II. Quality Assurance Systems

The adoption of quality assurance systems in the agrifood sector has been examined from a variety of perspectives. Karipidis et al. (Reference Karipidis, Athanassiadis, Aggelopoulos and Giompliakis2009) examine factors affecting the adoption of quality assurance systems in small food enterprises, identifying a set of internal and external benefits and costs of quality assurance systems at the firm level. Much of the prior literature has a heavy emphasis on the adoption of quality assurance systems as a component of food safety management systems (see, e.g., Herath and Henson, Reference Herath and Henson2006; Hobbs, Reference Hobbs2004). This literature often frames the analysis within the context of requirements by food retailers for the adoption of quality assurance requirements by suppliers (see, e.g., Hobbs, Reference Hobbs1996; Mainville et al., Reference Mainville, Zylbersztajn, Farina and Reardon2006; Northen, Reference Northen2001). Other work has examined the adoption of private-sector quality assurance systems within a broader international trade context (Hooker and Caswell, Reference Hooker and Caswell1999). Researchers have also examined quality certification within an environmental context. For example, in an application to the wine sector, Delmas and Grant (Reference Delmas and Grant2014) find that ecocertification leads to a price premium, whereas the use of an ecolabel on wine does not. Yiridoe and Marett (Reference Yiridoe and Marett2004) examine how organizations seek to reduce the costs of ISO14001 (environmental) certification with an application to a consortium of seven wineries in New Zealand. Unlike eco- or environmental certification, which focuses on a specific set of credence attributes, VQA is a broader signal of wine quality and assurance of authenticity across a set of recognized appellations.

Previous research examining collective reputation and regional quality signals in the wine industry has tended to focus on whether a price premium exists for wine from a particular region or wine with a specific quality designation, often through use of hedonic regression analysis (see, e.g., Carew and Florkowski, Reference Carew and Florkowski2008; Costanigro, McCluskey, and Goemans, Reference Costanigro, McCluskey and Goemans2010; Kwong et al., Reference Kwong, Cyr, Kushner and Ogwang2011; Levaggi and Brentari, Reference Levaggi and Brentari2014; Rabkin and Beatty, Reference Rabkin and Beatty2007; Rickard, McCluskey, and Patterson, Reference Rickard, McCluskey and Patterson2015; Schamel, Reference Schamel2009; Steiner, Reference Steiner2004). In an analysis of the Italian wine market, for example, Levaggi and Brentari (Reference Levaggi and Brentari2014) find that consumers value only verifiable characteristics as a proxy for quality attributes (including DOCGFootnote ¹ and the appellation) rather than generic but not externally verifiable attributes such as labels indicating “Reserve” or “Superior.” Castriota and Delmastro (Reference Castriota and Delmastro2012) examine the effect of individual winery factors, collective reputation, and institutional regulation on the individual firm reputation of Italian wineries.

Although also examining a quality assurance program and a collective reputation signal (VQA), this article differs in motivation and approach relative to the existing literature. Typically, in the quality assurance literature, a decision to adopt quality assurance certification is examined at the firm level, and this decision affects the entire output of the firm. The wine industry is different. The industry is characterized by a high degree of product differentiation, such that even the same winery will produce a range of wine qualities, targeted at different markets. Thus, the firm's VQA certification decision is made on a wine-by-wine basis. The article also differs in its use of a retail (revealed preference) data set to examine the adoption of quality certification on a product-by-product basis rather the use of firm-level stated preference survey data to examine a whole firm adoption decision. Given that collective and regional reputation signals tend to be associated with price premiums (e.g., Levaggi and Brentari, Reference Levaggi and Brentari2014; Rickard, McCluskey, and Patterson, Reference Rickard, McCluskey and Patterson2015), and this has been found to be the case for VQA wines in Canada (Rabkin and Beatty, Reference Rabkin and Beatty2007; Ugochukwu, Reference Ugochukwu2015), this article seeks to understand the factors affecting VQA certification and why wineries seek certification for some wines but not others.

III. The Canadian Wine Sector and Vintners Quality Alliance

Wine production in Canada began in 1862 in southern Ontario, and Canadian wine became available in the international market in 1981 (VanSickle, Reference VanSickle2011). Currently, there are 650 wineries operating in Canada.Footnote ² Two provinces account for the vast majority of wine production in Canada—Ontario and British Columbia—which together produce about 98% of Canadian wine. Quebec and Nova Scotia account for most of the remaining 2% of production (VanSickle, Reference VanSickle2011). The major wine-producing subregions in Canada include the Okanagan Valley in British Columbia and the Niagara Peninsula (Niagara-on-the-Lake) in Ontario, as well as Lake Erie North Shore and Pelee Island and the Niagara Escarpment in Ontario. A number of emerging wine regions exist including, among others, Prince Edward County and Georgian Bay in Ontario and Similkameen Valley and the Gulf Islands in British Columbia.Footnote ³ In Ontario, there are 190 registered wineries, of which 151 (79.5%) are registered with VQA and have some wines certified as VQA (VQA Ontario, 2015). In British Columbia, 173 of 270 wineries (64%) are registered with the British Columbia VQA (BC VQA).

The importation, distribution, and retailing of both domestic and imported wines in Canada is organized and regulated on a province-by-province basis through provincial liquor control boards (LCBs) (Carew and Florkowski, Reference Carew and Florkowski2012). The primary (in some cases, sole) retail outlet for wine in most Canadian provinces is the provincial LCB stores. The exception is Alberta where retail distribution of wine was privatized in 1993 (Alberta Gaming and Liquor Commission, n.d.). The LCBs also control which wines can be imported and distributed in each province, making the Liquor Control Board of Ontario (LCBO) one of the world's largest buyers and retailers of alcoholic beverages, including wine (LCBO, 2016). In Ontario, the LCBO uses a standard price markup for all domestic and imported wines irrespective of whether the wine has VQA certification. The regulation of wine distribution on a province-by-province basis has created interprovincial barriers to the movement of wine between provinces in Canada, reducing the extent to which wines produced outside of a province are sold in another province (Madill, Riding, and Haines, Reference Madill, Riding and Haines2003). This means that Ontario-produced wines dominate the retail market for domestic wines in Ontario, and the same is true of British Columbia–produced wines on the British Columbia retail market.

Although initially dependent on imported ingredients, in 1988 a watershed event for the Canadian wine industry saw the production of 100% Canadian wine from homegrown vinifera grapes. Since that time, the vineyard and winery sector in Canada has continued to grow, with the volume of grape production increasing from 61,000 metric tons in 2007 to 78,000 metric tons in 2015 (Agriculture and Agri-Food Canada, 2016). Assuming approximately 700 liters of wine produced per metric ton of grapes, this equates to approximately 54.6 million liters of wine annually from domestically produced grapes (Agriculture and Agri-Food Canada, 2016). Further impetus for growth and quality improvements was provided by the 1989 U.S.-Canada Free Trade Agreement (expanded to the North American Free Trade Agreement in 1994), which shifted the competitive landscape for Canadian wines with new competition from more established U.S. wine-producing regions. The enhanced competition is widely credited with creating the impetus for investments in higher-quality grape varieties and improved wine-making techniques.

Canadian wines have been recipients of a number of international awards in recent years. Examples include Double Gold for a 2012 Merlot from Burrowing Owl Estate Winery at the San Francisco International Wine Competition in 2015Footnote ⁴ and World's Best Pinot Noir under £15 for a 2011 Pinot Noir from Mission Hill Family Estate winery at the 2013 Decanter World Wine Awards (Montgomery, Reference Montgomery2013), as well as numerous international awards for Canadian ice wines.Footnote ⁵

Canada is the largest producer of ice wine in the world because of its relatively cold winter weather that favors production of ice wine, a dessert-style wine, which is typically sold at a premium relative to other types of wine. The harvesting of grapes used for ice wine production occurs during the colder months when the temperature is below −10 °C (14°F). At this temperature, the grapes are “naturally frozen on the vines” and contain less juice, which makes the wine sweeter.Footnote ⁶ Production of ice wine requires intensive and specialized labor, and output is usually small relative to other wines. For example, a 375 mL bottle of ice wine is produced using 3.5 kg of grapes, whereas this quantity of grapes would produce about six 375 mL bottles or three 750 mL bottles of ordinary table wine.Footnote ⁷ These factors, along with a growing demand for this premium wine, explain the price premium for ice wine. This premium has served as an incentive for counterfeiting, especially in China where there have been reports of fake ice wine labeled “Made in Canada” available on the Chinese market (O'Donnell, Reference O'Donnell2011).

A number of agencies have played important roles in the development of the Canadian wine industry including the Canadian Wineries Association, which represents wineries and promotes the adoption of production-enhancing innovations that improve grape and wine production. The VQA has also been a key driver of quality improvements. Established in 1988 as a nonprofit organization in Ontario, VQA was given the responsibility of implementing the provincial Vintners Quality Alliance Act enacted in May 1999 and passed into law in June 2000. VQA is Ontario's wine regulatory authority responsible for maintaining the integrity of local wine appellations and enforcing wine-making and wine-labeling standards.Footnote ⁸ It is similar to wine regulatory agencies in other countries, such as appellation d’ origine contrôlée in France, denominazione di origine controllata in Italy, and Qualitätswein Mit Prädikat in Germany. The VQA Act established a designation of origin, which consumers can use to identify quality wines produced in Ontario using homegrown grapes and certified to meet specific production method and other standards.

The VQA system was adopted by the province of British Columbia in 1990. Although based on similar standards, both VQAs are managed under separate provincial legislations (Rabkin and Beatty, Reference Rabkin and Beatty2007). Registration and use of VQA is voluntary for wineries. Wines that pass VQA authenticity tests bear the VQA logo on the bottle label to signal quality. Only wines produced in Ontario and certified to VQA standards can bear the “VQA” designation, whereas wines produced in British Columbia to VQA standards are labeled “BC VQA.”

The provincial VQAs have the mandate of ensuring authenticity of all VQA wines produced in Canada. They are also responsible for the development of standards and regulations that govern Canadian VQA wines. For example, in Ontario VQA establishes quality and process standards for grapes used by wineries, including the grape varieties used in production and the level of ripeness at which the grapes must be harvested to ensure high-quality wine. It carries out tests to ensure authenticity (i.e., ensuring that the wine is made from 100% Canadian grown grapes in designated regions). It regulates techniques used in wine making and establishes sensory/chemical criteria for finished wine. VQA also reviews wine labels to ensure that the wine inside the bottle correctly reflects what is on the label before the wine is placed on the shelf (VQA Ontario, 2004). Monitoring of vineyards is usually carried out by the representatives of the Grape Growers’ Association within each province.Footnote ⁹ VQA (BC VQA) is also responsible for approving and regulating new appellations recognizing wine-producing subregions in Ontario (British Columbia).

As a rule, a wine receives VQA approval and is certified as of Canadian origin after passing a series of quality assurance tests including sensory, laboratory analysis, and packaging review tests. VQA carries out on-site audits of all wineries every 5–8 months. The organization also conducts random wine inspections at Canadian retail stores in order to ensure that the VQA packaging and labeling standards are strictly observed.

VQA certification is achieved on a wine-by-wine basis. Focusing on the Ontario retail market for Canadian-produced wines, both VQA and non-VQA, this article explores the factors that determine a winery's decision to seek VQA status for a wine.

IV. Data and Estimation Strategy

To examine who chooses VQA certification and the factors influencing a winery's decision whether to seek VQA certification for a specific wine, a probit regression analysis is conducted. The analysis uses a rich data set sourced from the LCBO comprising data on Canadian wines sold in Ontario between 2007 and 2012. The data set comprises 1,537 observations of different types of wine produced in Ontario (89.86%) and British Columbia (10.14%) over the 6-year time period. Data on the sale of imported wines in Ontario during this period were not available. Wines from British Columbia are excluded from the analysis because there are some wineries (e.g., Mission Hill) that produce a large volume of wines in British Columbia that are underrepresented in the data set and, therefore, may be erroneously regarded as small wineries.

Wines produced using grapes that are not grown in Canada do not qualify for VQA certification. International-Canadian Blends may be composed of combinations of Canadian grapes and internationally sourced grapes or juice and are not eligible for VQA status. These wines tend to be low priced and high volume. In the case of blended wines, the data set does not specify whether a blended wine is an International-Canadian Blend or consists of blended Canadian varietals. For this reason, all blended wines in the data set were dropped from the analysis, leaving a final sample size of 1,023 observations for the probit analysis.

Each observation in the data set contains information on the LCBO retail price, whether or not the wine is VQA certified, the name of the wine (which was used to identify the winery and the region or origin), volume of wine supplied (measured as the number of cases of each wine supplied by a winery to the Ontario retail market from 2007 to 2012), and the type of wine. Additional information included the percentage alcohol content, sweetness (sugar or Brix level), vintage, and varietal. Product score (ranking by wine experts) was only available on a relatively small subset of the data and therefore could not be used in this analysis.

In a probit model, the dependent (response) variable is binomial and is based on the assumption that the functional form follows a normal (cumulative) distribution (Green and Hensher, Reference Green and Hensher2010, p. 20). The dependent variable in this case is whether the wine held VQA status. In the context of this article, a winery's decision to seek VQA certification for a specific wine is based on a vector of factors, X, including the price of wine (VQA certified and non-VQA base wine), a winery's characteristics (volume of sales, region), wine type/color, and the error term, which is normally distributed with zero mean.

Following Adamowicz et al. (Reference Adamowicz, Boxall, Williams and Louviere1998) and Olynk, Tonsor, and Wolf (Reference Olynk, Tonsor and Wolf2010), assume that the benefit to the ith winery is the following:

(1) $$U_{i1} \left( X \right) = \beta _1 X_i + \varepsilon _{i1} \; \; \; \; \left( {{\rm if\; it\; uses\; VQA}} \right)$$

(2) $$\; U_{i0}\left( X \right) = \beta _0X_i + \varepsilon _{i0}\; \left( {{\rm if\; it\; does\; not\; use\; VQA}} \right){\rm \; }$$

The decision by the ith winery to apply for VQA certification for a specific wine is such that

(3) $$Y_i = \left\{ {\matrix{ {1\; if\; U_{i1} \gt U_{i0}} \cr {0\; if\; U_{i1} \le U_{i0}} \cr}} \right.,\; \; $$

where Y _i  = 1 if the winery has VQA certification for a specific wine, and Y _i  = 0 otherwise.

Therefore, the probability of using the VQA system for a specific wine is given by

(4) $${\rm Pr}\left( {Y_i = 1} \right) = {\rm Pr}\left( {U_{i1} \gt U_{i0}} \right) = {\rm Pr}\left( {\beta _1 X_i + \varepsilon _{i1} \gt \beta _0 X_i + \varepsilon _{i0}} \right).$$

Considering the assumption of a normal distribution of the error term in probit analysis (Green, Reference Green2008, p. 777), and following Gujarati and Porter (Reference Gujarati and Porter2009, p. 566), equation (4) can be rewritten as

(5) $${\rm Pr}\left( {Y_i = 1} \right) = {\rm Pr}\left( {\varepsilon _{i0} - \varepsilon _{i1} \lt \beta _1 X_i - \beta _0 X_i} \right) = {\rm Pr}\left( {\varepsilon _i \lt \beta X_i} \right),$$

(6) $${\rm Pr}\left( {Y_i = 1} \right) = {\it \Phi} \left( {\beta X_i} \right).$$

This implies that

(7) $${\rm Pr}\left( {Y_i = 1} \right) = {\it \Phi} \left( {\mathop \sum \limits_{k = 1}^K \beta _i X_i} \right)$$

and

(8) $${\rm Pr}\left( {Y_i = 0} \right) = 1 - {\it \Phi} \left( {\mathop \sum \limits_{i = 1}^I \beta _i X_i} \right),$$

where Φ is the cumulative distribution function of the standard normal distribution, β _i represents parameters estimated using maximum likelihood estimation; and X′ is the vector of independent variables influencing a winery's decision to secure VQA certification for a specific wine.

Therefore, the probability of a winery having VQA certification for a specific wine is given by the following:

(9) $$P\left( {Y_i = 1 \vert X} \right) = \mathop \int \limits_{ - \infty} ^{X^{\prime}\beta} \displaystyle{1 \over {\sqrt {2\pi}}} e^{ - \displaystyle{{Z^2} \over 2}} dz = {\it \Phi} \left( Z \right),$$

where Z = β ₀ + β ₁ X _1i  + … + β _k X _ki .

Thus, the probit model for the use of VQA by a winery is explicitly stated as follows:

(10) $$E\left( {Y_i} \right) = \beta _0 + \beta _1 Price + \mathop \sum \limits_{k = 1}^K \beta _k Region + \mathop \sum \limits_{l = 1}^L \beta _l Volume + \mathop \sum \limits_{n = 1}^N \beta _n Color/Type,\; $$

where Y _i is the response (dependent) variable, VQA; β ₀ is the constant; and β ₁ , β_k, β_l , and β _n represent the coefficients of explanatory variables. Following Green (Reference Green2008, p. 778), the likelihood function used to estimate the parameters is given by

(11) $$L\left( \beta \right) = \mathop \prod \limits_{i = 1}^N P\left( {Y_i} \right) = \mathop \prod \limits_{i = 1}^N {\it \Phi} (\beta X_i)^{Y_i}{\it \Phi} \left( { - \beta X_i} \right)^{1 - Y_i}.$$

The marginal effects of the parameter estimates measure the impact of a change in an independent variable on the expected change in the dependent variable.Footnote ¹⁰ Marginal effects evaluated at the mean values of the covariates (X) (i.e., holding other variables constant at their sample mean values) were estimated. Following equation (6) and Green (Reference Green2008, p. 782), the marginal effect of the continuous explanatory variable (i.e., price) was estimated as follows:

(12) $$ME_c = \displaystyle{{\partial {\rm Pr}\left( {Y = 1} \right)} \over {\partial X_k}} = \phi \beta _k \left( {X_i^{\prime} \beta} \right),$$

where ϕ(Xβ) is the standard normal density evaluated at Xβ such that $\phi \left( {X\beta} \right) = \displaystyle{1 \over {\surd 2\pi}} e\displaystyle{{ - X\beta ^2} \over 2}$ , and φ is the probability density function of a standard normal variable.

Following Green (Reference Green2008, p. 781) and equations (7) and (8), the marginal effects of the dummy variables were calculated as follows:

(13) $$\Delta = {\it \Phi} \left( {X\beta {\rm \vert} X_i^k = d = 1} \right) - {\it \Phi} \left( {X\beta {\rm \vert} X_i^k = d = 0} \right),$$

where d represents dummy.

Table 1 Definition and Measurement of Variables

Note: LCBO, Liquor Control Board of Ontario; VQA, Vintners Quality Alliance.

The definitions and measurement of variables used in the analysis are provided in Table 1. Of the wines in the data used for analysis, 77.4% are VQA wines (314 red, 370 white, 42 ice wine, 37 rosé, and 29 sparkling), whereas 231 wines (7.43%) do not have VQA certification (88 red, 112 white, 22 rosé, and 9 sparkling). All ice wines held VQA certification. As expected, ice wine had the highest average price per bottle (Can$115.38), whereas rosé wines had the lowest average price (Can$15.70). In terms of subregional representation, of the 1,023 wines in the data used for analysis, Niagara Peninsula dominates (727 wines), followed by Prince Edward County (99), Lake Erie North (101), and “other Ontario” (96). A total of 147 Ontario wineries are represented in the data set, 73% (107) of which fall into the “very small” category (less than 200 cases distributed in Ontario over the 6-year time period), 20% (30) are small volume suppliers (200–499 cases), 5% (7 wineries) are medium volume suppliers (500–999 cases), and three are large volume wineries, meaning they supply more than 1,000 cases over the period captured by the data set. Table 2 summarizes the composition of wineries across the five size categories in terms of the number of wines produced and VQA certification.

Table 2 Distribution of Wineries by Size

Note: VQA, Vintners Quality Alliance.

Source: Liquor Control Board of Ontario (2013).

Table 2 indicates that of the 1,023 wines in the final data set sold in Ontario within the period under consideration, 9.87% (101) were supplied by 3 large wineries; 17% (175), by 7 medium-sized wineries; 20%, by 30 small wineries; and 50.7%, by 107 very small wineries. The table also shows the distribution of VQA and non-VQA wines by size of winery.Footnote ¹¹

Table 3 shows the breakdown of wineries in the data that used the VQA system for all their wines (89%), some of their wines (4.76%), and none of their wines (6%). The seven wineries that certify some (but not all) of their wines are particularly of interest and appear to be fairly diverse, comprising one large (Jackson-Triggs), two medium, two small, and two very small wineries. In addition, compared with other large wineries in the data set (Henry of Pelham, Cave Spring), wines supplied by Jackson-Triggs have the lowest average price of Can$19.34.

Table 3 Classification of Wineries Based on Use of VQA Certification

Note: VQA, Vintners Quality Alliance.

Source: Liquor Control Board of Ontario (2013).

V. Probit Results and Hausman Test

A probit analysis was carried out to identify the factors that influence a winery's decision to apply for VQA certification for a specific wine. The parameter estimates are shown in Table 4, and the marginal effects in Table 5. The marginal effects of the explanatory variables measure the impact (effect) of a change in the each explanatory variable (X _i ) on the expected change in the dependent variable (Y). The base wine in the regression is a rosé wine, produced by a very small winery in the “other Ontario” category.

Table 4 Probit Regression for the Use of VQA Certification by Wineries

Notes: Asterisks (*, **, and ***) indicate P values of 0.1, 0.05, and 0.01, respectively. LENS, Lake Erie North Shore; PEC, Prince Edward County; VQA, Vintners Quality Alliance.

Source: Liquor Control Board of Ontario (2013).

Table 5 Marginal Effects

Notes: Asterisks (*, **, and ***) indicate P values of 0.1, 0.05, and 0.01, respectively. LENS, Lake Erie North Shore; PEC, Prince Edward County.

Source: Liquor Control Board of Ontario (2013).

Price was not a significant determinant of a winery's decision to seek VQA certification for a specific wine. Winery region, volume supplied by the winery, and type of wine all influence the probability of a wine being VQA certified. Positive and statistically significant coefficients for each of the winery subregions relative to the base category “other Ontario” show that wines from these regions are more likely to be VQA certified. The marginal effect for Niagara Peninsula suggests that, all else held constant, the probability of a winery seeking VQA certification for a wine increases by 54% for wineries located in the Niagara Peninsula, and by 35% for wineries located in Lake Erie North Shore relative to the base category (“other Ontario”). Wines from Prince Edward County are also more likely to be VQA certified, although the effect is smaller. These results are not surprising. Niagara Peninsula is a well-established wine-producing subregion in southern Ontario, and 71% (727 out of 1,023) of the wines in the data set are produced in the Niagara Peninsula, whereas 9.67% (99 out of 1,023) and 9.87% (101 out of 1,023) of the wines are produced in Prince Edward Country and Lake Erie North, respectively. “Other Ontario” (the base subregion) including Pelee Island, South Georgian Bay, and Lake Erie South Coast are newer viticultural areas within Ontario and produced about 9.3% of the wines in the data set used for analysis. Of the 172 wineries in Ontario in March 2013, 140 (81.4%) were registered with VQA (VQA Ontario, 2013). Therefore, the probability of a winery located in one of the major Ontario wine-producing regions seeking VQA certification would be high.

Wine from large and medium-sized wineries (relative to very small wineries) is more likely to be VQA certified. The size of these effects is similar, with the probability of VQA certification increasing by 18% for wine from large wineries and by 16% for wine from medium-sized wineries. The type of wine also affects the probability of VQA certification, particularly for ice wine, which was positive and significant at 5%, with a marginal effect of 14%. This result is consistent with a priori expectations. Ice wines tend to be high-priced wines, and wineries may see VQA certification as being particularly important to signal quality and authenticity, both in the domestic market and in international markets. The coefficients for red and white wines were positive and weakly significant (P < 0.1), suggesting that, relative to the base category (rosé), these wines were more likely to be VQA certified, although the effect is much weaker than for ice wines.Footnote ¹²

Wineries incur costs in seeking and maintaining VQA certification for a wine. These costs include pecuniary costs in the form of testing and membership fees, as well as costs imposed by production restrictions to meet the VQA standards. For example, only 100% regionally grown grapes can be used for VQA wines, which are monitored and harvested at a particular level of ripeness. Addition of artificial sugar is not permitted. The VQA membership fee is a flat rate irrespective of the volume of wine supplied by a winery and therefore imposes some fixed costs, which may favor certification by larger volume suppliers, explaining the significance of the winery size variable for the two largest size categories.

Other opportunity costs are more difficult to evaluate, for example, the search costs of learning about VQA requirements and production methods. An attempt was made to control for the fixed costs of seeking VQA certification (e.g., costs of learning about the VQA system, costs of documenting compliance with VQA certification). A new dummy variable was defined as “experience with VQA.” Wineries in the data set that have all or some of their wines certified by VQA take a value of 1, and wineries with no experience with VQA (have none of their wines certified) take a value of 0. The variable was included in the probit model as a check; however, the coefficient was not significant.

A. Endogeneity Test for Price

Previous studies find that there is a premium for VQA certification (Rabkin and Beatty, Reference Rabkin and Beatty2007; Ugochukwu, Reference Ugochukwu2015); however, the probit results show that the price of the wine is not a significant factor that drives a winery's decision to seek VQA certification for a specific wine. A potential endogeneity issue exists and deserves attention—that is, do wineries seek VQA certification for higher-priced wines, or does VQA certification lead to higher wine prices? An instrumental variable (IV) estimation is one way in which to address this issue. Use of the IV approach requires a valid instrument (z) for the endogenous independent variable (x), which must meet two conditions (Green, Reference Green2008, p. 316), including the following:

1. 1. Instrument exogeneity: the instrument must be uncorrelated with the error term—that is, Cov(z, ε) = 0.

2. 2. Instrument relevance: the instrument must be highly correlated with the endogenous independent variable—that is, Cov(z, x) ≠ 0.

In the present analysis, however, IV estimation was not feasible because the independent variables in the model were not highly correlated with price, and therefore, none of these variables could serve as a valid instrument. The correlation analysis of the independent variables and price is reported in Appendix 1 and shows that the independent variables would perform as weak instruments. A weak instrument often results in coefficients that are biased and thus may lead to misleading inferences (Stock, Wright, and Yogo, Reference Stock, Wright and Yogo2002).

In the absence of an IV estimation strategy, to determine whether price and VQA are endogenous, a test of endogeneity (Hausman specification or t-test) was carried out following Gujarati and Porter (Reference Gujarati and Porter2009, p. 703) and Pindyck and Rubinfeld (Reference Pindyck and Rubinfeld1990, p. 304). Considering equation (10), assume the following models:Footnote ¹³

(14) $$V_{qa} = \beta _0 + \beta _1 P_i + \beta _2 R_g + \beta _3 V_S + \beta _4 C_l + \varepsilon _i, \; $$

(15) $$V_{qa} = \beta _0 + \beta _1 P_i + \varepsilon _{2i}, $$

whereV _qa is VQA certification, P _i is wine price, R _g is winery region, V _S is volume of wine supplied by a winery to the Ontario retail market, C _l is wine color/type, and ε _i is the error term.

Assume P _i and V _qa are endogenous (i.e., P _i is correlated with ε _i ), and R _g , V _S , and C _l are exogenous (uncorrelated with ε _i ). To verify if there is endogeneity between price and VQA, a two-stage least square regression was run, and a Hausman's or t-test for endogeneity was conducted. First we rewrite equation (14) in a reduced form as follows:

(16) $$P_i = \alpha _0 + \alpha _1 R_g + \alpha _2 V_S + \alpha _3 C_l + \mu _i, $$

where μ _i is the reduced form of the error term. Equation (16) was estimated using ordinary least squares (OLS), and the regression residuals ( $\widehat{{\mu _i}} )$ and predicted values of price ( $\widehat{{P_i}} $ ) calculated. The results of the OLS regression are shown in Table 6.

Table 6 Ordinary Least Squares Regression Results of the Reduced Form of Equation 14 (Equation 16)

Notes: Asterisks (*, **, and ***) indicate P values of 0.1, 0.05, and 0.01, respectively. Omitted (base) categories include “rosé” for wine color/type, “very small” for volume supplied to the Ontario market, and “other Ontario” for winery region. LENS, Lake Erie North Shore; PEC, Prince Edward County.

Source: Liquor Control Board of Ontario (2013).

$P_i = \widehat{{P_i}} + \widehat{{\mu _i}} $ , where $\widehat{{P_i}} $ represents the estimated price, and $\widehat{{\mu _i}} $ is the residual from the OLS regression. Substituting this for P _i in equation (15), we have

(17) $$V_{qa} = \beta _0 + \beta _1 \widehat{{P_i}} + \beta _1 \widehat{{\mu _i}} + \varepsilon _{2i}. $$

The second stage involves estimation of equation (17) using OLS and performing a test on the coefficient of $\widehat{{\mu _i}} $ . Table 7 presents the results.

Table 7 Regression of VQA against Estimated Values of Price and Residuals of the Error Term

Notes: Asterisks (*, **, and ***) indicate P values of 0.1, 0.05, and 0.01, respectively. VQA, Vintners Quality Alliance.

Source: Liquor Control Board of Ontario (2013).

Testing the null hypothesis of no endogeneity between VQA and wine price versus the alternative, Table 7 shows that at the 1%, 5%, and 10% levels, the coefficient of the residual ( $\widehat{{\mu _i}} $ ) is not significant based on the P value. The null hypothesis that there is endogeneity between VQA certification and the price of wine is therefore rejected. This suggests that VQA certification leads to higher wine prices and not the other way around.