable, taking the value of 1 indicating the alternative was chosen and 0 otherwise. Therefore each choice scenario yields five data observations for every respondent,
corresponding to the four competing product alternatives plus opt-out. The right hand side
variables reflect the attribute levels of each product that is available for purchase. To avoid
the assumption of equal marginal trade-offs, certification enters the model as an effectscoded variable. For the purposes of this study, we allowed USDA to act as the base-case
since consumers are likely most familiar with USDA certification and labeling. Thus, two

55

variables were included for certification, for self and private third party, respectively. Each
certification variable took a value of -1 to indicate USDA certification, 1 to indicate certification accordingly and 0 otherwise. The remaining right hand side variables are alternative
specific dummy variables corresponding to each of the four label attributes. Price was coded
as a continuous variable. Therefore, the logistic regression to be estimated for participant i
for each alternative j is as follows,

Choiceij = βP riceij + δ1 CertPij + δ2 CertSij
+ γ1 LabelSij + γ2 LabelOij + γ3 LabelLij + γ4 LabelTij

(8)

The parameter on price (β) approximates mean marginal utility of income and the parameters on each certification variable (δ1 and δ2 ) indicate the marginal (dis)utility associated
with a change from USDA certification to Private Third Party certification or Self certification, respectively. The parameters on each label variable (γ1 , γ2 , γ3 , and γ4 ) indicate the
marginal utility gained from the labeling claim (or lack thereof) on each product relative to
opt-out.
Thus, average willingness to pay for each label attribute, ceteris paribus, can be calculated as,
γ
W T P (Labeli ) = i
β

(9)

and the price premium that can be captured on average for each labeling scheme, relative to
the typical (unlabeled) product is,

P remium(Labeli ) =

(γi − γ4 )
β

(10)

The parameters of multinomial models are not directly interpretable in general. The es-

56

timated parameters will be transformed into marginal effects measures (WTP) to reflect
the increased (or decreased) probability of choice accompanying a one unit change in the
attribute level of each variable. Thus, it is anticipated that price and certification variables
will exhibit negative coefficient estimates, as increased price and certification other than
USDA should provide disutility to the consumer under our model assumptions.

Multinomial Logit
The multinomial logit model is best for datasets that contain only case-specific variables. It
is the simplest model computationally and yields parameter results that are the easiest to
interpret. To ensure model identification, the coefficient on one of the explanatory variables
is set to zero and all other coefficients interpreted with respect to that category. For example,
in our specification we have identified the base category as the no purchase option to make
all coefficients relative to market participation. The multinomial logit requires data to be in
wide-form, as explained in Chapter 2.

Conditional Logit
A closely related technique, the conditional logit, is a model also suited for behavioral modeling of polychotomous choice situations. Developed by McFadden in 1973, the conditional
logit model is an arguably preferable technique for estimating the class of models in which
choice among alternatives is a function of the characteristics of the alternatives rather than
characteristics of the individual making the choice, such as quality measures for the alternatives, which is the case here. The conditional logit model requires that the data be in long
form, as discussed in the previous Chapter. All case-specific variables will appear as a single
variable that takes the same value for the five alternative outcomes. The two models are
similar and perfectly symmetric in their statistical specification. The only difference is that
the explanatory variables are open to assume different values in each alternative in the CL
model, although the impact of a one unit change in an explanatory variable is assumed to be

57

constant across alternatives. It should be noted that multinomial logit and conditional logit
models assume that all respondents share the same coefficients for a given attribute, an assumption of consumer homogeneity across preferences. This assumption is likely unrealistic
if there are expectations that consumer preferences are in fact heterogeneous.

Random Parameters Logit
There are two primary strategies for addressing consumer heterogeneity in random utility models that differ in their assumptions about the distribution of preferences (Sagebiel
(2011)). The first is to employ a random parameters logit (RPL) estimation on the same
data set. The RPL accommodates heterogeneity in the data as a continuous function of
the parameters and is free of the independence of irrelevant alternatives (IIA) assumption.
The RPL is a more flexible method for modeling heterogeneity, compared to the latent class
model employed in the previous Chapter, because it can induce any behavioral assumption
in terms of preference distribution (Sagebiel). The conditional logit parameters are assumed
fixed, taking the same value for all respondents, whereas the RPL is characterized by randomness in parameters. The unconditional RPL, used here, measures choice probability as
a weighted average of all possible coefficients for the attribute parameters. The maximum
likelihood estimates are achieved through simulation.
In contrast, the conditional logit specification can model heterogeneity by incorporating interaction terms of case-specific demographic variables or perception rankings. The
drawback with this method is that the interaction terms introduced are motivated by the
researcher’s choice, and therefore are more exploratory in nature and not included in this
analysis. The conditional and random parameters logit specifications are compared with the
multinomial logit model by means of a likelihood ratio test to determine the best fit to the
data. Furthermore, coefficient choice probabilities and willingness to pay estimates will be
contrasted.

58

Results and Discussion
Information Treatment Tests
To specify our final model, we first estimated two separate multinomial logit models corresponding to each of the information treatments. Then the pooled data was used to estimate
the same model, where utility parameters were constrained to be equal across information
treatments. The test for parameter equality yields a test statistic that is calculated as follows
(Lusk (2006)),
λ = −2(LLp − (LLu + LLf ))

(11)

where LLp is the log likelihood value of the pooled model after controlling for scale, LLu
is the log likelihood value of the USDA information treatment model and LLf is the log
likelihood value of the Food Alliance information treatment model. The test statistic, λ is
distributed χ2 with K(M − 1) degrees of freedom, where K is the number of restrictions
(seven) and M is the number of information treatments (two).

Table 21: H0 : Pooling Across Information Treatments: Apple Survey
Subsample Modeled
n
LL
p-Value
All Respondents
502
11198.3910
USDA Information Treatment
251
-5568.8387
Food Alliance Information Treatment
251
-5623.3510
H0 : Pooling across Information Treatments is Okay
0.8807

Table 22: H0 : Pooling Across Information Treatments: Beef Survey
Subsample Modeled
n
LL
p-Value
All Respondents
500
10517.9060
USDA Information Treatment
248
-5320.7855
Food Alliance Information Treatment
252
-5193.1726
H0 : Pooling across Information Treatments is Okay
0.3419

59

Comparing the two likelihood function values using the likelihood ratio test resulted with
associated two-tailed p values of 0.8807252 for apples and 0.34187185 for steak, indicating
that the null hypothesis of parameter equality across information treatments cannot be rejected at any standard level of significance.
The likelihood ratio test results imply that information had no statistically significant
effect on choices made in the hypothetical choice experiment across labeling schemes. Consumers with more specific information about the production standards required for third
party certification exhibited preferences similar to those that received more general information about sustainable principles from the USDA. An interesting extension of this analysis
could have included a third subsample, randomly assigned to receive no information, as was
done in Lusk, Norwood and Pruitt (2006). The insignificance of the information treatment
may be due to competing preconceptions about sustainability attributes, or it may be the
persistence of general confusion over how these sustainable principles are implemented in
practice.
Tables 23 and 24 present our primary estimation results. The MNL results are as expected, as price coefficients are negative and significant, certification parameters are negative
and significant (with the exception of CertS for the USDA Apple sample), and label parameters are positive and significant.

60

Table 23: Information Treatment Comparison: Apple Survey
Treatment Treatment
Variable
Pooled
USDA
FA
Price
-2.140
-2.118
-2.161
(0.537)*
(0.076)*
(0.076)*
CertP
-0.250
-0.206
-0.291
(0.031)*
(0.043)*
(0.044)*
CertS
-0.076
-0.157
-0.058
(0.030)**
(0.043)
(0.042)*
LabelS
1.803
1.820
1.784
(0.079)*
(0.113)*
(0.113)*
LabelO
1.884
1.811
1.949
(0.079)*
(0.113)*
(0.113)*
LabelL
2.071
1.984
2.152
(0.080)*
(0.114)*
(0.113)*
LabelT
0.603
0.547
0.656
(0.085)*
(0.121)*
(0.119)*
(*) and (**) significant at 0.01 and 0.05 levels, respectively

Table 24: Information Treatment Comparison: Beef Survey
Treatment Treatment
Variable
Pooled
USDA
FA
Price
-0.379
-0.358
-0.402
(0.010)*
(0.013)*
(0.014)*
CertP
-0.321
-0.305
-0.337
(0.033)*
(0.047)*
(0.048)*
CertS
-0.191
-0.155
-0.228
(0.032)*
(0.046)*
(0.047)
LabelS
1.853
1.685
2.029
(0.085)*
(0.119)*
(0.122)*
LabelO
1.847
1.702
1.998
(0.085)*
(0.119)*
(0.122)*
LabelL
1.866
1.694
2.048
(0.085)*
(0.119)*
(0.123)*
LabelT
0.818
0.672
0.972
(0.089)*
(0.124)*
(0.127)*
(*) and (**) significant at 0.01 and 0.05 levels, respectively

61

Pooled Model Results
Moving forward with the pooled sample, three models were run on the data: the multinomial
logit, the conditional logit, and the random parameters logit. In general, the results indicate
that on average, as determined by the relative magnitudes of the coefficient estimates, the
preference ordering for labeling claims on apples are as follows in descending order: 1)
Local 2) Organic 3) Sustainable and 4) Typical. All parameters on Label are positive and
significant at the 1% level in all three models. For beef, the results indicate that on average,
as determined by the relative magnitudes of the coefficient estimates, the preference ordering
for labeling claims on ribeye steaks are such that local is the most preferred, followed by
an equal preference across organic and sustainable labels, and the typical product least
preferred. Similar to the apple results, all parameters on Label are positive and significant
in all three models, again at the 1% level.
The negative and statistically significant estimates on private third party certification and
self certification indicate that consumers receive disutility from either of these certification
schemes relative to USDA certification, ceteris paribus. Accordingly, the relative magnitude
of the two certification coefficients reveals that for both apple and beef samples, Self certified
products are least favorable, holding all other quality attributes constant. Also, as expected
the price coefficient is negative and significant for both apple and steak samples. The discrete
response profiles of choice frequencies across the five alternatives and coefficients for all three
models are reported in the following tables.

62

Table 25: Discrete Response Profile: Apple Survey
Choice
Frequency Percent
Sustainable
1064
23.25 %
Organic
1123
24.86 %
Local
1267
28.04 %
Typical
415
9.19 %
No Purchase
649
14.36 %

Table 26: MNL, CL, and RPL Parameter
Variable
MNL
Price
-2.14*
CertP
-0.25*
CertS
-0.07**
LabelS
1.80*
LabelO
1.88*
LabelL
2.07*
LabelT
0.60*

Estimates: Apple Choice Experiment
CL
RPL
-2.13*
-2.35*
-0.25*
-0.30*
-0.05
-0.11*
3.56*
3.61*
3.64*
3.75*
3.82*
3.93*
2.39*
2.73*

Log Likelihood
-11198.39
-8982.39
-5963.00
(*) and (**) significant at 0.01 and 0.05 levels, respectively

Table 27: WTP Estimates for Apple Labeling Scheme Attributes
Model
Model
Model
Used
Used
Used
Welfare Measure MNL
CL
RPL
WTP LabelS
$0.84/lb
$1.67/lb
$1.54/lb
WTP LabelO
$0.88/lb
$1.71/lb
$1.60/lb
WTP LabelL
$0.97/lb
$1.79/lb
$1.67/lb
WTP LabelT
$0.28/lb
$1.12/lb
$1.16/lb
Premium(LabelS) $0.56/lb
$0.55/lb
$0.37lb
Premium(LabelO) $0.60/lb
$0.59/lb
$0.43/lb
Premium(LabelL) $0.69/lb
$0.67/lb
$0.51/lb

63

Using Log Likelihood as a measure of fit, we see that the Random Parameters Logit
out-performs the multinomial and conditional logit models for apple responses. This result
indicates significant evidence of heterogeneity in consumer preferences for our sample. The
RPL model yields absolute WTP estimates in between those of the MNL and CL models,
but lower price premiums overall. Applying a simple welfare measurement to the parameter
estimates, we find that consumers are willing to pay $1.67/lb on average for locally produced
apples, $1.60/lb on average for organic apples, $1.54/lb on average for sustainably labeled
apples, and $1.16/lb on average for typical, unlabeled apples. These results yield price
premiums of $0.51/lb on local apples, $0.43/lb on organic apples and $0.37/lb on sustainably
produced apples, relative to their typical, unlabeled counterpart.

64

Table 28: Discrete Response Profile: Steak Survey
Choice
Frequency Percent
Sustainable
971
21.58 %
Organic
967
21.49 %
Local
979
21.76 %
Typical
424
9.42 %
No Purchase
1159
25.76 %

Table 29: MNL, CL, and RPL Parameter
Variable
MNL
Price
-0.38*
CertP
-0.32*
CertS
-0.19*
LabelS
1.85*
LabelO
1.85*
LabelL
1.87*
LabelT
0.82*

Estimates: Steak Choice Experiment
CL
RPL
-0.38*
-0.38*
-0.32*
-0.34*
-0.16*
-0.23*
2.89*
2.97*
2.88*
2.99*
2.91*
2.99*
1.88*
2.12*

Log Likelihood
-10517.91
-9027.11
-5985.00
(*) and (**) significant at 0.10 and 0.05 levels, respectively

Table 30: WTP Estimates for Steak Labeling Scheme Attributes
Model
Model
Model
Used
Used
Used
Welfare Measure MNL
CL
RPL
WTP LabelS
$5.00/lb
$7.61/lb
$7.81/lb
WTP LabelO
$5.00/lb
$7.58/lb
$7.86/lb
WTP LabelL
$5.02/lb
$7.66/lb
$7.86/lb
WTP LabelT
$2.22/lb
$4.95/lb
$5.58/lb
Premium(LabelS) $2.78/lb
$2.66/lb
$2.24lb
Premium(LabelO) $2.78/lb
$2.63/lb
$2.29/lb
Premium(LabelL) $2.81/lb
$2.71/lb
$2.29/lb

65

Using Log Likelihood as a measure of fit, we see again that the Random Parameters Logit
out-performs the multinomial and conditional logit models for the steak responses. Applying
the same welfare measurement to the steak participant subsample, we find that consumers
are willing to pay $7.86/lb on average for locally or organically produced steak, $7.81/lb for
sustainably produced steak, and $5.58/lb for typical unlabeled steak. These results yield
price premiums of $2.29/lb on local and organic steaks, and $2.24/lb for sustainable steaks
relative to their typical, unlabeled counterpart.
All welfare measurements should be interpreted relative to opt-out. That is, the willingness to pay on each label attribute is the amount of money the average consumer is willing
to pay for each labeled product after choosing to participate in the market, holding all else
constant and assuming no uncertainty regarding choice. Accounting for heterogeneity, by
employing the random parameters logit to model this data, WTP estimates fall in the range
of true market value for these products at 2010 price levels.
The positive willingness to pay estimates on all labels imply that all of the products individually are preferred to not buying any. On average, consumers are willing to pay the most
for locally produced apples. These results imply that consumers prefer locally produced
and organically produced steaks equally. Following the preference for locally grown and
produced apples and steak, organic and sustainably labeled products have the next highest
willingness to pay, for apples and steak respectively. However, in both cases the willingness
to pay welfare measurements are close in magnitude, which could easily lead to a reversal of
the preference ordering for these two label attributes if this experiment was replicated with
another population.

WTP comparison
The WTP estimates reported above are calculated from the RPL mean valuations of model
covariates. The assumptions inherent in this model ignore the distribution of preferences
around the mean of random parameters. We relax the strength of this assumption by fol-

66

lowing the simulation techniques prescribed in Hensher et al (2006) and similarly utilized
in Tonsor et al (2009). Specifically, we consider the entire distribution of WTP and the
statistical variability in parameter estimates to capture heterogeneous preferences and to
empirically test the null hypothesis of equal WTP preferences across (alternative specific)
labels.
95% confidence intervals of WTP are derived from generating a distribution of 999 WTP
estimates using a parametric bootstrapping technique proposed by Krinsky and Robb (1986).
The parametric bootstrapping method essentially draws 999 observations from multivariate
normal distributions around the coefficients from our logit model by resampling from the
original sample with replacement. For tests at the α level or for 100(1−α)% confidence there
are reasons for choosing the number of draws, B, such that α(B +1) is an integer. Therefore,
we follow the advice of Cameron and Trivedi (2010) and use B = 999 for confidence intervals
and hypothesis tests when α = 0.05. Mean WTP estimates and 95% confidence intervals
are identified incorporating both statistical and preference variability. An initial evaluation
of overlapping 95 % confidence intervals yields a general assessment of WTP differences.

67

Table 31: Hypothesis Test of Equal WTP Across Credence Labels: Apple Survey
Average
95% interval
Label
WTP LabelS
$1.61/lb
($1.45/lb, $1.79/lb)
WTP LabelO
$1.73/lb
($1.57/lb, $1.93/lb)
WTP LabelL
$1.78/lb
($1.60/lb, $1.97/lb)
WTP LabelT
$1.14/lb
($1.00/lb, $1.30/lb)

Table 32: Hypothesis Test of Equal WTP Across Credence Labels: Steak Survey
Average
95% interval
Label
WTP LabelS
$7.75/lb
($6.87/lb, $8.74/lb)
WTP LabelO
$7.75/lb
($6.87/lb, $8.74/lb)
WTP LabelL
$7.83/lb
($6.92/lb, $8.82/lb)
WTP LabelT
$4.97/lb
($4.26/lb, $5.80/lb)

The bootstrapped estimates reveal that the 95% confidence intervals on all three credence
labeling claims overlap for both apple and steak survey respondents. This result indicates
that consumers do not differentiate between sustainable, organic, and local labels. However,
because none of the three credence labeling attribute intervals overlap with the typical,
unlabeled product interval we can conclude that each of the labels individually is successfully
differentiated from the unlabeled product.

68

Conclusions
The objectives of this paper were to employ a stated preference approach, utilizing a choice
experiment framework, for measuring value associated with quality changes in sustainable
agricultural production practices, and to determine if providing information about sustainable agricultural practices affects willingness to pay estimates for sustainably labeled food
products. Our results support previous studies of food attribute valuation (Bond et al (2008),
Lusk and Briggeman (2009)) and illustrate the higher preference for locally grown and produced foods (Detoni and Tonsor (2009)). Moreover, our analysis provides evidence of weak
market differentiation between sustainable, organic and locally labeled products, which may
further exacerbate uncertainty regarding the attributes associated with these credence labeling schemes. Abram et al (2010) found similar results when evaluating perceptions of
“all natural” claims against organic. Providing varying levels of information on sustainable
agricultural practices did not yield significantly different preferences across respondents of
our survey.
Our analysis provides significant evidence of consumer heterogeneity in our sample and
demonstrates that positive price premiums can be captured by sustainably produced labeling
claims, relative to similar unlabeled and conventionally produced food products. However,
the price premiums calculated in our model reveal that there is a comparable tradeoff in quality associated between local, organic and sustainably labeled food products. Furthermore,
detailed information about sustainable certification guidelines had no significant impact on
choosing the sustainably labeled products. Thus, we conclude that, based on our results,
consumer demand for sustainably produced food may not be distinctly differentiable from
its local and organic counterparts. Overall, these findings suggest that profitable marketing
opportunities may exist for firms interested in selling sustainably produced food products,
however there needs to be considerable effort put into leveling information asymmetries
about product quality if sustainable label claims are to be differentiated from the local food
movement or more recognizable organic principles.
69

APPENDICES

70

USDA Information Treatment
Please consider the following information [provided by the United States Department of Agriculture]:
The word âsustain from the Latin sustinere (sus, from below and tenere, to hold), to keep
in existence or maintain, implies long-term support or permanence. As it pertains to agriculture, sustainable describes farming systems that are capable of maintaining their productivity and usefulness to society indefinitely. Such systems must be resource-conserving, socially
supportive, commercially competitive, and environmentally sound. Sustainable agriculture
was addressed by Congress in the 1990 Farm Bill and the Food, Agriculture, Conservation,
and Trade Act of 1990 (FACTA). Under that law, the term sustainable agriculture means an
integrated system of plant and animal production practices having a site-specific application
that will, over the long term:
1. Satisfy human food and fiber needs
2. Enhance environmental quality and the natural resource base upon which the agricultural economy depends
3. Make the most efficient use of nonrenewable resources and on-farm resources and integrate, where appropriate, natural biological cycles and controls
4. Sustain the economic viability of farm operations
5. Enhance the quality of life for farmers and society as a whole
Consumers are increasingly demanding on the ethical dimension of food quality. This
relates to the process of production and trade and its broad impacts on society and the
environment. It includes a wide range of social, environmental or cultural issues such as
the treatment of workers, a fair return to producers, environmental impacts and animal
welfare. Guidelines about what specific practices meet long-term environmental, economic
and social goals and constitute sustainable agriculture is still under debate. However, a
71

handful of groups have attempted to develop standards and/or provide certification services
based upon their own guidelines.

72

Food Alliance Information Treatment
Please consider the following information [provided by Food Alliance, an independent third party certifier of sustainably produced foods]:

The impacts of food production have become a mainstream concern. Expectations for
traceability, transparency and accountability in agriculture and the food industry are increasing. Sustainable agriculture comprises the ability to produce safe, healthy, delicious,
and affordable food to meet needs without degrading agricultural lands, the quality of life in
our communities, or the resiliency of the broader ecosystems on which we all depend. Farms
employing sustainability practices place important on issues including safe and fair working
conditions, humane treatment of animals, and protection of the environment. Sustainable
farms should be held to the following standards:
1. Protect, enhance, and conserve soil resources, water resources, and biodiversity
Food production improves soil productivity, protects water quality and supply, and
supports healthy native plant and animal communities.
2. Conserve energy, reduce and recycle waste
Waste streams from food production are minimized while reuse, recycling, and composting of resources is maximized. Businesses invest in innovation and improvement
to ensure efficient use and management of natural resources for energy and packaging,
transport, and daily operations.

73

3. Reduce use of pesticides, and other toxic and hazardous materials
Food businesses avoid use of chemicals that have adverse impacts on the health of
ecosystems. Agriculture relies on a biologically based system of Integrated Pest Management. Materials used for sanitation, pest control, waste treatment, and infrastructural maintenance are chosen to reduce overall negative consequences.
4. Maintain transparent and sustainable chain of custody
Farmers and food industry workers have secure and rewarding jobs that provide a
sound livelihood. Throughout the entire supply chain, food is produced and handled
in accordance with these Principle Values. Transparency is maintained independent
standards, third-party audits and clear labeling.
5. Guarantee product integrity, no genetically engineered or artificial ingredients
Foods are not produced using synthetic preservatives, artificial colors and flavors, genetically modified organisms (GMOs), or products derived from livestock treated with
sub-therapeutic antibiotics or growth-promoting hormones.
6. Support safe and fair working conditions
Employers respect workersâ rights and well-being, make safety a priority, maintain a
professional workplace, and provide opportunities for training and advancement.
7. Ensure healthy, humane animal treatment
Animals are treated with care and respect. Living conditions provide access to natural
light, fresh air, fresh water, and a healthy diet, shelter from extremes of temperature,
and adequate space and the opportunity to engage in natural behaviors and have social
contact with other animals. Livestock producers minimize animal fear and stress during
handling, transportation and slaughter.

74

8. Continually improve practices
Food businesses are committed to continually improving management practices. Improvement goals are integrated into company culture, regularly monitored, and acknowledged when achieved. Food buyers are proactively engaged in the food system, and support companies that are transparent about their improvement goals and
progress.

75

Choice Experiment Instructions, Definitions, and Cheap Talk Script
In the next section you will be presented with multiple different alternative packages of beef
ribeye steak that could be available for purchase in a retail store where you typically shop.
Besides the attributes listed below, each product possesses the same characteristics (e.g.,
similar color and freshness) and is produced in the U.S. Prices vary for each product and are
all in $/lb. units. Please consider the following information to help you interpret alternative
products.

Label: The package that contains the beef ribeye steak for your purchase may be labeled as
follows:
• Sustainable: This beef was produced using sustainable practices.
• Organic: This beef was produced using organic practices.
• Local: This beef was produced for distribution and sale locally.
• Typical: This beef is not labeled to suggest it was produced using any of the criteria
listed above.
Certification: The typical product has no certification label. Each labeled product can be
certified in one of three ways:
• USDA: The processes used and all claims made by the product label have been verified
by the USDA.
• Independent Third Party: The processes used and all claims made by the product label
have been verified by a third party unrelated to the farm of origin or retailer.
• Self: The processes used and all claims made by the product label have been verified
by the farmer producing the food.

76

The experience from previous similar surveys is that people often state a higher willingness to pay than what one is actually willing to pay for the good. For instance, a recent
study asked people whether they would purchase a new food product similar to the one you
are about to be asked about. This purchase was hypothetical (as it will be for you) in that
no one actually had to pay money when they indicated a willingness to purchase. In the
study, 80% of people said they would buy the new product, but when a grocery store actually
stocked the product, only 43% of people actually bought the new product when they had to
pay for it. This difference (43% vs. 80%) is what we refer to as hypothetical bias.

Accordingly, it is important that you make each of your upcoming selections like you would
if you were actually facing these exact choices at a store; noting that buying a product means
that you would have less money available for other purchases.

77

REFERENCES

78

REFERENCES
Aadland, D. and Caplan, A.J. 2006. Cheap talk reconsidered: new evidence from CVM.
Journal of Economic Behavior and Organization 60, 562-578.
Abrams, Katie, Courtney Meyers, and Tracy Irani, 2010. Naturally Confused: Consumers’
Perceptions of All-Natural and Organic Pork Products. Agriculture and Human Values 27:
365-374.
Adamowicz, W., Boxall, R., Williams, M., and Louviere, J., 1998. Stated preference approaches for measuring passive use values: Choice experiments and contingent valuation.
American Journal of Agricultural Economics 80, 64-75.
Batte, M.T., 2010. Consumer driven changes in food marketing channels: organics and
sustainable food systems in the united states: discussion. American Journal of Agricultural
Economics 93, 604-605.
Bond, C.A., D. Thilmany, and J. Kelling-Bond, 2008. Understanding consumer interest
in product and process based attributes for fresh produce. Agribusiness 24, 231-252.
Boxall, P. and Adamowicz, W., 2008. Understanding heterogeneous preferences in random
utility models: A latent class approach. Environmental and Resource Economics 23, 421-446.
Clonan, A., Holdsworth, M., Swift, J., and Wilson, P., 2010. UK consumers priorities
for sustainable food purchases. http://purl.umn.edu/91948
Darby, K., M. T. Batte, S. Ernst, and B. Roe, 2008. Decomposing Local: A conjoint
analysis of locally produced foods. American Journal of Agricultural Economics 90, 476-486.
Detoni, D., and G.T. Tonsor, 2009. The direct and indirect effects of ’locally grown’ on
consumers attitudes towards agri-food products. Agricultural and Resource Economics Review 38, 384-396.
Liljenstolpe, C. 2010. Demand for value-added pork in Sweden: A latent class model approach. Agribusiness 27, 129-146.
Lusk, J.L. and Briggeman, B., 2009. Food values. American Journal of Agricultural Economics 91, 1-13.
Lusk, J.L., Norwood, F.B. and Pruitt, J.R. 2006. Consumer demand for a ban on antiobiotic
drug use in pork production. American Journal of Agricultural Economics 88, 1015-1033.

79

Onozaka, Y. and D. Thilmany McFadden, 2010. Defining sustainable food market segments:
do motivations and values vary by shopping locale? American Journal of Agricultural Economics; doi:10.1093/ajae/aaq152
Onozaka, Y. and Thilmany-McFadden, D. 2011. Does local labeling complement or compete
with other sustainable labels? a conjoint analysis of direct and joint values for fresh produce
claims. American Journal of Agricultural Economics 93, 693-706.
Sagebiel, J. 2011. Comparing the latent class model with the random parameters logit a choice experiment analysis of highly heterogeneous electricity consumers in Hyderabad,
India. Working paper. julian.sagebiel@hu-berlin.de
Saunder, C. et al. 2010. Consumer attitudes towards sustainability attributes on food
labels. Presented paper at the New Zealand Agricultural and Resource Economics Society.
Shulz, L.L. and Tonsor, G.T., 2009. Cow-calf producer preferences for voluntary traceability systems. Journal of Agricultural Economics 61, 138-162.
Tonsor, G.T., Schroeder, T.C., Pennings, J.M. and Mintert, J. 2009. Consumer valuations
of beef steak food safety enhancement in Canada, Japan, Mexico, and the United States.
Canadian Journal of Agricultural Economics 57, 395-416.
Tonsor, G.T., and Shupp R. S. 2011. Cheap talk scripts and online choice experiments:
looking beyond the mean. American Journal of Agricultural Economics 93, 1015-1031.
Tonsor, G., and Shupp, R., 2009. Valuations of ’sustainably produced’ labels on beef, tomato,
and apple products. Agricultural and Resource Economics Review 38, 371-383.

80

Chapter 4: Willingness to Pay for Sustainably Labeled Steak and Apples: An
Application in Experimental Auctions
Introduction
A current trend in the US food system toward organic, sustainable, and local foods has
created a need for more research on the valuation of such credence attributes. Of these
trends, sales in the organic food sector have grown the fastest, encouraged by the strength
of its broad promotion and marketing. Most of the organic industry’s growth has occurred
in the years since the establishment of the USDA’s National Organic Program, in 2002,
creating a system of standards for certifying and labeling organic products. However, unlike USDA-certified organic products, foods designated as sustainably produced carry no
government-endorsed certification and bear no standardized label. The USDA National Institute of Food and Agriculture provides limited information on the purported sustainability
of different agricultural practices and warns that, “guidelines about what specific practices
meet long-term environmental, economic, and social goals and constitute sustainable agriculture are still under debate”.
“Economic sustainability” is a term used to identify strategies for using available resources
in a way that promotes efficiency and responsibility, with a goal of providing long-run benefits. Many interpret this to mean that a sustainable system supports and sustains local
economies by investing in community businesses, including but not limited to, agriculture.
Local foods have become more available following a dramatic increase in demand and succeeding a significant expansion of farmers markets across the country. The term “local” remains
undefined, with individual interpretations abound. However, interest in locally grown foods
continues to rise, matched, in response, by the inception of many state-sponsored marketing
campaigns. Understanding these growing trends and how they interact is therefore worthy
of further examination.
If sustainable agricultural systems possess similar environmental management practices

81

as organic and support local farms and the communities they feed, can the consumer differentiate between these competing claims? Comprehending consumer attitudes towards and
preferences between these three trends, as well as determining WTP for these production
attributes is of importance to producers and marketers in determining the effectiveness of
their labeling schemes and marketing programs.
The primary goal of this research is to determine consumer preferences and WTP for
foods labeled as sustainably produced when offered alongside their local and organic counterparts. The goal of this paper is complemented by three additional objectives: (1) compare
hypothetical and non-hypothetical WTP estimates to determine the extent of hypothetical
bias (2) use this combination of hypothetical and non-hypothetical choice data to understand
the relationship between sustainable, organic, and local label preferences, and (3) develop
a set of recommendations designed to guide industry-specific producers in the development
and implementation of a local marketing plan.
To accomplish these objectives, data gathered from a series of experimental auctions held
in the fall of 2010 will be utilized in several economic models of choice. The data obtained
in experimental auction procedures is believed to be more representative of actual behavior
as demand-revealing, non-hypothetical methods are used. The experimental auction participants form a subset of respondents to a nation-wide survey with a hypothetical choice
experiment component. Collectively, the use of experimental auctions augmenting hypothetical choice experiment data yield a more detailed and reliable understanding of consumer
behavior. The use of multiple sources of data will allow for a more robust understanding of
how consumers define and value the Òsustainably producedÓ label.

Research Methodology
Experimental Auctions
A growing literature exists on the design, implementation and evaluation of experimental
auctions with agricultural or food related applications. Due to the advantage of experimental
82

auction methods in isolating the effect of information provision, it has become an increasingly
popular avenue for investigating the impact of labeling schemes on consumer WTP for food
products. Applications of experimental auction procedures, used to evaluate information
provision or labels, range most recently across health and nutritional information (Hellyer
et al (2012) and Hobbs et al (2006)), country of origin labeling (Chern and Chang (2012)) ,
traceability and food safety (Lee et al (2011) and Nayga et al (2006)), genetic modification
and biotechnology (Colson and Huffman (2011), Corrigan and Rousu (2011), Lusk et al
(2004)), and other quality attributes such as Ògrass-fedÓ (Evans et al (2011) and Umberger
(2002)), and Òhormone-freeÓ (Alfnes and Rickertsen (2003)).
A small number of known studies have used experimental auctions in the valuation of
macro-level food system credence attributes. Rousu and Corrigan (2008) utilize experimental
auctions to compare several alternative fair trade labels in order to determine the welfare
loss from labels that inadequately inform consumers. More recently, Briggeman and Lusk
(2011) use a model of inequality aversion and altruism paired with experimental auctions
to investigate consumer WTP for organic foods and to better understand preferences for
fairness and equity in the food system. To the authors’ knowledge, our study is the first
to employ experimental auctions to investigate consumer preferences and WTP for food
products labeled with broadly interpreted sustainable production claims.

Data Collection
Experimental Design and Implementation
A series of five experiment sessions, held in October 2010 and attended by a total of seventysix participants, informs this study. The participants were recruited by the lead author from
three different grocery stores in the greater Lansing, Michigan area with the aid of a promotional poster. Grocery stores were targeted for recruitment for the purpose of attracting
primary household food shoppers. Demographic statistics on the experiment participants
are reported below.
83

All experiment sessions were held in a classroom space at Michigan State University’s
Pavilion for Agriculture and Livestock Education. At the time of their recruitment, each
participant signed up for a specific experiment session by providing a name and phone number. In return, each recruit was given information about the session, including: date, time,
compensation, directions to the facility, and a token green MSU pencil to incentivize followthrough. The recruited participant information sheet is available upon request. Participants
were called one day before the session with a reminder.
Upon arrival at the Pavilion, participants were given a unique ID for anonymity purposes
and then were instructed to begin an online survey at one of the mobile laboratory laptops.
The survey accompanying the experiment was identical to our nationwide survey, disseminated earlier in the summer of 2010, eliciting information about perceptions and preferences
towards sustainable farming practices and sustainably labeled foods. As each participant
completed the online survey, they signed the informed consent and received a copy of the
auction instructions. Auction instructions were read aloud, including an example bidding
strategy from a proposed auction scenario, and then participants were invited to view each
of the products that were to be auctioned.
Each experiment session consisted of a series of eight auctions corresponding to the eight
food products available; a 1 lb bag of apples labeled as sustainable, organic, local, or unlabeled and a 0.5 lb rib-eye steak labeled as sustainable, organic, local, or unlabeled. The
order of the auctions was randomized for each food product in every session. All products
were roughly identical and sourced from the same farm that was certified organic, local to
the participant population, and made farm-level claims of sustainable production. However,
the products were stripped of their original labeling and given only one of the following labels
for presentation to experiment participants: “Sustainable”, “Organic”, “Local” or “Typical”.
It should be noted that the new labels for experimental purposes made no false claims, but
allowed the researchers to isolate the effect of the Òsustainably producedÓ credence label
while controlling for other observed quality attributes such as color, consistency, or size.

84

This experiment utilized the second-price Vickrey auction with full bidding. This choice
of auction mechanism is popular in the literature for its relative performance compared to
alternatives. Additionally, this auction is easy to explain to participants and has been shown
to adequately measure “on margin” bidding (Lusk and Shogren 2007). No reference prices
were provided and no bids were revealed as auction rounds progressed. Participants were
randomly assigned to a group of 6-8 people, against whom they would be bidding in each
auction. The identity of group members was never revealed. Smaller groups were used to
retain the incentive compatible nature of the auction, while keeping participants engaged.
When the auctions were undertaken, each participant wrote down a WTP bid on a sheet of
paper provided to them, labeled with the product auctioned in that round. Bid sheets were
collected by one researcher and immediately recorded into an electronic spreadsheet. At the
conclusion of all eight auctions, one auction for each food product was randomly selected as
binding for purposes of payment.
Each participant was endowed with $25 to bid on the four 0.5 lb rib-eye steaks and $5 to
bid on the four 1 lb bags of apples, knowing that only one round, for each product, would
require binding payment. Participants were informed that this money was theirs and if they
chose not to bid on any product, or did not win the binding auction they would go home
with their full endowment. However, if they won the binding auction, they would go home
with the product and their endowment less the second highest bid. After the binding auction was announced for each product, participants were called up, one at a time, to receive
their endowments less any payments made for products won in the binding auctions, the
corresponding food products if applicable, and a $25 gift card to the grocery store where
they were recruited to compensate them for their time. Full disclosure of farm-of-origin
information was provided to participants that took home food products.
In total, 76 participants were recruited from Lansing area grocery stores for the described series of experimental auctions. A comparative summary of demographic statistics
can be found in the following table. In general, the participant population was more female,

85

younger, more frequently white, ate less beef in a given week, shopped at food co-ops and
farmers’ markets more often, were less likely to have dropped out of college before finishing
an associate’s degree, and had lower household income than the larger survey sample of 1002
households collected earlier in 2010.

86

Table 33: Summary Statistics of Survey and Experiment Samples
National
Experiment
Survey
Sample
Sample
Variable
Definition
Mean
or Mean
or
Frequency Frequency
Gender
1= Male
48 %
37 %
2 = Female
52%
63 %
Age
Average Age in Years
51.48
31.43
Adults
Number of Adults in Household
2.06
2.35
Children
Number of Children in Household
0.48
0.41
Meals
Number of Meals/ week with Apples
3.19
3.98
Meals
Number of Meals/ week with Beef
6.64
4.32
Shop
% of Total Shopping at Location
Grocery Store
81.97
56.53
Health Food Store
8.49
7.89
Food Co-op
2.05
9.07
Convenience Store
3.0
3.0
Farmers Market
3.72
16.16
Restaurant
4.38
5.46
Assistance 1 = On Food Assistance; 2 = Otherwise
1 = On Food Assistance
13.2 %
11.84 %
2 = Not on Food Assistance
86.8 %
88.16 %
Education Highest Level Completed
1 = Did not graduate from high school
2.59 %
3.95%
2 = Graduated from high school, no college
18.12 %
18.42 %
3 = Attended college, no degree earned
33.67 %
7.89 %
4 = Attended college, associates or trade degree 12.55 %
40.79 %
earned
5 = Attended college, Bachelors degree earned
22.52 %
18.42 %
6 = Graduate or advanced degree earned
10.56 %
10.53 %
Household Range of Pre-tax Income
Income
1 = Less than $20,000
19.12 %
51.32 %
2 = $20,000 - $39,999
28.88 %
14.47 %
3 = $40,000 - $59,999
21.51 %
14.47 %
4 = $60,000 - $79,999
11.95 %
11.84 %
5 = $80,000 - $99,999
7.57 %
1.32 %
6 = $100,000 - $119,000
4.98 %
3.95 %
9 = More than $120,000
5.99 %
2.63 %

87

Analysis
The main focus of the analysis presented here is to identify predictors of bid behavior by
estimating several regression models with the auction data. For all regression models the
dependent variable is the WTP bid. In keeping with the literature, this work employs a
double-censored tobit specification, as bids obtained in the auction are censored on the left
by zero and at the right by the endowment; $5 for apples and $25 for steaks. The regression
∗
of interest is specified as an unobserved latent variable, yi ,
∗
y i = xi β + i

(12)

where i N (0, σ 2 ) and xi is the (KX1) vector of exogenous and fully observed regressors
∗
(Cameron and Trivedi 2010). The observed variable yi is related to the latent variable yi
through the following rule:

yi =










L:









U:

y∗ ≤ L

y∗ : L < y∗ < U
y∗ ≥ U

where [L, U ] is the censored interval of observed values. The foregoing analysis uses maximum likelihood estimation under the assumptions that the regression error is homoskedastic
and normally distributed. ML estimates of (β, σ 2 ) solve the first-order conditions from maximization of the log likelihood based on the density function of censored observations. These
equations are nonlinear in parameters and therefore the solution uses an iterative algorithm.
A detailed summary of the data on the dependent variables, apple bids and steak bids
respectively, provides insight into potential problems with using MLE on the parameters of
the tobit model under the strict assumptions of homoskedasticity and normalcy of errors.
Examining apple bids further reveals only moderate skewness of 1.24 and slight kurtosis of
−0.14 after appropriate correction. Similar examination of steak bids yields a skewness of

88

1.14 and kurtosis of −0.52. These tests offer encouraging evidence of the consistency of the
MLE estimation, as the dependent variable appears to be approximately normal. Therefore,
we continue with results of the linear tobit without any transformation of the dependent
variable.

89

Results and Discussion
Pooled Sample Comparison
An identical analysis to the one utilized in Chapter 3 was employed on the choice experiment
survey data for the auction participant subsample. Before settling on the RPL model, the
multinomial logit and conditional logit were considered. The MNL was deemed a worse fit
by a log likelihood comparison. Overall, model fit of the CL was stronger and consistent
with earlier results. However, the MNL and CL models dropped the local label variable
because of collinearity with the organic label variable. Standard logit models do not take
into account any unobserved factors that may persist across alternatives for a given decision
maker. To take into account correlation in unobserved factors over alternatives, the RPL
allows the preference coefficients to be different for each respondent. The collinearity issues
in the MNL and CL models indicate that the choice of the local and organic products were
highly correlated for this much smaller and much less diverse sample, perhaps partially due
to inadequate sample size (n=40 for apples and n=36 for steak).
We first explored the difference between the choice experiment results from the national
survey sample and the auction participant sample by testing the null hypothesis of parameter
equality using a likelihood ratio test, as used in Chapter 3 to test the effect of the information
treatment. Comparing the likelihood function values using the likelihood ratio test resulted
with an associated two-tailed p value < 0.001 for both apple and steak surveys indicating
that the null hypothesis of parameter equality across participant samples is rejected at any
standard level of significance.
This result indicates that the auction participant sub-sample has significantly different
(hypothetical) preferences for apples and steak than the larger national survey sample. Moreover, the auction participant sample results yielded a positive coefficient on price in both
surveys. This can be interpreted to mean that higher prices led to a higher probability of
choice across alternatives. It may be that this small sample views price as an indicator of

90

quality moreso than the national sample. Also, the auction participant sample included a
large number of people recruited from a Health Foods grocery store that may be accustomed
to paying higher prices for health differentiated foods.
Additionally, the coefficient on the typical label is negative in the apple survey of experiment participants, suggesting that this sample prefers to opt out of the market over
purchasing the typical, unlabeled product. This result is supported by the fact that many
of the participants were recruited directly from a Health Foods specialty grocery store. By
their choice of shopping locale, those participants are essentially making the choice in reality
that was indicated by their hypothetical purchasing data; they are opting out of the regular
super market that carries the typical (unlabeled) products.

91

Table 34: Pooled Sample Comparison: RPL Estimates in Apple Survey
Sample
Variable
Pooled
Survey
Survey and
Only
Auction
Price
-1.90*
-2.13*
1.33**
CertP
-0.26*
-0.25*
-1.32*
CertS
0.00
-0.048
-0.264
LabelS
3.34*
3.55*
2.46*
LabelO
3.44*
3.63*
2.52*
LabelL
3.55*
3.82*
2.41*
LabelT
2.09*
2.39*
-1.68*
(*) and (**) significant at 0.01 and 0.05 levels, respectively

Table 35: Hypothesis Testing of Pooling Across Participant Sample: Apple Survey
Subsample Modeled
n
LL
p-Value
Pooled Sample
542
-9848.4917
Survey Only Sample
502
-8982.3905
Auction Participant Sample
40
-676.7408
H0 : Pooling across Participant Samples is Okay
< 0.001

92

Table 36: Pooled Sample Comparison: RPL Estimates in Beef Survey
Sample
Variable
Pooled
Survey
Survey and
Only
Auction
Price
-1.90*
-0.37*
1.29**
CertP
-0.26*
-0.32*
-5.99*
CertS
0.00
-0.16*
3.18*
LabelS
3.34*
2.88*
9.39*
LabelO
3.44*
2.88*
9.41*
LabelL
3.55*
2.90*
9.55*
LabelT
2.09*
1.88*
0.99**
(*) and (**) significant at 0.01 and 0.05 levels, respectively

Table 37: Hypothesis Testing of Pooling Across Participant Sample: Beef Survey
Subsample Modeled
n
LL
p-Value
Pooled Sample
536
-9743.3977
Survey Only Sample
500
-9027.1083
Auction Participant Sample
36
-643.1427
H0 : Pooling across Participant Samples is Okay
< 0.001

93

Raw Data Comparison
Eight participants self-identified as vegetarian and bid $0.00 in all four rounds of steak auctions and therefore were removed from the steak data analysis that follows. Vegetarians
made up 13.16% of this sample population. According to a 2006 study, employed by the
Vegetarian Resource Group, in a national poll only 6.7% of the research population identified
as vegetarian or never ate meat, making our sample twice as vegetarian as the purported
United States population. It may be hypothesized that the nature of the self-selected participant group may lead vegetarians, or other people with restrictive diets, to be more interested
in food related studies. Furthermore, two participants were deemed “unengaged bidders” as
a result of a $0.00 bid in all eight auctions but were not removed from the sample for analysis
because the authors are treating these observations as a choice to “opt out” of the market.
Table 38: WTP Estimates and Associated Price Premiums for Apple Attributes
Sample
Survey Only
Survey and Experiment
Welfare Measure
Survey Apple WTP Survey Apple WTP Auction Apple WTP
WTP LabelS
$1.67/lb
$1.85/lb
$1.49/lb
WTP LabelO
$1.71/lb
$1.89/lb
$1.55/lb
WTP LabelL
$1.79/lb
$1.81/lb
$1.39/lb
WTP LabelT
$1.12/lb
-$1.26/lb
$0.88/lb
Premium(LabelS)
$0.55/lb
$3.12/lb
$0.61/lb
Premium(LabelO)
$0.59/lb
$3.17/lb
$0.67/lb
Premium(LabelL)
$0.67/lb
$3.08/lb
$0.51/lb

Table 39: WTP Estimates and Associated Price Premiums for Steak Attributes
Sample
Survey Only
Survey and Experiment
Welfare Measure
Survey Steak WTP Survey Steak WTP Auction Steak WTP
WTP LabelS
$7.61/lb
$7.27/lb
$6.45/lb
WTP LabelO
$7.58/lb
$7.29/lb
$6.67/lb
WTP LabelL
$7.66/lb
$7.40/lb
$5.87/lb
WTP LabelT
$4.95/lb
$0.77/lb
$4.17/lb
Premium(LabelS)
$2.66/lb
$6.50/lb
$2.28/lb
Premium(LabelO)
$2.63/lb
$6.52/lb
$2.50/lb
Premium(LabelL)
$2.71/lb
$6.63/lb
$1.70/lb
From initial comparison of the table columns, several observations are made. First, compar94

ing the hypothetical WTP in the national survey sample with the experiment participant
sample yields a different preference ordering for each population on both products. While
the national survey participants preferred local apples over organic apples over sustainable
apples, the experiment participants preferred organic apples over sustainable apples over
local apples. Additionally, the mean hypothetical WTP estimates on all apple labels were
slightly higher for the experiment sample than for the national survey sample. The experiment sample results produced a willingness to accept measure on the typically unlabeled
apples, leading to much higher price premiums on the credence attribute labels relative to
the typical, unlabeled product.
While the national survey sample preferred local steak over sustainable steak over organic
steak, the experiment participants preferred local steak over organic steak over sustainable
steak. Additionally, the mean hypothetical WTP estimates on all steak labels were slightly
lower for the experiment sample than for the national survey sample. However, because the
hypothetical WTP for typical, unlabeled steak in the experiment participant population was
so low, the resulting price premiums on the three credence labeled products are much higher
for the experiment participants.
Now comparing the hypothetical WTP and the non-hypothetical WTP estimates for the
experiment participant sample reveals potential evidence of a degree of hypothetical bias as
hypothetical WTP estimates are higher for all credence labeled products than the associated mean auction bids. Additionally, the experiment participants bid a positive amount on
average for the typical unlabeled apples, even though the hypothetical results suggested a
willingness to accept measure instead. The auction bids induce the same preference ordering
as the hypothetical WTP data for apples overall.
The experimental auction bids on steak produce a different preference ordering than the
hypothetical data. Hypothetically, the experiment participants preferred local steak over
organic steak over sustainable steak. However, according to the non-hypothetical auction
bids, the participants preferred organic steak over sustainable steak over local steak. Addi-

95

tionally, the mean auction bid on the typical, unlabeled steak far exceeded the hypothetical
WTP estimate on the same product. Although the experiment participants indicated a general disinterest in the typical unlabeled products in the hypothetical survey scenarios, in the
non-hypothetical auction situation they consistently valued the typical, unlabeled steak more
than no steak at all. This may be partially due to the sensory experience associated with
the auction, that is, seeing the “typical unlabeled” steak (as it was identical in size, texture,
and color to all other steaks) persuaded participants that, it too, looked “good enough to
eat”.
Rounding out this analysis with the final pairwise comparison of the hypothetical WTP
from the national survey sample with the non-hypothetical WTP from the experiment participant sample yields interesting results. The hypothetical WTP estimates from the national
survey sample are much closer to the mean auction bids received in the experiment than the
hypothetical WTP estimates from the experiment participants themselves. This result is encouraging, because it suggests a much smaller magnitude of hypothetical bias if it is present,
when considering a sufficiently large sample size (n=500 versus n=40). This result could
be primarily context-driven. That is, the experiment participants that followed through
and came were recruited primarily from a Health Food specialty grocery store (there was
a much higher rate of no-shows for participants recruited from the generic grocery store),
they were informed that they would be participating in a food study on sustainability, and
they took the survey before participating in the auctions. It is possible that these factors
led experiment participants to answer the hypothetical shopping scenario questions more
dogmatically, so as to appear consistent with their choice of shopping locale, to promote
sustainable or organic agricultural products similar to the ones sold in their grocery store of
choice, or because they had not viewed the food products yet.
It is worth noting that the price premiums indicated by the mean auction bids were
highest for the organic label for both apples and steak, although we did not make any certification claims on any of the products. We cannot be certain whether or not the participants

96

automatically inferred that the organic label was backed by USDA certification, as that
perception information was not elicited at any point.
Hypothesis Testing of Equal WTP Across Samples
A similar bootstrapping technique was utilized to compare hypothetical and non-hypothetical
WTP estimates for the experiment participant sample as outlined in Chapter 3.
Table 40: Comparison of Hypothetical and Non-hypothetical WTP: Apple Survey
Hypothetical WTP
Non-hypothetical WTP
Average
95% interval
Average
95% interval
Label
WTP LabelS $1.85/lb
($1.12/lb, $2.61/lb)
$1.49/lb
($1.29/lb, $1.70/lb)
WTP LabelO $1.89/lb
($1.23/lb, $2.54/lb)
$1.55/lb
($1.32/lb, $1.78/lb)
WTP LabelL $1.81/lb
($1.01/lb, $2.56/lb)
$1.39/lb
($1.17/lb, $1.61/lb)
WTP LabelT -$1.26/lb
(-$2.10/lb, $0.87/lb)
$0.88/lb
($0.73/lb, $1.04/lb)

Table 41: Comparison of Hypothetical and Non-hypothetical WTP: Steak Survey
Hypothetical WTP
Non-hypothetical WTP
Average
95% interval
Average
95% interval
Label
WTP LabelS $7.27/lb
($5.86/lb, $8.70/lb)
$6.45/lb
($5.33/lb, $7.57/lb)
WTP LabelO $7.29/lb
($5.91/lb, $8.81/lb)
$6.67/lb
($5.52/lb, $7.81/lb)
WTP LabelL $7.40/lb
($5.98/lb, $8.96/lb)
$5.87/lb
($4.83/lb, $6.92/lb)
WTP LabelT $0.77/lb
($0.16/lb, $1.37/lb)
$4.17/lb
($3.25/lb, $5.08/lb)
The bootstrapped estimates reveal that the only 95% confidence intervals that do not
overlap between hypothetical and non-hypothetical estimates are the WTP measures on
typical, unlabeled steaks. This result suggests that the context in which the participants
took the survey and the cheap talk script had the effect of mitigating hypothetical bias in the
experiment participant sample for the three credence labels. That is, a comparison of 95%
confidence intervals leads us to the conclusion that the hypothetical WTP estimates were
not significantly different than the non-hypothetical WTP bids from the auction procedure,
with the exception of the downward bias on the hypothetical valuations of the typical,
unlabeled steak. The evidence of downward bias in the hypothetical typical steak purchases
97

is in contradiction with accepted theory. Usually, hypothetical bias exhibits a positive trend
towards hypothetical responses, that is, H0 : W T PHyp > W T PN on−hyp .

98

Tobit Model
The results of the base linear tobit model on non-hypothetical auction bids indicate that all
dummy variables corresponding to production label are statistically significant and positive.
The tobit MLE coefficients should be interpreted as the magnitude of the change in bid
corresponding to a one unit increase in the explanatory variable. Therefore, the coefficient
on each of the production labels can be loosely interpreted as the potential price premium
captured by that respective labeling scheme. The regression coefficients are consistent with
the summary of mean bids, yielding a preference ordering of Organic > Sustainable > Local
for both food products. In the apple auction, only age and number of children were statistically significant demographic explanatory variables. Age had a very small, yet positive
relationship with bids. Number of children has a slightly larger, negative, relationship with
bids. In the steak auction all included demographic variables were statistically significant
with the exception of age. Women, generally, bid higher than men, income and education
both had moderate positive effects on bids and again the number of children in a household
had a negative relationship with auction bids.

99

Table 42: Tobit Regression Results for Apple Bids
Explanatory Variable MLE Coefficient P-value
Gender
-0.12
-0.30
Age
0.01 *
0.047
Income
-0.02
0.542
Education
-0.01
0.830
Children
-0.16**
0.01
Label S
0.67**
0.00
Label O
0.73**
0.00
Label L
0.55**
0.01

Table 43: Tobit Regression Results for Steak Bids
Explanatory Variable MLE Coefficient P-value
Gender
-2.98**
0.00
Age
0.04
0.22
Income
0.46*
0.04
Education
0.56*
0.03
Children
-1.18**
0.00
Label S
2.58**
0.00
Label O
2.82**
0.00
Label L
2.07**
0.02

100

Conclusions
This paper discusses the results from a series of experimental food auctions, used to examine consumer responses to a variety of production labels on apples and ribeye steaks. This
work also examines the extent of hypothetical bias in estimates by comparing hypothetically elicited WTP with non-hypothetical auction bids for the same population. Overall,
this study consistently found that food labeled as sustainable, organic, or local influenced
participant willingness to pay.
It was confirmed using a LR test that the experiment participant sample preferences in
the hypothetical shopping scenarios contained in the survey were significantly different than
the preferences of the national survey sample. This could be attributed to a number of
factors, including difference in demographic make-up and context in which the survey was
taken. Additionally, sign reversals on the price coefficient and typical label coefficient were
observed for the experiment participant sample on hypothetical preferences.
Pairwise comparisons of the hypothetical WTP from the national survey sample, the
hypothetical WTP from the experiment participant sample, and the non-hypothetical WTP
bids from the experiment participants yielded interesting results. Across samples the preference ordering on credence labels was different. Specifically, the national survey sample
preferred local apples over organic apples over sustainable apples, while the experiment
participant sample preferred organic over sustainable over local in both hypothetical and
non-hypothetical treatments for apples. The national survey sample preferred local steaks
over sustainable steaks over organic steaks, while the experiment participant sample preferred local over organic over sustainable in the hypothetical treatment and organic over
sustainable over local in the non-hypothetical treatment for steaks.
A comparison of the hypothetical versus non-hypothetical WTP estimates for the experiment participant sample gave insufficient evidence of positive hypothetical bias, although
all hypothetical WTP estimates were higher than the mean WTP bids from the auction on
both products. Additionally, the experiment participant sample valued the typical, unla101

beled food products much less in the hypothetical shopping scenarios than in the auctions.
The hypothetical WTP from the national survey sample was much closer to the auction bids
than the hypothetical WTP from the experiment participants themselves. This result could
be indicative of the much smaller sample size and is encouraging evidence that hypothetical
bias dissipates slightly for larger, more demographically diverse samples.
The price premiums indicated by mean auction bids were highest for the organic label on
both apples and steak, although no certification claims were made on any auctioned products. We cannot be sure if the experiment participants inferred USDA certification from the
organic label or not. If the organic label was perceived to be endorsed by the USDA, our
results suggest that the USDA Organic certification translates a significantly higher value
to consumers. While, sustainable production claims may not yield any more purchase value
than organic certification or local production, sustainable labels appear to successfully differentiate food products from their conventional or unlabeled alternatives. This work provides
evidence that third-party sustainable certification has potential profitability for producers,
especially if USDA Organic certification is prohibitively expensive or otherwise intangible.
Furthermore, these results suggest that consumers do not distinguish between locally grown
and sustainably grown food products, inferring that the sustainable food label may not be
effectively communicating value beyond traditional Òeconomic sustainabilityÓ and the support of local economies. Thus, measuring other environmental and social consequences of
sustainable production practices may be worth investigating further for their marketing potential.
Finally, our paper supports the findings in the field literature that sustainable, organic,
and local food labels produce positive price premiums on food products. However, due to
the makeup of our participant population, we do not suggest taking the price premiums
estimated here to be representative of the general United States, or Lansing, MI population.
Instead, we hope this study has produced insight into the relationship between these three
credence-labeling schemes and relative consumer preferences in the market. Additionally, we

102

consider this a contribution to the growing literature on experimental auction applications
and methods.

103

APPENDIX

104

Survey Instrument

This is a survey designed to obtain information from consumers regarding food consumption
habits and related issues. Your participation in this survey is entirely voluntary and your
responses will be kept in strict confidence. The data gathered from this survey is needed for
analysis in the dissertation work of an MSU graduate student. If you have any questions,
comments, or concerns regarding this survey, please contact Dr. Robert Shupp by email
(shupprob@anr.msu.edu) or by phone (517-432-2754).

105

Notes for Coders: Please ensure the following are incorporated into the online survey.
Thank you.
1. Please randomize the order of statements in question number 13.
2. Please randomize the order of the issues in question number 14.
3. Each participant should randomly receive either Block A or Block B of questions
numbered 15 through 20. Please track which Block was received by each respondent.
4. Please randomize the order of issues in questions 15 - 20 for each Block.
5. Please randomize the order of the issues in question number 21.
6. Please randomize the value of X in question 24 between 0 and 100. Please replace
equation with monetary value once X has been chosen.
7. Please randomize the order of the issues in question number 25.
8. Each participant should randomly receive either Block A or Block B of the questions
numbered 26 through 34. Please track which was received by each respondent.
9. Please randomize the order of issues in questions 26-34 for each Block.
10. Moreover, for all questions with randomized designs please create data variables tracking the order of presentation for each respondent.

106

1. I am:
2 Male
2 Female
2. I am

years old. (Please fill in the blank.)

3. I live in the

zip code area (Please fill in the blank.) and my annual pre-tax,

household income is:
2 Less than $20,000
2 $20,000 - $39,999
2 $40,000 - $59,999
2 $60,000 - $79,999
2 $80,000 - $99,999
2 $100,000 - $119,999
2 $120,000 - $139,999
2 $140,000 - $159,999
2 $160,000 or more

107

4. The best description of my educational background is:
2 Did not graduate from high school
2 Graduated from high school, Did not attend college
2 Attended College, No Degree earned
2 Attended College, Associates or Trade Degree earned
2 Attended College, BachelorâĂŹs (B.S. or B.A.) Degree earned
2 Graduate or Advanced Degree (M.S., Ph.D., M.D., Law School)
2 Other (Please explain.)
5. There are

adults and

children living in my household. (Please fill in the two

blanks.)
6. What best describes your race?
2 White or Caucasian
2 Black or African-American
2 Asian or Pacific Islander
2 Mexican or Latino
2 American Indian
2 Other (Please describe.)
7. Are you the primary food/grocery shopper in your household? ( e.g. the individual
most often purchasing food) ?
2 Yes
2 No

108

8. Do you currently receive food assistance?
2 Yes
2 No
9. If you receive food assistance, please check off all forms of food assistance you receive:
2 Food Stamps
2 WIC
2 Project Fresh
2 Food Bank
2 Soup Kitchen
2 Other (Please specify.)
10. Most households consume on average 21 meals in a typical week. How many of these
21 meals consumed by your household normally include:
a. Apples:
b. Other fruits, besides apples:

109

11. Consumers purchase food from many sources. Please allocate the proportion of your
householdâĂŹs total food purchases across each of the following (Sum to 100%):
a. Supermarket Retailers (e.g. Kroger, Meijer, Wal-Mart)

%

b. Targeted Retailers (e.g. Whole Foods, Foods for Living, Better Health)
%

c. Food Cooperative (e.g. ELFCO)

d. Convenience Stores (e.g. Quality Dairy, 7-Eleven)
e. Farmers Markets/ CSAs
f. Butcher

%

%

%

%

g. Other (Please describe.)
12. Please rate the following statement in terms of your agreement:
“I rarely think about the sustainability of production practices when making food
purchases”.

Strongly Disagree
1

2

3

4

110

5

6

Strongly Agree
7

13. Please rate the following statements in terms of your agreement (Please circle only
one in each row.): “I believe that”

111

Strongly Disagree
farmers face lower levels of productivity
when using sustainable food production practices.
farmers could sustainably produce
food
without much additional monetary expense.
farmers currently participate in
sound
sustainable food production practices.
organic products are from farms
currently participating in sound
sustainable food production practices.
locally grown products are from
farms
currently participating in sound
sustainable food production practices.
smaller farms are more likely to
currently participate in sound
sustainable food production practices.
corporate farms are more likely to
currently participate in sound
sustainable food production practices.
all products labeled sustainably
produced
are from farms currently participating
in sound sustainable food production practices.

Strongly Agree

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

1

2

3

4

5

6

7

112

14. Please rank the following societal issues in order of importance to you (1 being most
important and 8 being the least important):
a. Human Poverty
b. U.S. Health Care System
c. Food Safety
d. Sustainable Agriculture
e. The Environment
f. Financial Well-Being of U.S. Farmers
g. Consumer Food Prices
h. Animal Well-Being and Welfare

113

In the next section you will be asked to choose which aspects of sustainable
apple farming are most and least important to you. Please take the time to read
the following definitions as related to sustainable production practices to better
help you in your responses.

Ground Cover and Area Management Practices are Employed: Adjacent areas are planted
with hedgerows, windbreaks, or other low-maintenance plantings to encourage specific beneficial organisms. Within tree rows, ground cover or mulch are selected and maintained to
improve soil microbial activity, organic matter levels and nutrient cycling.

Fertilizer and Nutrient Materials are used minimally: Soil quality, including organic matter
content, is established at planting and maintained at an optimum level to minimize commercial fertilizer needs.

Little to No Chemical Pesticides are Used for Pest Management: Chemical pesticides are
not used. Alternative strategies are employed, including biopesticides, mating disruption,
trap out and/or augmentation with beneficial organisms.

Little to No Chemical Herbicides are Used for Weed Management: Soil quality and ground
cover in the orchard and adjoining areas are planned and managed to prevent weeds and
weed seed immigration into the orchard. Cultural, mechanical or biological methods are
used to control weeds.

Pollinator Management is Employed: Bees are not placed in the orchard until blossoms
are open. Pesticides hazardous to bees are not used, or if needed in an emergency, are applied such that they are not hazardous to bees.

114

Other Pests are controlled using preventative measures, and habitat controls: Habitat is
modified around orchards to reduce nesting and perching sites for pest birds.

115

BLOCK A
15. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Little to No Chemical Pesticides are Used for Pest
Management
Other Pests are controlled using preventative measures, and habitat controls
Farm Size is smaller than average and Corporate Involvement is limited
Michigan farmers are financially stable

16. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Little to No Chemical Herbicides are Used for Weed
Management
Pollinator Management is Employed
Farm Size is small and Corporate Involvement is limited
Production, distribution, and sale is done locally
Michigan farmers are financially stable

116

17. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Fertilizer and Nutrient Materials are used minimally
Little to No Chemical Pesticides are Used for Pest
Management
Little to No Chemical Herbicides are Used for Weed
Management
Pollinator Management is Employed
Other Pests are controlled using preventative measures, and habitat controls
Farm Size is small and Corporate Involvement is limited
Production, distribution, and sale is done locally
Consumer food prices are affordable
Michigan farmers are financially stable

117

18. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Fertilizer and Nutrient Materials are used minimally
Little to No Chemical Pesticides are Used for Pest
Management
Little to No Chemical Herbicides are Used for Weed
Management
Other Pests are controlled using preventative measures, and habitat controls
Farm Size is small and Corporate Involvement is limited
Production, distribution, and sale is done locally

118

19. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Fertilizer and Nutrient Materials are used minimally
Pollinator Management is Employed
Other Pests are controlled using preventative measures, and habitat controls
Farm Size is small and Corporate Involvement is limited
Consumer food prices are affordable
Michigan farmers are financially stable

20. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Fertilizer and Nutrient Materials are used minimally
Farm Size is small and Corporate Involvement is limited
Production, distribution, and sale is done locally
Consumer food prices are affordable

119

BLOCK B
15. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Fertilizer and Nutrient Materials are used minimally
Little to No Chemical Herbicides are Used for Weed
Management
Pollinator Management is Employed
Other Pests are controlled using preventative measures, and habitat controls

16. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Little to No Chemical Herbicides are Used for Weed
Management
Other Pests are controlled using preventative measures, and habitat controls
Production, distribution, and sale is done locally
Consumer food prices are affordable
Michigan farmers are financially stable

120

17. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Little to No Chemical Pesticides are Used for Pest
Management
Pollinator Management is Employed
Other Pests are controlled using preventative measures, and habitat controls
Production, distribution, and sale is done locally
Consumer food prices are affordable

121

18. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Fertilizer and Nutrient Materials are used minimally
Little to No Chemical Pesticides are Used for Pest
Management
Pollinator Management is Employed
Production, distribution, and sale are done locally
Michigan farmers are financially stable

122

19. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Ground Cover and Area Management Practices are
Employed
Fertilizer and Nutrient Materials are used minimally
Little to No Chemical Pesticides are Used for Pest
Management
Little to No Chemical Herbicides are Used for Weed
Management
Consumer food prices are affordable
Michigan farmers are financially stable

20. Which one of the following aspects of apple farming do you believe are the most and
least important in a sustainable apple production system? Please check only one in
each column.
Least Important

Most Important
Little to No Chemical Pesticides are Used for Pest
Management
Little to No Chemical Herbicides are Used for Weed
Management
Pollinator Management is Employed
Farm Size is small and Corporate Involvement is limited
Consumer food prices are affordable

123

21. To the best of your knowledge, for each of the apple production practices indicated,
production involves which of the following restrictions? Check all that apply.
Sustainable

Organic

Local

Typical

Require ground cover and area management
Limited use of fertilizer and nutrient materials
Prohibited use of herbicides for crops
Prohibited use of insecticides for crops
Prohibited use of genetically modified organisms
Limited corporate involvement
Limited farm acreage

22. Which one of the following best describes the proximity from your home you consider
âĂIJlocally grownâĂİ apples to originate from?
a. within 10 miles
b. within 50 miles
c. within 100 miles
d. within 500 miles
e. within Michigan
f. Other (Please describe.)
23. Would you be willing to pay a premium for apples labeled as âĂIJsustainably produced?âĂİ
2 Yes
2 No
24. If Yes, Would you buy apples labeled as âĂIJsustainably producedâĂİ if it cost $1.49
+ 1.49*X% ?
2 Yes
2 No
124

24b. If No, Would you buy apples labeled as âĂIJsustainably producedâĂIJ if it cost $1.49
- 1.49*X%?
2 Yes
2 No

125

25. Please rank the emphasis you believe farmers producing âĂIJsustainably produced
applesâĂİ place on the following issues/concerns.
(1 being most emphasis and 6 being the least emphasis):
a. Profitability of their farm
b. Food prices faced by consumers
c. Quantity of their farmâĂŹs production
d. Impact their farm has on the environment
e. Food safety of products produced on their farm
f. Quality of apples produced

126

Please consider the following information [provided by the United States Department of Agriculture]:

The word âĂIJsustain,âĂİ from the Latin sustinere (sus-, from below and tenere, to hold),
to keep in existence or maintain, implies long-term support or permanence. As it pertains to
agriculture, sustainable describes farming systems that are capable of maintaining their productivity and usefulness to society indefinitely. Such systems must be resource-conserving,
socially supportive, commercially competitive, and environmentally sound. Sustainable agriculture was addressed by Congress in the 1990 Farm Bill and the Food, Agriculture, Conservation, and Trade Act of 1990 (FACTA). Under that law, the term sustainable agriculture
means an integrated system of plant and animal production practices having a site-specific
application that will, over the long term:
1. Satisfy human food and fiber needs
2. Enhance environmental quality and the natural resource base upon which the agricultural economy depends
3. Make the most efficient use of nonrenewable resources and on-farm resources and integrate, where appropriate, natural biological cycles and controls
4. Sustain the economic viability of farm operations
5. Enhance the quality of life for farmers and society as a whole
Consumers are increasingly demanding on the ethical dimension of food quality. This
relates to the process of production and trade and its broad impacts on society and the
environment. It includes a wide range of social, environmental or cultural issues such as
the treatment of workers, a fair return to producers, environmental impacts and animal
welfare. Guidelines about what specific practices meet long-term environmental, economic
127

and social goals and constitute sustainable agriculture is still under debate. However, a
handful of groups have attempted to develop standards and/or provide certification services
based upon their own guidelines.

128

In the next section you will be presented with multiple different alternative apples that
could be available for purchase in a retail store where you typically shop. Besides the attributes listed below, each product possesses the same characteristics (e.g., similar color and
freshness) and is produced in the U.S. Prices vary for each product and are all in $/lb. units.
Please consider the following information to help you interpret alternative products.

Label: The display that contains the apples for your purchase may be labeled as follows:
• Sustainable: These apples were produced using sustainable practices.
• Organic: These apples were produced using organic practices.
• Local: These apples were produced for distribution and sale locally.
• Typical: These apples are not labeled to suggest they were produced using any of the
criteria listed above.
Certification: The typical product has no certification label. Each labeled product can be
certified in one of three ways:
• USDA: The processes used and all claims made by the product label have been verified
by the USDA.
• Independent Third Party: The processes used and all claims made by the product label
have been verified by a third party unrelated to the farm of origin or retailer.
• Self: The processes used and all claims made by the product label have been verified
by the farmer producing the food.
The experience from previous similar surveys is that people often state a higher willingness to pay than what one is actually willing to pay for the good. For instance, a recent
study asked people whether they would purchase a new food product similar to the one you
are about to be asked about. This purchase was hypothetical (as it will be for you) in that
129

no one actually had to pay money when they indicated a willingness to purchase. In the
study, 80% of people said they would buy the new product, but when a grocery store actually
stocked the product, only 43% of people actually bought the new product when they had to
pay for it. This difference (43% vs. 80%) is what we refer to as hypothetical bias.

Accordingly, it is important that you make each of your upcoming selections like you would
if you were actually facing these exact choices at a store; noting that buying a product means
that you would have less money available for other purchases.

130

BLOCK A
26. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Private 3rd Party
$0.99 /lb
2

Self
$0.99 /lb
2

USDA
$1.99 /lb
2

$1.99 /lb
2

I choose not to
purchase any of
these options.

2

27. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Private 3rd Party
$1.99 /lb
2

USDA
$1.49 /lb
2

USDA
$0.99 /lb
2

$0.99 /lb
2

I choose not to
purchase any of
these options.

2

28. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

3rd Party
$1.49 /lb
2

Self
$1.99 /lb
2

3rd Party
$0.99 /lb
2

$1.49 /lb
2

131

I choose not to purchase any of these options.

2

29. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Private 3rd Party
$1.49 /lb
2

Private 3rd Party
$1.49 /lb
2

Self
$1.99 /lb
2

$0.99 /lb
2

I choose
not
to
purchase
any
of
these
options.

2

30. Which one of the following apple displays listed below would you choose to purchase
from?
Label
Certification
Price

Sustainable

Organic

Local

Typical

Self
$0.99 /lb
2

Self
$1.49 /lb
2

Self
$0.99 /lb
2

$1.49
2

I choose not to purchase
any of these options.

2

31. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

USDA
$1.99 /lb
2

Self
$1.49 /lb
2

Private 3rd Party
$1.99 /lb
2

$1.99 /lb
2

132

I choose not to
purchase any of
these options.

2

32. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

USDA
$1.99 /lb
2

Private 3rd Party
$1.99 /lb
2

Self
$0.99 /lb
2

$1.99 /lb
2

I choose not to
purchase any of
these options.

2

33. Which one of the following apple displays listed below would you choose to purchase
from?
Label
Certification
Price

Sustainable

Organic

Local

Typical

I choose not to purchase
any of these options.

USDA
$1.49 /lb
2

Self
$0.99 /lb
2

Self
$1.49 /lb
2

$0.99 /lb
2

2

34. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

USDA
$0.99 /lb
2

USDA
$0.99 /lb
2

Private 3rd Party
$0.99 /lb
2

$0.99 /lb
2

133

I choose not to
purchase any of
these options.

2

BLOCK B
26. Which one of the following apple displays listed below would you choose to purchase
from?
Label
Certification
Price

Sustainable

Organic

Local

Typical

Self
$1.99 /lb
2

Self
$1.99 /lb
2

USDA
$1.49 /lb
2

$0.99 /lb
2

I choose not to purchase
any of these options.

2

27. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Private 3rd Party
$0.99 /lb
2

USDA
$1.99 /lb
2

Self
$1.49 /lb
2

$1.99 /lb
2

I choose not to
purchase any of
these options.

2

28. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

3rd Party
$1.99 /lb
2

3rd Party
$0.99 /lb
2

3rd Party
$1.49 /lb
2

$1.49 /lb
2

134

I choose not to purchase any of these options.

2

29. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

USDA
$0.99 /lb
2

Private 3rd Party
$1.49 /lb
2

USDA
$1.49 /lb
2

$1.49 /lb
2

I choose not to
purchase any of
these options.

2

30. Which one of the following apple displays listed below would you choose to purchase
from?
Label
Certification
Price

Sustainable

Organic

Local

Typical

I choose not to purchase
any of these options.

USDA
$1.49 /lb
2

USDA
$1.99 /lb
2

USDA
$1.99 /lb
2

$1.49 /lb
2

2

31. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Self
$1.49 /lb
2

Private 3rd Party
$0.99 /lb
2

USDA
$0.99 /lb
2

$1.99 /lb
2

135

I choose not to
purchase any of
these options.

2

32. Which one of the following apple displays listed below would you choose to purchase
from?
Label
Certification
Price

Sustainable

Organic

Local

Typical

I choose not to purchase
any of these options.

Self
$1.99 /lb
2

USDA
$0.99 /lb
2

Self
$1.99 /lb
2

$1.49 /lb
2

2

33. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Self
$0.99 /lb
2

3rd Party
$1.99 /lb
2

3rd Party
$1.99 /lb
2

$0.99 /lb
2

I choose not to purchase any of these options.

2

34. Which one of the following apple displays listed below would you choose to purchase
from?
Label

Certification
Price

Sustainable

Organic

Local

Typical

Self
$1.49 /lb
2

USDA
$1.49 /lb
2

Private 3rd Party
$1.49 /lb
2

$1.99 /lb
2

I choose not to
purchase any of
these options.

2

Thank you for your time in completing this survey. Your input will strengthen
our research and help us obtain more accurate conclusions. If you wish to add
any comments please feel free to do so here:

136

REFERENCES

137

REFERENCES
Alfnes, Frode, and Kyrre Rickertsen. “European Consumers’ Willingness to Pay for U.S.
Beef in Experimental Auction Markets.” American Journal of Agricultural Economics 85.2
(2003): 396-405.
Briggeman, Brian C., and Jayson L. Lusk. “Preferences for Fairness and Equity in the
Food System.” European Review of Agricultural Economics 38.1 (2011): 1-29.
Cameron, A. C. and P.K. Trivedi. Microeconometrics Using Stata. College Station: Stata
Press, 2010.
Chern, Wen S., and Chun-Yu Chang. “Benefit Evaluation of the Country of Origin Labeling in Taiwan: Results from an Auction Experiment.” Food Policy 37.5 (2012): 511-9.
PAIS International.
Clonan, A., M. Holdworth, J. Swift, and P. Wilson, 2010. UK Consumers Priorities for
Sustainable Food Purchases. Conference Proceedings. http://purl.umn.edu/91948
Colson, G. and W. E. Huffman. ÒConsumersÕ Willingness to Pay for Genetically Modified Foods with Product-Enhancing Nutritional Attributes.Ó (2011) Amer. J. Agr. Econ.
93(2): 358Ð363.
Corrigan, J. R., and M. C. Rousu. 2011. ÒAre Experimental Auctions Demand Revealing when Values are Affiliated?Ó American Journal of Agricultural Economics 93, January.
Evans, Jason R., et al. “Determining Consumer Perceptions of and Willingness to Pay
for Appalachian Grass-Fed Beef: An Experimental Economics Approach.” Agricultural and
Resource Economics Review 40.2 (2011): 233-50.
Hellyer, Nicole Elizabeth; Fraser, Iain; Haddock-Fraser, Janet. ÒFood Choice, Health Information and Functional Ingredients: An Experimental Auction Employing BreadÓ Food
Policy 37. 3 (Jun 2012): 232-245.
Hobbs, Jill E., Kim Sanderson, and Morteza Haghiri. “Evaluating Willingness-to-Pay for
Bison Attributes: An Experimental Auction Approach.” Canadian Journal of Agricultural
Economics 54.2 (2006): 269-87.
Lee, Ji Yong, et al. “Valuing Traceability of Imported Beef in Korea: An Experimental Auction Approach.” Australian Journal of Agricultural and Resource Economics 55.3
(2011): 360-73.
Lusk, J. et.al. ÒEffect of Information About Benefits of Biotechnology on Consumer Acceptance of Genetically Modified Food: Evidence from Experimental Auctions in the United
138

Stated, England, and France.Ó European Review of Agricultural Economics Vol 31 (2) (2004)
pp. 179-204.
Lusk, J. et.al. ÒConsumer Welfare Effects of Introducing and Labeling Genetically Modified
Food.Ó Economics Letters 88 (2005) 382Ð388.
Lusk, J. and J. F. Shogren, 2007. Experimental Auctions: Methods and Applications in
Economic and Marketing Research. Cambridge University Press.
Nayga, R. M., R. Woodward and W. Aiew. ÒWillingness to Pay for Reduced Risk of
Foodborne Illness: A Nonhypothetical Field Experiment.Ó Canadian Journal of Agricultural Economics 54 (2006) 461Ð475.
Onken, Kathryn A., John C. Bernard, and Pesek,John D.„Jr. “Comparing Willingness to
Pay for Organic, Natural, Locally Grown, and State Marketing Program Promoted Foods in
the Mid-Atlantic Region.” Agricultural and Resource Economics Review 40.1 (2011): 33-47.
Onozaka, Yuko, Gretchen Nurse, and Dawn Thilmany McFadden. “Defining Sustainable
Food Market Segments: Do Motivations and Values Vary by Shopping Locale?” American
Journal of Agricultural Economics 93.2 (2011): 583-9.
Onozaka, Yuko, and Dawn Thilmany McFadden. “Does Local Labeling Complement Or
Compete with Other Sustainable Labels? A Conjoint Analysis of Direct and Joint Values
for Fresh Produce Claims.” American Journal of Agricultural Economics 93.3 (2011): 693706.
Rousu, Matthew C., and Jay R. Corrigan. “Estimating the Welfare Loss to Consumers
when Food Labels do Not Adequately Inform: An Application to Fair Trade Certification.”
Journal of Agricultural and Food Industrial Organization 6.1 (2008)
Saunders, C. et.al., 2010. Consumer Attitudes Towards Sustainability Attributes on Food
Labels. Conference Proceedings. http://ageconsearch.umn.edu/handle/96944
Umberger, Wendy J. “U.S. Consumer Preference and Willingness-to-Pay for Domestic CornFed Beef Versus International Grass-Fed Beef Measured through an Experimental Auction.”
Agribusiness 18.4 (2002): 491-504.

139

Chapter 5: Summary
Surprisingly little economic research has focused on sustainability in the context of agricultural production practices before this work. Generally, the literature on other value-added
food attributes, such as organic, local, natural, or GMO-free have found that credence attributes have a positive impact on consumer valuations and lead to higher WTP, despite
the inability to directly identify the attribute either before purchase or after consumption.
“Sustainably produced” is an attribute with no absolute definition, leaving sustainable food
labels much more open to consumer perceptions and inferences that go beyond the information provided on packaging or marketing labels. Understanding the perceptions and
preferences consumers hold regarding what a “sustainably produced” labeling scheme implies is crucial to determining if the product will succeed in the market.
Best-worst analysis was applied in this research to investigate the degree of importance
consumers give to ten sustainable farming production attributes and in particular was used
to determine behavioral differences across clustered subgroups of the population sample.
While consumers are generally familiar with organic standards outlining land-use and environmental impact variables of production, this study indicates that environmental indicators
of sustainability are less important to consumers than economic dimensions. Based on our
initial results, size, scale, and geographic scope capture the attributes of sustainability that
are most important to consumers. Therefore, differentiating food claims on the level of locality provides a marketing avenue worth exploring further, as supported by the growing
literature on the local foods movement.
Our study also provides supporting evidence that the term “sustainable” could be causing more confusion than it is adding value in credence labeling schemes. The latent class
assessments on both beef and apple consumer samples indicated that 50 - 55% of the evaluated population is highly price-driven or significantly indifferent to varied sustainability
attributes. However, these results could also imply that consumers perceive sustainability as
a bundle of attributes, and thus ranked them equally important in this setting. Given the
140

associated heterogeneous economic welfare impacts that would come as a result of acrossthe-board market adjustments, such as government endorsed sustainability labeling, or bans
on select farming practices, this is an important result.
Our analysis of hypothetical purchase choices demonstrates that positive price premiums can be captured by sustainably produced labeling claims, relative to similar unlabeled
and conventionally produced food products. However, the price premiums calculated in our
model reveal that there is a comparable tradeoff in quality associated between local, organic
and sustainably labeled food products. Furthermore, detailed information about sustainable certification guidelines had no significant impact on choosing the sustainably labeled
products. Thus, we conclude that, based on our results, consumer demand for sustainably
produced food may not be distinctly differentiable from its local and organic counterparts.
Overall, these findings suggest that profitable marketing opportunities may exist for firms
interested in selling sustainably produced food products, however there needs to be considerable effort put into leveling information asymmetries about product quality if sustainable
label claims are to be differentiated from the local food movement or more recognizable organic principles.
The experimental auction data provided consistent evidence that food labeled as sustainable, organic, or local influenced participant willingness to pay. However, it was confirmed
using a LR test that the experiment participant sample preferences in the hypothetical shopping scenarios contained in the survey were significantly different than the preferences of the
national survey sample. This could be attributed to a number of factors, including difference
in demographic make-up and context in which the survey was taken.
A comparison of the hypothetical versus non-hypothetical WTP estimates for the experiment participant sample, using bootstrapping methods, gave insufficient evidence of positive
hypothetical bias, although all hypothetical WTP estimates were higher than the mean WTP
bids from the auction on both products. Additionally, the experiment participant sample
valued the typical, unlabeled food products much less in the hypothetical shopping scenarios

141

than in the auctions. The hypothetical WTP from the national survey sample was much
closer to the auction bids than the hypothetical WTP from the experiment participants
themselves. This result could be indicative of the much smaller sample size and is encouraging evidence that hypothetical bias dissipates slightly for larger, more demographically
diverse samples.
While, sustainable production claims may not yield any more purchase value than organic
certification or local production, sustainable labels appear to successfully differentiate food
products from their conventional or unlabeled alternatives. This work provides evidence
that third-party sustainable certification has potential profitability for producers, especially
if USDA Organic certification is prohibitively expensive or otherwise intangible. Furthermore, these results suggest that consumers do not distinguish between locally grown and
sustainably grown food products, inferring that the sustainable food label may not be effectively communicating value beyond traditional Òeconomic sustainabilityÓ and the support
of local economies. Thus, measuring other environmental and social consequences of sustainable production practices may be worth investigating further for their marketing potential.
It is hard to conclude whether sustainable food labels will or will not follow the same
trajectory as organic. However, it is encouraging to see more emphasis on the differences
between organic and sustainable in the food system rhetoric, at both academic and popular
levels. Inevitably, more information is becoming available for consumers to distinguish between food labels. Therefore, a clear extension of this work is needed in comparing consumer
preferences for sustainably labeled foods with other bundled labels, such as Organic + Local.
Furthermore, a study should be conducted on the effect of using “sustainable” attributes as
a marketing tool, in mediums other than labels, on consumer valuation.

142