Recipient Organization
UNIVERSITY OF VERMONT
(N/A)
BURLINGTON,VT 05405
Performing Department
Community Development and Applied Economics
Non Technical Summary
While some may have thought the GM labeling issue waned in the early 2000s, a simple Google search on April 22, 2017 using the terms "food system policy GMO labeling" returned 1,880,000 items. At the time of the previous Hatch project, the same search returned 280,000 items. The first ten pages included media reports, links organizations promoting both pro and anti gm labeling viewpoints, Land Grant University websites, and scholarly journal articles. The issue of labeling continues to receive attention.Vermont is the only U.S. state with mandatory labeling experience. The Vermont law was superseded by Federal legislation that will put into place a federal standard for labelling foods that have been made using genetic engineering (GM). This law will not be fully operational for up to five years from its passage (July, 2016). Vermont's experience is important and can inform the process of design and implementation of the new law. The comment period has yet to be determined. Vermont is also unique in that it houses 14 years of representative consumer data with information on attitudes, beliefs, knowledge and behaviors related to GM labeling. This Hatch project will utilize this unique data set to answer six important questions about consumer perceptions, knowledge, trust, and use of GM labels. This project fits under augmenting dimensions of my current extramurally-funded research program. The research questions are:Do positive GM labels create preferences or reveal existing consumer preferences?Are different types of GM perceived equally by consumers?What sources of information are trusted or mis-trusted by consumers?What happens consumer demand for products in a positive (contains) or negative (GM free) labeling environment?What are the consumer typologies related to purchases of GM or GM free foods?What are the preferred national labeling structures preferred by consumers?
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Goals / Objectives
The research questions are:Do positive GM labels create preferences or reveal existing consumer preferences?Are different types of GM perceived equally by consumers?What sources of information are trusted or mis-trusted by consumers?What happens consumer demand for products in a positive (contains) or negative (GM free) labeling environment?What are the consumer typologies related to purchases of GM or GM free foods?What are the preferred national labeling structures preferred by consumers?
Project Methods
The social-ecological and economics of information models provide the framework for the project. Originally developed by Broffenbrenner (1979) in the context of human development, the model has been expanded to be used by many disciplines. It serves as an excellent framework for policy decisions as it links individual decisions with organizations "up the chain," including "big P" policy initiatives. The economics of information originally developed by Stigler (1961) uses an economic framework to aid in the analysis of how information leads to optimal market decisions. The model focuses on individual decision making as it is influenced by knowledge. Linking the two models provides a comprehensive basis on which to investigate how public policy and the influence of the media as a knowledge source impact individual decisions about public policy initiatives. An example of the Ecological Model is provided below.The first five objectives of this project require quantitative methods utilizing cross-section time series primary data. The analysis utilizes frequency analysis, bi-variate statistics, and multivariate analysis. The project overall includes both descriptive and predictive results that can inform business strategy, consumer education materials and policy implementation. The P.I. has extensive experience using all of the methods required for analysis and has published widely using them.Objectives 1-5 will be completed using primary data collected during the years 2000-2017 and 2018/19 to be collected. These data sets contain cross-section, individual level data. For the GM labeling analysis, various years of data from 2000-2019 will be pooled. Because questions ascertaining public support/opposition for GM labeling were not asked in an identical manner in comparable over time. Similar checks must be done on all demographic variables to be included in the merged data sets. Once all variables have a common name and coding scheme, the eight years of data, can be stacked and a variable for "year" created.Objective 1 requires an endogenous model specification. There is no research in the peer reviewed literature that has addressed this issue using revealed preference data (actual behavior). The issue of whether preferences are endogenous is has caused a very large controversy in the policy arena. Currently, all assertions are conjecture.Objectives 2 and 3 require and ordinal probit specification because the dependent variables of level of support for various types of GM technologies (Obj. 2) and levels of trust (Obj. 3) are measured on ordinal scales: level of support (Strongly support to strongly oppose, 5 point scale) and level of trust (1-10, in tertiles or quartiles).The ordered probit model was developed by Zavoina and McElvey (1975). It applies in applications such as surveys, in which the respondent expresses a preference with the ordered ranking given above. Since the µs are free parameters, there is no significance to the unit distance between the set of observed values (e.g., strongly support, support, neutral, oppose, strongly oppose) of y. They merely provide the ranking. Estimated are obtained by maximum likelihood and both individual and grouped data can be included.The specification for trust is analogous.Objective 4 requires a multinomial logit specification. Multinomial logistic regression is a classification method that generalizes logistic regression to multiclass problems, i.e. with more than two possible discrete outcomes. The model is used predict the probabilities of the different possible outcomes of a categorically distributed dependent variable, given a set of independent variables.We will estimate, given an individual noticed a GM label, whether they did not use it in their decision making, had prior preferences for a GM product and the label helped them identify the product, had prior preferences to avoid a GM product and the label helped them identify the product, had no prior preferences and the label increased their preference for the GM product, or had no prior preference and the GM label decreased their preference for the GM product. The estimates will give an indication of the characteristics of consumers that are associated with making any of the listed choices. Currently, there is no information in the U.S. about actual behavior when positive labels are available on products.Objective 5 requires a cluster analysis in order to identify typologies of consumers. The method of two-step cluster analysis using the log-likelihood distance measure and Schwarz's Bayesian Criterion (BIC) in the Statistical Package for Social Sciences (SPSS) will be used to identify consumer patterns. When the relationships among several variables are unknown and several variables may be considered dependent, cluster analysis can give an indication of complex patterns within a data set, not easily accomplished with economic, sociological, or even social-ecological models, which tend to lead to regression type approaches to empirical analysis (Lakdawalla and Philipson 2007; Moussavi et al. 2008; Pickett et al. 2005). Only behavioral variables will be used to identify consumer typologies. These variables include support/opposition to various types of GM technologies, trust in information sources, use of GM labels to make decisions, and preferences for labeling initiatives. Once the typologies are determined, chi-square analysis will be used to identify whether typologies differ by demographic characteristics. The last step it to identify the marginal impact of these demographic variables on group membership. This model requires a multinomial logit specification, as outlined for Objective 4.Objective 6 requires a content analysis approach where the outcome is a set of themes related to consumer preferences for labeling support and various labeling schemes. Within the three types of content analysis, the summative approach will be employed. A summative content analysis involves counting and comparisons, usually of keywords or content, followed by the interpretation of the underlying context. A list of keywords will be developed using the language of the bill S. 764, and keywords obtained from several key literature reviews. As of July 2016, the labeling choices outlined include QR codes, a simple statement (contains genetically modified ingredients; produced using genetic engineering; partially produced using genetic engineering), telephone numbers, and websites. However, simply counting and summarizing these approaches to labeling will not provide enough information about the context as to why stakeholders prefer a particular approach. Therefore, a list of key words will be developed based on a review of several reviews of literature and the Legislative language (National Academies, 2016; Klumper and Qaim, 2014; Krimsky, 2015; Cuhra, 2015; Nicolia et al., 2013; Domingo and Bordonaba, 2011; Fischer et al., 2015; Bawa and Anilakumar, 2013; Sweet and Kostov, 2014). NVIVO qualitative analysis software will be used to identify nodes (themes) and relationship of context of citizen comments to labeling preferences. Depending on the number of comments included in the transcript, it is quite possible that quantitative analysis comparing "type" of citizen, preference for a particular labeling initiative, and reasoning provided in the comment may be undertaken. It is important to note that organizations can comment as citizens as was the case with the company Sarah Lee in comments on whether sweeteners produced using GM sugarcane should be considered 'natural.' The list of keywords will be developed during year two of the project. Analysis of the transcripts of citizen comment will commence when they are released.