DISSERTATION THE VALUE AND ROLE OF FOOD LABELS: THREE ESSAYS EXAMINING INFORMATION FLOWS IN THE FOOD SYSTEM FOR EXPERIENCE AND CREDENCE ATTRIBUTES Submitted by Oana C. Deselnicu Department of Agricultural and Resource Economics In partial fulfillment of the requirements For the Degree of Doctor of Philosophy Colorado State University Fort Collins, Colorado Summer 2012 Doctoral Committee: Advisor: Marco Costanigro Co-Advisor: Dawn Thilmany-McFadden Stephan Kroll Kenneth Manning ii ABSTRACT THE VALUE AND ROLE OF FOOD LABELS: THREE ESSAYS EXAMINING INFORMATION FLOWS IN THE FOOD SYSTEM FOR EXPERIENCE AND CREDENCE ATTRIBUTES This dissertation investigates the role of food labels as means of conveying information about food product characteristics, with particular attention to experience and credence attributes. Unobserved product characteristics such as taste, food safety, nutrition, or quality are inherently difficult to quantify but are frequent determinants of demand. Since not all these characteristics are measurable (e.g., food safety) or directly observable (e.g., nutrition), there exists information asymmetry in the market between firms and consumers. Product labeling is a way for information that is initially hidden to eventually be disseminated in the marketplace. Different labeling schemes serve different roles in the marketing system. For example, nutrition information is critical in consumption decisions, while other product characteristics (such as “organic”, or “fair trade”), may be valued by consumers but not essential for decision-making. Across three essays, we provide an assessment of how different types of labels are used in the food system. We focus equally on labels that have a long and rich history of usage in the food system (such as nutrition labels, and more recently, geographical iii indication (GI) labeling which denote a relationship between the product origin and specific product characteristics), but also labels that address emerging, public-minded issues which may be increasingly relevant in the future (such as environmental impact labeling and Corporate Social Responsibility (CSR) labeling). First, we meta-analyzed the literature regarding GI valuation to generate a set of guidelines, independent of any particular study, outlining the factors that are instrumental for a GI product to capture a price premium. Our findings across many studies indicate that agricultural produce and minimally processed foods such as grains, fresh meats, fruits and vegetables, benefit the most from association with GIs. These product categories generally do not develop own private reputations (brands), and thus, the premia received from association with GI collective reputations is relatively high. On the other hand, in addition to GIs, products with high value-added and longer supply chains such as wines and olive oils may also use private brands more effectively for differentiation. This suggests that brands and GIs have at least a partial substitute relationship. So, as the most broadly framed of the studies here, this cross-sectional analysis would suggest a further exploration of targeted labeling strategies, used jointly or independently of specific brand-name products, is warranted. Next, using original survey data and looking at nutrition label information, we find that truncated nutrition searches (looking only at the front label), or misleading product claims (such as “organic) are among a broad set of reasons current nutrition labeling practices may be ineffective in uniformly conveying information to consumers. We find that a nutrition index summarizing the information on the back nutrition panel, coupled with the information on the front label, may help to mitigate the incomplete iv information problems presented above. Moreover, we find that the environmental impact of food production is hard to identify by consumers if there is a lack of proper certification. But, until more consensus about key outcomes is framed by relevant government or consumer-oriented NGOs, a similar “informational index” solution will not be possible, so policy options are more limited. Finally, using original survey data we identify consumer preferences for CSR actions in the dairy industry. We find animal welfare to be the most preferred CSR activity and a top priority for most consumers. Sustainable agricultural practices, energy consumption, and waste management are second, third, and fourth, respectively, in importance for consumers; while company involvement in the community has the lowest priority amongst consumers. Furthermore, we monetize the value of animal welfare claims, identified as the most important CSR activity by consumers, in the context of a trusted third-party certification such as the Validus animal welfare certification program. Together, these empirical analyses provide a diverse set of findings on consumer perceptions, use of information, part-worth valuation of specific characteristics, as well as how these findings may vary by segments of consumers and product categories. By exploring these issues from a variety of perspectives and methods, the studies make both market-relevant and methodological contributions to the food labeling field. v TABLE OF CONTENTS ABSTRACT……………………………………………………………………….. ii CHAPTER ONE: Introduction ………………………………………………….. 1 The Economics of Information: a Literature Review……………………………... 10 References ………………………………………………………………………… 15 CHAPTER TWO: A Meta-Analysis of Geographical Indication Food Valuation Studies: What Drives the Premium for Origin Based Labels? .......................... 19 Introduction ……………………………………………………………………..... 19 Background ……………………………………………………………………….. 22 Methodology and Data Description ………………………………………………. 26 Model and Estimation Methods ………………………………………………….. 34 Results …………………………………………………………………………….. 36 Discussion ………………………………………………………………………… 39 Conclusions and Future Research ………………………………………………… 42 Tables and Figures …………………………………………………………………45 References ………………………………………………………………………… 54 CHAPTER THREE: Assessing Consumer Response to Nutrition Labeling Information and Environmental Product Cues …………………………….... 60 Introduction ……………………………………………………………………..... 60 Background ………………………………………………………………………. 64 Methodology ……………………………………………………………………... 70 Sample Demographics …………………………………………………………… 77 Empirical Methodology ………………………………………………………….. 79 Results ……………………………………………………………………………. 87 Discussion ……………………………………………………………………...… 95 Conclusions ……………………………………………………………………... 101 Tables and Figures ……...………………………...…………………………….. 104 Online Access to Survey Instrument ...………………………………………….. 120 References ………………………………………..……………………………... 121 CHAPTER FOUR: Corporate Social Responsibility Initiatives and Consumer Preferences in the Dairy Industry…………………………………………… 127 Introduction ….………………………………..……………………………….. 127 Background ….…….…………………………...…………………………..….. 129 Survey Methodology ….……………………………………………………..… 134 vi Data Description and Survey Participants Characteristics …….………………. 137 Data Analysis ...……………………………………………….……………….. 140 Results ………...…………………………………………….………………..... 143 Conclusions and Marketing Implications ...………………….……………….... 148 Tables and Figures ……..……………………………………………………… 151 References ………………..……………………………………………………. 167 CHAPTER FIVE: Concluding Remarks ...…...……………………………… 170 Limitations and Directions for Future Research………………………………... 174 References …...…………………………………………………………………. 179 1 CHAPTER ONE Introduction Consumers make an average of 200 to 300 decisions regarding food consumption in any given day (Wansink et al., 2007). However, many product attributes, especially in the food industry, can be hard to assess at the moment of purchase. Whether there is uncertainty and imperfect information about the product characteristics, prices, or quality across the universe of products available, food choices are generally made in an incomplete informational environment. When firms have more information than consumers about the products in the marketplace, there is a loss of efficiency and an overall lower total economic surplus achieved in that market (Caswell, 1996). This lower surplus may be due to the lower utility achieved by consumers from transactions in the presence of incomplete information, an overall lower number of transactions and/or higher overall transaction costs. While information asymmetry can be manifested on a multitude of levels, this research focuses on information flows for experience and credence attributes at the retail level within the food system. Starting with Nelson (1970) and a subsequent contribution by Darby et al. (1973), the literature identifies three categories of product attributes classified depending on how easy it is for consumer to acquire information about them. Some product characteristics such as color or aspect are search attributes. Consumers can easily identify them by visiting and comparing across multiple stores. Experience 2 attributes, like taste or quality are revealed to consumers only after consumption. Generally, firm reputations develop as a response to the incomplete information associated with experience attributes (Caswell, 1996). Reputations are viewed as an expectation of high quality (Shapiro, 1982) when they lead to returning (as opposed to one-time) customers whose loyalty offers sufficient returns to incentivize investments in quality. The most difficult food attributes to collect information on are credence attributes (e.g., food safety, fair taste, or nutrition). The outcomes associated with credence attributes are very difficult or impossible to assess even after consumption. In this case, the government often chooses to play a role in making it feasible for consumers to assess credence-based qualities by requiring informational labeling (Caswell, 1996). Generally, four approaches to government labeling can be identified (Caswell et al., 2011). First, consumers may “need to know” specific information (such as nutrition, environmental sustainability, or food safety) when making purchase decisions. For example, disclosing nutrition facts in a standardized fashion is mandatory in North America because of governmental priorities related to public health. Second, information the public has the “right to know” is frequently regulated by mandatory or voluntary labeling because of popular, consumer-driven demands on policymakers. For example, GMO labeling (now required in Europe, but not the U.S.) is the most popular example of right-to-know labeling. Third, information consumers generally “want to know” about the products and production process (such as organic farming) is administered through minimum requirements that serve as the basis for voluntary labeling, and because they are voluntary, are more commonly used by food companies that feel they can effectively target consumers seeking such attributes. Fourth, product information relevant to the 3 regulatory oversight mission of “prevention of fraud” or deception of consumers is also subject to labeling. This dissertation investigates the role of food labels as means of conveying information about food product characteristics, with particular attention to experience and credence attributes. Across three essays, we provide an assessment of how different types of labels are used in the food system. We focus equally on labels that have a long and rich history in the food system (such as nutrition labels or geographical indication labeling, which denote a relationship between the product origin and specific product characteristics), but also labels that address novel, current issues and may be have a widespread implementation in the future (such as environmental impact labeling and corporate social responsibility labeling). In the context of government’s role in labeling, “need-to-know” labels such as nutrition information are implemented to correct market failures associated with credence nutrition information that cannot be asses even after consumption of the product. Nutrition labeling is mandatory in the United States; however, low rates of use are reported with respect to these labels (Viswanathan, 2002; Roe, 1999; Black, 1992; Higginson, 2002). Research shows that consumers generally do not check nutrition labels or use only one or two nutrition attributes (such as sugar or fat) when they do consult them. This results in an ongoing debate in the literature regarding the effectiveness of these labels in relaying information within the framework of their low use. Are nutrition labels accurately interpreted by consumers? Do they transmit uniform nutrition information to consumers? What is the effect of low (or truncated) usage of nutrition labels? 4 Similar to nutrition labels, geographical indication (GI) voluntary labels were implemented in the early 1990s. GIs have been successful (valued by consumers) in signaling a specific link (referred to as terroir) between the origin of production and product characteristics (Hermann et al., 2011). Prominent examples of GIs are Parmigiano-Reggiano cheese and Champagne wine, which are believed to be of higher quality, or display increased heritage connectedness. GIs are “want-to-know” labels signaling experience attributes that are valued by consumers, but given this designation, there is interest in how valued they are, particularly given that they are used in the same retail market environment as brand names. A very large volume of research quantifies the values consumers have for GI labeling for a variety of products in different regions of the world. However, the high variation in price premia associated with GIs raises the question concerning what factors drive GI valuation. Do institutional characteristics in each country play a role in GI valuation? Are some product categories associated with a higher price premium than others? This dissertation attempts to answer these questions. In addition to already existing labels with an established history in retail markets, such as nutrition and GI, current world trends and events give rise to consumer demand for new product information. The environmental impact of food products may qualify as “as “want-to-know” information. The environmental impact of food products has mostly been studied in the context of eco-labeling. Research finds that demand for eco-labeled products exists and consumers are willing to pay a price premium for more environmentally-friendly products (Johnston et al., 2006, Blend et al., 1999). However, in the lack of standardized labeling, it is hard to assess the environmental impact of food products. Individuals tend to develop personal norms (such as favoring product 5 packaging like cardboard, or buying organic foods) when choosing environmentally friendly products, and these norms are a significant predictor of their propensity to choose environmentally friendly options in the supermarket (Thǿgersen, 1999). This suggests that when credence attributes, such as the environmental impact of a product, are hard to assess, experience or search product attributes can be used as proxies for harder to assess credence attributes. However, how successful are environmental product cues in conveying information to consumers? Are they a substitute for standardized labeling? Can these cues bridge the information gap due to lack of standardized environmental labeling? Similar to environmental labeling research, research related to corporate social responsibility (CSR) information labeling is especially relevant today given the current growing importance of CSR claims for consumers, producers, NGOs, and the media. Company CSR actions are meant to internalize negative firm externalities on society and the environment and work on decreasing them. The popularity of CSR initiatives have been increasing, however, it has been documented that only a limited number of consumers use it as a purchase criterion. Out of the consumers who are likely to make a CSR-based purchase, only a minority (21%) actually use a company’s CSR position as a purchase criterion (Mohr et al., 2001). The wedge between the popularity of CSR and its use in product selection is minimally addressed in the literature. Some of the first attempts to explain it call it “the paradox” of CSR in consumer behavior (Öberseder et al., 2011). Also, since CSR information qualifies as “want-to-know” information, it means that only a subset of the population may be interested in it. Identifying consumer preferences and values for CSR actions is part of identifying whether demand for CSR 6 information exists. Our research attempts to clarify this issue further by addressing unanswered questions in the literature: do consumers value CSR actions enough to pay a price premium for them? What are the most valued CSR actions by consumers in the dairy sector? Given the lack of standardized CSR information labeling, how does CSR information reach consumers? Across three essays, this dissertation attempts to provide answers to the questions outlined above for nutrition labeling, voluntary geographical indication certification, environmental impact labeling, and CSR labeling. Together, these empirical analyses provide a diverse set of findings on consumer perceptions, use of information, part-worth valuation of specific characteristics, as well as how these findings may vary by segments of consumers and product categories. By exploring these issues from a variety of perspectives and methods, the studies make both market-relevant and methodological contributions to the food labeling field. The first essay presented in Chapter II, “A Meta-Analysis of Geographical Indication Food Valuation Studies: What Drives the Premium for Origin Based Labels?” uses a meta-analytical approach of the empirical literature on geographical indications (GIs) in order to establish a link between the price premium received by the evaluated products and specific product, industry, or institutional characteristics. Presumed higher quality, specific sensory attributes, heritage production methods or other particular characteristics of these products (associated with the region geographical microclimate) are primarily experience attributes that can be evaluated through consumption (Hermann et al., 2011). GIs as voluntary government certifications have developed in this case to signal the presence of product characteristics associated with specific geographical 7 origins. The presence of GI certification is generally valued positively by consumer, but some types of foods may benefit from associations with GI names more than others. In cases where GIs coexist with other forms disclosing hidden product information (such as product brands), we explore the dynamics between private (brands) and collective means (GIs) of signaling quality for experience goods. Chapter III, “Exploring Product Differentiation through Environmental Impact Claims and Metrics”, focuses on two important characteristics of food products: nutrition and environmental impact. Nutrition and environmental characteristics are food credence attributes, whose effect cannot be immediately determined even after consumption. Government regulation is generally the most appropriate means of resolving information asymmetry problems in markets for credence attributes (Caswell, 1996). In the US, there is a long history of nutrition regulation culminating with the Nutrition Labeling and Education Act (NLEA) passed by Congress in 1990 (Drichoutis et al., 2011). The NLEA regulates the uniform transmission of nutrition information through standardized nutrition facts and serving sizes on all packaged foods. However, increases in obesity rates (Berning et al., 2008, 2010) and reported low levels of use of information on the regulated product label (Higginson, 2002) suggest the need for improvement of product label standards or format. On the other hand, the use of rudimentary product cues (such as packaging material) by consumers to assess the environmental impact of food products (Thǿgersen, 1999) may be an indication that government intervention is this area is necessary. But, beyond the appropriate role of the government in labeling, this study will focus on how consumers currently process the information they receive in the retail marketplace for dairy products. 8 Chapter IV, “Corporate Social Responsibility Initiatives and Consumer Preferences in the Dairy Industry”, investigates how information about ethical product claims of food can be transmitted to consumers. Ethical product claims, such as those featured in corporate social responsibility reports about commitments towards increased air quality, low energy use, or animal welfare, are credence attributes whose outcome cannot be immediately determined by consumers even after consumption. In this case, government regulation, or that from another trustworthy third-party certification program, is necessary to make these claims credible (Caswell, 1996, 2011). In lack of specific government regulations, CSR information may be communicated through indirect channels, such as labels instituted by the government for other purposes. For example, Organic product labels may be indicative of higher standard for livestock animal welfare in dairy products, even though organic production does not imply the adoption of the most rigorous animal welfare protocols held as standards by some certification programs. However, trusted third-party certification (such as Validus animal welfare certification) is also suitable, in this case, for highlighting credence attributes, but may be less commonly known or understood by buyers because of its relatively smaller scope in the marketplace. Each of the essays included in this dissertation provides an original contribution to the literature on information asymmetry of intrinsic product characteristics. Experience attributes derived from the connection with a specific geographical region are regulated under geographical indication voluntary schemes. The current dissertation contains the only meta-analysis in the literature investigating the reasons behind the premium for GI valuation. We use published GI valuation studies to generate a set of guidelines, 9 independent of any particular study, outlining the factors that are instrumental for a GI based product differentiation scheme to capture a price premium. In addition to identifying the reasons behind GI valuation, we provide consumer valuation for product CSR credence attributes in the dairy industry. An original survey instrument is developed to elicit consumer preferences for CSR activities in dairy and value these activities in the context of current milk labels. This is the only study that we are aware of that identifies CSR preferences in dairy and provides a monetized value for them. Another contribution of this dissertation lies in measuring the information gap arising from lack of environmental product labeling. We use original survey data to statistically measure the environmental information gap by comparing it to nutrition, an area which currently benefits from standardized labeling. The boundaries of current knowledge in each of these labeling areas are identified and an original contribution is presented for each of them. 10 The Economics of Information: a Literature Review Imperfect information has profound effects upon the market structure of consumer goods (Nelson, 1970) and on consumer behavior in the market. The market structure may change, for example, when asymmetric product information leads to product differentiation and the creation of monopolistic competition (Wolinsky, 1984; Stiglitz, 1979; Schultz, 2004). Across several disciplines like economics, psychology, sociology, social psychology, and anthropology, researchers have attempted to explain individual human choice behavior under imperfect information (Hansen, 1972). Information has economic value because it allows individuals to make choices that yield higher expected payoffs or expected utility than they would obtain from choices made in the absence of information. The food industry provides an especially suitable example of the effect of asymmetric information on markets. Many food attributes and characteristics can be hard to assess by consumers. Unobserved product characteristics such as taste, style or quality are inherently difficult to quantify but are frequent determinants of demand. In some markets, products may be physically similar but differ in consumers' perceptions about quality, durability, or status (Berry, 1994). Lancaster (1966, 1991) proposes that consumers are not interested in goods per se, but in their properties or characteristics. In Lancaster’s approach, the major food product attributes related to quality include food safety (e.g., levels of microbial pathogens, residues), nutrition, value (e.g., compositional integrity, taste), package, and process (e.g., animal welfare, environmental impact) attributes (Hooker et al., 1996). However, not all these characteristics are measurable 11 (e.g., food safety) or directly observable (e.g., nutrition). In other words, there exists an information asymmetry in the market between firms (who are more knowledgeable than consumers about, for example, food safety or nutrition of a product), and consumers. Caswell (1996) suggests that the distinction developed by Nelson (1970, 1974, 1976, 1981) between search and experience goods, when applied to product attributes, is powerful in understanding how information that may be initially hidden, can be eventually disseminated in the marketplace. Nelson (1970) proposes two actions which consumers can take to assess quality and overall utility derived from a product: search (inspecting prior to purchasing), and experience (consuming the good). Search attributes (or goods) are defined by product attributes for which full information can be acquired prior to purchase. Clothing, footwear and furniture fare typically cited as examples of search goods (Seigel, 2006). Search attributes related to food are color, smell, and physical appearance. Experience goods are dominated by attributes that cannot be evaluated until purchase and consumption of the product. Examples of experience goods and services are automobiles, appliances, or weight control programs (Seigel, 2006). For food, experience attributes relate to taste, cooking properties, or texture of product when consumed. In 1973, Darby et al. added credence goods (or attributes) to this classification. Claims associated with credence attributes are difficult or impossible to determine even after consumption. For example food safety, nutrition, or ethical product claims (such as fair trade) fall into this category. Each of these types of attributes has specific information asymmetry problems in the market and solutions that alleviate these problems. In the market for search goods, 12 consumer information is easier to obtain. Search goods are more susceptible to substitution, as consumers can more easily identify and evaluate alternatives by visiting other outlets and stores. For food products, most search attributes (e.g., color) are not related to life-altering events associated to safety and nutrition so the “cost of being wrong” is not high (Caswell, 1996).The market for search goods has relatively minor regulatory activities, because consumers are in a position to provide direct incentives to firms to produce the search attributes that are most popular (Caswell, 1996). In markets for experience goods, quality information is the most important product characteristic (Caswell, 1996). Akerlof (1970) provides an example of market failure due to information asymmetry by describing the “lemon” problem in the market for used cars. A lack of credible quality signals creates incentives for sellers to misrepresent the quality of their goods. Buyer’s willingness to pay for high quality decreases. This creates the problem of adverse selection where high quality is crowded out by low quality resulting in a collapse of the market for high quality (Akerlof, 1970). In the market for experience goods, there exists a moral hazard problem of firms to misrepresent their products as high-quality and sell them to a one-time customer. One way firms navigate the moral hazard problem, reveal information and signal quality to consumers is through labeling, advertising, warranties, and building reputations. Reputations are costly to build and they require returning (as opposed to one-time) customers. Developing reputations is a good solution to alleviate information problems related to experience goods. The outcomes related to credence attributes are hard to assess by consumers even after consumption. Reputations rarely develop in response to credence attributes because 13 the consumer cannot learn it from his or her previous experience in consuming the product and cannot form a quality expectation related to a particular brand or name (Caswell, 1996). Reputational models of quality do not apply here, but quality signaling may still be used if it involves a third-party reputable certification agent whom consumers trust (Caswell, 1996). The government can play a role in increasing the number of informed consumers by facilitating communication through official and consistent labeling and certification. Labeling changes the amount or type of information that is available in the market and has the advantage to certify the effect of individual product attributes (the Lancasterian approach) as opposed to entire goods and services (Caswell et al., 2011). While information asymmetry is an important factor affecting consumer product selection and the product purchasing process, consumers’ tastes and preferences also affect market behavior. Individual preferences determine the relative importance given by each consumer to various product attributes, and different consumers make different choices based on their unique preference map. Consumer purchase behavior (buying a specific product when other substitutes are available) can be used to infer what product attributes consumers value (revealed preferences) and what their underlying preferences are (McFadden, 2001). However, while consumer choices for market goods can change based on the situation surrounding the decision-making process (Fishhoff, 1993), it has been shown that people’s values are more stable (Lusk et al., 2005). Values are defined as meta-preferences (Lusk et al., 2005), or “underlying preferences” (Becker, 1976) that people hold with respect to the essential aspects of human life. The desire to have a healthy lifestyle, be compassionate towards others, respect the environment, or achieve 14 fame and prestige are some of these values (Lusk et al., 2005). These values motivate the choice of products individuals make more so than individual preferences over a set of attributes, which can be circumstantial and contextual. Also, while consumers may not have specific preferences for individual product characteristics, they do hold underlying values that help them make decisions. For example, while people may not have specific preferences for vitamin A relative to vitamin B12 content of a specific food, they are likely to know whether “nutrition”, as a value, is important for them and make choice that support outcomes that are nutrition-friendly (Lusk et al., 2005). The means-end chain literature pioneered by Gutman (1982) upholds the idea that goods are the means (objects, or activities) in which people engage in order to achieve desired end-states such as “happiness, security, or accomplishment”. These recent developments of consumer theory provide a more profound insight into consumer choice. 15 References Akerlof, G. A. 1970. The Market for 'Lemons': Quality Uncertainty and the Market Mechanism. Quarterly Journal of Economics 84:488-500. Becker, G.S. 1976. The Economic Approach to Human Behavior. University of Chicago Press. Berning, J.P., Chouinard, H. H., McCluskey, J. J. 2008. Consumer Preferences for Detailed versus Summary Formats of Nutrition Information on Grocery Store Shelf Labels, Journal of Agricultural & Food Industrial Organization, 6(1). Berning, J.P., Chouinard, H. H., Manning, K. C., McCluskey, J. J., Sprott, D. E. 2010. Identifying consumer preferences for nutrition information on grocery store shelf labels, Food Policy, 35(5): 429-436. Berry, S. T. 1994. Estimating Discrete-Choice Models of Product Differentiation, The RAND Journal of Economics, 25(2): 242-262. Black A, Rayner M. 1992. Just Read the Label: Understanding Nutrition Information in Numeric, Verbal and Graphic Formats. London: HMSO. Caswell, J.A., Mojduszka, E. M. 1996. Using informational labeling to influence the market for quality in food products. Food Marketing, 7(5): 1248-1253 Caswell, Julie A. and Sven Anders. 2011. Private vs. Third Party vs. Government Labeling. In The Oxford Handbook of the Economics of Food Consumption and Policy, ed. Jayson Lusk, Jutta Rosen, and Jason Shogren. Oxford, United Kingdom: Oxford University Press. Costanigro, M, McCluskey, J. J., Goemans, C., 2010. “The Economics of Nested Names:Name Specificity, Reputations, and Price Premia". American Journal of Agricultural Economics 58 (3), 454–466. Darby M., R., Karni, E. 1973. Free Competition and the Optimal Amount of Fraud, Journal of Law and Economics, 16 (April): 67–86. Drichoutis, A. C., R. M. Nayga, and Lazaridis, P. 2011. Nutritional Labeling, Oxford Handbook on the Economics of Food Consumption and Policy. Eds. Jayson Lusk, Jutta Roosen, and Jason Shogren. Oxford University Press. Fischhoff, B. 1993. Value Elicitation: Is There Any-thing There?, in M. Hetchter, L. 16 Nadel, and R.E. Michod, eds. The Origin of Values. New York: Aldine de Gruyter, 187–214. Gutman, J. 1982. A means-end Chain Model Based on Consumer Categorization Processes, Journal of Marketing 46(2):60-72. Hansen, F. 1972. Consumer Choice Behavior: A Cognitive Theory. New York, The Free Press. Herrmann, R. and R. Teuber. 2011, Geographically Differentiated Products. In: Lusk, J., J. Roosen and J. Shogren (eds.), Oxford Handbook on the Economics of Food Consumption and Policy. Oxford: Oxford University Press 2011, Chapter 33. Higginson CS, Rayner MJ, Draper S, Kirk TR. 2002. The nutrition label – which information is looked at? Nutrition and Food Science, 32: 92–9. Hooker, N. H., J. A. Caswell. 1996. Regulatory Targets and Regimes for Food Safety: A Comparison of North American and European Approaches, in The Economics of Reducing Health Risk From Food, J. A. Caswell, ed., 1-17. Storrs, CT: Food Marketing Policy Center. Johnston, Robert J., and Cathy Roheim. 2006. A Battle of Taste and Environmental Convictions for Eco-labeled Seafood: A Contingent Ranking Experiment. Journal of Agricultural and Resource Economics 31(2): 283-300. Klein, L.R. 1998. Evaluating the Potential of Interactive Media through a New Lens: Search versus Experience Goods, Journal of Business Research, 41(3): 195–203. Lancaster, K.J. 1966. A New Approach to Consumer Theory, Journal of Political Economy, 74(2): 132-157 (April). Lancaster, K.J. 1991. Modern Consumer Theory. Edward Elgar Publishing Limited, England. Lusk, J., Briggeman, B. 2009. Food Values, American Journal of Agricultural Economics, 91(1): 184-196. McFadden, D. 2001. Economic Choices. American Economic Review, 91:351–78. Mohr, L.A., Webb, D.J., Harris, K., 2001. Do consumers expect companies to be socially responsible? The impact of corporate social responsibility on buying behavior. Journal of Consumer Affairs 35(1). 17 Nelson, P. 1970. Information and consumer behavior. Journal of Political Economy, 78(2): 311-329. Nelson, P. 1974. Advertising as Information, Journal of Political Economy, 82 (4): 729– 754. Nelson, P. 1976. Economic Value of Advertising, Yale Brozen (Ed.), Advertising and Society, New York University Press, New York, pp. 109–141. Nelson, P. 1981. Consumer Information and Advertising, M. Galatin, R.D. Leiter (Eds.), Economics of Information, N. Nijhoff Publishers, Boston, MA, pp. 42–77. Öberseder, M., Schlegelmilch, B. B., Gruber, V. 2011. Why Don’t Consumers Care About CSR? A Qualitative Study Exploring the Role of CSR in Consumption Decisions. Journal of Business Ethics 104:449-460. Schultz, C. 2004. Market transparency and product differentiation, Economics Letters 83: 173–178. Shapiro, C. 1982. Consumer information, product quality, and seller reputation, The Bell Journal of Economics, 13(1):20-35. Siegel, Donald S. and Vitaliano, Donald F. 2006. An Empirical Analysis of the Strategic Use of Corporate Social Responsibility (April). Available at SSRN: http://ssrn.com/abstract=900521 or http://dx.doi.org/10.2139/ssrn.900521 Stigler, G. 1961. The economics of Information. Journal of Political Economy, 69 (October): 213-225. Stigler, G. 1962. Information in the labor market. Journal of Political Economy, 70 (October): 94-105. Stiglitz, J. E. 1979. Equilibrium in Product Markets with Imperfect Information, the American Economic Review, 69(2): 339-345. Thǿgersen, 1999. The Ethical Consumer. Moral Norms and Packaging Choice. Journal of Consumer Policy 22(4): 439-460. Wansink, B., J. Sobal. 2007. Mindless Eating: the 200 Daily Food Decisions we Overlook. Environment and Behavior, 39(1): 106-23. http://dx.doi.org/10.2139/ssrn.900521 18 Wolinsky, A. 1984. Product differentiation with Imperfect information. Review of Economic Studies, 51(1):53-61. 19 CHAPTER TWO A Meta-Analysis of Geographical Indication Food Valuation Studies: What Drives the Premium for Origin Based Labels? 1. Introduction Agricultural and food products have long been associated with unique characteristics and heritage aspects associated with their origin. Geographical names have been used since classical times to identify products of exceptional quality; for example, historical documents reveal the notoriety of olive oils from Baetica in Rome (Blasquez et al., 1992). Through the ages, a number of products identified by their origins emerged and, more recently, have established a niche in food and beverage markets. Well-known examples of Geographical Indications (GIs) are the wines of Bordeaux and Porto, the cheeses of Parma and Rochefort, and the hams from Parma and Bayonne. In general, the association of food products and geographic names identifies distinct agro-ecological conditions, typically raised animal breeds and plant varieties, and human capital uniquely suited to the region. These conditions are often associated with the definition of terroir (Joslin, 2006). In addition, the names of GI products may signal specific modes of production, and commonly emerged based on the collective reputation of numerous producers. 20 In an increasingly industrialized and standardized food market, GI labels seem to suggest consumers a more genuine, unique and higher quality food (Broude, 2005); while offering producers an opportunity to differentiate their products and, perhaps, obtain higher prices. Thus, firms may use a GI to signal intrinsic quality attributes to consumers, and thus, capture a reputation rent (Menapace et al., 2011). A measure of a GI label’s success might, then, be partially evaluated by the price differential between a GI product and its branded or commodity competitors in the market. Based on this criterion, and using the empirical literature documenting how GI products’ valuations measure up relative to commodities in the same product category, it is possible to identify which food categories have secured higher premia. To this end, we compiled a pool of 25 empirical studies analyzing GI labels and observed that the statistical and economic significance of estimated price differentials vary substantially. Using these studies we aim to provide preliminary answers to the following questions: what critical factors determine price premiums? Do these factors vary across products and countries? When do GIs add more value to products than alternative differentiation strategies? Food producers, namely those producing processed products, such as cheeses or wines, have, and use, alternative marketing strategies to differentiate their products based on individual reputations. In fact, there is a wide variety of products in the food and beverage sectors that have achieved widespread recognition and popularity using brand names rather than geographic designations. Thus in some product categories there may co-exist GIs and branded products (or trademarks). While indications of origin have mainly been used in Southern European countries, they are becoming increasingly common in Northern Europe, the New World and in developing countries. Examples 21 include wines from specific viticultural areas in North America, Australia and New Zealand; Jamaica’s Rum and Blue Mountain coffee, as well as India’s Basmati rice and Darjeeling tea (Costanigro et al., 2009; Schamel et al., 2006; Das 2006; Gautam et al., 2010; Deppeler et al. 2011). To the best of our knowledge there is no previous study attempting to compare GI price premiums across product categories. Still such information could help both producers and policy makers decide in which cases the labeling of origin might be suitable marketing instrument. A suitable methodology to compile and investigate common patterns in the published work is meta-analysis regression. This technique is quite common in the medical science to establish common patterns in related studies and reconcile possible conflicting evidence (Hunt, 1997). It is also increasingly used in economics to perform “a more formal and objective process of reviewing an empirical literature” (Stanley, 2001, pp. 147-148). Our intent is to generate a set of guidelines, independent of any particular study, outlining the factors that are instrumental for a GI based product differentiation scheme to capture a price premium. More specifically, the primary objective of this study is to (meta-) analyze the empirical literature on GIs in order to establish a link between the GI premium and specific product, market characteristics and/or institutions. We consider three major dimensions of each product examined: 1) broad food categories, degree of food processing and product prices; 2) existence/absence of an alternative differentiation mechanism (i.e. branding) and; 3) the institutions and laws regulating the use of GIs. We proceed by summarizing the relevant literature on why consumers and producers may 22 value GI labels, describing the data and methodology employed here, presenting the findings and discussing marketing and policy implications in the following sections. 2. Background Economists suggest that GIs are used in food markets to signal intrinsic qualities of foods that consumers attribute to certain origins (Menapace et al., 2011). Indeed, Costanigro et al (2010) emphasize how GIs may essentially provide a means to broadly categorize food choices, thereby facilitating consumer learning and the articulation of quality expectations (a reputation effect). However, the reasons behind consumers’ and producers’ use of GIs are likely to be complex and multi-faceted. Scarpa et al (2005) suggest one potential rationale, arguing that consumers’ ethnocentric preferences or home bias may explain some of the preferences for origin labeled foods. In other words, the argument is that consumers tend to prefer products from the region or country with which they identify. Another reason, suggested by Broude (2005), is that GIs may counteract the perception that increased globalization has led to overly standardized food choices imposed by international brands. Still another argument is that GIs reveal and represent some sort of authenticity, cultural heritage or the ability to trace food choices to their origin (Herrman et al., 2010). In short, there seems to be a renewed interest by some segment of consumers in “authentic,” “traditional,” “wholesome,” and “traceable” food which seems related to a range of factors such as increased awareness of food safety, the socio-cultural status of 23 consuming certain foods and renewed interest in, or nostalgia of, one’s culinary heritage (Ilberry et al., 2000). Farmers may use GI designation to differentiate their products and avoid competition in commodity markets, where brand-based product differentiation is otherwise impractical. That is, farmers and primary food processors using GI labels may have easier or cost effective access to niche markets, and have the ability to extract premium prices (Bramley et al., 2009). A theoretical framework explaining the use of GIs and trademarks has been proposed by Menapace et al, (2011), extending an earlier model from Shapiro (1983) on the relationship between minimum quality standards, reputation, and price premia. In both articles, premia for high quality are modeled as (lagged) returns from investment in quality. Since reputations develop slowly over time, a price premium (above cost of production) is necessary to induce firms to produce at any quality level above the minimum standard imposed on all firms. The farther away a firm moves from the minimum standard in the quality spectrum, the longer it will take to build the reputation, the larger the premium needs to be. Thus, for producers with limited resources who are located in a GI area, it may be a sensible strategy to use the origin labels rather than to develop their own reputation through a brand. Along these lines, policy-makers have long acknowledged consumer interest and the potential of GIs to impact product valuation, international trade flows and farm policy (Herrmann et al., 2010). Most importantly, GIs may represent a key option to raise farmers’ incomes and promote rural development (Josling, 2006). After a long period of spontaneous and informal development, designations of origin have been the object of 24 increasing policy and regulatory efforts, most notably in Europe. In the early 1990’s, the European Union conferred legal protection to foods and foodstuffs with a GI through Regulation (EEC) 2081/92 (EEC Council, 1992). At the core of this regulation is the idea that products originating from certain regions are sui generis, in that there is a direct link that can be demonstrated between the product origin and its final quality (Herrmann et al., 2010). This link occurs either via a set of standardized processing practices typical of a region or by the concept of terroir. The varying strength of this link is the rationale behind the use of two labels: in the case of a PGI, either production, processing or preparation of a product need to occur in the geographical area; while for a PDO all stages must occur in the same region (O’Connor, 2007). In other words, PDOs have more stringent standards of production and signal a stronger link between origin and the product’s attributes. Finally, this regulation confers protection from “abusive” or unwarranted use of a protected designation of origin (PDO) or a Protected Geographical indication (PGI). While the EU legislation on GIs is perhaps the most fully articulated and comprehensive (Josling, 2006), other countries have their own systems. In the US, GIs are protected within the standard trademark system, and most often simply verify the geographical origin of a product (Menapace et al., 2009). Names or signs, which otherwise would be considered primarily geographically descriptive, can be registered as quality assurance programs (USPTO, 2011). The process of establishing and using such a verification process is straightforward. First, an agency (at the state or regional level) establishes the standards governing a GI based trademark (e.g.: Idaho Potatoes must be grown in Idaho, and must be of a specific variety, e.g. Burbank, see O’Connor, 2007). It 25 is up to the agency to choose how strict these standards are based on their perceptions of the existence of differentiation opportunities in the marketplace. Then, anyone who meets these standards is permitted to use the geographical name to market their product. In the case of GIs, the geographical origin is usually the main attribute that is regulated by the quality assurance program or trademark (USPTO, 2007). However, the allowance of multiple criteria suggests that trademark programs may display a weaker link between origin and product attributes than the PGI and the PDOs, and instead, require a broad set of practices to truly differentiate the product in the consumer’s eyes. In short, both food producers and consumers seem to benefit from the use of GI labels. From a producer’s perspective, origin can be an inexpensive way to differentiate a product and obtain a price premium. For consumers, it is a way to reduce search costs, as GI can incorporate a heuristic with which consumers limit the number of options on a choice set for a product category of interest (Costanigro et al., 2010). When origin is a valuable attribute, there will be a strong incentive to free ride. Consequently consumers may distrust the origin label unless there is some form of assurance that the product they face is genuine. This is why some origin labels, notably those regulated and recognized by the EU, have a standard, third-party monitoring and certification scheme to which all producers using the label must comply. 26 3. Methodology and Data Description As already mentioned, this research employs a meta-analysis to determine what factors influence the variation of price premium across products using GI labels. This methodology is increasingly popular in economics and recent examples of its application include Lusk et al (2005) on the valuation of genetically modified foods; Brander et al, (2011), on the value of urban open space; and Lagerkvist et al, (2011), on consumer willingness to pay for animal welfare. The meta-analysis methodology entails to quantitatively analyze the results of empirical studies that investigate the same topic. This is a popular analysis in social studies, medical and clinical research, and psychology (Hedges et al., 1998). The objective of meta-analyses is to provide an overview of the research on a particular topic by summarizing and synthesizing the results in the field, as well as testing theoretical and practical hypotheses that cannot be tested in the primary research (Brannick et al., 2008). Generally, this is accomplished by estimating the mean of the distribution of effect sizes (coefficient estimates) from multiple studies, and estimating and explaining the variance in the distribution of these coefficient estimates (Brannick et al., 2008). Examining the variance in coefficient estimates explains how study characteristics affect research results and helps draw overarching relevant conclusions about the topic of study. Some advantages and disadvantages of meta-analyses emerge when compared to traditional literature reviews. One advantage is that, while traditional literature reviews may only selectively include studies based on the reviewer’s own subjective view of the quality of the study, meta-analyses include studies based on clearly defined rules and thus 27 are less biased (Wolf, 1986). In addition, the subjective weighting of studies or the failure to examine study characteristics as explanations of results across studies are addresses in meta-analyses compared to traditional literature reviews (Wolf, 1986). On the other hand, one of the main criticisms of meta-analyses is that it includes published research that is biased in favor of significant findings because insignificant findings are rarely published (Rosenberger et al., 2009; Wolf, 1986). However, this is also the pitfall of traditional literature reviews, since publication bias affects both. Statistical issues arise when analyzing data compiled from numerous studies that use different methods to generate their own estimates. Some of the ways biased conclusions may be obtained in meta-analyses include a strong bias towards publishing positive but not negative results (Rosenthal, 1979), weighing equally the results of all studies even through there may be qualitative differences among them, or including multiple results from a single study. This latter problem of within-study correlation of estimates is one of the main analytical criticisms of meta-analyses. Two statistical models have been historically used to examine this type of data: fixed-effects and random-effects models (Hedges et al., 1998; National Research Council, 1992). In the fixed-effects model, it is assumed that the underlying effect of each study is the same. The variation in investigated outcome will therefore reflect only the random variation within each study but not any potential heterogeneity across studies (Schulze et al., eds., 2003). Random- effect models have been used to account for within-study correlation of estimates (Lusk et al, 2009). The random effects model incorporates variation between the models. It is assumes that each study has its own effect. 28 In other words, if there is reason to believe that the effect sizes are homogeneous in nature and the researcher wishes to make inferences only about the parameters in the set of studies that are observed, then fixed effects model is appropriate. In contrast, if estimates are not homogeneous and inferences need to be generalized beyond the observed studies, random effects model can be used (Hedges et al., 1998). Recently, however, meta-analysis studies test for the existence of fixed or random effects and may choose neither (Lusk et al, 2009; Ehmke, 2006). In these cases, a simple OLS model may be appropriate. It can also be argued that not all studies synthesized in a meta-analysis should be given equal weight (Wolf, 1986). Some studies may be based on very small or unrepresentative samples of subjects. Assigning equal weights may lead to less representative studies contributing equally to results as more well-designed studies (Wolf, 1986). Using the sample size as weight gives higher weights to studies that provide “more evidence” and more precise parameter estimates (Schulze et al., 2003). Most meta-analysis methodologies originate from the psychology literature, where most meta-analyses are done. While psychology data may be different than, for example, economics data, the general rules and framework developed in psychology also applies to social sciences studies. While the debate about the usefulness of meta-analyses is on-going (Hunter et al., 1996; Feinstein, 1995), it a useful tool frequently adopted by researchers (Schulze et al., eds., 2003). Meta-analysis is not a strictly standardized technique and criticisms originate not only on statistical grounds but also on conceptual and philosophical grounds (Schulze et al., eds., 2003). However, the technique is helpful in highlighting gaps in the literature and providing insights into new directions for research (Wolf, 1986). 29 In order to compile the database used in this study, we searched several applied economic and food industry databases for studies estimating consumers’ willingness to pay (WTP) or market premium for GIs in a variety of food products. More specifically, EconLit, Web of Science, EBSCO Business Source Premier, and Google Scholar were consulted in early 2011. Studies published after this date or in other databases may not be included. Since the first transnational regulation on GI products was introduced in the EU in 1992, we only included studies dated from 1990 onwards. To identify relevant studies we used the following keywords and keyword combinations: “geographical indication”, “protected designation origin”, “protected geographical indication”, “PDO”, “PGI”, “trademark”, “WTP label”. To be included in the sample, the studies had to meet two general criteria: 1) GI valuation estimates were reported as a premium/discount with respect to a generic, non-GI, product, and, 2) the product has a strong geographical connotation, identifying a specific region of production. To be precise, the first criterion implied including only articles for which it was possible to obtain valuation estimates (either directly or as a function of the reported estimates) calculated with respect to a generic (non-GI) reference product or a superordinal product categorization 1 (for example, Bordeaux wine valued with respect to a pool of other European wines, or other French wines). As for the second criterion, all estimates relative to products carrying a PDO, PGI, or trademarked geographical label were included, as well as products originating from a very specific region that may not have an official GI label (e.g. wine from Hunter Valley, Australia). Studies estimating consumer valuation of country of origin labels (COOL) were excluded from the sample 1 Examples of GIs studies excluded under this criterion include Mtimet, 2006; Santos, 2005; Schamel, 2003; Ali, 2007; Combris, 1997. 30 because the link between geographic name and specific growing conditions (the concept of terroir) was considered too weak. That is, a WTP differential for similar food products made in U.S. vs. made in China might have more to do with perceived differences in food safety standards than differences in growing conditions. Finally, we did not consider studies estimating the premium for locally-grown products, as products marketed as “local” rarely identify specific enough characteristics of the region of production. For local products, the geographic connotation relates more to the distance (rather than product origin) between location of production and the location of consumption, and is therefore a relative concept. In short, what is perceived as local by a New York consumer is certainly not local for a San Francisco one, and vice-versa. In total, 25 studies were identified and relevant information was compiled in a dataset for further analysis. These studies often report estimates for more than one GI, leading to a total sample size of 141 product-specific estimates. The sample was adjusted to exclude extreme outliers, yielding a final sample size of 134 observations collected from 22 papers. Table 2.1 lists each study, the food product involved, the broadly defined methodological approach of each study, as well as the number of GI estimates collected. (See Table 2.1) As in other meta-analysis studies involving valuation of labeled attributes (Ehmke, 2006; Lusk et al., 2005), estimates of the GI premia were normalized across articles as the percentage price (or valuation) difference between labeled and unlabeled products. Thus, to construct our dependent variable, we use the formula: 31 Price of GI Product- Price of Reference Product % Premium= *100 Price of Reference Product       . This specification normalizes the estimates across the different years, units of measure (i.e., kilograms, pounds, cc, etc) and currencies reported. It should be noted that several challenges emerged in compiling the data. In a study using an experimental design where a reference price was not given, (Groot et al., 2009), the median of the price treatments is used as reference price (following Lusk et al., 2005). Furthermore, many studies (more than 30% of our sample) reported only point estimates, and not the associated standard errors. Even for the cases in which some measure of the precision of the estimates was provided, we found them to be extremely heterogeneous 2 . Another limiting data issue regarded the demographics of the sample, and particularly income, which were either missing or reported inconsistently across studies (for example, “high” vs. “low” income instead of income categories or levels) 3 . While we acknowledge these limitations, the compiled dataset contains a wealth of information that does allow for some useful comparisons and analysis including: location and period covered by the study, type of GI scheme (PDO, PGI, GI-based trademarks or generic geographical references), sample size and type of data used in the original study (i.e. survey, experiment, scanner data, etc), and methodology used to 2 The metrics used included standard errors, t-statistics, exact p-values or cutoff p-values (e.g., 0.01, 0.05, and 0.1). While all these measurements could be transformed into a uniform variable, for 44 out of a total of 141 observations (31.2% of our sample size) no measurement of precision of the WTP estimate was reported. 3 Income was considered an important variable a priori since studies that include a larger proportion of more affluent consumers may have inflated willingness to pay estimates. 32 estimate the price premium (hedonic methods, contingent valuation, other) 4 . The valuation estimates were also categorized by broad food classes (cheese, meat, fruit, etc) and three super-categories based on the level of processing that the base agricultural commodity underwent (highly processed for cheese and wine; low/intermediate for olive oil, grain, coffee, meat; and fresh produce for fruits and vegetables). A final categorization was based on the perceived propensity for firm branding within each product market, which we consider as another important product differentiation mechanism. Wine and olive oil where characterized as markets in which brands are almost always present, while cheese and meat both may be branded or generic, and at least in this time frame, branding was more rare for grain, fresh fruits and vegetables. A description of the variables and their descriptive statistics is provided in Table 2.2. (See Table 2.2) The percentage premium for all GIs varies widely from a minimum of -36.73% for Provolone Valpadana Cheese (Galli, 2010) to +181.92% for Valle d’Aosta Fromadzo Cheese in Italy (Galli, 2010). The average percentage premium for GIs is 15.12% once extreme outliers 5 were removed. While the mean WTP is positive, indicating that consumers are generally willing to pay more for GI products, there is a great deal of variability in the reported premia(a estimated standard deviation of 35.5%). It should be 3 Methodologies coded as “other” include simple reporting of a price differential between the labeled product and an unlabeled substitute (Galli et al.), auctions/ bids (Stefani, 2005; Akaichi et al., 2009), random utility models (Botonaki et al., 2004), and contingent valuation methods (Skuras et al, 2002). 5 To reduce the effect of extreme (and perhaps suspicious) observations on our estimates, we eliminated 7 observations falling outside a +/- 2 standard deviations from the mean estimated percentage premium. (see Table 2.1 for excluded studies). One std.dev. in this sample is 38% and the mean is 21.3%, so estimates outside the -54% and +94% range were excluded. 33 also noted that the majority of studies in this sample (55%) are based on valuations by European consumers, followed by North and Central American studies (31%) and, finally, Australian and New Zealand studies (14%). Figure 2.3 shows the broad product categories represented in our sample by the GI scheme (PDO, PGI, or trademark). (See Figure 2.3) From a statistical viewpoint, it would be ideal to have all product categories represented within each GI-based quality assurance scheme, with similar frequencies. Instead, PDO-protected products are mostly cheese, followed by wine, olive oil, fruits and vegetables, and meat. The majority of PGI certified products in our sample are meats, followed by grains and olive oil; while GI trademarks are mostly used with wine products 6 (73%), and fruits and vegetables, such as Washington apples and Idaho potatoes. Comparing PDO and PGI product lists, it appears that, with the exception of fresh produce, the more processed products such as cheese, wine, and olive oil self-select into the more complex PDO quality assurance, while the less processed meats and grain products are mostly certified by the less onerous process associated with a PGI. 6 Wines are coded as trademarks when the original study specifies that they are produced in a specific American Viticultural Area (AVA) 34 4. Model and Estimation Methods The main advantage of meta-regression analysis is the ability to describe the variation existing in the selected studies (Stanley 2001), but there are still several options for model specification which depend on priors about what variables may explain the variation. We estimate three model specifications, the most descriptive of which (Model 1) takes the form: (1)                         ij 0 1 i 2 i 3 i 4 i 5 i 6 i 1 i 2 i 3 i 1 j 2 j 3 j ij %Premium = α +α Wine +α Cheese +α Meat +α Grain +α OliveOil +α FruitVeggie +β PDO + β PGI + β CertMark + +γ Primary Data + γ Conjoint + γ Hedonic + ε ; where ij %Premium indicates the i th estimated premium from the j th study. Thus, the general modeling framework assumes that the percentage WTP/price premium for GI certified food products depends on product/market specific characteristics (as captured by the alpha coefficients), the quality assurance scheme (beta coefficients), and a series of study-specific controls (gamma coefficients) accounting for the data and methods used in each original study. The reference categories for each set of dummy variables are respectively coffee, unregulated regional designations of origin, and studies using methods “other” than conjoint and hedonic analyses. Model 2 and 3 aim to abstract from specific product categories and investigate general product and market characteristics which may explain variations in GI premia. In Model 2 we replace the product category dummies with variables quantifying the level of processing, to obtain the specification: 35 (2)       ...ij 0 1 i 2 i 1 i ijPremium = α +α HighlyProcessed +α Fresh Produce + β PDO   In Model 3 we focus on the degree of firm branding observed for each product: (3)       ...ij 0 1 i 2 i 1 i ijPremium = α +α FullBrand +α MixedBrand + β PDO   Admittedly, these two ”umbrella” categories are somewhat collinear, as longer supply chains seem to be typical of markets in which brand names have developed. As it was not possible to directly include reliable measures of the variance of the estimates in our meta-analysis, our approach was to designate statistically insignificant estimates as zero. For the remaining estimates, we follow the approach of Lusk et al (2005) and use the sample size of the original study as a measure of precision. The argument is that, as long as a study employed a consistent estimator, we expect the variance to decrease as the sample size increases. Thus, all three models are first estimated via ordinary least squares and then by weighted least squares, where the weights are proportional to the sample size of each study. This implies that estimates of GI premia generated from a larger sample size will have a greater effect on our estimated coefficients than estimates coming from a smaller sample. Regarding the error term of our model, it seems reasonable to assume that the residuals are uncorrelated across studies, but some degree of correlation should be expected when premium estimates are obtained from the same study. As a cautionary measure, we use a robust (clustered on the individual study) estimator of the variance- 36 covariance matrix. Random and fixed effect (panel) models were also estimated. For the fixed effects model, the null hypothesis that all fixed effects are jointly equal to zero cannot be rejected with a joint F-stat (prob>F=0.943). For the random effects model, the null hypothesis that within-study variances are zero, tested with the Breusch-Pagan LM Test, cannot be rejected (prob>Chi 2 =0.218). This suggests that the weighted OLS regression estimation method may be appropriate. 5. Results Estimation results are reported in Table 2.4. Both un-weighted and weighted results are provided for Model 1, while Model 2 and 3 are presented only in the weighted version. As a robustness check, Model 1 was also estimated (via WLS) using only the data from Europe-based studies. For Model 1, the weighted model is superior to the un- weighted model in that it provides more precise estimates (lower standard errors), and overall model fit (R-squared increases from 0.241 to 0.666). Thus, we focus the discussion on the results estimated via WLS. (See Table 2.4) The first notable result is that GI labeling for grain, meat and fresh produce commands the highest price premium, 121.5%, 72% and 64%, respectively. Cheese follows with a percentage increase in premium of 43.5%. In contrast, the lowest 37 percentage price increase for GI labeling are associated with olive oil and wine, with 31% and 21.5% premia, respectively. All these estimates are statistically significant at the 1% level. It should be noted that, as average prices are quite different across product categories, this ranking of premia may change if they are considered in absolute monetary terms. However, we find the percentage representation preferable as it normalizes for differences in cost of production and added value. When only European studies are used in the estimation, the magnitude of the premia changes (and statistical significance is lost because of the smaller sample size), but the ordinal ranking is generally preserved (see Figure 2.5). (See Figure 2.5) Controlling for product-specific differences, a European product with a PDO certification commands a price premium 21% higher than one using a non-regulated regional name. In short, the PDO percentage premium is higher than the average PGI value, which aligns with our expectations, considering that the PDO certification process is more complex and requires a stronger connection between raw materials, stages of production, final product characteristics and the geographical area of production. While this ordinal ranking in premium for PDO and PGI certifications appears clear, little more can be said regarding the magnitude of the PGI premium since Table 2.3, shows that the point estimate for PGI certification is imprecise, with very large standard errors, weak significance and changing signs. 38 In the US, the presence of a GI trademark is associated with an even higher price premium than the PDO, 39%. This finding is worthy of further discussion given that the process surrounding these designations is relatively unregulated, which would suggest weaker quality assurance. Moreover, in terms of methodology, valuation methods such as conjoint analyses and hedonic models tend to generate higher premia estimates than the reference group of “other” methods, by an average of 54% and 64%, respectively. Results from Model 2 suggest that the categorization by level of processing is not informative with respect to cross-product differences in price premia observed in Model 1. GIs in fresh produce provide the largest premium (27.8%), but the processing intercept shifters have weak significance and most of the product-specific premia seem to transfer to the PDO and PGI estimates, which increase to 30.7% and 10%, respectively. Model 3 is slightly superior in fit (see adjusted R 2 ) to Model 2, and produces results that are more consistent with those obtained with the more product-driven Model 1. According to Model 3, the GI premium for fully branded products (wine and olive oil) is 34.5% lower than products not generally carrying a private label. Products that sometimes display brand names (meats, cheeses) also register a decrease in their price premium, albeit a smaller and insignificant one. 39 6. Discussion Findings from this study may provide an interesting survey of the field’s understanding of location-based price dynamics. Based on a meat-analysis, GIs constitute an effective differentiation instrument in food markets. However, the magnitude of the price premium associated with GIs varies rather significantly across products. Comparing high (percent) premium (grain, meats, fruits, vegetables and produce) and low premium products (wine, olive oil, cheese), a set of key differentiating characteristics emerge. (See Table 2.6) The prevalence of high GI premia seem to correspond to minimally processed foods with short supply chains, and a large number of atomistic, undifferentiated producers. In contrast, price premia are smaller when the products are processed, the supply chain is long and the firm brands are known to consumers. This result is in line with the theoretical prediction of Menapace and Moschini (2011) and Costanigro et al (2010). Given the nature and collinearity of the existing literature’s valuation studies (which is the data available for analysis), it is hard to determine which factor is the most critical in triggering some pricing power for affiliated agriculture and food producers. However, our results are consistent with the hypothesis that the extent and importance of firm branding is one of the most important factors. Indeed, the inversely proportional 40 relationship between the presence of firm branding in a product category and the price premium that GIs can capture is quite evident (see Figure 2.5), and robust to the type of consumers (rest of the world vs. European only). An interpretative framework for this finding is provided by Costanigro et al (2010) who found that, at parity of quality, shifting from cheap to expensive wines induces reputation premia to migrate from collective names (viticultural areas) to brand names (specific wineries). When interpreting this finding, one must consider the economic tenet of search costs: when buying cheap products (such as grains, fruits and vegetables), it may not be worth it for the consumer to critically differentiate across many individual producers. GIs are therefore the main product differentiation tool because they provide a simple categorization of the available choices. When purchasing more expensive products (such as wine and olive oil), the incentive to learn about differences in quality across brand names is more pronounced. Indeed, the quality of individual firms is likely more consistent than the quality of groups of producers, and therefore, firm reputations provide a better assurance of quality and consistency than GIs. This reasoning does not necessarily imply that GIs have little use in markets for expensive food products. As a matter of fact, the ubiquitous presence of denominations of origin in wine and cheese (see Figure 2.3) is a proof to the contrary. A possibility is that, for expensive food products, consumers may use GIs to narrow down the large choice set of competing firms to a specific group(s) of producers for which learning about individual firm differences is worth the time. Then, consumers can investigate the subset of selected brands (identified by the GI) more thoroughly, or invest in directly 41 experiencing a specific product. This hypothesis is worthy of future investigation, as it is not testable given the summary nature of the current analysis. The institutional framework regulating GIs and its effect on price premium is interesting to consider given its implications for marketing policies. In Europe, more stringent regulations for the PDO appear to secure a higher price premium than its less cogent quality assurance counterpart (PGI). Stricter regulations may signal increased benefits to consumers in the form of food safety, quality assurance, stronger cultural/ heritage connection, etc., prompting a higher willingness to pay for products that are more closely regulated. It is therefore surprising that the GI trademarks in the United States, representing a less stringent accreditation process than the PDO or PGI, command a premium (39%) higher than both the PGI and PDO. Even though the results is robust to alternative econometric specifications of the model (see Table 2.4), one caveat is that the product classes carrying PDO or PGI labels are much more heterogeneous than what we report for trademarks. Also, country-specific factors and sample demographic controls which could not be controlled for in the model (especially sample income), may make GI estimates across such diverse countries not directly comparable. In summary, our work confirms the work of Shapiro (1983) and Menapace and Moschini (2012) regarding the relationship between minimum quality standards, reputation price premia, and use of GI labels. In both articles, premia for high quality are modeled as (lagged) returns from investment in quality (see Figure 2.7). (See Figure 2.7) 42 Since reputations develop slowly over time, a price premium (above cost of production) is necessary to induce firms to produce at any quality level above the minimum standard imposed on all firms ( 1 0q in Figure 2.7, upper panel). The farther away a firm moves from the minimum standard in the quality spectrum, the longer it will take to build the reputation, and the larger the premium needs to be to work as an incentive for producing higher quality. The economic rationale for the lower reputation premium is that the presence of an additional label shortens the lag between producing at high quality and developing a corresponding reputation. In short, GI labels would benefit consumers by lowering the reputation costs for buying high quality food products. 7. Conclusions and future research Agricultural and food products have long been associated with unique quality attributes strongly associated with the agro-ecological characteristics and culinary traditions of their origin. GIs formalize this connection in the marketplace, typically leading to positive price premia. In this study, we investigate this market dynamic further by analyzing how price premia for GIs vary by product, regional designation, and intrinsic product characteristics. In terms of percentage price premium, agricultural produce and minimally processed foods benefit the most from GI differentiation. We interpret this finding in light of the fact that, in addition to GIs, products with valued 43 added characteristics and longer supply chains may use private brands to capture reputation premia. In other words, brands and GIs may play a similar role in product differentiation, and thus, be substitutes for each other. The institutional framework for the GI was found to matter: within the same country, quality assurance schemes with higher quality standards such as the PDO receive a higher premium than less stringent ones (PGI). Moreover, when multiple labeling schemes with different minimum quality standard coexist (as for PDOs and PGIs in Europe), the price premium associated with the labels is lower than when a single label is used (as for the GI trademark in the US). Our interpretation is that reputations for high quality are easier to achieve (and thereby less costly for the consumer) when multiple quality assurance schemes segment the quality spectrum. This analysis identified a number of possibilities for future research both from a consumer’s and producer’s perspective. As mentioned above, consumers may be using a GI label to narrow the set of choices when searching for certain (branded) types of food. We envision using experimental methods to test this hypothesis, varying the labels across products and labeling options. This may even provide information to retailers who continue to fine-tune their sourcing and point-of-purchase strategies in efforts to maintain market share among an increasingly diverse set of customers that seek attributes aligned to their specific preferences. In considering producer strategies and decisions, it would be interesting to explore what motivates or prevents a producer from using a GI available in their location, given that these designations seem to be an accessible way to differentiate their output and secure a premium. Another would be to formally evaluate GI use and branding in the 44 context of alternative product and advertising strategies by individual producers or regional producer associations. 45 8. Tables and Figures Table 2.1. Summary of GI valuation studies included in the final analysis: No. Authors Year Food Category Methods No. of Estimates 1 *Akaichi et al. 2009 Fruit-Veggie Other 1 2 Bombrun et al. 2003 Wine Hedonic 12 3 Bonnet et al. 2001 Cheese Other 1 4 Botonaki et al. 2004 Wine Other 1 5 Costanigro et al. 2009 Wine Hedonic 7 6 Fotopoulos et al. 2001 Olive Oil Conjoint 1 7 Fotopoulos et al. 2003 Fruit-Veggie Conjoint 2 8 **Galli et al. 2010 Cheese Other 31 9 *Groot et al. 2009 Fruit-Veggie Conjoint 1 10 Hassan et al. 2006 Cheese/ Meat Hedonic 2 11 Ittersum et al. 2007 Cheese/ Fruit- Veggie/ Meat Other 6 12 Loureiro et al. 2000 Meat Hedonic 6 13 McCluskey et al. 2007 Fruit-Veggie Conjoint 1 14 Menapace et al. 2011 Olive Oil Conjoint 3 15 Mesias et al. 2010 Meat Other 1 16 Mueller-Loose et al. 2011 Wine Hedonic 11 17 Oczkowski et al. 1994 Wine Hedonic 20 18 Quagrainie et al. 2003 Fruit-Veggie Other 5 19 Sanjuan-Lopez et al. 2009 Fruit-Veggie Hedonic 3 20 Santos et al. 2005 Olive Oil/ Cheese/ Wine Hedonic 9 21 Schamel et al. 2006 Wine Hedonic 6 22 *Skuras et al. 2002 Wine Other 1 23 Stefani et al. 2005 Grain Conjoint 3 24 Stefani et al. 2006 Grain/ Meat/ Fruit- Veggie Other 3 25 Teuber et al. 2010 Coffee Hedonic 4 *Excluded from final sample due to outlier estimates **Four estimates excluded from final sample due to outlier estimates 46 Table 2.2. Description of variables: Variable Description Mean St. Dev. Min Max WTP (%) Value of the product in percentage price premium (+/ - ) % 21.32 37.8 -36.73 181.9 WTP no outliers Observations lying outside +/- 2 standard deviations from the mean are excluded 15.12 26.13 -36.73 90.6 WINE Binary variable coded 1 if the product is in Wine Category, 0 otherwise 0.47 0.50 0 1 CHEESE Binary variable coded 1 if the product is in Cheese Category, 0 otherwise 0.24 0.43 0 1 COFFEE Binary variable coded 1 if the product is in Coffee Category, 0 otherwise 0.03 0.17 0 1 MEAT Binary variable coded 1 if the product is in Meat Category, 0 otherwise 0.07 0.25 0 1 FRUIT/VEGGIE Binary variable coded 1 if the product is in Fruit/Veggie Category, 0 otherwise 0.10 0.31 0 1 OLIVE OIL Binary variable coded 1 if the product is in Olive Oil Category, 0 otherwise 0.05 0.22 0 1 GRAIN Binary variable coded 1 if the product is in Grain Category, 0 otherwise 0.04 0.19 0 1 PDO Binary variable coded 1 if product is PDO, 0 otherwise 0.45 0.50 0 1 PGI Binary variable coded 1 if product is PGI, 0 otherwise 0.09 0.28 0 1 TRADEMARK Binary variable coded 1 if product is defined as a Trademark or AVA (for wines) in original paper, 0 otherwise 0.21 0.41 0 1 REGIONAL Binary variable coded 1 if product is regional (no specific geographic regulation), 0 otherwise 0.35 0.44 0 1 PRIMARY DATA Binary variable coded 1 if primary data, 0 if secondary data sources are used 0.18 0.38 0 1 CONJOINT Binary variable coded 1 if methodology is Conjoint, 0 otherwise 0.07 0.26 0 1 47 HEDONIC Binary variable coded 1 if methodology is Hedonic, 0 otherwise 0.60 0.49 0 1 OTHER Binary variable coded 1 if methodology is not Conjoint, Hedonic; 0 otherwise 0.33 0.47 0 1 LOW/INTERMEDI ATE PROCESSED Binary variable coded 1 if product involves low to intermediate processing, 0 otherwise(meat, grain, olive oil, coffee) 0.19 0.39 0 1 HIGHLY PROCESSED Binary variable coded 1 if product involves a high level of processing, 0 otherwise (cheese, wine) 0.71 0.45 0 1 FRESH PRODUCE Binary variable coded 1 if product is retailed fresh, 0 otherwise (fruit/ veggies) 0.10 0.31 0 1 FULL-BRAND Binary variable coded 1 if product is most likely to have a brand (wine, olive oil), 0 otherwise 0.52 0.50 0 1 MIXED-BRAND Binary variable coded 1 if product could have a brand (meat, cheese), 0 otherwise 0.31 0.46 0 1 NO BRAND Binary variable coded 1 if product most likely does not have a brand (fruit/veggie, grain, coffee), 0 otherwise 0.17 0.38 0 1 48 Figure 2.3. Product categories by quality assurance scheme 49 Table 2.4. Estimation Results a Variable Model 1 OLS all Model 1 WLS all Model 1 WLS Europe Model 2 WLS all Model 3 WLS all Wine 22.96* 21.57*** (12.17) (0.69) Cheese 26.6 43.48*** 19.59*** (16.47) (5.03) (2.48) Meat 32.26 72.03** 66.01** (19.68) (25.97) (21.95) Fruit/Veggie 24.88* 63.88*** 18.06 (14.83) (16.44) (19.21) Olive Oil 26.30 31.19*** 0.66 (16.74) (6.47) (2.54) Grain 51.76** 121.54*** 107.33*** (21.80) (22.12) (17.72) Full Brand -34.49* (17.09) Mixed Brand -14.02 (17.01) Highly Processed -3.09 (10.32) Fresh Produce 27.76 (18.15) 50 PDO 12.03* 20.69*** 8.58*** 30.69*** 21.91*** (6.63) (4.13) (1.78) (7.96) (3.53) PGI 5.77 -37.23 -69.07*** 10.29 -7.65 (14.89) (25.41) (20.48) (12.78) (4.62) Trademark 35.05*** 39.01*** 39.08*** 39.56*** (6.11) (0.92) (0.93) (1.03) Primary Data -10.05 -1.28 -0.99 -0.95 1.82 (9.83) (9.65) (10.55) (9.36) (11.07) Conjoint 17.57 53.75*** 60.41*** 44.67*** 58.29*** (13.64) (15.87) (18.02) (15.37) (15.94) Hedonic 1.43 63.78*** 65.36*** 51.68*** 62.65*** (10.18) (3.5) (2.46) (7.98) (4.20) Constant -23.45 -85.81*** -50.28*** - 49.05*** -29.07* (15.28) (3.5) (2.64) (15.42) (17.02) Adjusted-R2 0.241 0.666 0.814 0.636 0.656 F-stat 4.51 - 319.4 344.58 330.3 (0.000) - (0.000) (0.000) (0.000) Obs. 134 134 71 134 134 a: robust clustered SE in parentheses, ***significant at 1%, **significant at 5%, *significant at 10% 51 Figure 2.5. Price premia across product groups (comparison between all data and European data) 52 Table 2.6. Product Characteristics influencing GI price premium Characteristic High Percent Premium Low Percent Premium Product Grain, fruits, vegetables, agricultural produce Wine, olive oil, cheese Length of Supply Chain Short Long Numbers of Producers More (farmers) Less (Food Industry) Brand Names Generally No Generally Yes Processing level Generally Low Generally High Product/ Quality Differentiation Lower, depends on product variety cultivar Higher, depends on food processor 53 Source: adapted and simplified from Shapiro (1983) and Menapace and Moschini (2012) Figure 2.7. Equilibrium reputation premium  Pr for producing at quality level  1q with single ( 1 0q , upper panel) and double ( 1 0q and 2 0q , lower panel) minimum quality standards.  C q represent cost of production,  P q is the market price. 54 9. References Akaichi, F., Gil, J., 2009. Factors Explaining the Consumers ' Willingness-to-Pay for a ”Typical" Food Product in Spain: Evidence from Experimental Auctions. Contributing paper to the International Association of Agriculture Economists Conference in Beijing, China. Ali H.H., Nauges C., 2007. The Pricing of Experience Goods: The Example of en primeur Wine. American Journal of Agricultural Economics, 89(1):91-103. Blasquez, J. M., 1992. The Latest Work on the Export of Baetican Olive Oil to Rome and the Army. Greece & Rome, Second Series, 39(2), 173-188. Bombrun, H., Sumner D. A. 2003. What determines the price of wine, AIC Issues Brief; 18(18). Bonnet, C., Simioni M., 2001. Assessing consumer response to protected Designation of Origin labeling: a mixed multinomial logit approach. European Review of Agricultural Economics, Vol. 28 (4), 433-449. Botonaki A, Tsakiridou E., 2004. Consumer response evaluation of a Greek quality wine. Acta Agriculturae Scandinavica, Section C - Economy; 1(2):91-98. Bramley, C., Biénabe, E., and Kirsten, J., 2009. The Economics of Geographical Indications: Towards a Conceptual Framework for Geographical Indication Research in Developing Countries. In: The Economics of Intellectual Property. Suggestions for Further Research in Developing Countries and Countries with Economies in Transition; Geneva, 109-141. Brannick, M. T., Yang, L., and Cafri, G. 2008. Comparison of Weights in Meta-analysis Under Realistic Conditions. Poster presented at the 23 rd annual conference of the Society for Industrial and Organizational Psychology, San Francisco, CA. Broude, T., 2005. Taking “Trade and Culture” Seriously: Geographical Indications and Cultural Protection in the WTO. Journal of International Economic Law, 26(4), 633-692. Centner T.J., Turner S.C., Bryan JT., 1989. Product differentiation protection: developing a strategy for multiple producers of regional specialty crops. Journal of Food Distribution Research; 20(2):13–9. Combris P., Lecocq S., Visser M., 1997. Estimation of a hedonic price equation for Bordeaux wine: does quality matter? The Economic Journal, 107(441):390–402. Costanigro M., Mittelhammer R., McCluskey J., 2009. Estimating Class-specific Parametric Models under Class Uncertainty: Local Polynomial Regression 55 Clustering in a Hedonic Analysis of Wine Markets. Journal of Applied Econometrics, 24:1117-1135. Costanigro, M, McCluskey, J. J., Goemans, C., 2010. The Economics of Nested Names: Name Specificity, Reputations, and Price Premia. American Journal of Agricultural Economics 58 (3), 454–466. Das, K., 2006. International Protection of Geographical Indications with Special Reference to Darjeeling Tea” The Journal of World Intellectual Property, 9(5): 459-495. Deppeler, A., Stamm, H., Thevenod-Mottet, 2011. Evaluating the Effects of Protecting Geographical Indications: scientific context and case studies. In Swiss Federal Institute of Intellectual Property (Eds.): The Effects of Protecting Geographical Indications, Ways and Means of their Evaluation, Publication No. 7; second revised edition, 09.11. Available at: http://www.ige.ch. Ehmke, M., 2006, International Differences in Consumer Preferences for Food Country- of-Origin: A Meta-Analysis. Selected paper at the Annual Meeting of the American Agricultural Economics Association, Long Beach, California, July, 23- 26. Available at: http://ageconsearch.umn.edu/bitstream/123456789/14638/1/sp06eh01.pdf. European Commission., 2007. European Policy for Quality Agricultural Products. Available at: http://ec.europa.eu/agriculture/publi/fact/quality/2007_en.pdf. European Union Directorate General for Agriculture and Rural Development (EUDG), 2011. Background note: 1000th quality food name registered. MEMO/11/84. Brussels, 15 February. Available online at: http://ec.europa.eu/agriculture/quality/schemes/index_en.htm (consulted on 25/04/2011). EEC Council, 1992. COUNCIL REGULATION (EEC) No 2081/92, on the protection of geographical indications and designations of origin for agricultural products and foodstuffs of 14 July 1992. Official Journal of the European Community L 208: 1- 8. Feinstein A. R. 1995. Meta-analysis, statistical alchemy for the 21st century. Journal of Clinical Epidemiology, 48:71-79. Fotopoulos, C., Krystallis, A., 2001. Are Quality Labels a Real Marketing Advantage? A conjoint application on Greek PDO Protected Olive Oil. Journal of International Food & Agribusiness Marketing, Vol. 12, Issue 1, 1-22. Fotopoulos, C., Krystallis A., 2003. Quality labels as a marketing advantage: the case of http://ec.europa.eu/agriculture/quality/schemes/index_en.htm 56 the 'PDO Zagora' apples in the Greek market. European Journal of Marketing, Vol. 37, no.10: 1350-1374. Galli F., Carbone A., Caswell J.A., Sorrentino A. 2010. A Multi-Criteria Analysis for the Assessment of the Performance of Italian PDO Cheeses, 22-25. Gautam, K. and Bahl, N., 2010. Geographical Indications of India, Socio-Economic and Development issues, AIACA Policy Brief, October 2010. Available at: http://www.aiacaonline.org/pdf/policy-briefs-geographical-indications-india- socio-economic-development-issues.pdf. Gehlhar M., Coyle W., 2001. Global Food Consumption and Impacts on Trade Patterns, in Changing Structure of Global Food Consumption and Trade. Agriculture and Trade Report No. WRS-01-1, U.S. Department of Agriculture, Economic Research Service, May. Groot, E, Luis A., 2009. MaxDiff Approaches for PDO "Calanda" Peaches (Spain). 113th EAAE Seminar, Sept 3-6. Hassan, D., Monier-Dilhan S., 2006. National Brands and Store Brands: Competition through Public Quality Labels. Agribusiness 22, no. 1, 21-30. Hedges, L. V., and Vevea, J. L. 1998. Fixed- and random-effects models in meta- analysis. Psychological Methods, 3, 486-504. Herrmann, R., Teuber, R., 2010. Geographical Differentiated Products. In: Lusk, J., J. Rosen and J. Shogren (eds), Oxford Handbook on the Economics of Food Consumption and Policy. Oxford University Press, Chapter 36. Oxford, UK. Hunter, J. E., Schmidt, F. L. 1996. Cumulative research knowledge and social policy formulation: The critical role of meta-analysis. Psychology, Public Policy, and Law, 2(2): 324-347. Ilbery, B., Kneafseym M., 2000. Registering Regional Specialty Food and Drink Products in the United Kingdom: The case of PDOs and PGIs’, Area, 32(3): 31. Ittersum (van), K., Meulenberg, Matthew T. G., van Trijp, Hans C. M., Candel, Math J. J. M., 2007. Consumers' appreciation of Regional certification labels: A pan- European Study. Journal of Agricultural Economics 58, no. 1, 1-23. Josling, T., 2006. The War on Terroir: Geographic Indications and Transatlantic Trade Conflict. Journal of Agricultural Economics, 57(3), 337-363. Loureiro, M. L., McCluskey, J. J., 2000. Assessing Consumer Response to Protected Geographical Identification Labeling. Agribusiness 16(3), 309-320. http://www.aiacaonline.org/pdf/policy-briefs-geographical-indications-india-socio-economic-development-issues.pdf http://www.aiacaonline.org/pdf/policy-briefs-geographical-indications-india-socio-economic-development-issues.pdf 57 Levidow, L., Bijman, J., 2002. Farm Inputs under Pressure from the European Food Industry. Food Policy 27(1), 31-45. Lusk, J.L., Jamal, M., Kurlander L., Roucan, M., Taulman L., 2005. “A Meta-analysis of Genetically Modified Food Valuation Studies”, Journal of Agriculture and Resource Economics 30(1), 28-44. McCluskey J.J., Mittelhammer R.C., Marin A.B., Wright K.S., 2007. Effect of Quality Characteristics on Consumersʼ Willingness to Pay for Gala Apples. Canadian Journal of Agricultural Economics/Revue canadienne dʼagroeconomie, 55(2): 217-231. Menapace, L., Colson, G., Grebitus, C., Facendola, M, 2011. Consumers' preferences for Geographical Origin Labels: evidence from the Canadian Olive Oil Market. European Review of Agricultural Economics, 1-20. Menapace, L., Moschini, G. 2012. Quality Certification by Geographical Indications, Trademarks and Firm Reputation Department of Economics. European Review of Agricultural Economics. Forthcoming. Online version available at: http://erae.oxfordjournals.org/content/early/2011/11/17/erae.jbr053.abstract Mesías F.J., Gaspar P., Pulido F. 2010. “The Role of Protected Designation of Origin in Consumer Preference for Iberian Dry-cured Ham in Spain”. Italian Journal of Food Science, 22(4):367-377. Mtimet, N., Albisu, L. M., 2006. Spanish wine consumer behavior: A choice experiment approach. Agribusiness, Vol. 22, Issue 3, 343-362. National Research Council 1992. Combining information: Statistical issues and opportunities for research (pp. 5-15). Washington, DC: National Academy Press. O’Connor, B., 2004, reprinted 2007. The Law of Geographical Indications” Cameron May Ltd. Publishing, 132. Okzkowski, E.,1994. A Hedonic Price Function for Australian Premium Table Wine. Australian Journal of Agricultural Economics, Vol. 38, no.1, 93-110. Rosenberger, R.S., and Johnston, R.J. 2009. Selection Effects in Meta-Analysis and Benefit Transfer: Avoiding Unintended Consequences, Land Economics, August 1, 85(3): 410-428. Rosenthal, R. 1979. The