The wine industry faces new market opportunities created by factors such as increasing trade liberalization and globalization, emerging competition from new market players, and declining consumption in traditional wine-producing countries ( Anderson et al ., 2001 ; Wittwer et al ., 2003 ). Increasing export volumes and added value creation through differentiation strategies are the key ways to maintain profitability in the Spanish wine industry ( Albisu & Zeballos, 2014 ; Del Rey, 2014a ). The United States (US) is one of the world's main wine importers, with growing domestic demand. Spain's wine exports to the US have shown significant growth rates, but Spanish market share still remains below that of many competing countries.

In this context, it was relevant to analyse the way in which the various attributes of Spanish quality wines influenced demand behaviour in the US. Knowledge of how US consumers assessed these characteristics could be used by producers, exporters and distributors to design the necessary strategies to improve their market positioning. To be aware of what factors might influence the wine prices could help producers to take decisions about where to position their product on the market ( Cousido Cores, 2017 ).

Hedonic price modelling represents a valid empirical methodology ( Cacchiarelli et al ., 2016a ) and it is widely used to explain the decision-making process leading up to the purchase of a specific product in terms of its combined attributes, by allowing the estimation of the implicit price of each attribute. It has been applied to this end in various sectors and markets. Research on the agriculture and food industry considers products as diverse as livestock ( Troncoso et al ., 2012 ), apples ( Dinis et al ., 2011 ) or saffron ( Sanjuán-López et al ., 2009 ). The high degree of differentiation makes the wine suitable for the application of hedonic methodology, taking into account the multiple intrinsic characteristics that provide consumers with a large range of wines ( Oczkowski & Doucouliagos, 2015 ). The method has, nevertheless, received some criticism because of the difficulty of selecting the appropriate variables or the need to assume that the quality attributes of wines are unrelated to their price ( Unwin, 1999 ). The selection of the attributes included as determinant of the wine price was based on the review of the literature and the information provided by several sources.

The planning and implementation of long-term investment strategies could benefit from having information about the attributes more valued by consumers. The results can help in assessing the need to include specific attributes in the final product, by enabling comparison between their implicit prices and their marginal cost. This explains the widespread use of hedonic price models to determine the relative importance of the intrinsic characteristics of wine ¹ . Some studies are focused on traditional markets, such as France ( Landon & Smith, 1997 ; Lecocq & Visser, 2006 ; Ali & Nauges, 2007 ; Cardebat & Figuet, 2009 ; Carew & Florkowski, 2010 ), Italy ( Brentari et al ., 2011 and 2015 ; Caracciolo et al ., 2013 ; Roma et al ., 2013 ; Levaggi & Brentari, 2014 ; Cacchiarelli et al ., 2014 ) or Germany ( Schamel, 2003 ), as well as new producers, such as Australia ( Oczkowski, 1994 , 2015 ; Steiner, 2004 ; Davis & Ahmadi-Esfahani, 2005 ), New Zealand ( Bicknell et al ., 2005 ; Bicknell & MacDonald, 2012 ), US ( Yang et al ., 2012 ; Cuellar & Claps, 2013 ; Delmas & Grant, 2014 ), Chile ( Melo et al ., 2005 ; Troncoso & Aguirre, 2006 ; Ortuzar & Alfranca, 2010 ) or Argentina (San Martin et al ., 2008; Estrella, 2014 ).

The domestic market for Spanish wines has been analyzed from different perspectives. Angulo et al . (2000) estimate a hedonic price function for red wines, Morilla & Martínez (2002) for Spanish wines with Designation of Origin , Rodríguez & Castillo (2009) for red wines from Castilla La Mancha , and Perza (2010) for Catalan wines. The acceptance of Spanish wines has also been assessed in foreign markets such as the United Kingdom (UK) ( Steiner, 2009 ), France ( Cardebat & Figuet, 2014 ) and, jointly with wines from other origins, in the US ( Schamel, 2006 ).

It is interesting to point out that Spain is currently the country with largest vineyeard area in the world and the top wine exporter but its wine is sold for its low price ( Parga-Dans & Alonso González, 2017 ). Considering the evolution of the US wine market, it is important to have some knowledge about which wine characteristics can contribute to achieve higher prices in that market. So far, no study offers a specific valuation of prices in wine's US markets with respect to Spanish wines, despite the importance of exports for the Spanish wine industry and the relevance of the US market. This research aimed to meet this gap, by trying to determine which features of quality Spanish wines were most valued by US consumers. Our analysis took a representative sample of Spanish wine exports to identify the attributes influencing consumer willingness-to-pay in each segment. Our hypothesis was that there are some specific attributes that can influence wine's prices. These attributes has been clustered in six groups of attributes are shown to influence wine's prices: vintage, aging, quality rating, appellation, type of wine and size of winery. Any business decision involving one of these attributes was relevant for the positioning of a wine in the US market.

Hedonic price theory enables product differentiation analysis based on the utilities provided by product characteristics or attributes, through the estimation of an implicit price or hedonic price function. Pioneer studies in the literature on the hedonic price theory include Waugh (1928) and Court (1939) , but it was not until the 60s that it came into more extensive use in research conducted by Griliches (1961) , Chow (1967) and Muth (1969) , and the study by Rosen (1974) , founded on the work of Lancaster ( 1971) , which estimates a perfect competition model optimizing hedonic price theory.

A hedonic price function is a mathematical function relating the price of a good to its characteristics or attributes. Any variable influencing the consumer decision or production cost can be considered in the function and, if proved significant, it can be assumed to have consumer value or relevance in the production process. Two key parameters need to be specified: the functional form and the product characteristics to be included as explanatory variables.

Equation [1] describes a general hedonic function model

where p denotes the price of the i-th bottle of wine and the j-th attribute. The market is assumed to be in equilibrium and consumers and producers are assumed to maximize their objective functions.

Linear, log-linear, double logarithmic or logarithmic functional forms can be adopted ( Brachinger, 2002 ). As usual in hedonic price models, the attributes of interest were represented exclusively by dummy variables ( Morilla & Martínez, 2002 ; Steiner, 2004 ; Rodríguez & Castillo, 2009 ), which reduces the risk of impacts from measurement errors in variables and prevents from heteroscedasticity. This option also restricts the choice of functional form to linear or log-linear models. Most studies opt to estimate the log-linear model ( Steiner, 2004 ; Melo et al ., 2005 ; Luppe et al ., 2009 or Carew & Florkowski, 2010 ), for the variance-stabilizing properties derived from the logarithmic transformation. Equation [2] describes the log-linear model estimated by ordinary least squares ²

The results of the estimated models are presented in Table 2. Except for appellation, the observed pattern for each attribute was similar across price categories, although the magnitudes varied.

Aging was the most influential attribute. A positive relationship was found between aging and wine's prices in all price categories, except, importantly, the ultra premium category, where crianzas had more negative impact than joven wines. In the popular premium and super premium categories, gran reserva wines had more impact on prices, increasing the mean price in the category by 47.6% and 16.2% respectively. In the ultra premium category, boutique wines had the greatest impact on prices, raising them to 156% of the average price. It is worth noting that aging had less relative influence in the lower price categories and that the highest willingness-to-pay values were for boutique wines .

A similar result was obtained for the quality rating attribute, which played a bigger role in the ultra premium category than in the others. However, the effect had the same sign in all price categories: impact on price was less positive, or becomes negative, as the quality rating decreased. It is worth noting that the quality ratings of the wines in the super premium category showed very little variation (only 9.77 points between the significant variables). This suggested that potential efforts to improve quality ratings would not significantly influence the demand for these wines. In the ultra premium category, there was a linear relationship between the quality rating and the price impact, which was positive only for wines that qualify as classic. These results highlighted the importance of an excellent quality rating (outstanding or classic), given that lower ratings had a negative or zero price impact.

The results also revealed an inverse relationship between winery size and wine price. The impact of the exclusiveness factor was eroded in the lower price category, while it was stronger in the super premium category, where large firm size had a negative impact 4.14 percentage points higher than that of small firm size (-2.7%), while micro-firm size had a positive impact of 2.7%. In the ultra premium category, although large firm size was penalized (-6.7%), micro-wineries did not obtain higher positive impact than they did in the super premium category (2.6%) and, in fact, small wineries had more negative impact than medium-sized ones. A possible explanation for this is that the exclusiveness factor was captured, not by firm size, but by the boutique variable, which was the most influential variable in this price category.

The type of wine was not significant in the popular premium category, while, in the super premium and ultra premium categories, the most negative impact was for rosés (-6% and -23.9% respectively) and reds had more positive impact than whites.

Impacts on prices of the most recent vintages were more positive than the older ones in all three price categories. In the super and ultra premium categories, a linear relationship could be observed. However, in the popular premium category, three different behaviors could be identified. 80s vintages had the most negative impacts, while most of the 90s vintages were below the average price and their effect, while significant, was not very strong. The more recent vintages had a positive influence on sale price. Thus, as vintages approach the current year, impact on prices improved from negative scores to the top score which was assigned to vintage 2010 which had an impact of 13.4% ⁷ .

With regard to appellation, in the popular premium category, the most positive impacts on prices were Bierzo, Penedès, Rioja, Toro, Catalunya region, Rueda, Somontano, Navarra, La Mancha, Jumilla and Calatayud , that is, a wide spread across the whole country. The most favorable impact labels among the super premium wines, however, were those from north and northwest Spain, while the majority of those with negative impacts were from the south, although there were also some from northern regions ( Aragón, Navarra and Catalunya ). In the ultra premium category, most appellations (73%) had positive impact, the top going to wines from Ribera del Duero (+24.2%). The appellation with the greatest negative impact was Valdepeñas (-40.9%). It is worth noting that some appellations ( Yecla, Utiel-Requena, Cariñena, Vinos de la Tierra de Castilla or Valdepeñas ) had no positive impact in any price category. Furthermore, the impacts for Valdepeñas wines decreased as their retail price increased, reflecting poor consumer-perceived quality, whereas some appellations, such as Rioja, Ribera del Duero, Priorat, Penedès, Rueda and Toro, had positive impact in all price categories where they were significant.

It is also important to mention some of the results obtained for the two best-known Spanish appellations, Rioja and Ribera del Duero , which were also the two most represented in the sample. While the positive impact for the Rioja appellation diminished as the sale price increased, Ribera del Duero obtained its highest positive impact in the ultra premium price category, where most of the wines with this appellation were concentrated (84% of wines with this appellation class as ultra premium and 1.3% as popular premium). This concentration was a prestige factor that promoted willingness to purchase a Ribera del Duero in the upper price categories.

Rías Baixas, Calatayud, Navarra or Montsant had negative impact in the higher price categories but positively in lower price categories. This appeared to suggest disinclination to purchase beyond a certain price threshold. Positive impact on prices of some appellations, such as Campo de Borja or Vinos de la Tierra de Castilla y León , on the other hand, increased with the price of the wine. The influence of other appellations was erratic and the price effect could be positive or negative.

The first conclusion arising from the results of this research was that all the wine attributes considered in the hedonic function were valued by US consumers and had an impact on the sale price. So any business and marketing strategy directed at any of these characteristics is relevant to the positioning of a specific wine on the US market. The super premium category is a more complex decision-making environment, because the number of attributes influencing demand was greater. In other words, there are various potential product-positioning options. Special care must be taken with decision-making in the top and bottom price categories, however, because there were fewer attributes to enable market success, and therefore fewer factors on which to work.

US consumers set great store by the aging of wine, which was, overall, the factor with most influence on the sale price (a similar result is obtained by Morilla & Martínez, 2002 ; Melo et al ., 2005 ; and Perza, 2010 ), except in the super premium category, where other attributes, particularly vintage year and appellation, were more highly valued. Significantly, the US consumer valued length of aging and high-end price placement which reflects specific characteristics, as in the case of boutique wines, especially in the ultra premium price category. In this respect, it would be appropriate to analyze the associated costs of aging wines for longer periods, or producing a boutique wine targeted to this market, in order to determine whether it would increase profits. It is surprising to note, however, that crianzas were valued similarly or even lower than joven wines, so there may be no profit from aging wine to qualify as crianza , as suggested by Rodríguez & Castillo (2009) .

On the other hand, it can be observed that the US consumer appreciated the chance to buy a quality wine for less than expected in the two lowest price categories, which suggests that marketing strategies based on discounts or special promotions might enable product positioning in these markets. This, combined with factors such as the good value for money that Spanish wine offers the average consumer ( ICEX, 2014 ), the high and increasing willingness of the US consumer to discover new wines ( Del Rey, 2014b ), and the growing interest of young consumers in imported wines ( Wine Market Council, 2012 ), constitutes an incentive to design customer-attracting promotions and establish distributor/retailer agreements.

The exclusiveness factor, measured by firm size, can serve as a market differentiation tool in the US because it represents added value for US consumers, who, according to results reported in the literature ( Morilla & Martínez, 2002 ; Rodríguez & Castillo, 2009 ; Kwong et al ., 2011 ; Delmas & Grant, 2014 ), attach greater value to wines produced by micro-enterprises than to those produced by large wineries. Wine price decreased with size of winery in all price categories, but the effect was more marked in the super premium category. High-end wine market consumers showed greater appreciation of micro-wineries and greater rejection of large wineries than was the case with the consumers of less expensive wines.

With regard to the type of wine, reds were more highly valued than whites or rosés, which were negatively valued by US consumers, which confirms the results obtained by Steiner (2009) for Spanish wines in the British market. This picture appears to mirror Spanish export volume and market share trends for the different types of wine in the US, where red wine exports have gained ground in recent years at the expense of rosés, which have shrunk significantly. In any event, this emerged as the least relevant attribute for consumers making a purchase decision, and it had no significance whatever in the lowest price category, despite having been reported as a relevant factor in other studies ( Panzone & Simões, 2009 ; Yoo et al ., 2011 ).

US consumers were found to attach very high value to expert quality ratings, especially at the high end of the market, where the importance of other factors decreases ( Schroeter et al ., 2011 ; Cuellar & Claps, 2013 ). This justifies efforts to improve classifications, especially in the ultra premium category. Our results showed a proportional relationship between quality rating and consumer valuation, as reported in past studies ( Schamel & Anderson, 2003 ; Cacchiarelli et al ., 2014 ). However, it is worth noting that only quality scores above 90 were significantly effective when it came to compensating the investment effort they required. Furthermore, strategies needed to be directed not only at achieving high quality scores but also at making them known to the public in order to stimulate demand and support higher prices ( Hilger et al ., 2011 ; Cardebat & Figuet, 2014 ).

Vintage year was generally considered a factor of medium relevance in decision-making, and thought to have less importance at the high end than at the middle and low end of the wine market. Our analysis showed the most recent vintages to be the most highly valued, possibly due in part to the increased availability and familiarity of Spanish wines in the US market in the last few years. In any case, there would be little to be gained from exporting older vintages, especially from the cheaper end of the market, since the costs could outweigh the profits.

Appellation and soil and climate conditions played a relevant role in consumer choices, confirming the findings of other authors ( Brentari et al ., 2011 ; Amrusch & Wirl, 2013 ; Caracciolo et al ., 2013 ; Roma et al ., 2013 ). The location of a winery can therefore either help or hinder its positioning in the US market. Potential investment in new wineries should therefore be carefully assessed, as the disadvantages of an erroneous location could not be reversed by enological treatments or short-term investment ( Troncoso & Aguirre, 2006 ) or by brand strengthening tactics ( Boatto et al ., 2011 ).

It is remarkable that, the various appellations in the upper price category received mostly positive ratings, supporting the idea of a certain prestige factor and a positive overall image in association with Spanish wines. However, there were few significant appellations in this category, so further promotion and positioning efforts are needed to increase awareness and acceptance by US consumers and thus improve that prestige factor.

Overall, the highest ratings went to wines from northern Spain, while those from the southern, surrounding regions and Aragón received negative ratings. According to these findings, certain appellations, such as Y ecla, Utiel-Requena, Cariñena, Vinos de la Tierra de Castilla and, particularly, Valdepeñas , need to modify their promotional strategies, and work intensively to improve their positioning.

Higher market penetration was found to enhance consumers' wine knowledge and raise their "liking" assessment, which is consistent with the findings of Steiner (2009) for Spanish wines in the British market. Wines from Rioja, Ribera del Duero, Priorat, Penedès, Rueda and Toro , all with strong presence in the sample, received positive ratings. The exceptions to this pattern, Navarra and Rías Baixas , need to intensify their promotion efforts and adjust their marketing strategies, as consumers showed no willingness to pay high prices for these wines. The ratings for wines from Campo de Borja or Vino de la Tierra de Castilla y León , on the other hand, increased with higher prices, which seems to suggest that they enjoy better positioning, especially in the upper price categories.

In relation to this, there are public programs to promote firms exporting to foreign markets available to wineries and wine producers and a growing role is played by EU promotion programs in third countries ( MAGRAMA, 2012 ).

Our results suggested that strategies for the Spanish wineries to obtain higher prices in the US wine market were dependent on the wine price category. In the popular premium category, the most influential attribute was aging. In the middle category (super premium), moreover of aging, the exclusivity factor measured through the size of the firm could play an important role. Marketing tools could be used in this area by wineries trying to improve their relative position. Additionally, wineries could take some advantage of the region of origin in this category, because this attribute had more influence in this segment. Finally, in the ultra premium category aging wines for longer periods or producing a boutique wine and try to obtain a good quality rating are the two main objectives to reach by wineries.

Future studies might consider regional climate variables and their evolution over time, with a view to revealing possible influences and changes in buyer behavior. Further research could explore US consumer behavior and factors influencing willingness to purchase each of the appellations considered in this work.