Vertical differentiation via multi-tier geographical indications and the consumer perception of quality: The case of Chianti wines
Introduction
In 2013 the Consortium of wine producers in the Chianti Classico Geographical Indication (GI henceforth) agreed to introduce the “Gran Selezione” (Great Selection), a quality certification label, to promote an “upward expansion of the oenological offer of the Chianti territory”.1
As worthwhile as this endeavour may seem, the move raised controversy among the members of the Consortium, as a Decanter article reported (Brook, 2014). After all, designations of origin tied to the Tuscan region already featured multiple tiers of quality labels, including Chianti, Chianti Classico, and Chianti Classico Riserva. Would the new label bring real value to Chianti producers or just increase bureaucratic costs? Would the Gran Selezione benefit some producers while damaging others?
Having established a large body of theoretical and empirical work on minimum quality standards and GIs,2 economists are well set to address such questions. Briefly, in markets for experience (Nelson, 1970) and credence (Darby and Karni, 1973) goods, quality labels can increase the number of consumers’ choices and, at the same time, allow producers to price above marginal cost (Zago and Pick, 2004). The empirical task, therefore, is to determine the share of consumers willing to pay for the higher quality label, evaluate substitution patterns, and consider costs.3
As it often turns out however, not all questions raised by the winemakers are so straightforward. Giovanni Poggiali’s of Felsina Winery for example was concerned that “if we present our single-vineyard Rancia as Gran Selezione, then consumers will assume our other top wines such as our pure-Sangiovese Fontalloro are not as good” (Brook, 2014). Mr. Poggiali’s thought points to two important concepts typically not considered in the GI literature: first, that even though quality standards linked to Protected Designation of Origin (PDO) are an objectively-defined set of practices, consumers will interpret the label subjectively. Second, and perhaps more crucially, that perceived quality is a contextual and comparative concept—production processes and sensory characteristics may remain unaltered, but the Chianti Classicos not making the cut for the Gran Selezione may lose prestige, just like the release of a “new and improved” phone cheapens the feel of the one we hold in our pocket.
The objective of this article is to establish a theoretical framework to study vertical differentiation via multi-tier GI quality certifications, test whether restructuring of perceptions effects of the type suggested by Mr. Poggiali can be detected and quantified, and understand the pros and cons of the strategy adopted by the Chianti Classico consortium. As the question is fundamentally empirical, our approach relies on an online experiment where consumers’ quality perceptions and product choices are elicited under different labeling scenarios. The type of heuristics we have in mind are similar to those considered in the product line and umbrella branding literature, where upward or downward brand extensions affect consumers’ perceptions of all the products marketed by the same brand (Chintagunta, 1996, Heath et al., 2011). The introduction of higher quality products is generally thought to increase brand equity (Randall et al., 1998) and possibly market power (Kadiyali et al., 1998), but it is also possible for the strategy to backfire and damage the differentiating brand (Caldieraro et al., 2015).
Chianti producers have been pursuing a strategy of quality standards and product differentiation for centuries4; and this is no hyperbole. The Lega del Chianti was founded in Florence in the thirteenth century to regulate administrative relations with the leading producers of a red wine made with Sangiovese grapes from the Chianti region, and the first notarial document in which the name Chianti is used to refer to wines produced in that region dates to 1398. In 1716 the Grand Duke Cosimo III de’ Medici issued a decree in Florence specifying the boundaries of the areas in which Chianti wines could be produced, and set up a Congregation to oversee the production, shipping, fraud-control and marketing of wine.
As the popularity of Chianti wines increased through the centuries, the acreage and region of production expanded beyond the traditional boundaries, but in 1932 the modern Consortium protecting the authenticity of Chianti wines established the Chianti Classico label and its distinctive red rooster trademark to identify the wines produced within the historical (1716) region, differentiating them from the more generic wines produced in the broader Chianti. In 1984, Chianti obtained the DOCG designation (Denominazione di Origine Controllata e Garantita - Denomination of Controlled and Guaranteed Origin), which is awarded to wines of certified origin with codified production processes and guaranteed wine quality; while in 1996 the Chianti Classico became a DOCG independent of the broader Chianti DOCG.
Table 1 portrays the main production differences between the two denominations. Chianti Classico wines with higher alcohol content (>12.5%) with at least two years of aging may be further qualified as Reserve (Chianti Classico Riserva). The “Gran Selezione” label (Great Selection) identifies a limited number of wines meeting very stringent quality standards (including approval by a tasting commission). It is important to note that this type of multi-tier labelling strategy is not unique to Chianti. Rosso di Montalcino and Brunello di Montalcino for example are wines produced exclusively in the territory of the Municipality of Montalcino, but quality standards are much more stringent for Brunello than for Rosso, so that the two wines are sold at very different price points (see Table 1).
The theory of vertical differentiation via GI (Zago and Pick, 2004, Moschini et al., 2008, Menapace and Moschini, 2011, Mérel and Sexton, 2012) generally assumes that certification of production processes communicates a certain quality level (e.g. high vs. low), and the interpretation of a signal is not affected by the presence/absence of other labels. This is reasonable when each region certifies a single label, segmenting the market into unlabeled vs. GI product, but less so for the case of the multi-tier strategy adopted in the Chianti and other viticultural areas (e.g. Montalcino, Bordeaux). A contribution of this article is therefore to examine how multi-tier quality labels affect the distribution of the quality marketed under a given GI, and then model how consumers’ perceptions may change in response.
The context and modeling assumptions are tied to the regulations surrounding the European Protected Designations of Origin, where a causal link between labeled production practices and product quality is established (EU 1151/2012).5 In addition to informing consumers and supporting a diversified food production, the stated objectives of European regulations include guaranteeing the profitability for producers of traditional agricultural products as a rural development strategy (Gragnani, 2013). The analysis of consumer perceptions and choices we developed here allows us to offer some considerations on the conditions under which multi-tier GI certifications can promote such objectives, but also raise some caveats related to distributional effects.
The first hypothesis we raise regarding consumer perceptions follows the standard assumption that consumers are both rational and well informed about the certification standards, so that perceptions match the expected quality marketed under each label (e.g. Zago and Pick, 2004). The second and third hypotheses waive rational expectations to integrate Steenkamp’ (1990) model of quality perceptions, positing that the interpretation of food labels results from a subject-object interaction, and is “neither completely subjective nor wholly objective (p.312)”. In this context food labels and production practices are essentially quality cues which consumers use to infer important quality traits (Messer et al., 2017) according to an individual’s knowledge and beliefs.
Even though wine consumption can certainly inform quality perceptions, the signaling role of labels remains central for two main reasons. First, in a market with a vast number of differentiated products (such as wine), the number of consumers having direct experience is generally small compared to first-time, uninformed buyers. Second, and perhaps more importantly, ample evidence shows that labels and other signals can influence (or bias) quality perceptions, even after consumption. For example, information about the high price of a wine not only makes people express higher quality ratings after tasting (as in Almenberg and Dreber, 2011), but has been shown to influence the very neural processes activated by pleasant experiences (Plassmann et al., 2008). Similarly, identical cookies seem to taste better when labeled as organic (Lee et al., 2013).
As Verdú et al. point out (2004), “[consumers] are not experts and probably do not know anything about the processes involved in the production and ageing of wines”. Teuber (2011) finds that, even among the very consumers of a GI product (Hessian apple cider) “consumers’ awareness and knowledge about GI is very limited”. One can therefore hypothesize that a Gran Selezione may be interpreted as a signal of higher quality, but consumers unfamiliar with GI regulations may naïvely interpret the label as a newly developed high-quality product (rather than, at least in part, a reclassification of existing wines). Following this logic, a testable hypothesis is that perceptions of other Chianti wines may be independent of the presence/absence of the Gran Selezione. Lastly, if we accept that quality perceptions are inherently subjective, then systematic and heuristic biases (see for example Kahneman et al., 1991) may play an important role. The “comparative stigma” idea raised by Mr. Poggiali follows this logic, suggesting that the presence of a Gran Selezione may cheapen the perceived quality of all other wines marketed under the Chianti umbrella, and is the third hypothesis we test.
Having relaxed the assumption that consumers’ perceptions necessarily match the expected quality output under each labeling standard, we derive a model of demand for experience goods based on the Mussa and Rosen (1978) framework, expanding the GI literature to consider the case of multi-tier certifications with quality perceptions endogenous to the labeling regime. The model is used to show how changes in market shares following the introduction of a new label can be parsed into two separate effects: choice availability, linked to the fraction of consumers who prefer higher quality, and perception restructuring, whose effects cannot be signed a priori. Then, we use data from a choice experiment to estimate the model, predict market shares, and quantify each effect.
The importance of distinguishing preferences from beliefs/perceptions to understand food choices has also been raised by recent experimental work on food labels (Lusk et al., 2013, Costanigro et al., 2015), even though not in the context of GIs. This distinction may seem purely academic to some: whatever the cause, an ex-ante investigation (e.g. via lab or choice experiment) of the share of consumers who will buy the Gran Selezione, and the implied substitution patterns (i.e. cannibalization of other Chianti wines vs. market gains against competitors), would provide all the information wine makers need to support it or oppose it. This view, however, misses some important implications.
The overarching issue is the identification problem raised by Manski (2004): if observed choices are consistent with multiple combinations of subjective beliefs (in this case, about product quality) and preferences (i.e. WTP for quality), then identifying preferences from choice data requires knowing or observing consumers’ beliefs. Were the new label unsuccessful, the consortium would want to know if consumers are just happy with the quality of a lower tier Chianti Classico (suggestive of market saturation), or because they don’t think that a Gran Selezione will differ that much from a Classico (implying a need for further differentiation or consumer messaging), and this cannot be ascertained by simply observing or estimating market shares.
To measure how multi-tier labels influence quality perceptions we designed four between-subject scenarios with an increasing number of competing quality standards, and then asked participants to rate product quality of each GI label via Likert scales. The use of psychometric or other subjective (Likert-type) constructs to explain WTP for labels has been proposed before (Costanigro et al., 2014a, Costanigro et al., 2014b, Malone and Lusk, 2017b) and has some clearly appealing features. For one, Likert scales are easily understood by survey participants, even in an online setting, and it is no coincidence that wine magazines (Wine Spectator, Wine Enthusiast, Decanter etc.) and consumer review websites (e.g. Vivino.com) use a similar system to report quality ratings (i.e. 1–100 scores or 1–5 stars). The representation of quality differences on a linear spectrum is also consistent with the vast body of theory on asymmetric information, from Akerlof, 1970, Mussa and Rosen, 1978, to the work on minimum quality standards (Leland, 1979, Shapiro, 1983, Bockstael, 1984, Boom, 1995) and GI labeling (Winfree and McCluskey, 2005, Moschini et al., 2008, Menapace and Moschini, 2011).
One significant drawback of Likert scales, however, is that they are notoriously susceptible to framing effects. Here, we borrow a remedy from the multiple price list experimental literature (see Andersen et al., 2006) by randomly assigning participants to two different types of elicitation scales, controlling econometrically for framing effects. The use of stated (rather than revealed) preference methods also comes with the standard caveats (see for example Brooks and Lusk, 2010), but choice experiments are particularly well suited to investigate the research questions examined here. Just like vertical differentiation models abstract from quantity (assuming that consumers purchase only one product) to study price-quality relationships, choice experiments require participants to select one preferred product within a choice set, isolating the tradeoffs central to a model of vertical differentiation. Indeed, the close analogy between the theoretical and empirical model is a prominent strength of the approach we propose. With measurements of quality perceptions in hand, one can estimate a choice model where product quality appears as a right-hand-side variable, as it is customary in virtually all vertical differentiation models.
While the advantages of controlling for quality perceptions are substantive, it is also true that the use of subjective constructs as regressors begs the question of endogeneity, and the design of a careful instrumentation strategy. After all, perceptions are by definition idiosyncratic, and therefore potentially correlated with the residual. Acknowledging this problem, we implement a two-pronged instrumentation strategy. First, a series of subject-specific, randomized information treatments designed to induce exogenous variation in quality perceptions, a methodology similar to the one adopted by Teisl and Roe (2010) to control for endogenous beliefs while estimating WTP to avoid foodborne pathogens. Second, we use perceptions relative to one label as instruments for another, following the strategy proposed by Gergaud et al. (2017, for endogenous collective reputations) and Malone and Lusk (2017a, for beer quality perceptions). A model of choice with subjective perceptions and heterogeneous (random) preferences for quality is then estimated adopting a control function approach (Petrin and Train, 2010) to correct for endogeneity. Estimated parameters are then used to predict market shares before and after the introduction of the new quality tier, while separating the effects of perception restructuring from choice availability effects induced by the introduction of the Gran Selezione. We conclude by examining the policy implications of the multi-tier differentiation strategy from the perspective of a GI vs. the individual firm.
Section snippets
Conceptual model and theoretical framework
In this section we develop a formal model of how multi-tier labels tied to the adoption of increasingly stringent quality standards may affect product quality, consumer perceptions and purchasing choices. The general framework draws from Moschini et al. (2008), with some major modifications. On the supply side, we consider the more realistic case where some producers surpass the minimum quality standards imposed by the GI regulation, so that a label is tied to a distribution of quality, rather
Experimental design and survey description
To estimate the model in Eq. (8), we conducted an online choice experiment where both quality perceptions and product choices were elicited. The experimental design was first obtained for a scenario including a full palette of differentiated Chianti wines (C, CC, CCR CCGS), a competitor wine (either Rosso di Montalcino DOC or Brunello di Montalcino DOCG, at random), and a “none of the above” option. The Montalcino wines are ideal competitors, as they also are Sangiovese-based Tuscan wines with
Empirical estimation and results
Table 2 presents the average quality perceptions (elicited with the tools presented in Fig. 2 and projected to a 0–100 numerical scale) across labeling regime. A first result is that there is little doubt that quality perceptions for Chianti wines change across scenario. A nonparametric k-sample test for equality of medians7 strongly
Conclusions, policy implications, limitations and future resarch
Motivated by the case of the Chianti Classico wine consortium strategy, we study vertical differentiation via multi-tier quality certification labels tied to a Geographical Indication (GI). First, we model how multi-tier labeling affects the quality output of a GI, and empirically test three competing hypotheses describing how consumer perceptions are altered by the introduction of a new quality tier. The fundamental point we raise is that consumers may interpret quality signals (including food
Acknowledgments
we are thankful to Alessndro Bonanno, Fabio Boncinelli and Caterina Contini for their useful feedback and comments. We also acknowledge the contribution of three anonymous referees who gave insightful critique of an early version of the article.
References (64)
- et al.
Consumers wine preferences according to purchase occasion: personal consumption and gift-giving
Food Qual. Prefer.
(2019) - et al.
The wine headache: consumer perceptions of sulfites and willingness to pay for non-sulfited wines
Food Qual. Prefer.
(2014) - et al.
Is it love for local/organic or hate for conventional? Asymmetric effects of information and taste on label preferences in an experimental auction
Food Qual. Prefer.
(2014) - et al.
The signaling effect of mandatory labels on genetically engineered food
Food Pol.
(2014) - et al.
Evaluating the net benefits of collective reputation: the case of bordeaux wine
Food Pol.
(2017) - et al.
Consumer knowledge affects valuation of product attributes: experimental results for wine
J. Behav. Experim. Econ.
(2016) - et al.
Product line extensions and competitive market interactions: an empirical analysis
J. Econ.
(1998) - et al.
You taste what you see: do organic labels bias taste perceptions?
Food Qual. Prefer.
(2013) - et al.
The cost of making wine: a tuscan case study based on a full cost approach
Wine Econ. Pol.
(2017) - et al.
Cognitive biases in the assimilation of scientific information on global warming and genetically modified food
Food Pol.
(2015)