Credence good labeling: the efficiency and distributional implications of several policy approaches.

Increasingly, consumer goods are differentiated by process-attributes, e.g., organically produced food, dolphin-safe tuna, free-range poultry, genetically modified organism (GMO)-free food, low-emissions electricity, irradiated food, etc., as well as by use-attributes, e.g., taste, texture, performance. Important implications may arise for various sectors of agriculture, for the environment, and for international trade as consumers shift purchases among goods produced by different methods; for example, many importing countries having chosen to adopt mandatory labeling of foods that contain GMOs (Sheldon 2002; Lapan and Moschini 2004). However, many of these newly demanded process-attributes are not correlated with end-use attributes and, hence, an asymmetric information problem ensues: consumers cannot verify process-attribute claims, even after lengthy inspection or consumption of the good. Goods that suffer such ex post information asymmetries are a simplified version of credence goods (Darby and Karni 1973) and several authors have forwarded formal models and analysis on the topic, e.g., Dulleck and Kerschbamer (2006); Emons (1997, 2001); Feddersen and Gilligan (2001); Fong (2005); Pitchik and Schotter (1987); Taylor (1995); Wolinsky (1993, 1995). (1)

In practice, one method for addressing the credence good problem is the use of labeling. Credence good labeling in a particular market, however, requires a series of practical decisions concerning implementation. First, there is the choice between discrete and continuous labeling: specifically, a label either communicates that a good meets a certain quality threshold (the tuna-fish is dolphin safe) or the exact level of quality being produced (electricity costs for this appliance are $55 under normal operating conditions). Second, there is a choice between certification and labeling under the authority of a government agency (the U.S. Food and Drug Administration) or through a private firm, (Consumer Reports). Third, if government labeling is mandated, a choice has to be made between whether or not to allow private certifiers to further communicate quality differences. For example, electricity providers in 23 U.S. states must disclose information concerning the environmental profile of energy sources but they may also seek private certifications such as the Center for Resource Solution's Green-E[R] renewable electricity certification. In contrast, all organic certification in the United States is now overseen by federal or state government entities and private firms may not establish alternative definitions of organic.

In this article we analyze how these practical decisions impact the size and distribution of surplus created in a market featuring goods differentiated along a single, vertically differentiated credence dimension. The underlying quality of process attributes, and credence goods more generally, is often vertically aligned; i.e., if different goods were offered at marginal cost all consumers would prefer the same high-quality item. Quality has often been modeled in the product differentiation literature as horizontal or spatial differentiation where not all consumers agree upon the ranking of differentiated goods offered at marginal cost, e.g., Dixit and Stiglitz (1977).

Our modeling efforts build on the vertical product differentiation model of Shaked and Sutton (1982, 1983), which in turn draws on the earlier work of Gabszewicz and Thisse (1979, 1980). We show that, so long as certification is not too expensive, firms will hire private certifiers to label goods that surpass a discrete quality threshold rather than participate in government certification programs, even when private certifiers have no cost advantage over government certifiers. This result is driven by the assumption that discrete labeling is less expensive than continuous labeling and by the assumption that private certifiers validate firm-chosen levels of quality while public certifiers may validate quality levels that are sub-optimal for firms.

We show that, on average, consumers gain from private labeling, as the alternative in our game structure is a monopoly involving the lowest quality good. The average consumer would prefer government labeling with no opportunity for additional private labeling, but only if it is a discrete label featuring a quality threshold higher than that chosen by firms and so long as it does not drive the high-quality firm from the market. If the government imposes mandatory continuous labeling, the firms' best response is to hire no private certification. However, if the government imposes mandatory discrete labeling and chooses a quality standard quite different from the firm's optimal quality level, the firm may hire a private firm to certify its preferred quality level.

Several articles in the agricultural economics literature address the issue of labeling and regulation in credence good markets. Marette, Crespi, and Schiavina (1999) show that a cartel providing information through a common certified labeling scheme increases welfare if labeling costs are high, even if producers collude to reduce competition, while Marette, Bureau, and Gozlan (2000) show in a simple monopoly setting that, unless the fixed costs of safety effort are high, both minimum standards and certified labeling can resolve the credence good problem in the case of food safety. McCluskey (2000) demonstrates that credence good markets with probabilistically accurate certification are transformed into experience good markets when consumers engage in repeat purchases, while McCluskey and Loureiro (2005) demonstrate that improved monitoring in a credence good market can improve quality. Crespi and Marette (2001) investigate how the method of funding of credence good certification can affect social welfare. In addition, other articles in the literature draw inferences concerning policies governing credence attributes of foods. Giannakas and Fulton (2002), and Fulton and Giannakas (2004) show that the relative welfare ranking of labeling versus no labeling of GMOs depends on factors such as consumer aversion to GMOs and the level of segregation costs under mandatory labeling. Moschini and Lapan (2005) show that mandatory labeling of GMOs is an economically wasteful regulation as suppliers of GMO-free food have to undertake costly segregation costs. Zago and Pick (2004) show that the labeling of credence characteristics such as a good's region of production can leave consumers and high-quality producers better off while harming low-quality producers.

With the exception of McCluskey (2000), and McCluskey and Loureiro (2005), all of the above articles draw upon a model of vertical product differentiation introduced by Mussa and Rosen (1978), where the number of goods and the level of quality of each good are exogenously imposed as part of the analysis. In contrast, by using Shaked and Sutton's (1982, 1983) model of vertical product differentiation, we can attack several issues not previously analyzed in the literature on credence good labeling because we endogenize firms' choice of both entry and level of quality production. Specifically, we explicitly allow for a continuum of quality levels rather than only two levels of quality that are exogenous to the firms. Such considerations are nontrivial in practice; the absolute level of quality has been the focus of intense debate in areas such as the labeling of GMO and organic products, while firms constantly alter attributes such as energy efficiency on appliances as new technology becomes available.

Furthermore, we analyze rigorously the difference between governmental and nongovernmental labeling regimes (see also Crespi and Marette 2001; Segerson 1999). In addition, we examine other key labeling institutional details (continuous versus discrete, voluntary versus mandatory, exclusive versus nonexclusive government labeling). Such practical matters of implementation may have great impacts on the market. For example, a key point of contention surrounding organic labeling among some in the organic production community was USDA's appropriation of the term organic, which effectively precluded private firms from developing and certifying standards higher than those that emerged from federal or state rule-making process. (2)

Our model is closest to Boom (1995), Scarpa (1998), Lutz (2000), and Lutz, Lyon, and Maxwell (2000). (3) These authors introduce the related concept of minimum quality standards into models of vertical product differentiation, but none consider the possibility of imperfect information and product quality in the presence of credence goods. In addition, our use of the term quality standard differs from the term minimum quality standard as used in these articles in that not all firms are required to produce at or above the standard; rather firms must do so to receive the discrete label.

The remainder of the article is structured as follows. We introduce the basic model in the next section and discuss possible labeling regimes. We then derive a solution to the model under the case of perfect information about quality. Next we consider the credence good case under five different labeling regimes and derive several propositions concerning the welfare and distributional implications of the various labeling interventions. We end with concluding remarks and discussion of possible results under less restrictive assumptions.

The Model