Literature Review

As consumers’ interest in local food is growing steadily, so is the number of studies on topics and issues related to local foods (see Martinez et al. [Reference Martinez, Hand, Pra, Pollack, Ralston, Smith and Vogel2010] for an overview). However, there is not a universally accepted definition of “local food” (Martinez et al. Reference Martinez, Hand, Pra, Pollack, Ralston, Smith and Vogel2010). Definitions of local food are usually based on one or more of the following features or characteristics: geographic proximity (DePhelps et al. Reference DePhelps, Williams, Foltz, Potter and Faunce2005; Hu et al. Reference Hu, Batte, Woods and Ernst2012), political boundaries (e.g., Washington apples, Idaho potatoes, California peaches, and Florida citrus) (Zepeda and Leviten-Reid Reference Zepeda and Leviten-Reid2004), how local food is retailed (e.g., farmers’ market and community-supported agriculture), and length of the supply chain (Marsden, Banks, and Bristow Reference Marsden, Banks and Bristow2000), among others. In this article, we adopt the political boundary definition, which is “grown and available for purchase within a State's borders” (Martinez et al. Reference Martinez, Hand, Pra, Pollack, Ralston, Smith and Vogel2010, 13). This definition is also the most popular definition among the top 10 grocery retailers (Martinez et al. Reference Martinez, Hand, Pra, Pollack, Ralston, Smith and Vogel2010).

Most local food studies focus on consumer preferences and WTP for local products. Martinez et al. (Reference Martinez, Hand, Pra, Pollack, Ralston, Smith and Vogel2010) summarize a series of studies on WTP for a wide range of locally produced food in the United States. Not surprisingly, they find that consumers are willing to pay higher prices for local foods. A number of studies examine the effect of multiple factors that influence consumer preferences and WTP for local food, including quality perception (Brown Reference Brown2003; Carpio and Isengildina-Massa Reference Carpio and Isengildina-Massa2009), nutritional reasons (Eastwood, Brooker, and Gray Reference Eastwood, Brooker and Gray1999; Loureiro and Hine Reference Loureiro and Hine2002), better value for the price (Wolf, Spittler, and Ahern Reference Wolf, Spittler and Ahern2005), support for the environment and local economy (Darby et al. Reference Darby, Batte, Ernst and Roe2008), and demographic characteristics (Brown Reference Brown2003). Other studies analyze how the values of “local” and “organic” interact and influence consumer WTP (James, Rickard, and Rossman Reference James, Rickard and Rossman2009; Yue and Tong Reference Yue and Tong2009; Roosen, Kottl, and Hasselbach Reference Roosen, Kottl and Hasselbach2012).

Several methods have been used by researchers to study consumer preferences and WTP for local food. Earlier studies tend to use hypothetical approaches such as personal interviews, as well as online, mail, and telephone surveys (Eastwood Reference Eastwood1996; Brown Reference Brown2003; Zepeda and Leviten-Reid Reference Zepeda and Leviten-Reid2004). In a hypothetical survey, respondents answer WTP questions where the payment of the stated WTP is hypothetical. These studies have been criticized for not being incentive-compatible to reveal the real consumer WTP (Wertenbroch and Skiera Reference Wertenbroch and Skiera2002). In recent years, experimental auctions have become increasingly popular to investigate the impact of labeling on WTP for food attributes (Dickinson and Bailey Reference Dickinson and Bailey2002; Umberger et al. Reference Umberger, Feuz, Calkins and Killinger-Mann2002; Lusk, Feldkamp, and Schroeder Reference Lusk, Feldkamp and Schroeder2004). Real money and real products are exchanged in an experimental setting so that participants have a greater incentive to reveal their true valuation of a product than in a hypothetical survey setting (Lusk Reference Lusk2003; Lusk, Feldkamp, and Schroeder Reference Lusk, Feldkamp and Schroeder2004). For example, Grebitus, Lusk, and Nayga (Reference Grebitus, Lusk and Nayga2013) use second-price auctions to study the effect of distance of transportation on consumer WTP for local food. Similarly, Shi, House, and Gao (Reference Shi, House and Gao2013) use a Becker-DeGroot-Marschak (BDM) auction to determine in what way purchase intentions affect WTP for organic and locally grown blueberries. Other methods of studying consumer preferences and WTP for local food include conjoint analysis (Darby et al. Reference Darby, Batte, Ernst and Roe2008) and choice experiments (Alfnes et al. Reference Alfnes, Guttormsen, Steine and Kolstad2006; Yue and Tong Reference Yue and Tong2009).

In this article, we employ BDM auctions of broccoli to study the effects of locally grown information on (1) consumer WTP and (2) consumer perception of product quality. In a BDM auction, subjects submit sealed bids for a good. A random price is then drawn from a predetermined distribution. Individuals with bids higher than the randomly drawn price “win” the auction and purchase a unit of the good at that randomly drawn price. Because the bids of respondents do not determine the purchase price, the BDM auction creates an optimal environment for rational respondents to reveal their actual WTP (Becker, DeGroot, and Marschak Reference Becker, DeGroot and Marschak1964; Lusk and Shogren Reference Lusk and Shogren2007).

Our article is related to a stream of literature that studies the effect of country of origin (COO) on consumer WTP and perception of product quality. According to Elliott and Cameron (Reference Elliott and Cameron1994), consumer attitudes about local and nonlocal products are similar to the effect of COO, which has long been discussed in the literature. Newman et al. (Reference Newman, Turri, Howlett and Stokes2014) provide an overview of research related to COO labeling and implications for food marketing systems. This literature generally agrees that consumers tend to perceive domestic food to be of superior quality than imported food products (Umberger Reference Umberger2005; Lobb and Mazzocchi Reference Lobb and Mazzocchi2007; Pouta et al. Reference Pouta, Heikkilä, Forsman-Hugg, Isoniemi and Mäkelä2010). Consumer WTP for COO information has also been widely researched by marketing and consumer behavior literature. For example, Lim et al. (Reference Lim, Hu, Maynard and Goddard2013) study U.S. consumer preference and WTP for COO-labeled beef steak and food safety enhancements. In many cases, consumers’ higher WTP for domestic food is associated with their perceptions of superior quality (Dickinson and Bailey Reference Dickinson and Bailey2002; Umberger et al. Reference Umberger, Feuz, Calkins and Killinger-Mann2002). Loureiro and Umberger (Reference Loureiro and Umberger2003) find that consumers are willing to pay an average of $1.53 and $0.70 per pound more for steak and hamburger labeled as “U.S. Certified.”

Despite the importance of quality perception in deciding consumer preference and WTP for local food (Durham, King, and Roheim Reference Durham, King and Roheim2009), little has been done to examine the effect of locally grown information on consumer perceptions of local food quality and how these perceptions relate to consumer WTP, which is the focus of this study. One exception is a study by Stefani, Romano, and Cavicchi (Reference Stefani, Romano and Cavicchi2006), who examine the impact of alternative definitions of the region of origin on consumer WTP and consumer evaluation of food quality. Another study that is closely related to our work is by Bi et al. (Reference Bi, House, Gao and Gmitter2012), who use experimental auctions to study the effect of sensory attributes (viewing, peeling, and tasting) on consumer WTP for two new tangerine varieties. They find that consumers change their WTP based on the different attributes of the tangerines and that internal fruit attributes (e.g., flavor, juiciness, and ease of peeling) are more important to consumers than external attributes (e.g., appearance).

Experimental Design

We designed and implemented an economic experiment with nonstudent subjects to examine consumer WTP for three broccoli varieties: a commercial variety from California (from here on referred to as “California variety”) and two newly developed NYS grown varieties (from here on referred to as “NYS 1” and “NYS 2”) that are undergoing field trials before being launched to market. The California variety was bought from a local grocery store and had a dark-green, firm, uniform, and domed head. The two NYS varieties were harvested from an agricultural experiment station where the field trial for the new broccoli varieties was conducted. NYS 1 had a light-green, flat, and nonuniform head. NYS 2 was designed by researchers to have a very similar appearance to the California variety. To maintain similar postharvest product attributes, we stored the two NYS varieties under the same conditions that the California variety was stored before arriving at NYS grocery stores. After harvested, the two NYS varieties were stored in a cooler, packed with flake ice around and on top of the broccoli heads. The cooler was then placed in a forced-air room for 5 to 7 days, which is the estimated time to ship broccoli from California to NYS. All three broccoli varieties were then parboiled at the same time and packed in small samples for use in the economic experiments.

We collected WTP information from subjects who were exposed to one of two treatments regarding the origin of the three broccoli varieties. In the first treatment (from here on referred to as the “no information” treatment), subjects did not receive information about the origin of the varieties. They revealed their WTP for the three varieties solely based on their evaluations of the appearance and the taste of the three broccoli varieties. In the second treatment (from here on referred to as the “information” treatment), subjects were told that the two NYS varieties were grown in NYS, and the third variety in California.

Subjects were recruited through a local experimental economics research laboratory's email system.Footnote ³ They were seated randomly at individual computer terminals with privacy shields, were informed that all decisions they made would be kept strictly confidential, and were given $25 for participating and told they might have the opportunity to actually purchase broccoli. A maximum of 24 computer terminals were available per session, and the number of subjects in each session ranged from 15 to 24. After signing a consent form, subjects were given a brief introduction of the experiment, which included the amount of money they would earn and the rules of the experiment. We began each session with two practice rounds to demonstrate how the BDM auction would be conducted. In the practice round, subjects submitted bids for a dollar bill and a chocolate bar so they would become familiar with the bidding process of the auctions.

We follow the principles of good practice to perform sensory evaluation of food suggested by Lawless and Heymann (Reference Lawless and Heymann2010) when conducting the broccoli sensory evaluation experiment. We use the affective test method, the most widely used sensory test method in the food industry to determine consumer preference of products (Chambers and Wolf Reference Chambers, Wolf, Chambers and Wolf1996). The affective test method asks untrained assessors to answer a set of questions to measure preference, liking, and impression of various sensory attributes (Kemp, Hollowood, and Hort Reference Kemp, Hollowood and Hort2011). At the beginning of the broccoli sensory evaluation experiment, subjects were verbally informed of the evaluation procedure. After that, lab assistants displayed one crown (approximately 1 pound) of each of the three broccoli varieties in a tray so that subjects could examine the appearance of the broccoli varieties closely. The three varieties were labeled broccoli “A,” “B,” and “C.” After observation, subjects were guided to record their evaluation of the appearance of the broccoli varieties on a 9-point scale (from 1 to 9, with 9 being most favorable) on an information sheet provided on their table. After all subjects completed the appearance evaluation, the subjects were then guided to proceed to the tasting evaluation portion. Participants at each station were provided small parboiled samples of the three broccoli varieties. The labels on the container of the parboiled samples matched the labels of the broccoli crowns shown during the appearance evaluation portion. Palate cleansers (water and unsalted crackers) were provided to subjects to minimize taste bud fatigue and to remove flavor carryover from the previous sample. After the tasting evaluation, participants were also asked to provide their evaluations (on the same 9-point scale used in the appearance evaluation portion) of the taste of the broccoli varieties.

After observing the appearance and having tasted the three broccoli varieties, subjects were asked to place bids for 1 pound of each variety in the auction. Each subject submitted bids between $0.00 and $5.00 for 1 pound of each variety.Footnote ⁴ The BDM auction method was used to elicit maximum WTP for the broccoli varieties. Although subjects bid for all three varieties, they were informed that only one of the three auctions would result in an actual transaction. After bids for all auctions were submitted, one out of the three auctions was randomly chosen to be binding. In this case, the subjects who won their bids for the randomly selected auction would be “required” to purchase 1 pound of broccoli at the market price, which would be deducted from their participation endowment. After the auctions, subjects completed a computerized survey asking demographic and purchasing habit information, including gender, age, education, income, cooking frequency, broccoli consumption in their meals, and whether the subject was the primary shopper in the household (please see supplementary Appendix for the survey used in the experiments).

Data and Empirical Model

We collected 240 observations from 80 nonstudent subjects in the broccoli tasting experiment sessions. Table 1 shows descriptive statistics of demographic information about experimental subjects based on responses to questions in the survey. We also compared the subjects from our experiments with the U.S. grocery shopper sample reported by the Food Marketing Institute (FMI). In summary, our sample is representative of the U.S. grocery shopper sample. The percentage of female subjects in our experiment (72.5%) is very close to that of the FMI sample (69.7%). Subjects (in our data) aged 25–34 are slightly less represented in our study than in the FMI sample, whereas subjects aged 35–49 are slightly more represented than in the FMI sample. This is because we excluded students from our study, so the age is biased upward. For the same reason, the percentage of subjects in the lowest income band is also smaller than that of the FMI sample. Regarding education, the percentage of subjects who received a bachelor's degree or more in our experiment (33.8%) is close to the FMI sample (33%). In regard to cooking frequency, 27.5% of subjects cook their meals 4–5 times a week, and 41.3% cook their meals more than 5 times a week. More than half of the subjects include broccoli in their meals at least once a week. On average, 81.25% of the subjects are the primary shopper in their households.

Table 1. Demographic Characteristics of Participants from the Broccoli Experiment Compared with Food Marketing Institute (FMI) Grocery Shopper Sample (Frequency: %)

^a FMI Grocery Shopper Trends Report 2013 (https://www.fmi.org/forms/store/ProductFormPublic/u-s-grocery-shopper-trends-2013-data-tables).

Table 2 provides subjects’ average WTP and their evaluation scores of the appearance and taste of the three broccoli varieties, by locally grown information treatment. Forty-one subjects participated in the no information treatment sessions, and 39 subjects participated in the sessions with locally grown information. Subjects in the two treatments have similar demographic characteristics, and there are slightly more female and primary shopper participants in the “no information” treatment (see Table A1 in the Appendix). Table 2 shows that the average WTP for all 80 subjects is $1.60 per pound of broccoli, which is comparable to retail prices in local supermarkets. For the no information treatment, the average WTP is $1.53, which is slightly lower than the average WTP for the whole sample. When subjects are given the locally grown information, the mean WTP becomes $1.68, which is slightly higher than the average WTP for the whole sample. Subjects’ evaluation scores of appearance and taste in the two treatments follow a similar pattern.

Table 2. Summary Statistics for Willingness to Pay (WTP) and Score of Appearance and Taste

Notes: Standard deviation in parentheses. NYS, New York State.

A simple comparison of the descriptive statistics among the three broccoli varieties within each treatment reveals intriguing information. For the no information treatment, subjects’ average WTP for the California variety is $1.62, whereas their WTP for NYS 1 and NYS 2 are much lower, $1.44 and $1.52, respectively. Meanwhile, the average score of appearance for NYS 1 is 4.22, which is much lower than the score for the California variety (7.83). For NYS 2, the difference in the scores for both appearance and taste are modest relative to the California variety. The average score of appearance for NYS 2 is 6.59, which is 1.24 lower than the California variety, and the average score of taste for NYS 2 is 6.4, which is 0.28 lower than the California variety.

In contrast, the subjects’ WTP and evaluations of appearance and taste are substantially different in the information treatment, in comparison with the no information treatment. First, both the WTP and the taste for the two NYS varieties become higher than for the California variety (recall they were lower for the no information treatment). Interestingly, appearance evaluations of the two NYS varieties in the information treatment (5.82 and 6.38) are lower than for the California variety (8.00), which is similar to the no information treatment. In addition, although evaluations of the appearance and taste of NYS 2 (6.38 and 7.59) are both higher than those of NYS 1 (5.82 and 7.21), the WTP for them is practically the same ($1.70 and $1.71).

When comparing data between the two treatments (information and no information), the descriptive statistics suggest that subjects’ WTP and evaluations of appearance and taste of the California variety are very close. For example, subjects’ WTP values for the California variety in the two treatments are both $1.62, and their score of appearance is only slightly higher when given information of locally grown (8.00) compared with when they are not provided information (7.83). In contrast, WTP and evaluation of appearance and taste of the two NYS varieties are all markedly higher in the information treatment than in the no information treatment. The only exception is that the appearance evaluation score for NYS 2 in the information treatment (6.38) is slightly lower than that of the no information treatment (6.59).

To test whether subjects’ WTP is affected by the provided locally grown information, we run two random effects models. We first run a simple generalized least squares (GLS) random effects model with and without demographic variables. We also run a Tobit model to account for the censored nature of the WTP data. The Tobit model has been widely used by agricultural economists to study consumer WTP for attributes of food products (e.g., Bernard and Bernard Reference Bernard and Bernard2009; Kanter, Messer, and Kaiser Reference Kanter, Messer and Kaiser2009). The latent value of individual i’s WTP for variety j, denoted as $WTP_{ij}^{\ast}$, is expressed as a function of the variety V _j, the dummy variable I to indicate whether the subject receives locally grown information treatment, and the subjects’ demographic characteristics X _i. Because individuals submitted bids for different broccoli varieties in the experiment, we employ a random effects Tobit model to account for the panel nature of the data. The parameter v _i is an individual-specific disturbance for subject i, and ε_ij is the error term that is assumed to follow a normal distribution with mean zero and standard deviation σ. In equation (1), we assume a linear functional form for the WTP equation. The relationship between the observed variable WTP _ij and the latent variable $WTP_{ij}^{\ast}$ is shown in equation (2). If we assume the observed WTP _ij and the latent $WTP_{ij}^{\ast}$, to be the same, then equation (1) collapses to the GLS random effects model.

(1)$$WTP_{ij}^{\rm {^\ast}} = {\rm \alpha} + {\rm \beta} _jV_j + {\rm \gamma} I + {\rm \delta} _jV_jI + {\rm \theta} X_i + v_i + {\rm \varepsilon} _{ij}$$

(2)$$WTP_{ij} = {\rm max}\lcub {0\comma \; \; WTP_{ij}^{^\ast}} \rcub $$

In the model specified previously, α is the WTP for 1 pound of the California variety when no information about the origin of the three types of broccoli are revealed; β_j captures the price premium that consumers are willing to pay for the NYS variety j (relative to the California variety) when the locally grown information is not provided; γ is the effect of locally grown information on consumer's WTP for the California variety; δ_j describes the interaction effects between varieties and information treatment, which captures the effect of locally grown information on the price premium that consumers are willing to pay for the NYS variety j (relative to the California variety); and θ is a vector of parameters of consumer characteristics.

It is also important to examine how subjects’ perceptions of the quality (appearance and taste) of the three broccoli types are affected by the locally grown information provided. We employ the SUR model to take into account that a subject's evaluations of the appearance and the taste of the broccoli varieties might be correlated. The regression equations (equations 3 and 4) in the SUR model are similar to equation (1) with the exception that the subjects’ evaluations of the appearance and taste of the three broccoli types are now the dependent variables. We use superscripts “A” and “T” to denote the parameters in the Appearance and the Taste equations, respectively. We assume that the error terms for different subjects are independently distributed. However, the two error terms ${\rm \varepsilon} _{ij}^A$ and ${\rm \varepsilon} _{ij}^T$ may be correlated for each subject i.

(3)$$Appearance_{ij} = {\rm \alpha} ^A + {\rm \beta} _j^A V_j + {\rm \gamma} ^AI + {\rm \delta} _j^A V_jI + {\rm \theta} ^AX_i + v_i^A + {\rm \varepsilon} _{ij}^A $$

(4)$$Taste_{ij} = {\rm \alpha} ^T + {\rm \beta} _j^T V_j + {\rm \gamma} ^TI + {\rm \delta} _j^T V_jI + {\rm \theta} ^TX_i + v_i^T + {\rm \varepsilon} _{ij}^T $$