Thursday, 9 August 2018
Do consumers prefer local animal products produced with local feed? Results from a Discrete-Choice experiment
Elsevier
Food Quality and Preference
Volume 71, January 2019, Pages 217-227
Food Quality and Preference
Author links open overlay panelAdrianoProfetaUlrichHamm
https://doi.org/10.1016/j.foodqual.2018.07.007
Get rights and content
Highlights
•
Do consumers prefer local feed for animal products labelled as ‘local’?
•
A survey and discrete choice experiment with 1602 consumers were conducted.
•
High preferences and WTPs are evident for all food products with local feed.
•
Animal products produced with local feed could open up a new market niche.
Previous article in issue
Next article in issue
1. Introduction
Many studies have revealed a high consumer preference and willingness-to-pay for local food (Feldmann and Hamm, 2015). Nonetheless, there is very little research on whether consumers accept the use of imported feed for animal products labelled as ‘local’. Europe is heavily dependent on protein-rich feed from South America and the US (Watson et al., 2017) and this dependency holds particularly for the animal production hot spots in the Netherlands and North-Western Germany (Van Grinsven, Spiertz, Westhoek, Bouwman, & Erisman, 2014), and the organic sector.
Soybean cultivation in Brazil and Argentina is linked to deforestation, savannah removal, and land grabbing (Boerema et al., 2016, Smaling et al., 2008). Most German consumers do not know that locally produced animal products in Germany are often produced with imported feed and that these imports are associated with negative environmental or social effects (Uhl and Schnell, 2014, Wägeli et al., 2016). In the EU, research and cultivation of domestic protein plants have been increasingly promoted. The German Federal Ministry of Food and Agriculture (BMEL) has started a protein strategy to reduce the proportion of imported soya (BMEL, 2016). Nonetheless, producing animal products with local feed in Europe is usually more expensive than using imported protein feed due to comparative disadvantages in production costs (Kaltenecker, Kemper, Schaack, & von Schenk, 2017). Therefore, local production chains are only economically feasible for a farmer if either public subsidies are paid to the farmers, or higher prices for animal products produced with local feed can be achieved in the market.
Currently, most animal products sold in Germany are not labelled with any information about the feed used in the production process and only a few studies with limited samples have considered placing emphasis on local feed origin. Wägeli et al. (2016) showed that there is a high potential demand for such labelling, at least in the organic market sector. The present paper is not limited to the organic food segment and analyses if German consumers prefer a local feed labelling on local food. For this purpose, a Discrete-Choice experiment (DCE) was applied for the product categories ‘eggs’, ‘milk’, ‘pork cutlets’ and ‘beef steaks’. In the framework of the DCE for labelling the local feed origin, the German label ‘Regionalfenster’ was used (see Chapter 3 and Fig. 1). In this paper, the term ‘local’ is used throughout rather than the word ‘regional’. ‘Regional’ in the word ‘Regionalfenster’ has not been changed since this is the proper name of this labelling.
Download high-res image (240KB)Download full-size image
Fig. 1. Regionalfenster (original and translated).
The study also aims to understand how two labels with distinct, but potentially complementary characteristics – local (product & feed) and organic - interact (Costanigro, Kroll, Thilmany, & Bunning, 2014). We also aimed to provide insights into reasons explaining consumer preferences for animal products produced with local feed. For this purpose, data regarding consumer behaviour (e.g. buying frequency of organic products) and attitudes (e.g. local consciousness) were collected and integrated in the applied logistic regression models. Hempel and Hamm (2016) showed that organic-minded consumers have strong preferences for local products. This study tests if these findings can be transferred to local feed and to conventional consumers as well.
2. Consumers’ preferences and motives for local food and scale development
From previous studies (e.g. Feldmann and Hamm, 2015, Köster, 2009), it is known that consumer preferences for local food are driven by a number of motives. According to Feldmann and Hamm (2015), some consumers criticize increasing imports in the national food market. This group regards local food as a more environmentally and climate friendly alternative. Furthermore, consumers also express greater trust in local food products, as local food was perceived as safer and easier to trace its origin (Burchardi et al., 2005, Darby et al., 2008, Nganje et al., 2011, Yue and Tong, 2009). Brown, Dury, and Holdsworth (2009) concluded that individuals who regularly make sustainable food choices often do so for more altruistic reasons. Altruistic attitudes towards local food dealt with support of the local economy and community through social relationships and/or proximity (Bean and Sharp, 2011, Burchardi et al., 2005, Dunne et al., 2011, Roininen et al., 2006, Yue and Tong, 2009). Wägeli et al. (2016) showed that the use of local feed in the production of animal products likewise addresses some of these motives, in particular “support of the local region”, “short distances of transport” and “food safety”. In this context, this study focuses on the impact of the construct ‘local consciousness’ on the preference for a local feed labelling.
Based on the concept of ethnocentrism, Shimp and Sharma (1987) introduced the concept of consumer ethnocentrism. This is defined as “the beliefs held by consumers about the appropriateness of purchasing products originating in a foreign country”. Consumers with a strong consumer ethnocentrism are more likely to buy local products (van Ittersum, 2001, Sharma et al., 1995, Shimp and Sharma, 1987). The construct ‘consumer ethnocentrism’ can be measured by the consumer-ethnocentrism-scale (CET-scale). This scale consists of 17 items which evaluate agreement versus disagreement on a 7-point Likert-scale (see e.g. Orth, & Firbasová, 2003). On this foundation, Balling, 2000, Staack, 2002 developed local consciousness scales that represent a condensed, modified variant of the CET-scale. Based on these two studies and the item battery applied by Wägeli (2014), we created an adapted local consciousness scale consisting of nine items as presented in Section 4.4.
Due to overlaps in the associations with organic and local food products, and the determinants for organic and local food purchases (e.g. environmental friendliness), consumers who regard one of the product attributes important are more likely to favour the other as well (Mirosa and Lawson, 2012, Robinson-O'Brien et al., 2009). Along these lines, Hempel and Hamm (2016) showed that organic-minded consumers have stronger preferences and higher estimated WTP-values for local products. For this reason, and to test if the mentioned findings can be transferred to local feed as well, this study looked at the preference for organic products. For this purpose, respondents’ answers for the buying frequency of organic food and the buying frequency at organic shops were combined to create an organic scale (OS) that displays the preferences for organic food (see Section 4.4.).
3. ‘Regionalfenster’ as carrier for a local feed labelling in the DCE
In 2014, the label ‘Regionalfenster’ (literally regional window) was introduced for locally produced food in Germany. Stakeholders and initiators were food producers, local food initiatives, retailers, and control bodies. These are organized in the ‘Regionalfenster e.V.’. In 2017, the ‘Regionalfenster’ had 760 licencees and more than 4000 products carried the label (Regionalfenster, 2017a). Furthermore, the German federal government plans to expand and strengthen the Regionalfenster approach (see https://www.cdu.de/koalitionsvertrag-2018). A recent study of the Thünen institute (Zander, 2018) revealed that consumers are willing to pay a price premium of about 20% for carrots and strawberry jam if these products are labelled with the ‘Regionalfenster’. In the mentioned study, brand awareness of the ‘Regionalfenster’ was about 30%.
The label is characterised by criteria that include a clear definition of the region of origin (namely administrative district, definition of a distance from the place of production, federal state, or natural boundary), a precise allocation of the ingredients to the region, and transparent control through a neutral, three-step inspection system (Regionalfenster, 2017b). In a Germany-wide survey (Hermanowski et al., 2014) consumers (N = 2018) were asked about when they consider a food product as local. About 37% indicated a German federal state, whereas 25% stated a definition of local food that refers to the distance between the point of production and the point of purchase, with specifications ranging from zero to 100 km. Fifteen per cent defined the term local by landscape boundaries and approximately 7% by the name of a city or town. Only a minority of 9% considered a German origin on the whole as local. The found consumer perception of the term local fits well together with the regulation of the ‘Regionalfenster’ association that stipulates that the use of the label is reserved to local regions that are smaller than Germany. In this context an additional labelling of the local feed origin is allowed only if 100% of all feedstuffs stem from local production (see Fig. 1). For the DCE, the ‘Regionalfenster’ was used as the carrier for a local feed labelling.
It is to highlight that this study did not only focus on the 100%-local feed labelling. Notwithstanding the ‘Regionalfenster’ guidelines, lower local feed shares of 90% and 75% were considered in the DCE as well. On one hand, it can be assumed that a local feed share of 75% appears within reach for many farms through minor efforts because the predominant part of the feeding rations can be covered by locally produced carbohydrate-rich feed grain. On the other hand, the last 10%, respectively 25% of the ration that is the protein-rich component, represents a distinct problem. Despite of the efforts made by the German ministry’s protein strategy, it is still not possible to fully cover the German poultry industry’s demand for soybeans with domestic protein sources (Witten, Paulsen, Weißmann, & Bussemas, 2017) So far, how consumers react to feed labellings that indicate local feed shares below 100% has not been considered. One could hypothesise that consumers may be satisfied with lower local feed shares. Therefore, we analyse if there is a turning point where the WTP no longer increases substantially.
4. Methodological background
4.1. Data collection methods
Cross-sectional consumer data was collected using a quantitative survey approach in which consumer choice experiments were conducted to measure the importance of different levels of local feed for eggs, milk, pork cutlets and beef steaks. Choice experiments have been shown to reduce social desirability bias (Huang, 2006, Kreuter et al., 2008, Tourangeau et al., 2003), as individuals often display socially desirable preferences to interviewers (Phillips and Clancy, 1972). Computer-Assisted Self-Interviews (CASI) were conducted with 1602 consumers in conventional retail shops in four different parts of Germany (Bavaria, Hessia, Lower Saxony, Brandenburg) to assure that only food shoppers were interviewed and that the shoppers had not been influenced by interviewers while answering. It is to highlight that respondents from discount shops were not part of the survey.
The study design and the practicability of the experiment were tested in a pre-test with 25 participants. The pre-test results led to slight changes in the questionnaire design, but not in the experiment. The local consciousness of respondents was measured by eight statements (see Section 4.4) from which ‘local scale’ was created. For this purpose, a seven-point Likert scale ranging from totally agree to totally disagree was applied. To capture the preferences for organic food, the buying frequency of organic products and the buying frequency in organic markets were collected. Based on this data an ‘organic scale’ was created (see Section 4.4).
4.2. Description of the choice experiment
DCE method is based on micro-economic theory according to which consumers always try to maximize their benefit (McFadden, 1974). In DCEs, consumers must choose from a set of different products offered at determined prices. The products differ regarding the tested product attributes (e.g. share of local feed, price, etc.). According to micro-economic theory, participants will choose the product with the highest benefit. By means of DCEs, consumers’ benefit for each tested product attribute can thus be revealed, as well as the influence of each product attribute on the probability of purchasing the product. The DCEs were conducted for eggs, milk, pork cutlets and beef steaks. The products varied by three attributes: local feed share, EU-organic-label, and price. The options were: no labelling at all, ‘Regionalfenster’ without indication of a local feed share, ‘Regionalfenster’ with indication of a 75% local feed share, ‘Regionalfenster’ with indication of a 90% local feed share and ‘Regionalfenster’ with indication of a 100% local feed share (see Table 1).
Table 1. Analysed attributes and their characteristics.
Parameter Characteristics
local feed share without ‘RF’ ‘RF’without indication of local feed share ‘RF’& 75% local feed share ‘RF’& 90% local feed share ‘RF’& 100% local feed share
price very low low medium high very high
milk (1 L) 0.49 € 0.69 € 0.89 € 1.09 € 1.29 €
pork cutlets (200 g) 1.19 € 1.79 € 2.39 € 2.99 € 3.59 €
eggs (6er-Pack) 0.89 € 1.19 € 1.49 € 1.79 € 2.09 €
beef steaks (200 g) 2.99 € 3.59 € 4.19 € 4.79 € 5.39 €
EU-Organic-Logo Without logo
For the survey, the indicated production region and the origin of the feedstuffs in the ‘Regionalfenster’-label were adapted to the survey region. For example, in Bavaria for the origin of the milk and for the origin of the feed, the name of the federal state Bavaria was indicated. As processing/filling location, the postal code and the name of a city near the survey location were chosen. E.g. for the survey in Nuremberg, the city of Fuerth was chosen, which is 8 km away Nuremberg. The EU-organic-label was included since previous studies had shown the importance of this aspect to consumers (Janssen and Hamm, 2014). Furthermore, it allows for the consideration of the interaction between the tested local feed labels and the organic label. All tested labellings were dummy-coded. The five price levels used in the choice experiment were within the price range that encompassed observed market prices at food retailers in Germany during the winter of 2016/2017 (see Table 1).
In each choice set, consumers had the choice between three product alternatives and a no-choice option. The no-choice option was included to get a more realistic purchase situation and thus raise the validity of the data (Hensher, 2010). A D-efficient unlabelled design (D-error = 0.949) was generated using the software Ngene (ChoiceMetrics, 2012). The priors used are based on the analysis of the study from Wägeli et al. (2016) and on expert judgement. For each product category, twenty-seven sets were grouped into nine blocks (three choice-sets per block). Each respondent received one block from each product category. The study participants thus had to make three different choices for each kind of product (milk, eggs, pork cutlets and beef steaks), resulting in twelve choices in total. The survey order of the choice sets within each block, as well as the positioning of the alternatives, were randomised to prevent ordering effects (Loureiro and Umberger, 2007). The DCE was included in a computer assisted survey in which the products were depicted in photographs (product packages) like in an online-shop (see Fig. 2).
Download high-res image (139KB)Download full-size image
Fig. 2. Choice-Set example.
Murphy, Allen, Stevens, and Weatherhead (2005) showed that conducting hypothetical experiments can lead consumers to overstate their true willingness-to-pay (WTP). While this choice experiment was conducted hypothetically, hypothetical bias must not necessarily impact the testing of our main hypotheses. If we assume that hypothetical bias is constant across participants and attributes, we can still confidently examine relative consumer WTPs for the attributes within the models. This study is interested in understanding marginal WTPs (mWTP), rather than the total WTPs for a product.
4.3. Theoretical background
A random parameter logit model (RPL) was used to analyse consumer preferences for the extrinsic parameters under examination. The RPL is based on random utility theory. The models estimated reveal the surveyed individuals’ preferences for the discrete set of eggs, milk, pork cutlets or beef steaks alternatives offered. As such, the observed outcomes display relative preferences for the modelled set of alternatives. The intensity of preferences is represented by the utility Unsj of alternative j perceived by respondent n in choice situation s. Preconditioned, the respondents act as utility maximizer; they choose the alternative with the highest utility. The utility Unsj is assumed to consist of a deterministic component Vnsj and an stochastic component εnsj (Hensher, Rose, & Greene, 2015):
(1)
The deterministic component Vnsj is assumed to be a linear function of observed attribute levels x of each alternative j and their corresponding weights (β), with a positive scale factor σ:
(2)
where βnk is the marginal utility associated with attribute k for respondent n (Hensher et al., 2015). Population parameter weights that vary randomly around a mean are estimated. Thus,
(3)
is the mean of the distribution of marginal utilities held by the sampled population, ηk represents a deviation of preferences among the respondents and znk represents random draws taken from a pre-specified distribution for each respondent n and attribute k (Hensher et al., 2015). The probability that respondent n in choice situation s will choose alternative j is given as the probability that outcome j will have the maximum utility:
(4)
A dummy coding was used for the no-buy alternative and the attributes organic label and local feed share, setting the non-labelling as reference category. Given the theoretical background, data was modelled according to the following utility expressions (model I). In the subsequent models and tables RF is used as abbreviation for ‘Regionalfenster’:
(5)
(6)
Within the RPL, the attributes ‘organic label’ and ‘local feed share’ were modelled as random components. It was assumed that the random parameters were normally distributed. Price was best modelled under a combination of a linear and quadratic price term. Both parameters were set as fixed, as recommended by Revelt and Train (1998). The no-buy alternative was modelled as an alternative specific constant (ASC). Halton draws with 250 replications were used for all estimations presented in this paper. To check the robustness of the results, the same models were also estimated based on random draws and shuffled Halton draws. All procedures came to similar results. To compare the model fits of different models, the log-likelihood ratio-test was used (Hensher, Rose, & Greene, 2005).
To account for possible interactions between the local feed labellings and the ‘organic label’, a second model (model II) was estimated. In the following equation, all local feed labels were interacted with the ‘organic label’ (O.L.):
(7)
In a further step, the local feed labellings were additionally interacted with the preferences for organic food (OS) which was measured by the developed ‘organic scale’ (model III). This approach allows consideration of how the preferences for organic food moderate the evaluation of the local feed labellings. Furthermore, the organic label was interacted with the ‘organic scale’ on one hand and the ‘local scale’ on the other hand. Moreover, the local feed labellings were interacted with the ‘local scale’ (LS) to check the impact of local consciousness on these labellings (see formula (8)).
(8)
When price and the effect of interest have linear functional forms, the marginal willingness to pay is calculated by −βeffect/βPrice. If price has a quadratic functional form, another approach must be applied. In the expression for utility, let βPrice and βPrice_Squared be the coefficients for Price and Price_Squared. Let C be the current price and Δu be the change in utility caused by the changes in the other attribute. This may be a change in a one level of a dummy variable to the other or perhaps a change in one unit of a continuous attribute. The marginal WTP is the price change needed to equalize utilities and is a solution to the equation (Onken, Bernard, & Pesek, 2011):
(9)
This can be formulated as the quadratic equation aWTP2 + bWTP + c = 0, where a = βPrice_squared, b = βPrice + 2CβPrice_Squared, and c = Δu. The quadratic formula is:
(10)
In the calculated models, βPrice and βPrice_Squared are negative. This implies that the desired solution is the one with the minus sign. Thus, the price shows an ideal point (negative parabolic) function. In this study, the WTPs for the local feed labels under consideration were calculated based on the described model II. For this purpose, the interaction effects between the organic and the local label were accounted for.
4.4. Description of sample
Data was collected at 16 supermarkets of conventional retailers in 13 locations in four different regions of Germany (North = 380, South = 403, East = 409, West = 410). Due to their restrictive policies, it was not possible to integrate discount shops in the survey. Screening questions ensured that participants regularly consumed at least one of the tested products (milk, eggs, pork cutlets and beef steaks). Furthermore, consumers had to be at least 18 years old. Consumers were approached by a random rule, i.e. the interviewers addressed each third consumer in the shops to participate in the study. In total, 1602 interviews entered the final analysis. Regarding age and gender, the sample composition was similar to the German average. The mean is 47.6 years compared to 50.4 years in the adult (over 18 years) German population. In the sample 53.4% of the participants were female and in the German population 50.7% (German Statistical Office, 2017).
The answers for the buying frequency of organic food and the buying frequency at organic shops (see Table 2) were added up to create an ‘organic scale’ that displays the preferences for organic food (see Table 3). Table 4 delivers an overview on the distribution of the ‘organic scale’ values and indicates the centred values (zero mean) that were used for the subsequent analyses.
Table 2. Buying frequency of organic food/at organic shops.
Mean (sd)
Buying frequency of organic food never = 1 3.02 (1.32)
less than once a month = 2
1–3 per month = 3
once per week = 4
repeated times per week = 5
Buying frequency at organic shops never = 1 1.68 (0.81)
sometimes = 2
often = 3
very often = 4
Table 3. Calculation scheme for the organic scale (OS).
Buying frequency of organic food
1 2 3 4 5
Buying frequency at organic shops 1 2 3 4 5 6
2 3 4 5 6 7
3 4 5 6 7 8
4 5 6 7 8 9
Table 4. Distribution organic scale (OS) values.
Organic scale values
OS
OS centred 2
(−2.7) 3
(−1.7) 4
(−0.7) 5
(+0.3) 6
(+1.3) 7
(+2.3) 8
(+3.3) 9
(+4.3)
N 247 235 267 322 230 179 72 50
% 15.4 14.7 16.7 20.1 14.4 11.2 4.5 3.1
OS8 OS7 OS6 OS5 OS4 OS3 OS2 OS1
To account for the impact of the individual local consciousness based on the statements in Table 5, a ‘local scale’ was created. For the applied item battery, a standardised Cronbach alpha of 0.71 could be calculated. The statements “When I buy food, aspects other than the origin are more important to me” and “The food origin is not important to me, the main thing is organic” were inverted. In a first step for each individual, the statement values were added up to create the ‘local scale’. Then the ‘local scale’ values were z-transformed, and the z-scores were used for the subsequent analyses.
Table 5. Statements local consciousness (n = 1,602).
Statements mean sd
I buy local foods to support the local economy 5.6 1.6
I have a greater trust in local food products 5.5 1.6
I’m willing to pay a higher price for local food products 5.5 1.6
Food from their own region is more eco-friendly 5.3 1.7
By buying local food products, I support animal welfare 4.7 1.8
Generally, local food is more expensive 4.6 1.7
When I buy food, aspects other than origin are more important to me 3.5 1.9
The food origin is not important to me, the main thing is organic 2.9 1.8
Likert scale ranging from totally disagree (1) to totally agree (7).
5. Results and discussion
5.1. Consumer preferences for local feed
In the models based on formula. (5), (6), the coefficients had a positive sign for all tested labelling attributes. The presence of the product attributes (level) had a positive influence on the probability of buying (see Table 7). For all products, the linear price effects and the price squared effects had negative coefficients. Hartl and Herrmann, 2009, Onken et al., 2011, or Troiano et al. (2016) who carried out DCEs for food, likewise found the combination of a linear and a squared price superior to a linear price parameter alone. In all product categories, the ‘would not purchase’ alternative had a significant negative effect.
For all four products, (milk, eggs, pork cutlets and beef steaks), consumers preferred the ‘Regionalfenster’ with 100%-local feed over all other options. The parameter coefficients for the feed labelling options decreased with falling local feed share. The coefficients between the different local feed shares were significantly different from each other (Wald-test, p ≤ 0.01). The drop in the coefficients from ‘Regionalfenster’ & 75% local feed share to ‘Regionalfenster’ without indicated feed share was significantly higher for all products than the differences between all other analysed local feed shares (Wald-test, p ≤ 0.01). The results demonstrate that the indication of feed origin delivered an added value to consumers as found by Wägeli et al. (2016) for the organic consumers segment. Furthermore, this holds not only for a 100% local feed labelling, but for 90% and 75% as well, at which the added value decreases. The coefficient ‘Regionalfenster’ without indicated feed share can be interpreted as a (pure) local product label. Thus, the corresponding coefficients indicate the net effect of the local product origin (see Table 6). For the calculation of the net effect of the different local feed (shares), the coefficient for ‘Regionalfenster’ without indicated feed share must be subtracted from the ‘Regionalfenster’ labellings with a local feed share (e.g. net effect 100% local feed share = βRF − βRF&Local Feed 100%).
Table 6. Net effects of local product origin and net effects of 100% local feed share.
Eggs Milk Pork cutlets Beef steaks
Net effect local product origin 0.58 0.46 0.38 0.60
Net effect 100% local feed share 2.41 2.05 2.68 2.25
Table 7. Estimation results - model I.
Eggs Milk Pork cutlets Beef steaks
Mean estimates Random Parameters
organic label 0.41 *** 0.48 *** 0.23 ** −0.04
‘RF’ without indicated feed share 0.58 *** 0.46 *** 0.38 * 0.60 ***
‘RF’& 75% local feed share 1.74 *** 1.45 *** 1.58 *** 1.63 ***
‘RF’& 90% local feed share 2.28 *** 1.75 *** 2.54 *** 2.25 ***
‘RF’& 100% local feed share 2.99 *** 2.51 *** 3.06 *** 2.85 ***
No-Buy-Option −4.02 *** −3.94 ** −4.66 *** 4.05 ***
Mean estimates Non-Random Parameters
price −0.86 *** −0.73 *** −0.48 *** −0.39 ***
squared price −1.95 *** −3.77 *** −0.46 *** −0.44 ***
Sd. mean estimates
Organic Label 2.54 *** 2.69 *** 2.46 *** 2.08 ***
‘RF’ without indicated feed share 1.30 *** 1.32 *** 1.75 *** 1.33 ***
‘RF’& 75% local feed share 0.85 *** 0.07 1.06 *** 0.87 ***
‘RF’& 90% local feed share 1.16 *** 1.50 *** 1.00 *** 0.89 ***
‘RF’& 100% local feed share 2.73 *** 3.14 *** 2.53 *** 2.57 ***
No-Buy-Option 6.71 *** 5.33 *** 9.15 *** 7.84 ***
Observations 1602 1602 1602 1602
Log-likelihood −5155.71 −5181.74 −4991.54 −5104.50
McFadden’s R2 0.23 0.22 0.25 0.23
Note: ***, **, *Significant at the 1%, 5% and 10% levels, respectively.
Table 6 shows that the net effect of a 100% local feed share is five to seven times greater than the effect of the local product origin. It can be hypothesized that the indication of different feed shares in the DCE heightened the awareness of the respondents to this attribute. This could have led to a depreciation of the local product origin. For eggs and milk, the effect size of the organic label was comparable to the labelling ‘Regionalfenster’ without indicated feed share (see Table 7). For pork cutlets, the organic label coefficient was lower in comparison to the local product origin, although still positive. No significant effect could be found for the product category beef steaks. The findings are in line with the literature, which found higher preferences for the aspect ‘local’ than for ‘organic’ (Bazzani et al., 2017, Costanigro et al., 2011, Costanigro et al., 2014, Denver and Jensen, 2014, Feldmann and Hamm, 2015, Hempel and Hamm, 2016, James et al., 2009, Menapace and Raffaelli, 2017, Wirth et al., 2011).
In model II, the feed labels were all interacted with the organic label (see Table 8). Likelihood-ratio-tests that compared model I and model II revealed significant model fit improvements in all categories. For all products, the corresponding parameters revealed negative interactions that are predominantly significant. That is to say, the joint effect of the analysed local feed labellings and the organic label were lower than the summed up main effects of these parameters. Yue and Tong (2009), who applied a choice experiment for locally and organically grown tomatoes in Minnesota, likewise found such a negative interaction effect. The same holds for the research of Costanigro et al. (2014) that analysed the effects of organic and local labels on milk in the US. They concluded that local and organic labels are partial substitutes. Wägeli et al. (2016) showed that the concrete use of local feed in the production of animal products addresses attributes like ‘support of the local region’ and ‘short transport distances’. Along this line, Lusk and Briggeman (2009) reported that consumers’ preferences for organic food are significantly related to certain food values, such as naturalness, fairness, and the environment, which are likely to affect the valuation of other sustainability claims. They argued that, if consumers value sustainability, they may feel that buying locally grown products is already addressing the issue of sustainability, and therefore, other sustainability claims added to local production contribute marginally less than that claim alone (i.e., they are substitutes). This hypothesis is supported by a qualitative work of Berlin, Lockeretzy, and Bellz (2009) that stated that concepts of local, small-scale and organic were often blended in people’s minds. Contrary to our results, Gracia et al. (2014) revealed a significant positive interaction between the attributes ‘organic’ and ‘local’ for eggs in Spain. It can only be hypothesized that this difference is caused by the interaction ‘free range’ (attribute in the study) and ‘organic’, which was not considered in the model.
Table 8. Estimation results - model II.
Eggs Milk Pork cutlets Beef steaks
Mean estimates Random Parameters
organic label 1.63 *** 0.72 *** 0.94 *** 0.53 **
‘RF’ without indicated feed share 1.29 *** 0.82 *** 0.94 ** 0.92 ***
‘RF’& 75% local feed share 2.63 *** 1.90 *** 2.36 *** 2.12 ***
‘RF’& 90% local feed share 3.31 *** 1.92 *** 3.05 *** 2.63 ***
‘RF’& 100% local feed share++ 4.01 *** 2.78 *** 3.80 *** 3.56 ***
no-buy-option −3.50 *** −3.88 ** −3.98 *** −3.72 ***
‘RF’ without indicated feed share × organic label −0.68 ** −0.36 −0.71 ** −0.10
‘RF’& 75% local feed share × organic label −1.51 *** −0.66 ** −1.33 *** −1.02 ***
‘RF’& 90% local feed share × organic label −1.51 *** −0.21 −0.72 ** −0.61 *
‘RF’& 100% local feed share × organic label −1.56 *** −0.35 −1.03 ** −1.06 ***
Mean estimates Non-Random Parameters
price −0.94 *** −0.46 *** −0.48 *** −0.42 ***
squared price −1.41 *** −4.48 *** −0.38 *** −0.33 ***
Sd. mean estimates
organic Label 6.08 *** 2.70 *** 2.54 *** 2.08 ***
‘RF’ without indicated feed share 3.08 *** 1.35 *** 1.78 *** 1.22 ***
‘RF’ & 75% local feed share 1.92 *** 0.04 1.03 *** 0.84 ***
‘RF’& 90% local feed share 3.17 *** 1.53 *** 1.15 *** 1.08 ***
‘RF’& 100% local feed share 4.97 *** 2.87 *** 2.44 *** 2.34 ***
no-buy option 16.60 *** 5.49 *** 9.23 *** 8.05 ***
‘RF’ without indicated feed share × organic label 1.03 0.30 1.10 * 0.07
‘RF’& 75% local feed share × organic label 0.38 0.01 0.25 0.17
‘RF’& 90% local feed share × organic label 1.72 0.43 0.53 0.31
‘RF’& 100% local feed share × organic label 6.14 *** 2.35 *** 1.43 *** 2.10 ***
observations 1,603 1,603 1,603 1,603
Log-likelihood −5122.04 −5164.57 −4976.08 −5079.01
McFadden’s R2 0.23 0.23 0.25 0.24
Note: ***, **, * Significant at the 1%, 5% and 10% levels, respectively.
Considering the main effects for the organic label and the different local feed labellings in model II, the parameters had a larger effect size in comparison to model I. Due to the integrated interactions, the parameter for the organic label revealed the pure effect of this attribute when it is displayed without any local feed labelling. Likewise, the local feed parameters in model II show the effects when displayed without the organic label. Interestingly, the negative interaction terms were all relatively large, and for the meat products even larger than the effect of the organic label. The total utility of a local feed labelled product decreased when the organic label was added. This result appears irrational and cannot be explained by the ‘substitution approach’ alone. An explanation is that a subpopulation of consumers displays polarized preferences (see Costanigro et al., 2014, Hurley et al., 2011) in favor of local versus organic or vice versa. Many studies have shown that respondents may not necessarily weigh up all the attributes and their levels yet base their choice on a limited selection of attributes (Hensher, 2006). According to Kragt (2013), this behaviour is referred to as attribute non-attendance (ANA) and has been observed in stated choice experiments as discontinuous preference ordering (e.g. McIntosh and Ryan, 2002), lexicographic choices (e.g. Sælensminde, 2002), or disinterest (Rigby and Burton, 2006). If ANA exists, respondents do not make complete trade-offs among all the attributes presented, violating the axiom of continuity. Failure to recognize ANA can lead to erroneous and biased WTP estimates (Scarpa, Gilbride, Campbell, & Hensher, 2009).
In this study, we focused on the stated buying frequency of organic products and the buying frequency in organic supermarkets (organic scale) as sources for lexicographic behavior. Only 15.4% of the respondents (segment OS8) declared they never bought organic food products or shopped in organic supermarkets. According to Jensen, Denver, and Zanoli (2011), who carried out a cross-country study (UK, Italy, US), such non-users exhibit general resistance to the idea of paying higher prices for organic products. Moreover, these consumers tended to distrust organic farmers and growers, as well as the underlying certification schemes. Such a lack of trust was frequently mentioned in the literature (Hughner et al., 2007, Padel and Foster , 2005, Wier et al., 2008, Yiridoe et al., 2005). Furthermore, Jensen et al. (2011) stated that this distrust also extended to consumers of organic food who were sometimes seen as constituting a strange or deviant group, or as having been duped into paying premium prices by a fad of the food industry. One can assume that such a polarized consumer segment contravenes against the basic assumption in the analysis of choice data of unlimited substitutability between the attributes (in this case organic vs. local).
To account for lexicographic behaviour in model III, the organic label was interacted with the ‘organic scale’ (see Table 10). This approach leads in the direction of the equality constrained latent class models (ECLC) proposed by Kragt (2013) because it likewise allows for a more detailed segment analysis. Instead of using the ECLC, we preferred our extended interaction model because, thereby, the developed ‘organic scale’ could be integrated in the model as the explanatory variable that directly affects/moderates the degree of ANA. Another advantage is that the model can be kept more concise and clear. In addition, in model III the developed ‘local scale’ and ‘organic scale’ were interacted with the different feed labellings and the organic label.
Likelihood-ratio-tests that compared model II and model III revealed significant model fit improvements in all categories. Upon examining the results for the interaction organic label × organic scale, a positive and significant relationship could be found in all product categories. For each individual, this coefficient must be multiplied with its individual-specific ‘organic scale’ value (OSi) ranging from 4.3 to −2.7 (meanOS = 0) for the calculation of the absolute interaction effect. If we focus on the mentioned non-users of organic products (OS8), the value −2.7 (see Table 4 in Section 4.4) must be used for this multiplication. The calculated value must then be summed with the main effect of the organic label to receive the total effect of the organic label for the non-user group (Total effect organic label = βorganic label + βorganiclabel × organic scale × organic scale). For the non-users, the total effect of the organic label was not significant (see Table 9) or revealed negative parameters for three out of the four product categories. Therefore, these results provide some evidence for the existence of ANA. For the product categories ‘milk’, ‘pork cutlets’ and ‘beef steaks’, it must be said that the total labelling effect was even significantly negative.
Table 9. Calculation of the total effect of the organic label for the organic non-user-group.
Eggs Milk Pork cutlets Beef steaks
Main effect organic label 1.93 0.89 0.96 0.82
Interaction effect organic label × organic scale 0.75 × (−2.7) = −2.03 0.74 × (−2.7) = −2.00 0.59 × (−2.7) = −1.59 0.49 × (−2.7) = −1.32
−0.10 −1.11*** −0.63*** −0.50***
Note: ***, **, *Significant at the 1%, 5% and 10% levels, respectively.
Table 10. Estimation results - model III.
Eggs Milk Pork cutlets Beef steaks
Mean estimates Random Parameters
organic label 1.93 *** 0.89 *** 0.96 *** 0.82 ***
‘RF’ without indicated feed share 1.48 *** 0.64 *** 0.66 *** 1.04 ***
‘RF’ 75% local feed share 2.92 *** 1.82 *** 2.28 *** 2.26 ***
‘RF’90% local feed share 3.63 *** 2.10 *** 3.04 *** 2.90 ***
‘RF’100% local feed share 4.32 *** 2.84 *** 3.76 *** 3.77 ***
no-Buy-Option −4.30 *** −4.64 *** −4.53 *** −4.25 ***
‘RF’ without indicated feed share × organic label −0.85 ** −0.07 −1.01 * −0.17
‘RF’ 75% local feed share × organic label −1.74 *** −0.45 −1.26 *** −1.19 ***
‘RF’ 90% local feed share × organic label −1.79 *** −0.37 −0.63 * −0.90 ***
‘RF’ 100% local feed share × organic label −1.74 *** −0.47 −0.92 *** −1.22 ***
‘RF’ without ind. feed share × organic scale −0.21 *** −0.30 *** −0.01 −0.24 ***
‘RF’ 75% local feed share × organic scale −0.08 −0.15 ** −0.15 −0.09
‘RF ’90% local feed share × organic scale 0.16 ** 0.11 0.02 0.13 *
‘RF’ 100% local feed share × organic scale 0.44 *** 0.35 *** 0.41 *** 0.32 ***
‘RF’ without indicated feed share × local scale 0.27 ** 0.13 −0.12 −0.01
‘RF’& 75% local feed share × local scale 0.35 *** 0.23 ** 0.15 0.14
‘RF’& 90% local feed share × local scale 0.76 *** 0.64 *** 0.52 *** 0.64 ***
‘RF’& 100% local feed share × local scale 1.29 *** 1.20 *** 1.07 *** 1.05 ***
organic label × organic scale 0.75 *** 0.73 *** 0.59 *** 0.49 ***
organic label × local scale 0.08 0.23 ** 0.05 0.12
price −0.93 *** −0.72 ** −0.37 *** −0.46 ***
squared price −1.67 *** −4.38 *** −0.48 *** −0.33 ***
Sd. mean estimates
organic label 1.90 *** 2.11 *** 1.89 *** −1.58 ***
‘RF’ without indicated feed share 1.18 *** 1.24 *** 1.66 *** 0.87 ***
‘RF’ 75% local feed share 0.79 *** 0.01 1.04 *** 0.70 ***
‘RF’90% local feed share 0.74 *** 0.58 0.69 ** 0.64 **
‘RF’100% local feed share 1.87 *** 1.76 *** 2.21 *** 2.13 ***
no-Buy-Option 8.30 *** 6.19 *** 9.94 *** 8.80 ***
‘RF’ without indicated feed share × organic label 0.40 0.31 2.34 ** 0.14
‘RF’ 75% local feed share × organic label 0.23 0.01 0.59 0.67
‘RF’ 90% local feed share × organic label 1.03 * 0.98 ** 0.67 0.78
‘RF’ 100% local feed share × organic label 1.89 *** 2.50 *** 0.09 1.58 ***
‘RF’ without ind. feed share × organic scale 0.31 0.16 0.42 *** 0.27 *
‘RF’ 75% local feed share × organic scale 0.13 0.01 0.16 0.05
‘RF ’90% local feed share × organic scale 0.71 ** 0.12 0.31 0.06
‘RF’ 100% local feed share × organic scale 0.83 * 0.01 0.32 * 0.26
‘RF’ without indicated feed share × local scale 0.15 0.16 0.04 0.01
‘RF’& 75% local feed share × × local scale 0.02 0.01 0.22 0.34
‘RF’& 90% local feed share × local scale 0.39 *** 0.12 0.20 0.64 **
‘RF’& 100% local feed share × local scale 0.35 0.01 0.32 0.96 ***
organic label × organic scale 0.52 *** 0.10 0.72 *** 0.56 ***
organic label × local scale 0.53 0.52 ** 0.01 0.07
observations 1,603 1,603 1,603 1,603
Log-likelihood −4,885.56 −4,929.31 −4,784.77 4,901.97
McFadden’s R2 0.27 0.26 0.28 0.26
Note: ***, **, *Significant at the 1%, 5% and 10% levels, respectively.
Table 11. Marginal WTPs for Eggs, Milk, Pork cutlets and Beef steaks.
P1 P2 P3 P4 P5
Eggs
Model II Main effects
Organic Label 0.40 *** 0.34 *** 0.29 *** 0.26 *** 0.22 ***
‘RF’ without indicated feed share 0.32 *** 0.27 *** 0.24 *** 0.21 *** 0.17 ***
‘RF’ & 75% local feed share 0.61 *** 0.52 *** 0.45 *** 0.40 *** 0.35 ***
‘RF’ & 90% local feed share 0.73 *** 0.64 *** 0.56 *** 0.49 *** 0.43 ***
‘RF’ & 100% local feed share 0.86 *** 0.75 *** 0.66 *** 0.59 *** 0.52 ***
Milk
Model II Main effects
Organic Label 0.13 ** 0.10 ** 0.08 ** 0.07 ** 0.06 **
‘RF’ without indicated feed share 0.14 *** 0.11 *** 0.09 *** 0.08 ** 0.07 **
‘RF’ & 75% local feed share 0.31 *** 0.25 *** 0.20 *** 0.17 *** 0.15 ***
‘RF’ & 90% local feed share 0.31 *** 0.25 *** 0.21 *** 0.17 *** 0.15 ***
‘RF’ & 100% local feed share 0.41 *** 0.34 *** 0.29 *** 0.25 *** 0.21 ***
Pork cutlets
Model II Main effects
Organic Label 0.53 *** 0.41 *** 0.32 *** 0.28 *** 0.24 ***
‘RF’ without indicated feed share 0.53 *** 0.41 *** 0.32 *** 0.28 *** 0.24 ***
‘RF’ & 75% local feed share 1.14 *** 0.93 *** 0.76 *** 0.66 *** 0.58 ***
‘RF’ & 90% local feed share 1.40 *** 1.16 *** 0.95 *** 0.84 *** 0.73 ***
‘RF’ & 100% local feed share 1.65 *** 1.38 *** 1.14 *** 1.01 *** 0.90 ***
Beef steaks
Model II Main effects
Organic Label 0.21 * 0.19 * 0.16 * 0.14 0.13
‘RF’ without indicated feed share 0.36 *** 0.32 *** 0.28 ** 0.25 ** 0.23 **
‘RF’ & 75% local feed share 0.81 *** 0.71 *** 0.63 *** 0.57 *** 0.52 ***
‘RF’& 90% local feed share 0.97 *** 0.86 *** 0.77 *** 0.70 *** 0.63 ***
‘RF’& 100% local feed share 1.28 *** 1.14 *** 1.02 *** 0.92 *** 0.84 ***
Note: ***, **, *Significant at the 1%, 5% and 10% levels on the base of Wald tests.
Considering the interaction ‘Regionalfenster’ 100% local feed share x ‘organic scale’ in all product categories, positive effects could be found. For each individual, this coefficient must be multiplied with the individual-specific ‘organic scale’ value. Thus, it can be analysed how the individual organic preference modifies the parameter ‘RF’ 100% local feed share. For consumers with above-average preferences for organic food and thus positive ‘organic scale’ values (segments OS1(+4.3), OS2 (+3.3), OS3(+2.3), OS4 (+1.3), OS5 (+0.3)) the effect size is reinforced, whereas below-average and thus negative ‘organic scale’ values (segments OS6(−0.7), OS7(−1.7), OS8(−2.7) reduce this parameter.
For the consumer segment OS1 the expressed high preference for organic food even overcompensated the negative interaction between the 100% local feed labelling and the organic label in all product categories considered (e.g. eggs for OS1: (0.44 * 4.3–1.74 = +0.15). The signs change, and the interaction becomes positive. For milk and pork cutlets, this holds for the segments OS2 and OS3 as well. Concerning the interaction ‘Regionalfenster’ 90% local feed share x ‘organic scale’, significant and positively moderating effects of the preferences for organic food could be found only for eggs and beef steaks. For all other feed labellings, the ‘organic scale’ showed either no or a negative impact.
Concerning the impact of local consciousness, a positive interaction effect for the local scale existed for all products for a 100% and a 90% local feed share. Furthermore, positive and significant effects could be found for milk and eggs for the 75%-level as well. In the product category eggs, even for the interaction ‘Regionalfenster’ without indicated feed share and ‘local scale’, a positive and significant effect could be identified. It is to highlight that only for eggs there was a significant and relatively small effect of the local consciousness on the organic label. This means that local consciousness went along with an increased preference for a local feed labelling, but not necessarily with an increased preference for organic labelled products. This independency is supported by the extremely low correlation (ρ = 0.01) between the local scale and the organic scale. The asymmetry that the preference for organic goes along with local preference, but not vice versa, is interesting and supports the findings of Denver and Jensen (2014).
5.2. Marginal willingness to pay
Marginal WTP (mWTP) estimates for the different local feed labellings can be taken from Table 11. The mWTPs were calculated for each of the five price levels (P1-P5) used in the DCE. The figures for the displayed mWTPs are based on model II. This means the mWTPs reflect the effect for the different local feed share labelling when displayed without the organic label.
The WTP measures follow similar interpretation of the part-worth utilities in the logit models, but they offer Euro values for the various attributes. The numbers show the additional amount a consumer would be willing to pay at different price levels for the given attribute change. As should be expected, the mWTP numbers decline as the price increases and with increasing organic preference, the WTP increased.
For a comparison of the results with the findings of Wägeli et al. (2016) that focused on a 100% local feed label, the mWTPs (main effects) were considered against the background of the base price (P1). To receive the net mWTP for the different local feed shares from the corresponding mWTP, the effect of the pure local product label (‘Regionalfenster’ without indicated feed share) must be subtracted. Thus, net mWTPs for the 100%-local feed labelling of 0.54 € (eggs), 0.27 € (milk), 1.12 € (pork cutlets) and 0.92 € (beef steaks) was obtained. Wägeli et al. (2016) found higher mWTPs for 100% local feed for eggs (0.80 €) and milk (0.84 €) for the analysed organic consumers, whereas the mWTP for pork cutlets (1.18 €) has a comparable level.
6. Conclusions and limitations
The present study suggests that animal products produced with local feed could open up a new market niche. In all product categories considered, the study revealed high consumer WTPs for animal products produced with local feed. Due to problems with the economically sound cultivation of high protein plants in Germany, local production chains are only economically feasible for farmers when higher prices for organic animal products produced with local feed can be achieved in the market. In this study, outstanding WTPs could be found. Nonetheless, future economic studies that focus on the additional production costs of the use of local feed in the supply chain must be compared with these results. It is to highlight that the respondents paid significantly more for 100% local feed than for the lower feed share labellings. Due to the existing positive WTPs for a 75% and a 90% local feed share, economic comparisons with the production costs should be carried out for these levels as well.
In general, a local feed labelling can be recommended to the ‘Regionalfenster’ licencees who already have a 100% local feed supply. Furthermore, the ‘Regionalfenster e.V.’ should consider allowing the labelling of local feed shares below 100%. In this context, due to the positive WTP, a 75%-level appears to be an accepted threshold.
Organic production and demand are still on the rise. However, competition on the market is also becoming fiercer. The production of animal products using locally produced feed could be a promising differentiation strategy in this sector. The results show that, specifically, consumers with a high preference for organic food (segments OS1–OS5) have above-average preferences for (organic) local food produced with local feed for the levels 100% (all product categories) and 90% (only eggs and beef steaks).
From a methodological point of view, the paper delivers valuable insights on the interaction between local (feed) labels and the attribute ‘organic’. The negative interactions found quantitatively confirm the hypothesis that concepts of local and organic were often blended in people’s minds. This means the attributes local and organic are partial substitutes. Interestingly, the evaluation of the ‘Regionalfenster’ is moderated by consumers’ preferences for organic products. The higher the preferences for organic food, the better the 100% local feed labelling is evaluated. This moderating effect of the organic preference even overcompensates negative interactions in all product categories for the consumer segment OS1. From these findings it can be concluded that a complete supply of local feed is seen more and more as a complement as the preference for organic increases. Therefore, in organic shops where consumers with a high organic preference can be found, the 100% local feed labelling can be recommended for skimming the additional WTP, whereas for distribution in conventional retailers, a local feed labelling with lower shares is more appropriate.
In this study, it was not possible to collect data in discount shops. Discount shops represent a large segment in Germany and account for about 43% of the retail turnover (GfK, 2003). Therefore, the representativeness of the chosen sample is limited to the conventional food retail market without discount shops. Future studies should consider discount shoppers as well. Nonetheless, it must be said that it is difficult to get access to this distribution channel to carry out surveys. Furthermore, in this paper we did not focus on the motives for considering the local feed origin in detail. We introduced the local consciousness scale, but we did not differentiate whether the preference depends more on environmental aspects or on food safety reasons. Another point for future research is the identification of a lower threshold for the local feed share that is still accepted by consumers. In this study, we found an additional WTP for the 75% local feed share labelling, but we did not analyse if there is a preference for lower local feed shares (e.g. 50%) as well. Furthermore, we did not test if consumers expect a certain local feed share for the ‘Regionalfenster’ without indicated local feed share or if she/he interprets it as a signal of no local feed share. It can be hypothesized that the indication of different local feed shares in the DCE increased the awareness of the respondents to this attribute. This could have led to a depreciation of the local product origin regarding the labelling ‘Regionalfenster’ without an indicated local feed share. We propose to evaluate this probable impact via an appropriate experimental design strategy in future research. From a methodical point of view, it would be interesting to compare the applied models with the equality constrained latent class models as proposed by Kragt (2013) or a random parameter model with correlated parameters. Since experiments such as the reported study are limited to only a few product categories, such experiments must be replicated numerous times using many additional product categories to obtain generalisations across product categories (Wells, 2001).
Appendix A. Supplementary data
The following are the Supplementary data to this article:
Research data for this article
Data not available / The authors do not have permission to share data
About research data
References
Balling, 2000
Balling, R. ‘Ergebnisse von Verbraucherbefragungen zur Bedeutung der regionalen Herkunft bei Nahrungsmitteln’, Dachverband Agrarforschung (ed.), Regionale Vermarktungssysteme in der Land-, Ernährungs- und Forstwirtschaft, Bd. 30, (2000) pp 19–37.
Bazzani et al., 2017
C. Bazzani, V. Caputo, R.M.R.M. Nayga, M. Canavari
Revisiting consumers valuation for local versus organic food using a non-hypothetical choice experiment: Does personality matter?
Food Quality and Preference, 62 (2017), pp. 144-154
ArticleDownload PDFView Record in Scopus
Bean and Sharp, 2011
M. Bean, J.S. Sharp
Profiling alternative food system supporters: The personal and social basis of local and organic food support
Renewable Agriculture and Food Systems, 26 (3) (2011), pp. 243-254
CrossRefView Record in Scopus
Berlin et al., 2009
L. Berlin, W. Lockeretzy, R. Bellz
Purchasing foods produced on organic, small and local farms: A mixed method analysis of New England consumers
Renewable Agriculture and Food Systems, 24 (4) (2009), pp. 267-275
CrossRefView Record in Scopus
BMEL, 2017
BMEL Beans, Peas and Co. The Federal Ministry of Food and Agriculture's Protein-Crop Strategy for promoting the cultivation of pulses in Germany. Berlin: Federal Ministry of Food and Agriculture. 2017, Available at:
