Research highlights
-
Federal and state programs promoting local food have grown significantly.
-
Estimates of consumers’ willingness to pay (WTP) premiums for local food are largely based on hypothetical surveys.
-
We investigate consumer WTP for local foods in a large-scale field experiment.
-
Our experiment finds no effect of a generic locally produced label on consumer WTP.
-
Consumer WTP declines as the perceived distance of local increases.
Introduction
Recent trends in the U.S. food system point to increasing interest among consumers in locally produced food (Carpio and Isengildina-Massa Reference Carpio and Isengildina-Massa2009; Grebitus et al. Reference Grebitus, Lusk and Nayga2013; Martinez and Park Reference Martinez and Park2021). While the 2008 Farm Bill defines a product as local if the total distance that the product is transported does not exceed 400 miles from the origin or if the food product is raised, produced, and distributed in a particular state (Food, Conservation and Energy Act of 2008; Thilmany McFadden Reference Thilmany McFadden2015; Li et al. Reference Li and Messer2020a) according to the U.S. Department of Agriculture (USDA), there is no formal or universally agreed-upon distance that defines a food as local (Thilmany McFadden Reference Thilmany McFadden2015; Low et al. Reference Low, Adalja, Beaulieu, Key, Martinez, Melton, Perez, Ralston, Stewart, Suttles, Vogel and Jablonski2015; Martinez Reference Martinez2016). In the absence of a formal definition, U.S. consumers have tended to associate the term with in-state and regional geographic boundaries and/or as their having a personal connection to the production system (Thilmany McFadden Reference Thilmany McFadden2015; Martinez Reference Martinez2016; Li et al. Reference Li and Messer2020a). In response to these perceptions, 47 states have invested in policies and marketing campaigns to support and promote local food (see Table 1) since their production typically is small in scale and involves greater costs per unit than industrial production. In addition, millions of federal dollars from the Farm Bill are supporting these systems through marketing and promotion, business assistance and agricultural research, rural and community development, and nutrition and education programs (Martinez et al. Reference Martinez, Hand, Da Pra, Pollack, Ralston, Smith, Vogel, Clark, Lohr, Low and Newman2010; Johnson and Cowan Reference Johnson and Cowan2019).
Table 1. State marketing campaigns for local foods
There is no question that federal and state initiatives promoting local food systems are growing. The question is whether U.S. consumers are currently willing to pay more for local food than for food produced and transported over long distances – whether they value the localness of the food they consume, and, as a result, whether producers of local food are likely to earn more money per item sold by marketing their foods as local. Understanding whether a price premium exists for local products is important to inform marketers and policymakers. The existence of premiums should entice producers to market local foods since they will generate greater revenue and capture higher profits. It should also encourage retailers to purchase and market local foods, even if they cost more than similar food produced in other locations. This increase in demand for local foods, if it exists, can be important for the viability of local food systems since the production of local food is often small-scale.
Our study investigates whether a local label inspires Mid-Atlantic consumers to pay a premium for mushrooms and oysters in the Mid-Atlantic. We elicit revealed preferences in a large-scale field experiment using the incentive-compatible Becker-DeGroot-Marschak (BDM) auction mechanism. This research contributes to the discussion about the definition of localness to reveal whether a generic local label can generate additional profits for producers and retailers.
Localness has been defined in various ways in the literature. The terms “locally grown” or “locally produced” have been used most often, but studies have also explored defining localness in “food miles” and “food distance” (Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015; de-Magistris and Gracia Reference de-Magistris and Gracia2016; Li et al. Reference Li and Messer2020a) and preferring to “buy the product from a farmer [I] know” (Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015). Li et al. (Reference Li, Messer, Mamadzhanov and McCluskey2020b) tested several definitions of local oysters and found that none produced significantly greater WTP. Kecinski et al. (Reference Kecinski, Messer, Knapp and Shirazi2017) inferred preferences for local foods using product attributes that specified oyster brand and harvest location. The authors found that consumers did not exhibit preferences for oysters harvested from specific locations. Rather than specifying a harvest location or distance, this study tests the impact of a generic label, “locally produced,” on WTP for mushrooms and oysters. This generic description reflects the current state of the market in which there is no universal definition of local food (Martinez et al. Reference Martinez, Hand, Da Pra, Pollack, Ralston, Smith, Vogel, Clark, Lohr, Low and Newman2010). Additionally, we elicit consumer perceptions of local by asking adult consumers to specifically state how far away food production can be (in miles) for them to consider the production as local.
Our study contributes broadly to the literature using revealed preference methods with a large sample size to study local labels. Printezis et al. (Reference Printezis, Grebitus and Hirsch2019) found that 80% of studies in their meta-analysis relied on hypothetical choice experiments. Most of the incentive-compatible experiments in the meta-analysis pointed to a premium on local foods. However, with the exceptions of a few studies that used a within-subject design (Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015; Kallas et al. Reference Kallas, Alba, Casellas, Berges, Degreef and Gil2019; Sanjuán-Lopez and Resano-Ezcaray Reference Sanjuán-López and Resano-Ezcaray2020; Li et al. Reference Li, Messer, Mamadzhanov and McCluskey2020b) many of those studies that used incentive-compatible mechanisms analyzed relatively small samples and many investigated markets outside of the U.S. (see Table 2). Printezis et al. (Reference Printezis, Grebitus and Hirsch2019) compared the original study results to results produced when controlling for publication bias and methodological variation and found that the WTP premiums identified in the studies did not disappear when controlling for those factors. Rather, the magnitudes of the premiums changed significantly. Overall, they found a range of premiums for local foods of $1.70 to $2.08 per pound. However, this range of premiums decreased to just $0.29 to $0.40 after correcting for publication bias. Thus, while much of the previous literature points to local premiums, current estimates may be exaggerated due to publication bias and the use of relatively small sample sizes in existing incentive-compatible studies (Ferraro and Shukla Reference Ferraro and Shukla2022; Printezis et al. Reference Printezis, Grebitus and Hirsch2019). Our research provides further evidence that the price premium for locally labeled foods may not be as large as previously claimed and perhaps this premium currently does not exist at all.
Table 2. Previous incentive-compatible studies investigating the local label
Note:
* indicates the study used a within-subject design.
Related literature
Comprehensive reviews of literature related to consumers’ WTP more for local foods have recently been conducted by Printezis et al. Reference Printezis, Grebitus and Hirsch2019 and Enthoven and Van den Broeck Reference Enthoven and Van den Broeck2021. In general, previous literature has found a premium for local food over nonlocal food (Loureiro and Hine Reference Loureiro and Hine2002; Carpio and Isengildina-Massa Reference Carpio and Isengildina-Massa2009; Campbell et al. Reference Campbell, Lesschaeve, Bowen, Onufrey and Moskowitz2010; Onozaka and McFadden Reference Onozaka and McFadden2011; Carroll et al. Reference Carroll, Bernard and Pesek2013; Hempel and Hamm Reference Hempel and Hamm2016; Pritnezis et al. Reference Printezis, Grebitus and Hirsch2019; Enthoven and Van den Broeck Reference Enthoven and Van den Broeck2021). Several studies have identified premiums for locally produced fruits and vegetables (Jekanowski et al. Reference Jekanowski, Williams and Schiek2000; Loureiro and Hine Reference Loureiro and Hine2002; Brown Reference Brown2003; Giraud et al. Reference Giraud, Bond and Bond2005; Darby et al. Reference Darby, Batte, Ernst and Roe2008; Sirieix et al. Reference Sirieix, Grolleau and Schaer2008; Thilmany et al. Reference Thilmany, Bond and Bond2008; Campbell et al. Reference Campbell, Lesschaeve, Bowen, Onufrey and Moskowitz2010; Nganje et al. Reference Nganje, Hughner and Lee2011; Onozaka and McFadden Reference Onozaka and McFadden2011; Grebitus et al. Reference Grebitus, Lusk and Nayga2013; Hempel and Hamm Reference Hempel and Hamm2016; Printezis and Grebitus Reference Printezis and Grebitus2018) and value-added processed products like maple syrup (Giraud et al. Reference Giraud, Bond and Bond2005), wine (Jekanowski et al. Reference Jekanowski, Williams and Schiek2000; Grebitus et al. Reference Grebitus, Lusk and Nayga2013), bread (Hasselbach and Roosen Reference Hasselbach and Roosen2015), honey (Wu et al. Reference Wu, Fooks, Messer and Delaney2015), beer (Hasselbach and Roosen Reference Hasselbach and Roosen2015), flour (Hempel and Hamm Reference Hempel and Hamm2016), blueberry jam (Hu et al. Reference Hu, Batte, Woods and Ernst2012), and strawberry preserves (Onken et al. Reference Onken, Bernard and Pesek2011). Other studies found local premiums for animal products (Wang et al. Reference Wang, Halbrendt, Kolodinsky and Schmidt1997; Jekanowski et al. Reference Jekanowski, Williams and Schiek2000; Doyon et al. Reference Doyon, Simard, Messer, Tamini and Kaiser2008; Umberger et al. Reference Umberger, Thilmany McFadden and Smith2009; Caroll et al. Reference Carroll, Bernard and Pesek2013; Chang et al. Reference Chang, Xu, Underwood, Mayen and Langelett2013; Tempesta and Vecchiato Reference Tempesta and Vecchiato2013; Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015 ; Hasselbach and Roosen Reference Hasselbach and Roosen2015; Hempel and Hamm Reference Hempel and Hamm2016; Willis et al. Reference Willis, Carpio and Boys2016) and seafood (Giraud et al. Reference Giraud, Bond and Bond2005).
There is some evidence that price premiums may be due to other factors associated with local products. For example, the premium was high for local foods only when consumers were not from farm households (Brown Reference Brown2003) or when consumers perceived local farmers as struggling, marginalized, and deserving of special attention (Toler et al. Reference Toler, Briggeman, Lusk and Adams2009). Local premiums sometimes reflected the status of value-added high-end luxury products (Giraud et al. Reference Giraud, Bond and Bond2005) or highly consumed products (Tempesta and Vecchiato Reference Tempesta and Vecchiato2013). Printezis and Grebitus (Reference Printezis and Grebitus2018) found that local products sold at the grocery store elicited premiums not found at farmer’s markets. Moreno and Malone (Reference Moreno and Malone2021) found that consumers’ preferences for local food varied depending on the type of food item considered and whether the item had a local identity – it was associated with the area in some way, such as being a crop the state is known for producing (i.e., maple syrup in Vermont).
Like much of the previous literature on local food systems, the aforementioned studies frequently rely on hypothetical choice experiments (Printezis et al. Reference Printezis, Grebitus and Hirsch2019; Enthoven and Van den Broeck Reference Enthoven and Van den Broeck2021). The literature on incentive-compatible revealed preference studies on preferences for local food products is limited. Such studies commonly used lab experiments, artefactual experiments, or nonhypothetical choice experiments. Most incentive-compatible studies were conducted in the United States (Yue and Tong Reference Yue and Tong2009; Costanigro et al. Reference Costanigro, Thilmany McFadden, Kroll and Nurse2011; Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015; Kecinski et al. Reference Kecinski, Messer, Knapp and Shirazi2017; Fan et al. Reference Fan, Gómez and Coles2019; Li et al. Reference Li and Messer2020a; Reference Li, Messer, Mamadzhanov and McCluskey2020b; Wu et al. Reference Wu, Fooks, Messer and Delaney2015) or Europe (Gracia et al. Reference Gracia, de Magistris and Nayga2012; de-Magistris and Gracia Reference de-Magistris and Gracia2016; Wägeli et al. Reference Wägeli, Janssen and Hamm2016; Bazzani et al. Reference Bazzani, Caputo, Nayga and Canavari2017; Sanjuán-López and Resano-Ezcaray Reference Sanjuán-López and Resano-Ezcaray2020) and rarely in other parts of the world like South America (Kallas et al. Reference Kallas, Alba, Casellas, Berges, Degreef and Gil2019). In many cases, the incentive-compatible research has also found that consumers are willing to pay a premium for local foods including fresh produce and nuts (Yue and Tong Reference Yue and Tong2009; Costanigro et al. Reference Costanigro, Thilmany McFadden, Kroll and Nurse2011; Grebitus et al. Reference Grebitus, Lusk and Nayga2013; de-Magistris and Gracia Reference de-Magistris and Gracia2016; Fan et al. Reference Fan, Gómez and Coles2019; Sanjuán-López and Resano-Ezcaray Reference Sanjuán-López and Resano-Ezcaray2020; Moreno and Malone Reference Moreno and Malone2021), processed foods such as applesauce (Bazzani et al. Reference Bazzani, Caputo, Nayga and Canavari2017) and wine (Grebitus et al. Reference Grebitus, Lusk and Nayga2013), and animal products (Gracia et al. Reference Gracia, de Magistris and Nayga2012; Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015; Gracia and de-Magistris Reference Gracia and De-Magistris2016; Wägeli et al. Reference Wägeli, Janssen and Hamm2016; Kallas et al. Reference Kallas, Alba, Casellas, Berges, Degreef and Gil2019; Li et al. Reference Li and Messer2020a; Reference Li, Messer, Mamadzhanov and McCluskey2020b). However, other studies have found that only small segments of consumers are willing to pay more (Adalja et al. Reference Adalja, Hanson, Towe and Tselepidakis2015; de-Magistris and Gracia Reference de-Magistris and Gracia2016; Li et al. Reference Li and Messer2020a).
To our knowledge, no incentive-compatible study has investigated consumer preferences for mushrooms. However, a stated preference study found evidence of a premium on local mushrooms (Chakrabarti et al. Reference Chakrabarti, Campbell and Shonkwiler2019). Recent studies on consumer WTP for oysters showed local premiums in stated preference studies (Chen et al. Reference Chen, Haws, Fong and Leung2017; Tian et al. Reference Tian, Croog, Bovay, Concepcion, Getchis and Kelly2021), but the evidence from incentive-compatible revealed preference studies is mixed (Kecinski et al. Reference Kecinski, Messer, Knapp and Shirazi2017; Li et al. Reference Li and Messer2020a).
Methods
Experimental design
The diagram in Fig. 1 illustrates the flow of the experiment used in this study. To elicit revealed preferences in response to labeling food as local, we implemented a large-scale field experiment that employed the incentive-compatible BDM auction mechanism to measure consumer WTP for local mushrooms and oysters (Becker et al. Reference Becker, DeGroot and Marschak1964). Participants were given an account balance of $10 and asked to make purchasing decisions for white button mushrooms and raw oysters by stating the highest amount of money they would pay for the product (aka., their WTP). Footnote 1 The products were presented in bundles that had a market value of approximately $5, and participants’ offers were bounded between $0 and $10. The upper bound was based on the market value of the food items ($5) and the project budget.
Figure 1. Diagram of experimental flow.
Note: Food item refers to oysters or mushrooms; the presentation of food items was randomized to avoid order effects.
We specifically chose to study premiums for mushrooms and oysters because of their histories and connections to the local region in our study. The mushroom industry is well established in the Mid-Atlantic states, which currently account for 64% of the total volume of national sales of fresh mushrooms (USDA NASS 2021). Despite its established presence (Kennett Square in southeastern Pennsylvania along the Delaware border is colloquially known as the “Mushroom Capital of the World”), little is known about consumers’ WTP for mushrooms labeled as local. Oysters from the Delaware Inland Bays, on the other hand, are a relatively new product as oysters produced in the area declined dramatically due to disease and overharvesting and are just recently coming back due to aquaculture methods (Kecinski et al. Reference Kecinski, Messer, Knapp and Shirazi2017). However, oysters are generally produced in great numbers in the region, especially in the Chesapeake Bay (primarily Maryland and Virginia).
This study was approved by the Institutional Review Board at the University of Delaware. Data were collected between June 1 and October 31, 2019, at six locations in the Mid-Atlantic region of the United States – a local creamery, a state fair, a ferry terminal, a university campus laboratory, Osher lifelong learning institute, and university-sponsored community event promoting coastal research. Participants were recruited through community flyers and convenience sampling in person at public events. In total, 1,050 adults aged 22 years or older participated in the study.Footnote 2
The experiment presented each participant with the opportunity to purchase mushrooms and oysters. We randomly assigned participants in a session to the local-label treatment and control groups. Participants in the treatment group viewed bundles labeled simply as “locally produced” with no definition of local provided. For example, when stating WTP for oysters, participants in the treatment group were asked to “please indicate the maximum amount you would be willing to pay (from $0.00 to $10.00) for 2 locally produced oysters” (see Appendix A). Identical items were presented to the control group with no labels.
As is common with economic experiments that use the BDM protocol, participants were informed Footnote 3 that, after the purchase decisions were made, the experiment administrators would randomly select one of the food items for implementation at a randomly generated price. Participants who expressed WTP equal to or greater than the randomly generated price purchased the selected item. They received the product and the balance of the funds in cash if they were successful bidders. Participants who expressed that WTP was lower than the randomly generated price did not purchase the food item and received the entire $10 balance.
Because we used a generic definition of local, we asked participants to define local food in a follow-up question after the experiment: “Up to what distance (in miles) do you consider food to be locally produced?” The post-experiment survey also collected demographic characteristics (see Appendix A).
Power analysis
This study is part of a larger experiment in which the label treatment was randomized by group to measure peer effects (Langer et al. Reference Langer, Davidson, McFadden and Messer2022). To determine the target sample size, we used a power simulation for a cluster-randomized crossover study with multiple regressions.
Footnote 4
Following Reich et al. (Reference Reich, Myers, Obeng, Milstone and Perl2012), power was determined using a simulation of Cohen’s
${F^2}$
, which relies on the use of a predicted R
2 – the proportion of variations in outcomes explained by the treatments in multiple regression (treatments).
Footnote 5
Based on our regression analysis of pilot data (N = 52) collected in the same manner as in the full experiment, we assumed R
2 = 0.45 in the cluster-randomized power analysis. The analysis determined that a sample of 1,060 participants was required to identify a statistically significant treatment effect for at least one treatment with power equal to 0.80 when groups were assigned to clusters of four. Because the power analysis was clustered in groups of four, the calculated sample size is a conservative estimate of the sample required for this portion of the larger study that assigned the local-label treatment to individuals. Therefore, we also conducted a statistical power analysis for two unbalanced sample t-tests.
Footnote 6
Of the 1,050 participants in the study, we assigned 455 participants to the treatment group and 595 to the no-label control group. In this analysis, the required statistical power of the local-label experiment was 0.89 with 95% confidence to detect a minimum effect of 0.2 for our two unbalanced sample groups. We chose the conservative estimate of 0.2 following Cohen’s small effect size for the t-test (Cohen Reference Cohen1988).
By implementing a between-subject design, our study evaluates the isolated decision to produce a locally labeled product that might occur in a retail setting for new and emerging products such as a restaurant or café where there is no “nonlocal” option available. Additionally, in most grocery store settings foods labeled as local are displayed alone and not alongside a similar product that was not grown locally.Footnote 7 We would note that other studies of consumer preferences have used within-subject designs because they tend to have stronger statistical power and because they can simulate a market setting where similar products that differ primarily by the production processes are displayed side by side, which can impact consumers’ reference points (e.g., conventional milk, rBST-free milk, and organic milk in Kanter et al. Reference Kanter, Messer and Kaiser2009).
Data
Table 3 presents summary statistics of the participants in the experiment. On average, participants perceived foods as local when they were produced within 42.3 miles. Almost 57% of the participants were women, and the average participant age was 45. About one-third (35%) of the participants preferred not to identify their political affiliations; the other 65% were distributed almost equally to conservative, moderate, and liberal. White participants represented 77% of the sample, a figure comparable to the weighted average white population of 74% in Delaware, Maryland, and Pennsylvania, the resident states of most of our participants (U.S. Census Bureau 2021). Almost 60% of the participants had at least a bachelor’s degree. Thus, our sample participants had more education on average than the populations of the three states studied: residents age 25 or older holding at least a bachelor’s degree made up 32.7% of the population in Delaware, 40.9% in Maryland, and 32.3% in Pennsylvania (U.S. Census Bureau 2021). The sample’s mean household income was $73,850, which is slightly higher than the median household income for Delaware ($69,110) and Pennsylvania ($63,630) but slightly lower than for Maryland ($87,060) (U.S. Census Bureau 2021). Most of the participants (71%) resided in Delaware.
Table 3. Summary statistics
To understand participants’ familiarity with the food items offered in this experiment, we asked how often they consumed oysters and mushrooms. The sample consisted of a greater number of nonconsumers of oysters (48.2%) than nonconsumers of mushrooms (20%) (see Figure B1 in Appendix B). The fact that a significant portion of the sample does not consider themselves to be consumers of oysters is not surprising, especially since in this research the oysters were being served raw (as is common in high-end restaurants and retail outlets). Interestingly, the percentages of participants who reported “almost never” consuming oysters (17.2%) were similar to the percentage of participants who reported “almost never” consuming mushrooms (13.4%). Future research on consumer demand for locally labeled foods could explore how these labels impact consumers who do and do not consider themselves regular consumers of these foods.
Figure 2 shows that the participants’ WTP values are frequently censored at the left (at $0) and the right (at $10) and that censoring is more common for oysters than mushrooms. Panel A represents all participants; Panel B represents the subset created by excluding nonconsumers (participants who reported never consuming the item). Panels A and B show that left censoring is more common than right censoring even when nonconsumers are excluded. Panels C and D present density plots of WTP for the no-label control group versus the treatment group.
Figure 2. Density plot of willingness to pay of participants for oysters and mushrooms.
Empirical methods
To measure the effect of the local label on WTP, we estimate a two-limit Tobit regression model (censored normal regression) since our data are truncated at the lower ($0) and upper ($10) bounds (Wang et al. Reference Wang, Halbrendt, Kolodinsky and Schmidt1997; Greene Reference Greene2012). For individual
$i$
$$y_i^* = \;{\bf{\beta '}}{{\bf{x}}_i} + {e_i}$$
where
$y_i^*$
is the latent variable of WTP, β is a vector of coefficients to estimate,
${{\bf{x}}_i}$
is a vector of independent variables including the label treatment, and
${e_i} \sim N\left( {0,\;{\sigma ^2}} \right)$
represents residuals that are assumed to be independently and normally distributed with mean zero and variance
${{\rm{\sigma }}^2}$
. By denoting observed WTP as
${y_i}$
, the model can be presented as
$${y_i} = \left\{ {\matrix{ 0 & {if\;y_i^* \le 0} \cr {{\bf\beta} '{{\bf{x}}_i} + {e_i}} & {if\;0 \lt y_i^* \lt 10} \cr {10} & {if\;y_i^* \ge 10} \cr } } \right\}.$$
We estimate the parameters using the corresponding maximum likelihood procedure (Maddala Reference Maddala1986; Wang et al. Reference Wang, Halbrendt, Kolodinsky and Schmidt1997):
$$L\left( {\beta ,\sigma } \right) = \prod\limits_{{y_i} = 0} \Phi \left( {{{ - \beta '{x_i}} \over \sigma }} \right)\prod\limits_{{y_i} = y_i^*} {{1 \over \sigma }} \phi \left( {{{{y_i} - \beta '{x_i}} \over \sigma }} \right)\prod\limits_{{y_i} = 10} {\left[ {1 - \Phi \left( {{{10 - \beta '{x_i}} \over \sigma }} \right)} \right]} $$
where
$\phi \left( \cdot \right)$
and
${\rm{\Phi }}\left( \cdot \right)$
represent the standard normal density function and distribution function, respectively.
The expected value of latent variable y* is
$E\left[ {y_i^*{\rm{|}}{\bf{x}}} \right] = \;{\bf{\beta }}'{\bf{x}}$
. The coefficient vector β thus contains the marginal effects of the independent variables on the latent variable y* and is represented as
${{\delta E[{y^*}|x]} \over {\delta X}} = \bf\beta $
. The marginal effect of the independent variables on dependent variable y (WTP) is the product of β and the probability that y* is greater than 0 but less than 10 (Greene Reference Greene2012).
$${{\delta E[y|x]} \over {\delta x}} = {\bf\beta} *Prob\;[0 \lt {y^*} \lt 10]$$
The Tobit estimates for participants who had the lowest and highest WTP could be inaccurate because marginal effects for the extreme quantiles are likely to be different than the conditional mean (Gustavsen and Rickertsen Reference Gustavsen and Rickertsen2011), which is assumed to represent the preferences of the entire sample. If the coefficients for various quantiles of participant choices are diverse, the results could fail to reflect the true heterogeneity of preferences.
Since some of the participants reported WTP at the lower and upper bounds (see Fig. 2), the conditional mean may not be truly representative of the sample. To investigate this further, we conducted censored quantile regressions (CQRs) to estimate a coefficient for each quantile (Buchinsky Reference Buchinsky1998), determining whether estimates of WTP at the extremes are different from estimates at mean and median WTP. Furthermore, unlike the Tobit estimator, CQR estimates are consistent in the presence of heteroskedasticity and nonnormally distributed errors (Powell Reference Powell1986). Thus, CQR Footnote 8 provides a further robustness check on the estimated WTP.
In addition to the base programs in R, we use the pwr and censReg packages for the empirical analysis (Team Reference Team2017).
Results
Table 4 presents the results of the Tobit regressions analyzing the effect of the local label on consumer WTP for oysters and mushrooms. Though we find no statistical evidence that consumers are willing to pay a premium for locally produced oysters and mushrooms, we find a statistically significant negative relationship between the perceived distance defining local production and WTP for oysters. A one-mile increase in the distance perceived as local decreases WTP for oysters by $0.005. Interaction of the local label with perceived distance has no effect on WTP.
Table 4. Tobit regression results: effect of local label on willingness to pay for oysters and mushrooms
Standard errors in parentheses
***p < 0.001, **p < 0.01, *p < 0.05.
We further find some associations between WTP for the oysters and several demographic characteristics. Individuals who politically identified as moderate were willing to pay $0.78 more for oysters on average than individuals who identified as liberal. Individuals with the highest levels of education were less willing to pay for oysters than individuals who had only a high school diploma, and Asian and Black consumers were willing to pay more for oysters than white consumers. The results for mushrooms reveal no statistically significant relationships between consumer demographics and WTP. We also find that income plays no significant role in WTP for oysters and mushrooms. In both cases, the marginal effects of the independent variables on WTP are similar to the coefficient estimates from the models.
Because perceptions of oysters as local can vary by attributes such as distance to the coast, we controlled for experiment-site fixed effects. We find no significant differences in WTP associated with the field sites, including the ones in coastal areas, relative to individuals who completed the experiment in lab settings.
To test the sensitivity of our results, we also ran Tobit regressions for the sample excluding participants who never consumed the food item and for a model in which all individual observations were pooled rather than disaggregated by food item. Those results are consistent; we once again find no evidence of an effect of the local label (see Tables B2 and B3 in Appendix B).
Robustness tests
Table 5 presents the CQR estimates for participants at the 5th, 25th, 50th (median), 75th, and 95th percentiles of WTP using a reduced-form model and pooling the data from all participants. As shown in the table, we find no difference in treatment effects for the lowest and highest quantiles and no significant treatment effect in any of the percentiles. Furthermore, we find that the distance perceived as local does not consistently predict WTP for all quantiles. The perceived distance is significant for the 25th and 95th percentiles but is not significant for the median participant or for the 75th percentile.
Table 5. Regression result from censored quantile regression: effect of local label on willingness to pay (pooled observations)
Standard errors in parentheses.
***p < 0.001, **p < 0.01, *p < 0.05.
Overall, the results of our statistically powered incentive-compatible field experiment and associated robustness checks consistently show that the generic “locally produced” label had no effect on Mid-Atlantic consumers’ WTP for mushrooms and oysters and that the distance perceived as local was negatively associated with WTP for them. We thus find that generic labeling of mushrooms and oysters as locally produced and potentially even labels identifying these products as produced in a consumer’s home state could fail to generate price premiums.
Implications of our findings for policymakers and producers
This experiment mimics the current policy environment where no standard definition of “local” exists, and consumers undoubtedly perceive local labels differently. The design also employs a between-subject design as most settings where food is labeled as local do not have a nonlocal version of the product also for sale in that setting. In this study, consumers were unwilling to pay a premium for “local” products, which suggest that such a generic label is not influencing their buying decisions. Thus, the results cast doubt on the efficacy of publicly funded labeling programs that provide only generic descriptions such as “Locally Produced.” This potential implication for promotion programs is further supported by the inverse relationship found between the WTP premium and the distance perceived as local. The wider the area described as local, the more WTP declines. Ultimately, local food producers potentially could benefit from policymakers establishing an agreed-on geographic or distance-related definition of “local” that does not extend beyond communities or, at most, regions within a state.
Because there is currently no universally agreed-upon definition of local, many consumers associate the term with geographic boundaries such as states or regions (Thilmany McFadden Reference Thilmany McFadden2015; Martinez Reference Martinez2016). This perception is supported by the large number of state-level programs that fund marketing and promotion initiatives for products grown within state boundaries. As shown in Table 1, the vast majority (47 of 50) of states currently engaged in at least one local food promotion program. This study shows that such investments in local food promotion policies may not be effective at garnering a market premium and potential premiums decline as the geographic distance widens. Without a market premium, producers may not be able to serve local markets in the long term since the production of local foods is typically small-scale and incurs higher costs per unit than industrial production.
Furthermore, the results could indicate that the initial boost in demand from local food promotion programs may have diminished over time. Trends are indeed shifting as we see recent sales of local edible farm products in the U.S. driven primarily by retailers, institutions, and intermediate markets rather than direct-to-consumer outlets (Martinez Reference Martinez2021). Further research on whether consumer preferences for locally produced food have shifted over time is certainly warranted.
Conclusions
Consumers are increasingly seeing foods labeled as locally produced in retail venues, but whether these labels are benefiting local food systems depends upon two important questions that remain: (1) Are consumers willing to pay a premium for local production and (2) What is the definition of local? Studies that found that consumers were willing to pay premiums for locally produced products were mostly based on hypothetical surveys and the definition of local varied by study. Further complicating our understanding of consumers’ WTP premiums is the fact that few incentive-compatible economic field experiments have been conducted so far and, with the exception of a notable few, most existing studies have relied on relatively small sample sizes.
We conduct an incentive-compatible nonhypothetical field experiment with a large sample size of adult consumers in the Mid-Atlantic region of the United States to assess consumers’ WTP premiums for locally produced mushrooms and oysters. We used a power analysis to identify the sample size needed to measure the effects of a generic local-label treatment. Participants were randomly assigned to either the local-label treatment group or the no-label control group and were asked to submit their highest WTP for mushrooms and oysters using funds endowed to them at the beginning of the experiment. We chose mushrooms because an extensive market for local mushrooms was established in the Mid-Atlantic region. We chose oysters because locally produced oysters are a relatively new product in the market.
We find no evidence that labeling these products as “locally produced” generates price premiums. We find that the lack of effect of labeling a food generically as local persisted even when we interacted with the effect of the label with consumers’ perceived definitions of local in terms of miles. The consideration that a price premium may not exist is important since the meta-analysis of studies of WTP for local food raised a concern about existing literature as there is evidence of publication bias in favor of nonnull results (Printezis et al. Reference Printezis, Grebitus and Hirsch2019). Publication bias can result in exaggerated effect sizes, especially when coupled with low sample sizes (Ferraro and Shukla Reference Ferraro and Shukla2022).
Our study analyzes a sample size that is large enough to provide adequate statistical power to detect economically significant treatment effects. These findings generally contradict the results of numerous prior studies of foods labeled as local. It is important to note that most of these other studies relied on hypothetical survey methods (Printezis et al. Reference Printezis, Grebitus and Hirsch2019) and, with a few exceptions, the revealed preference for economic experiments on the effects of labeling local production often employed designs with relatively small sample sizes (Table 2). The results are in line with results from a couple of previous studies; generic local food campaigns that fail to identify the outer boundary of the products’ local origins are likely to be ineffective in promoting WTP for the products. Because we find a negative relationship between distance perceived as local and WTP, labels indicating that foods come from nearby farms, such as in the county or region, could be more effective.
Note that this study uses a generic local label. Future research could test the effects of different kinds of local labels and scopes of definitions of local as consumers’ perceptions of localness are likely to vary in different regions and for types of products. States often promote particular products based on the production volume or quality of product in that region. Future research could also explore whether consumer perceptions vary based on the source of information (e.g., federal USDA marketing efforts, State-level marketing efforts, or directly marketed by retailers).
Finally, studies of the long-term viability of local food systems, in general, are needed. Most existing studies, including this one, have evaluated specific products in particular regions. A nationally representative survey could explore different definitions of local labels for a wide variety of food products to identify common and disparate characteristics of labels that successfully promote demand across regions and products. The continued study of consumer preferences for local foods is especially important for policymakers to understand how to best support local food systems now that preferences may have shifted due to the COVID-19 pandemic. As consumers faced supply chain disruptions, one solution was to turn to local markets. Huang et al. Reference Huang, Sant’Anna and Etienne2021 found that sales of local produce increased among higher and middle-income households during the pandemic; however, in many cases, COVID-19 restrictions led to the closure of farmers’ markets, a popular retailer of locally grown food. A longitudinal study tracking the effects of a local label for a series of food products over time would shed light on whether perceptions have changed over time and how the pandemic influenced preferences for local foods.
Ultimately, the results of this study have important implications for policymakers in terms of how best to promote markets for local foods, the likely effectiveness of the types of programs designed to date, and the level of government support potentially required to develop and sustain the systems. At least in the case of mushrooms and oysters in the Mid-Atlantic of the United States, labeling the products as “locally produced” does not appear to be an effective marketing strategy.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/age.2023.27
Data availability statement
The data that support the findings of this study are available from the corresponding author, [KAD], upon request.
Acknowledgments
The authors wish to acknowledge the University of Delaware Center for Experimental and Applied Economics for logistical support, equipment, and feedback on the experimental design of this study. Many thanks especially to former graduate student Melissa Langer for her tireless efforts in recruiting participants for this study. We appreciate the opportunity to present this research in the Online Agricultural and Resource Economics Seminar (OARES) series where we received valuable feedback from attendees and coordinators. We also express deep gratitude to the two anonymous reviewers and editor Trey Malone.
Funding statement
This work was supported by the National Science Foundation EPSCoR Grant 1757353.
Competing interests
The authors declare none.
Open access
