Coordinating for Quality and Organization: A Theoretical Model and Empirical Findings

Coordinating for Quality and Organization: A Theoretical Model and Empirical Findings Guenter Schamel Associate Professor School of Economics and Management Free University of Bozen-Bolzano Bozen-Bolzano, BZ, 39100, ITALY Telephone: +00390471013170 Fax: +00390471013009 Email: guenter.schamel@unibz.it Francisco Javier Santos-Arteaga Research Associate School of Economics and Management Free University of Bozen-Bolzano Bozen-Bolzano, BZ, 39100, ITALY Email: FranciscoJavier.Santos-Arteaga@unibz.it Presented at the Economics and Management of Networks Conference (EMNet 2013) (http://emnet.univie.ac.at/) Robinson Hotel and University Ibn Zohr Agadir, Morocco November 21-23, 2013 1

Coordinating for Quality and Organization: A Theoretical Model and Empirical Findings Abstract The theoretical section of the paper presents a formal model illustrating how cooperatives and private firms can coexist within a market while obtaining different quality rewards. The basic intuition follows from the literature on firm boundaries determined via incomplete contracts where organizational forms, when agreed upon competitively, condition the sense of entitlement of the parties. If feeling aggrieved by the outcome of the contract, parties may shade by underperforming, which creates deadweight losses. The main result obtained states that if the intensity of shading depends positively on the existing payoff imbalances between bosses and managers, then (non)integration with coordination is more plausible when the profits of bosses and benefits of managers are (dis)similar. Moreover, given some plausible parameter constraints, social surplus tends to be higher under integration whenever coordination takes place. These results illustrate that both organizational forms, an integrated cooperative and a nonintegrated private firm, may coexist in a coordinated equilibrium and how the former may even obtain a higher social surplus than the latter one. When applied to our empirical setting, this result leads us to conclude that similar interests between both parties, the boss and the unit managers, strengthened by the shared regional DOC denomination, would lead to a small value of shading, which encourages coordination within a cooperative (integrated) environment. On the other hand, privately emphasizing a dominant brand over the rest obtaining an IGT denomination premium results from dissimilar interests between the parties, which leads to a nonintegrated though coordinated equilibrium scenario. In the empirical section we analyze a data set for cooperative and private wineries from Northern Italy (Alto Adige and Trentino). We first use a hedonic model to test whether wines from private wineries receive a wine quality premium and/or a reputation premium for ownership structure relative to cooperatives. For Alto Adige, we estimate a significantly positive reputation premium and a higher quality premium for wines produced by cooperatives. For the Trentino, we estimate a significant reputation premium for private wineries but higher quality premium for wines produced by cooperatives. In a second model, we include interaction terms between quality denominations (IGT/DOC) and organizational form (i.e. Coop/Non-Coop) to see if there is any strategic orientation towards specific denomination rules and organizational forms. Our expected result is that cooperatives concentrate on regional DOC rules while private wineries increasingly turn to a strategy of marketing and branding distinctly different IGT wines. We test this hypothesis estimating a modified equation on the whole sample with DOC wines produced by cooperatives selected as base category wines. The estimated coefficients indicate that cooperatives may emphasize DOC production and use IGT denomination for lower quality grapes, while private producers tend to use IGT to market distinctly different wines often with specific brand names. 2

Introduction Privately owned firms and cooperatives represent ownership forms that can be found concurrently in many markets throughout the world (Sexton and Lavoie, 2001; Van Bekkum and van Dijk, 1997). Even in the U.S., cooperative enterprises are predominant in a number of industries including agriculture where they market as much as 1/3 of total production (Hansmann, 1996). Also in Europe, large share of agricultural produce is marketed by farmer-owned cooperatives, but increasingly concerns are raised about their effectiveness and efficiency. In Europe, recent changes of the common agricultural policy imply that production and marketing of these products must meet specific requirements (improved quality management and marketing conditions, production efficiency etc.) in order to comply with internal support regulations and to attain competitive positions within Europe as well as internationally. We propose a model of product quality and reputation of cooperatives versus privately owned firms and how they are reflected in market prices. As an application, we examine the wine sector in the province of Alto Adige in Northern Italy where about 70% of the total wine production is marketed through cooperatives. For comparison, this is more than twice as much as in Germany where about 1/3 of the total production is processed and marketed by cooperatives (DRV, 2009). Product quality and reputation crucially affect product prices. Depending on the producer, considerable price differences exist even between similar products. In the case of wine, we often observe that a bottle produced and marketed by a cooperative sells for less than a bottle of comparable quality from a privately owned firm. One explanation for this may be that cooperatives lack control over the entire production chain which can lead to larger variation in grape quality and hence higher uncertainty about wine quality further downstream. Thus, consumers face uncertainties regarding product quality and reputation for cooperatively produced wine. A cooperative s reputation for product quality depends on the contributions of individual members growing grapes subject to variation as well as on the cooperative management ability to produce high quality wine. In contrast, a privately owned firm can be characterized by a higher degree of control within the production chain and thus may also be able to gain higher levels of reputation. In Alto Adige, wine cooperatives are characterized by modern production conditions and utilization of cutting-edge vineyard management systems. For example, they require that growers are cutting back on grape tonnage at certain predefined times during the growing season in order to limit yields, thus raising grape quality and wine quality further down the production line. (Schamel and Schubert, 2012). Therefore, it would be interesting to see if the cooperatives are able to claim sufficient control of the production chain relative to non-cooperative producers such that the hypothesized difference in the wine quality and reputation premium disappears. While the focus in this paper is on vertical collaboration between growers and winemaking within a cooperative enterprise, inter-organizational (i.e. horizontal) collaboration has been identified as an important success factor for the Australian wine industry (Jordan, Zidda and Lockshin, 2007). Theoretical Environment This section of the paper presents a formal model illustrating how cooperatives and private firms can coexist within a market while obtaining different quality rewards. We build on the model developed by Hart and Holmstrom (2010) to analyze the strategic choice of organizational form among producers. We will particularly concentrate on the similarity of the payoffs received by the parties composing the units within an organization as the main determinant of the boundary 3

choices of firms. Payoff similarities will also be used to show how cooperatives may be more socially efficient than private producers within the current strategic setting. Basic Assumptions and Notation The basic environment is that of Hart and Holmstrom (2010) and we will therefore restate their initial assumptions and maintain their notation. The basic intuition follows from the literature on firm boundaries determined via incomplete contracts where organizational forms, when agreed upon competitively, condition the sense of entitlement of the parties. If feeling aggrieved by the outcome of the contract, parties may shade by underperforming, which creates deadweight losses. If ever at all, shading takes place after the organizational form has been chosen. The strategic environment is composed by two units, A and B, that have a lateral relationship, i.e. they interact within the same output or input market, such that each unit is operated by a manager who triggers external effects on the other unit. Units are presented with a binary decision, they must choose either Yes or No. Coordination occurs if and only if both units choose Yes. Otherwise, units face noncoordination. In this sense, coordination may be interpreted as the decision of both units to remain as active producers within a joint project while any of them leaving the project results in noncoordination. Two main organizational forms are considered: nonintegration, where units are separate firms whose managers are also the bosses, and integration, where units are part of a single firm with an outside manager acting as the boss and the managers of each unit as subordinates. We will identify nonintegrated organizational forms with private independent firms, while the integrated scenario will be assumed to correspond to a cooperative structure. Intuitively, one can consider wineries as being composed by growers and winemakers. Growers could delegate the winemaking process in an external winemaker or contribute to the winemaking process themselves. When growers delegate in an external manager [integration], their individual contributions to the process are not explicitly acknowledged, with the winemaker losing complete control over the quality of the production chain. However, if growers interact in the winemaking process themselves [nonintegration], they are able to highlight the quality of their individual contributions within the production chain. In this regard, a higher degree of quality control is exerted over the production chain. Two types of benefits are assumed to be generated by each unit: monetary transferable profits, v i, i = A, B, and private nontransferable benefits, w i, i = A, B, in the form of job satisfaction for the manager working in the corresponding unit. The boss of a unit can divert all profits from the unit to herself, leading to a nonintegrated payoff of v i + w i if she is also the manager of unit i = A, B. Private benefits always reside with the managers. Thus, if both units are integrated, the professional outsider acting as boss receives v A + v B. Note that, under nonintegration both bosses account for the private benefits generated by each unit, which are ignored under integration in favor of total profits. Social surplus is given in both cases by v A + v B + w A + w B. Given the previous setting, bosses and unit managers face the following coordination game Payoffs Yes No Yes (Δv A, Δw A ); (Δv B, Δw B ) (0, 0); (0, 0) No (0, 0); (0, 0) (0, 0); (0, 0) 4

where the row payoffs correspond to unit A and the column ones to unit B. Without loss of generality, profits and private benefits have been normalized to zero in both units under noncoordination. The following notation is introduced to simplify the presentation Δz A = Δv A + Δw A, Δz B = Δv B + Δw B. where Δz i, i = A, B, represents the change in total surplus in unit i derived from coordination, and Δz A + Δz B accounts for the change in aggregate social surplus absent shading costs. Henceforth, S will denote the social surplus obtained net of shading costs. Coordination is assumed to lead to a reduction in private benefits due to the independence lost by the managers and its effect on job satisfaction Δw A 0, Δw B 0. Shading will be used to force bosses to internalize the externalities generated on other parties. This may occur under integration and nonintegration, since the relationship between both units is assumed to persist in both settings after the strategic coordination decision is made. It will also be assumed that a party receiving k i less than his maximum payoff will be aggrieved by k i and shade to the point where the payoff received by the other party falls by θk i. Hart and Holmstrom assume that θ ϵ (0,1) is an exogenous parameter identical for all parties. We will endogenize this parameter as a function of the spread existing between coordination profits and private benefits within a given unit. Moreover, our results will require the upper bound of the image interval on which θ is defined to be widen. We will elaborate further on the intuition lying behind the results obtained in the following section. The non-integration coordination condition (NIC) defined for any Δz i value, i = A, B, is given by Δz i + θ Δz j 0, (1) where i j. If Δz i 0, i = A, B, then (1) is trivially satisfied. However, if Δz i < 0, and Δz j > 0, then this condition states that coordination will take place under non-integration only if the costs of shading imposed by manager j on manager i are larger than the losses derived by the latter from coordination. Social surplus in the (NIC) setting [with Δz i < 0, and Δz j > 0 ] is therefore given by S NIC = Δz A + Δz B + θδz i under coordination - θδz j under noncoordination Note that, if coordination occurs, then unit i will shade by θδz i, as it receives a payoff of Δz i < 0 from coordinating. The integration coordination condition (INT) defined for any Δv i value, i = A, B, is given by Δv i + Δv j +θ(δw i + Δw j ) 0. (2) Trivially, if Δv i 0, i = A, B, then (2) is violated. Thus, for this condition to hold we need at least one of the Δv i changes in private profits to be positive. 5

Social surplus in the (INT) setting [with Δz i < 0, Δz j > 0, and Δv i + Δv j > 0] is given by S INT = Δz A + Δz B + θ(δw A + Δw B ) under coordination - θ(δv A + Δv B ) under noncoordination If coordination takes place, then managers will shade by θ(δw A + Δw B ), as both Δw A and Δw B are negative. If, on the other hand, condition (2) does not hold and units do not coordinate, then the boss will shade by θ(δv A + Δv B ). In order to provide additional intuition for the analysis performed in the next section, we rewrite the respective (NIC) and (INT) coordination conditions as follows Δv A + Δw A + θ (Δv B + Δw B ) 0 (1 ) Δv A + θδw A + Δv B + θδw B 0 (2 ) It should be noted that the lower degree of quality control exerted over the production chain within the cooperative [integrated] environment implies that the contributions of the individual growers to the winemaking process cannot be explicitly acknowledged. As a result, when shading, they can only do so through their respective Δw A and Δw B values, as illustrated in equation (2 ). On the other hand, private [nonintegrated] wineries are able to recognize the contributions of the individual growers, which allows the latter ones to shade through their entire Δz B values, as described by equation (1 ). Main Results Intuitive Presentation The organizational form is chosen so as to maximize social surplus net of shading costs, which may be incurred after a given organizational form has been agreed upon. In this regard, units may either operate independently or delegate in an independent boss who maximizes her joint private profit. We start by illustrating how, given our definition of shading intensity, managers will be more willing to delegate if their Δw values are close to the respective Δv of the boss. Consider the coordination payoffs received by the managers and the boss within each unit. We endogenize the intensity of the shading parameter θ as a function of the difference in coordination incentives existing between the boss and the unit managers. The definition of the [finite] θ i (v, w) variable, i = A, B, is therefore given by θ i (Δv i, Δw i ) = Δv i + Δw i (3) Note that we have to account for the possibility that Δv i < 0, while knowing that Δw i < 0 under coordination. Thus, a substantial difference between both payoffs leads to an increase in the value of the shading parameter. That is, the strength or effort dedicated by a party to shade depends on the existing differences in objectives (and payoffs) with respect to the other one. This assumption follows directly from the guilt-envy (Fehr-Schmidt) inequality aversion literature based on comparisons of absolute differences in payoffs between the parties, see Camerer (2003) for a review of the literature on this topic. The importance that the shading parameter has in determining the coordination incentives of bosses and managers within both units can be easily illustrated numerically. Coordination Incentives with θ = 0. 6

INT > NIC always in the setting with Δz i < 0, and Δz j > 0, even if coordination does not take place, since Δv A + Δw A 0, (1 ) Δv A + Δv B 0, (2 ) and Δv B > 0 while Δw A < 0. Coordination Incentives with θ = 0.5, Δv A = Δv B = 5, and Δw A, Δw B ϵ [-10, 0]. Coordination Incentives with θ = 1. INT = NIC. It is easy to see how coordination is more likely to take place under integration (nonintegration) when the θ variable is relatively low (high). In this case, similar (dissimilar) interests between both parties in the form of coordination payoffs would lead to a lower (higher) shading intensity, which encourages coordination within an integrated (nonintegrated) organizational environment. 1 It immediately follows that Lemma 1. If the intensity of shading depends positively on the existing payoff imbalances between bosses and managers, we have the following 1 It should be noted that θ i (Δv i, Δw i ) could be normalized within the [0, 1] interval after defining (exogenously) some bounds for Δv and Δw. This constraint would keep the analysis within the parameter value limits considered by Hart and Holmstrom, where (1) (2). However, if θ is allowed to be defined above one, as is the case here, then we have that (2) (1) for θ > 1. In this regard, when comparing absolute differences in payoffs between both parties, we could also define ex-ante bounds for Δv and Δw determining a pair of θ i (Δv i, Δw i ), i = A, B, integration limit values. The analysis performed in the following section provides additional intuition on this option. 7

Integration with coordination is more plausible when the profits of bosses and benefits of managers are similar. Nonintegration with coordination is more plausible when the profits of bosses and benefits of managers are dissimilar. Corollary 1. Given Δz i < 0, Δz j > 0, and Δv i < 0, social surplus tends to be higher under integration whenever coordination takes place, due partly to the lower θ values generated by the respective units. Both these previous results are merely intuitive and the intuition justifying them follows from basic numerical examples. However, these results illustrate that both organizational forms, an integrated cooperative and a nonintegrated private firm, may coexist in a coordinated equilibrium and how the former may even obtain a higher social surplus than the latter one. Formal Analysis We turn now to a more formal analysis where the current model will be used to explain how both these organizational forms may coexist optimally within unequal coordinated equilibria. Cooperatives are more willing coordinate when the θ (Δv, Δw) values are small and similar for all units involved. At the same time, if heterogeneity is allowed for in the values of θ (Δv, Δw), then any unit with a sufficiently divergent payoff structure [leading to a high θ (Δv, Δw) value] has an incentive to avoid the integrated setting and impose a nonintegrated though coordinated organizational form. Thus, highly unequal θ (Δv, Δw) values between units favor the emergence of nonintegrated but coordinated structures. 2 In order to illustrate these points, we must allow for heterogeneous θ (Δv, Δw) values to be defined between both units. Consider two different θ i (Δv i, Δw i ) values, θ A and θ B, one for each unit, though the analysis can easily account for a larger number of units. We concentrate on the nontrivial Δz i < 0, and Δz j > 0 case, and assume that i=a and j=b. Thus, in order for a nonintegrated equilibrium with coordination to be more plausible than an integrated equilibrium with coordination we need (1 ) > (2 ), which, after some basic algebra, implies that (1 θ A ) Δw A > (1 θ B ) Δv B. (4) Note that Δw A < 0 and Δv B > 0. As a result, it is sufficient (though not necessary) for this inequality to hold that either θ A or θ B are higher than one with the other being at least as high. This implies that large payoff differentials between both parties within a unit may shift coordination to a nonintegrated environment. However, it is also possible for this inequality to hold when θ B > 1 and θ A < 1. In this case, the unit shifting faces larger payoff inequalities between its parties. Moreover, Δv B should be large enough to achieve coordination under nonintegration, an equilibrium which would not necessarily be plausible under integration. Clearly, for the sake of completeness, an integrated equilibrium with coordination would be more plausible than a nonintegrated equilibrium with coordination if (2 ) > (1 ), which implies that (1 θ B ) Δv B > (1 θ A ) Δw A. (4 ) 2 It clearly follows that a larger number of heterogeneous units would favor the nonintegrated [coordinated] setting over integration. 8

In this case, it is sufficient (though not necessary) for this inequality to hold that either θ B < 1 and θ A 1 or θ B 1 and θ A < 1. Social surplus could be higher under either one of these organizational structures. Note, however, that highly aligned and similar payoffs work in favor of an integrated organization [cooperative] due to the smaller value of θ generated by its units. We will show how, in the Δz i < 0, and Δz j > 0 case, there exist reasonable payoff and shading values that allow for an integrated organizational form to lead to a higher social surplus under coordination than the nonintegrated one. For this to be the case, we require that S INT > S NIC under coordination which simplifies to Δz A + Δz B + θ A Δw A + θ B Δw B > Δz A + Δz B + θ A Δz A θ B Δw B > θ A Δv A. (5) We know from equation (4) that θ B > θ A. 3 Thus, in order for (5) to hold we need 0 > Δw B >> Δv A. The main implications derived from equations (4) and (5) for the coexistence of both organizational forms within socially unequal coordinated equilibria are summarized as follows Δv B >> 0 Δw B >> Δv A Δw A. (6) These requirements state that the unit avoiding integration must exhibit considerably unequal payoffs between the boss and the manager. In this case, the unit avoids integration but keeps on coordinating under nonintegration. At the same time, the other unit must exhibit similar negative payoffs that prevent its shading from affecting coordination under integration. 4 Clearly, in order for coordination under integration to be more plausible and lead to a higher social surplus than coordination under nonintegration we simply require that Δv B > 0 Δw B > Δv A Δw A, which is implied by (6). Regarding our empirical setting, described through the following sections, we may conclude that similar interests between both parties, the boss and the unit managers, strengthened by the shared DOC denomination, would lead to a small value of θ, which encourages coordination within a cooperative [integrated] environment. On the other hand, privately emphasizing a dominant brand over the rest in order to obtain an IGT denomination premium results from dissimilar interests between the parties, which leads to a nonintegrated but coordinated equilibrium scenario. 3 Note that it is also possible for both θ values to be higher than one with θ B θ A, which would weaken the strength of the requirements derived from equation (5). 4 Note that Δz A < 0 is an essential requirement for social surplus to be higher under integration. If this were not the case and Δv A > 0, then social surplus would always be lower under integration since Δw B < 0 < Δv A, which violates (5). Unit A managers shade in both cases due to the benefits lost under coordination, but when Δv A > 0 the boss of the unit obtains positive profits that relatively increase the nonintegrated social surplus despite the intensity of his shading. 9

Data and Analysis We analyze a data set of wines evaluated in the annual Le Guide de l Espresso (I vini d Italia) for the Alto Adige and Trentino regions in Northern Italy. We obtained three years of data published in the guide for the years 2012-2014. The sample used in the estimation consists of 1265 wines from Alto Adige and 724 from Trentino. We employ a hedonic model to test whether cooperatives or private wineries can obtain higher implicit prices for specific parameters that directly impact welfare levels. A higher level of welfare may be reached through a price premium for the collective reputation indicator for cooperatives as quality wine producers or through a higher wine quality premium. The data guide lists a range of applicable retail prices per bottle from which we use the lower bound for estimation purposes. The price information used in the estimation is submitted prior to the quality evaluation (i.e. the point rating by the expert tasters). Thus, it does not reflect any direct effects due to a favorable quality rating. The experts rate the wines according to a 20-point scale in half-point steps. Information on the number of bottles of a particular wine produced is also provided. The age of the wines when evaluated ranged from 1 to 13 years. Le Guide de l Espresso also provides a starranking (between 0 and 3) for a winery s distinctiveness which could be regarded as a proxy for individual winery reputation. Moreover, the wine guide differentiates wine color, sweet or desert wines, DOC and IGT designated wines, biologically or bio-dynamically produced wine, wine variety and special recommendations such as value for money and best regional buys. In addition, the data set allows to categorize whether a wine was produced by a local cooperative or not. Cooperatively produced wines, red vs. white wines, sweet wines, IGT vs. DOC designated wines, bio-labeled wines and special recommendations are regular dummy variables while wine variety is a categorical dummy. As dependent variable we use the logarithm of the lower price bound [log(p)]. We employ a log-linear function in our estimation. In a first regression, we estimate the following equation without interaction terms (Model 1): log(p) = α + β 1 log(points)+β 2 log(bottles)+β 3 Age +β 4 Stars +γ 1 Red + γ 2 Sweet + γ 3 Coop + γ 4 IGT + γ 5 Bio + γ 6j Variety + θ 1 ValueRec + θ 2 BuyRec + ε where log(p) is the logarithm of price P, log(points) is the logarithm of the Gault Millau points and log(bottles) is the logarithm of the production quantity, Coop is dummy variable as an indicator for the collective reputation of cooperatives while ε is the error term with a zero mean and uniform variance. The regression equation stated above includes a number of variables to control for willingness to pay (price) effects due to: - production quantity (scarcity effect implied by the number of bottles produced) β 2 - wine age (storage premium due to age in years at the time of evaluation) β 3 - star ranking (winery reputation effect) β 4 - red vs. white wines (red wine premium) γ 1 - sweet or dessert wines (sweet wine premium) γ 2 - IGT classification effect (outside DOC rules) γ 4 - Bio-labeled wines (organic-premium) γ 5 - wine variety (varietal premium) γ 6j - value recommendation (ValueRec) θ 1 - best buy recommendation (BuyRec) θ 2. 10

Given its log-linear functional form, estimating the equation above yields price premiums and discounts relative to the contribution of the base category (non-sweet white DOC wine that is not bio-labeled and not a differentiated variety for the region). We differentiate seven varieties/wine types in the estimation. Five are in common for both regions (Gewürztraminer, Pinot Noir, Sauvignon Blanc, Riesling and Spumante). Lagrein and Schavia are specific for Alto Adige while Teroldego, Nosiola are specific for Trentino. According to our expectation that consumers face more uncertainty regarding product quality and reputation for cooperatively produced wine, we hypothesize that wines coming from privately owned producers receive a price premium relative to cooperatively produced wine. To test this hypothesis, we employ a hedonic pricing model differentiating cooperative vs. non-cooperative producers. To test the hypothesis that cooperatively produced wines receive a reputation discount relative to other producers, we expect a significant but negative coefficient for the cooperative dummy variable (indicating a negative collective reputation for cooperatives). Thus, we would look for a negative coefficient γ 3 for wine produced by cooperatives. Moreover, if we split the sample into wines produced by cooperatives and private wineries (non-cooperatives) we would expect a lower quality premium for cooperatively produced wines. Using our data set of two different wine regions in Northern Italy, we would also like to test if there are regional differences in terms of quality coordination and ownership structure. Our a priori expectation is that cooperatives in Alto Adige achieve a higher level of quality coordination and thus are in a better position to compete with private wineries in terms of wine quality and reputation. Thus comparing Model 1 results for Alto Adige and Trentino, we would expect that cooperatives in Alto Adige outperform the cooperative in Trentino in terms of reputation and quality premiums. In a second model, we include interaction terms between DOC and IGT denominations and ownership structure (Coop vs. NonCoop) in order to examine if there is any strategic orientation towards specific denomination rules with respect to ownership structures. Our expected result is that cooperatives are concentrating on DOC denominated wines while privately owned wineries are increasingly producing and marketing distinctly different wines outside DOC rules. The second regression equation estimated looks as follows (Model 2): log(p) = α + β 1 log(points)+β 2 log(bottles)+β 3 Age +β 4 Stars +γ 1 Red + γ 2 Sweet + γ 5 Bio + γ 6j Variety + θ 1 ValueRec + θ 2 BuyRec + η 1 IGT*Coop + η 2 IGT*NonCoop + η 3 DOC*NonCoop + ε Notice that in Model 2, the base category is a cooperatively produced DOC wine (a non-sweet white wine that is not bio-labeled and not a differentiated variety in the region). The three remaining interaction terms between denomination rules and ownership structure are: - IGT * Coop or IGT classified wine produced by cooperatives - IGT * NonCoop or IGT classified wine produced by privately owned wineries - DOC * NonCoop or DOC classified wine produced by privately owned wineries Finally, we tested the models and/or the residuals for normality (Jarque-Bera-Test) and heteroskedasticity (White-Test) and do not find any significant problems in the data. We also employed a RESET test which rejected other functional forms. 11

Results The results of our estimation for Model 1 are listed in Table 1 (for Alto Adige/AA) and Table 2 (for Trentino/TN). We cannot confirm the hypothesis that wines coming from privately owned producers receive a reputation premium relative to cooperatively produced wines for the Alto Adige region. Our estimation for AA Wines (Table 1) reveals a significant but positive coefficient for cooperative reputation. This would indicate that Alto Adige cooperatives receive a collective reputation premium (about 11%) relative to their local privately owned competitors. This is even more remarkable given the fact that the model corrects for individual winery reputation via the Stars variable. On the contrary, our estimation for TN Wines (Table 2) reveals a significant but negative reputation coefficient for the Trentino cooperatives (with a collective reputation discount of about 6%). --- Insert Tables 1 and 2 here --- Comparing the quality premium (coefficients on the points rating) in the cooperatively produced subsample relative to the non-coop subsample we can also confirm that the posted hypothesis is not correct for both Alto Adige and Trentino, i.e. that cooperatively produced wines receive a significant quality premium relative to non-coop wines (2.748 vs. 2.515 when comparing the respective columns in Table 1). Thus, we can confirm that cooperatives in Alto Adige and Trentino are characterized by modern production conditions, efficient horizontal coordination using cutting-edge vineyard management systems which in the end result in a significant quality premium relative to their local privately owned competitors. Moreover, in Alto Adige, the model also confirms a positive reputation premium for cooperatively produced wines. Thus, our analysis suggests that these cooperatives are able to claim sufficient control of the production chain relative to non-cooperative producers such that the hypothesized difference in the wine quality and reputation premium disappears. In practice, this is due to requiring growers to cut down the grape tonnage at certain predefined times during the growing season in order to limit yields (often way below the DOC maximum), thus raising grape quality and wine quality further down the production line. The remaining results on the control variables listed in Tables 1 and 2 are as expected in terms of sign and magnitude relative to other studies. For example, the storage effect (wine age) is relatively consistent across the subsamples ranging between 9% and 12%. There is a red wine premium for Alto Adige but not for the Trentino region. It is interesting to note that bio-labeled wine has a positive coefficient at least for Alto Adige which is in contrast to other studies that have claimed the opposite (e.g. Delmas and Grant, 2010). Moreover, we note that the premium on IGT wine is positive for both Alto Adige (10.5%) and Trentino (7.6%). This result suggests that wines produced outside the local DOC rules receive a price premium. Thus, the question is raised whether DOC regulations, established to guarantee quality wine production are not necessarily working in favor of receiving higher market prices. However, we notice that the coefficients for IGT wines are positive and significant in the noncooperative sub-samples for both regions. This would indicate that private wineries may emphasize production according to IGT classification to sell of higher quality grapes and to market own-branded wines while avoiding DOC rules. We argue that this strategic orientation is confirmed by our results. Cooperatives get a collective reputation premium for focusing on DOC rules while their non-cooperative competitors use an IGT strategy emphasizing branding. 12

Finally, in Table 3 we list the results including the interaction terms between IGT/DOC denominations and ownership structure (Coop/Non-Coop) for both Alto Adige and Trentino. Comparing IGT and DOC denominations, the estimated coefficients indicate that relative to a cooperatively produced DOC wine, private wineries receive a significant premium for their DOC wines in Trentino (+4.5%) but not so in Alto Adige (-11.3%). Moreover, in Trentino, privately owned wineries receive a significant premium for their IGT wine (relative to cooperatively produced DOC wine) while the coefficient is not significant for Alto Adige. The interaction term for IGT*Coop is not significant for both regions. --- Insert Table 3 here --- The results obtained with both Models corroborate our results obtained in our theoretical model. Cooperatives seem to outperform private wineries in Alto Adige in terms of wine quality and reputation. They do this by emphasizing DOC production and using IGT rules to market lower quality grapes. In Trentino, cooperatives are less successful and private wineries outperform them under DOC rules as well as through IGT when marketing distinctly different wines under specific brand names. In Alto Adige, cooperatives get a collective reputation premium focusing on DOC rules. In Trentino, cooperatives receive lower prices relative to private wineries who manage to receive a price premium of 4.5% for their DOC wines and 13.3% for their IGT wines (relative to a cooperatively produced DOC wine). In Alto Adige, it seems that cooperatives focus on DOC wines while private wineries at least to some degree void DOC rules to market and brand distinctly different IGT wines but according to our sample do not receive a price premium relative to DOC wine. On the other hand, in Trentino, private wineries receive price premia for DOC and IGT wines relative to their cooperative competitors. Summary and Conclusion In this paper, we have introduced a formal model and empirical evidence illustrating how cooperatives and private firms can coexist within a market while obtaining different quality rewards. The reputation of a cooperative for product quality depends on the contributions of its individual growers and its managerial ability to produce high quality wine. In contrast, a privately owned firm is characterized by a higher degree of control within the production chain. Therefore, the choice between a cooperative and a private firm organizational form depends on the difference between the objectives and payoffs obtained by the parties composing the organization and the resulting shading parameters. In Alto Adige, cooperative wineries manage to organize their production decisions such that they are able to compete with private wineries in terms of wine quality and reputation. This empirical result supports the theoretical model that cooperatives may reach a higher social surplus due to larger reputation and quality premiums. 13

Table 1: Model 1 Results for Alto Adige/AA (Dependent Variable: Log(Price) AA Wines AA Coops AA Non-Coops Variable Coeff. t-stat. Prob. Coeff. t-stat. Prob. Coeff. t-stat. Prob. Constant -4.458-9.915 0-4.530-5.639 0-3.926-7.361 0 Log(Points) 2.682 17.02 0 2.748 9.874 0 2.515 13.36 0 Log(Bottles) -0.071-8.616 0-0.058-4.007 0-0.071-6.95 0 Age 0.108 12.45 0 0.124 11.47 0 0.111 7.152 0 Stars 0.069 7.047 0 0.010 0.605 0.546 0.100 8.543 0 Red Wine 0.172 5.787 0 0.258 5.249 0 0.119 2.898 0.004 Sweet Wine 0.253 6.358 0 0.255 6.091 0 0.277 3.971 0 Cooperatives 0.108 6.517 0 IGT Wine 0.105 3.121 0.002-0.068-0.574 0.567 0.145 4.250 0 Bio-Wine 0.062* 1.648 0.099-0.168-1.387 0.166 0.069* 1.844 0.066 Lagrein -0.048-1.432 0.152-0.096-1.381 0.168 0.003 0.068 0.946 Schiava -0.283-7.291 0-0.399-7.137 0-0.217-4.247 0 Gewürztraminer 0.236 9.959 0 0.246 7.036 0 0.245 7.653 0 Pinot Nero 0.051 1.247 0.213-0.178-2.506 0.013 0.132 2.795 0.005 Sauvignon Blanc 0.096 4.522 0 0.132 3.794 0 0.078 2.864 0.004 Riesling 0.109 4.266 0 0.032 1.250 0.212 0.117 4.005 0 Spumante 0.139* 1.941 0.053 0.147* 1.798 0.073 Value for Money -0.362-19.73 0-0.361-12.28 0-0.344-14.89 0 Best Buy Region -0.219-7.604 0-0.183-3.519 0.001-0.220-6.841 0 F-statistic 134.97 0 67.05 0 86.54 0 Wald F-statistic 165.30 0 134.64 0 97.97 0 Adjusted R 2 0.656 0.738 0.621 Std. err. estimate 0.244 0.225 0.248 Sum sq. residuals 74.48 18.30 53.36 Observations 1265 377 888 Notes: Estimation Method: LS / White heteroskedasticity-consistent standard errors & covariance. The symbols,, and * denote significance at the 1%, 5%, and 10% level, respectively. 14

Table 2: Model 1 Results for Trentino/TN (Dependent Variable: Log(Price) TN Wines TN Coops TN Non-Coops Variable Coeff. t-stat. Prob. Coeff. t-stat. Prob. Coeff. t-stat. Prob. Constant -5.200-7.711 0-7.049-4.344 0-4.203-5.514 0 Log(Points) 2.874 11.638 0 3.582 6.091 0 2.489 8.954 0 Log(Bottles) -0.050-5.603 0-0.052-3.638 0-0.053-4.700 0 Age 0.090 9.411 0 0.099 5.300 0 0.100 7.997 0 Stars 0.144 7.610 0 0.109 2.602 0.010 0.157 7.371 0 Red Wine -0.002-0.079 0.937-0.155-3.359 0.001 0.031 1.142 0.254 Sweet Wine 0.217 3.909 0-0.176-2.025 0.045 0.320 5.765 0 Cooperatives -0.061-2.790 0.005 IGT Wine 0.076 3.566 0 0.013 0.264 0.792 0.098 4.077 0 Bio-Wine -0.014-0.379 0.705 0.371 8.588 0-0.038-1.042 0.298 Teroldego 0.130 3.005 0.003 0.069 0.796 0.428 0.184 3.645 0 Schiava -0.086-3.060 0.002-0.104-2.302 0.023-0.071* -1.948 0.052 Gewürztraminer 0.113 3.116 0.002 0.001 0.020 0.984 0.162 4.707 0 Pinot Nero 0.164 4.179 0 0.178 2.927 0.004 0.151 3.314 0.001 Sauvignon Blanc -0.081-1.548 0.122-0.206-1.685 0.094-0.046-0.808 0.420 Riesling 0.053 0.761 0.447 0.223 2.458 0.015 0.025 0.394 0.693 Spumante 0.167 3.548 0-0.079-0.879 0.381 0.207 3.610 0 Value for Money -0.363-16.536 0-0.411-10.815 0-0.331-12.960 0 Best Buy Region -0.172-5.300 0-0.219-3.310 0.001-0.161-4.468 0 F-statistic 109.41 0 26.29 0 90.89 0 Wald F-statistic 117.78 0 84.37 0 93.19 0 Adjusted R 2 0.730 0.725 0.732 Std. err. estimate 0.233 0.207 0.233 Sum sq. residuals 38.38 6.24 29.47 Observations 724 164 560 Notes: Estimation Method: LS / White heteroskedasticity-consistent standard errors & covariance. The symbols,, and * denote significance at the 1%, 5%, and 10% level, respectively. 15

Table 3: Model 2 Results for Alto Adige and Trentino (Dependent Variable: Log(Price) Alto Adige Wines Trentino Wines Variable Coeff. t-stat. Prob. Variable Coeff. t-stat. Prob. Constant -4.170-9.255 0 Constant -5.239-7.749 0 Log(Points) 2.656 16.94 0 Log(Points) 2.873 11.63 0 Log(Bottles) -0.071-8.613 0 Log(Bottles) -0.051-5.620 0 Age 0.109 12.44 0 Age 0.089 9.441 0 Stars 0.068 7.016 0 Stars 0.143 7.504 0 Red Wine 0.167 5.567 0 Red Wine -0.002-0.086 0.931 Sweet Wine 0.263 6.549 0 Sweet Wine 0.216 3.878 0 IGT*NonCoop 0.011 0.318 0.750 IGT*NonCoop 0.133 4.812 0 IGT*Coop -0.108-0.948 0.343 IGT*Coop 0.015 0.286 0.775 DOC*NonCoop -0.113-6.863 0 DOC*NonCoop 0.045* 1.863 0.063 Bio-Wine 0.064* 1.706 0.088 Bio-Wines -0.018-0.508 0.611 Lagrein -0.043-1.283 0.200 Teroldego 0.137 3.143 0.002 Schavia -0.279-7.175 0 Nosiola -0.086-3.039 0.003 Gewürztraminer 0.233 9.853 0 Gewürztraminer 0.116 3.276 0.001 Pinot Nero 0.056 1.373 0.170 Pinot Nero 0.163 4.090 0 Sauvignon Blanc 0.095 4.468 0 Sauvignon Blanc -0.084-1.594 0.111 Riesling 0.109 4.291 0 Riesling 0.053 0.759 0.448 Spumante 0.136* 1.934 0.053 Spumante 0.171 3.652 0 Value for Money -0.360-19.70 0 Value for Money -0.361-16.48 0 Best Buy Region -0.213-7.689 0 Best Buy Region -0.172-5.309 0 F-statistic 128.50 0 F-statistic 103.87 0 Wald F-statistic 157.30 0 Wald F-statistic 111.60 0 Adjusted R 2 0.657 Adjusted R 2 0.730 Std. err. estimate 0.244 Std. err. estimate 0.233 Sum sq. resid. 74.20 Sum sq. residuals 38.28 Observations 1265 Observations 724 Notes: Estimation Method: LS / White heteroskedasticity-consistent standard errors & covariance. The symbols,, and * denote significance at the 1%, 5%, and 10% level, respectively. 16

References BEVERLAND, M. 2007, "Can cooperatives brand? Exploring the interplay between cooperative structure and sustained brand marketing success," Food Policy 32: 480 495. BONUS, H. 1986, "The Cooperative Association as a Business Enterprise: A Study in the Economics of Transactions," Journal of Institutional and Theoretical Economics 142(2): 310-339. CAMERER, C.F., 2003, Behavioral Game Theory. Experiments in Strategic Interaction. Princeton University Press. DELMAS, M. and GRANT, L., 2010, "Eco-Labeling Strategies and Price-Premium: The Wine Industry Puzzle," Business and Society 49(4): pp 1-39. DRV, 2008, "Entwicklung der Winzergenossenschaften." Available at http://www.deutschewinzergenossenschaften.de/ FUJITA, M., KRUGMAN, P., and VENABLES, A., 1999, The Spatial Economy: Cities, Regions, and International Trade, Cambridge, MA: MIT Press. GCAC, 2000, The agricultural cooperation in the European Union: Tendencies and current issues. Brussels: General Committee for Agricultural Cooperation. HANSMANN, H., 1996, The Ownership of Enterprise. Belknap Harvard. HART, O. and HOLMSTROM, B., 2010, A Theory of Firm Scope, Quarterly Journal of Economics, Vol. CXXV: pp. 483-513. HELPMAN, E. and KRUGMAN, P., 1985, Market Structure and Foreign Trade. Cambridge, MA: MIT Press. HOFFMANN R., 2005, Ownership Structure and Endogenous Quality Choice: Cooperatives versus Investor-Owned Firms, Journal of Agricultural & Food Industrial Organization, 3(2), Article 8. Available at http://www.bepress.com/jafio/vol3/iss2/art8 JORDAN, R., ZIDDA, P.; and LOCKSHIN, L., 2007. "Behind the Australian wine industry's success: does environment matter?", International Journal of Wine Business Research, Vol. 19 (1), pp.14 32. ROSEN, S., 1974, Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competition, Journal of Political Economy, Vol. 82: pp. 34-55. SAN MARTIN, G.; TRONCOSO, J., and BRÜMMER, B., 2008, Determinants of Argentinian Wine Prices in the U.S., Journal of Wine Economics, 3 (1), 72-84. SCHAMEL, G., 2009, "Can German Wine Cooperatives Compete on Quality?" Paper presented at the 27 th International Association of Agricultural Economists Conference, Beijing, China. SCHAMEL, G., and SCHUBERT, S.F. (2012). "A Dynamic Optimal Control Model of Crop Thinning." Paper presented at the 2012 Conference of the International Association of Agricultural Economists in Foz do Iguacu, Brazil. SEXTON, R. J., and LAVOIE, N., 2001, Food Processing and Distribution: An Industrial Organization Approach. (Chapter 15, Volume 1B) in Handbook of Agricultural Economics, edited by Gardner B. and G. Rausser, Elsevier, North Holland, 863 932. SHAPIRO, C., 1983, "Premiums for high quality products as Returns to reputations," Quarterly Journal of Economics, 98: 659-679. TIROLE, J., 1988, The Theory of Industrial Organization. Cambridge, MA: MIT Press. TIROLE, J., 1996. "A theory of collective reputations (with applications to the persistence of corruption and to firm quality)," Review of Economic Studies, 63: 1-22. VAN BEKKUM, O., and VAN DIJK, G., 1997, Agricultural Cooperatives in the European Union, Trends and Issues of the 21st Century. Assen: van Gorcum. 17