The price penalty for red meat substitutes in popular dishes and the diversity in substitution

Life cycle assessments (LCA) often highlight the environmental and health benefits for consumers if western diets substitute red meat. However, the specific trade-off consumer face when asked to substitute a red meat dish is scarcely researched, often neglecting the bouquet of substitution options and/or the price component involved. Four substitution strategies are evaluated within an individually adapted choice based conjoint: the substitution by (1) the same red meat dishes with a halved meat portion size, (2) novel plant-based products that mimic the functionality and taste, (3) authentic plant-based components that just mimic the functionality, and (4) vegetarian dishes that just neglect the meat component if still familiar to consumers. The analysis is executed for three popular red meat dishes to account for consistency across meal scenarios, namely Meatballs, Spaghetti Bolognese and Sausage Buns. The analysis is sensitive to red meat consumption habits to better understand the preferences of consumers that can actually substitute a red meat intake.


Introduction
I enjoy eating beef, but I fear we have to eat less. Beef production requires roughly 36 (±13) times the feed mass than the edible meat generated [1]. The conversion rate points to the amount of land resources required to provide food for human consumption. The carbon footprint of 1 calorie derived from beef is estimated at 22 g CO2-equivalent in contrast to 0,05 g for 1 calorie derived from pulses [2]. The widely noted EatLancet report considered the health and environmental impacts of diets and recommends roughly 2.6 kg of beef or lamb per year [3], while within the EU-27 about 15 kg are available for consumption each year [4]. From the report follows that the current level of red meat consumption is not sustainable, threatens our health and depletes our means of production. Nevertheless, red meat consumption is deeply rooted in western dietary patterns [5]. Consumers are often faced with tasteful red meat variations inherent to western food cultures.
This status of red meat in western food cultures makes red meat prevalent, especially in food settings that allow only for a limited number of offerings such as when eating-out. About 7% of all EU-expenditures are spend on the out-of-home food market [6]. The share is growing each year [6]. This out-of-home market is characterized by suppliers that have to prioritize the meat component. Many variations of the option (c) exist for decades but have not been able to mainstream into the dietary patterns of meat-eaters in the west. Nevertheless, several consumer segments prefer the idea of meat substitution with authentic alternatives over meat resembling products [15]. Familiar vegetarian meal that simply neglects the meat (d): In some instances, the meat component is just neglected (4) and the dish remains familiar and authentic to consumers. Consumers are used to these dishes but they are often thought of as less appealing or low budget alternatives.
Conclusively, none of the substitution strategies is currently on an eye level with red meat consumption. The substitution categories have also been described in more detail in a mapping review by Boer and Aiking [5]. The review highlights how change strategies should build on the availability of a range of substitution options in order to give consumers flexibility. Unfortunately, consumer researchers have predominantly focused on just one category and denied consumers flexibility in the substitution task. A parallel analysis can highlight relative potentials, particularly with heavy meat eaters that are asked to substitute. Consumer research needs to better identify the type of consumers willing to substitute with any category, with a specific one and with none. Additionally, most substitution studies neglect the economic perspective, such as estimating WTP differences, which can make it difficult to communicate the results to policy makers and caterers.
This study fills the gap and accounts for a multitude of options to substitute. An utility framework for an individually adapted choice based conjoint (CBC) will be applied, where consumers are asked to decide between the set of substitution options and an applicable popular red meat dish. The analysis explores substitution preferences across three popular dishes, i.e. three case studies. Further, we calculate the carbon footprint of each evaluated dish to account for trade-offs between environmental benefits and consumers' WTP. The diversity of the substitution categories is typically not available in the out-of-home market that needs to prioritize dishes. A better understanding of the price penalty that substitutes invoke can help progressive food business and researchers to design more promising interventions on who and how to enable red meat substitution.

Research objective
Based on the described substitution pathways two main research objectives are addressed: 1. An estimation of the price penalty (lower average WTP) for the substitution options, 2. An estimation of the price penalty sensitive to past consumption levels of the red meat dish at hand. While the former is of interest to caterers and restaurants that want to maximize revenues, the latter emphasizes frequent eaters that can actually substitute the red meat dish. Thereby, it provides a societal perspective on the potential of substitution, if substitutes were readily available in the market. Empirical evidence on the WTP for substitution pathways is still limited. The out-of-home market predominantly supplies the red meat dishes. A WTP analysis estimates beef burgers at almost three times the WTP of plant based burgers [17]. Other studies have shown how authentic alternatives or reduced meat portions can reduce meal satisfaction or sensory liking [12,13,18]. Therefore, we hypothesize a price penalty for all the substitution options (Table 1, H1), which is more pronounced for heavy red meat eaters because higher consumption of red meat is associated with a lower acceptance of substitutes [15] (Table 1, H3).
As we have previously explained how the preferences between substitution options are hardly understood, yet, we speculate that the price penalty is not equal for all substitution options (Table 1, H2). This implies an ordinal consumer preference structure, which also holds for the preference structure of heavy red meat eaters ( Table 1, H4). An ordinal structure would be of interest to caterers and researchers faced with designing substitution interventions. However, it should be complemented by an idea of the type of consumer segments interested in a specific substitution category. We supplement the research objectives by exploratory research on the relationship of consumer characteristics with a preference for each substitution option. Each hypothesis is confirmed if a statistically significant difference (5%-level) is consistently found across all 3 cases of popular red meat dishes (Table 1).
Further, the carbon footprint of each dish is approximated, as one environmental indicator that urges the substitution of red meat. The consideration of the CO2-footprint helps to indicate whether a CO2 tax, which is broadly discussed, would be able to compensate for the price penalty between any of the dishes. Therefore, the estimated WTP results are complemented by a report of the CO2 tax-level demanded to level consumer preferences.

Study design
The proposed study is a cross sectional survey experiment in the form of an individually adapted choice based conjoint, where participants evaluate hypothetical food choices in an out-of home lunch setting. Participants evaluate substitution options against popular red meat dishes, while the design is sensitive to heterogenous WTP among consumers.

Three popular red meat dishes
To apply the previously presented substitution categories to tangible cases of food choices, we introduce three popular red meat dishes (RMD): (a) Spaghetti Bolognese, as a "mixed dish" for which meat is not the main meal component [5]. The quantity of meat is not overly visible and consumers do not necessarily recognize a reduction of meat within the sauce. (b) Meatballs with rice and peas, as a "meat-centered" dish where plant-based ingredients are typically considered a side-dish [5], and (c) buns with butter and sausage, which presents a snack option for lunch instead of a warm meal decision ( Table 2). The red meat dishes are well known. European consumers have developed a habitual use of them.
For the given dishes CO2 footprint estimations are available ( Table 2). The IFEU Institute supplied CO2-footprint averages for the major ingredients of the dishes based on comparable value chains, capturing emission from farm to retail [20]. This allows for a calculation on Is the price penalty less severe for any of the substitution options?
U(S1,S2,S3,S4)6 ¼ U(S1,S2, S3,S4) Model 1 confirmed if at least one substitution option provides consistently greater or lesser utility across all 3 RMD cases Is the price penalty more pronounced if consumers frequently eat the red meat dish?
Model 1 and 2 confirmed if consistently shown for all 3 RMD cases Is the price penalty less severe for any of the substitution if consumers frequently eat the red meat dish?

PLOS ONE
The price penalty for red meat substitutes in popular dishes and the diversity in substitution whether a CO2 tax-level can change the ordinal preference structure. Different CO2 tax-levels are globally discussed, e.g. 55 € per CO2 ton (5,5 cents per kg) is planned in Germany for the energy and transportation sector [21].

Sampling procedure
A quoted consumer survey will be launched online. Participants shall be representative of Germans with internet access by age, gender and income [22]. Participants are invited to the study via a market research company who deals out a small reimbursement fee. The recruitment process allows to not inform about the particular focus of the study to avoid self-selection bias. Data quality checks include completeness and streamlining. Streamlining ensures that participants use a minimum of time on a set of questions [15]. Moreover, it monitors how participants progress through question blocks in order to exclude participants with speeding and/or systematic answering behavior. The monitoring of overall time spend on the survey becomes unnecessary. No other data exclusion criteria are planned. Budgetary constraints allow for 1000 consumers. Power calculation have been executed with a Cox model in STATA comparing a regression slope (the marginal utility for the substitution options) to a reference (the optout). Under the assumption of a minimum detectable effect size of 0.2, standard deviation of

Choice procedure
After socio-demographic questions, consumers are introduced to the out-of-home setting. Consumers are asked to imagine a lunch routine where a facility was preselected and now they may select one out of five dishes. The dish components are named but the dishes are no further introduced because the unbiased consumer beliefs and opinions are of interest to the research objective. Before consumers face the choice sets, they are exposed to a cheap talk that explains how they should imagine the purchase situation, act as if real money would be asked of them and how consumers sometimes misinform on their preferences. Participants are required to confirm that they understood the cheap talk. The cheap talk reduces hypothetical bias [23].
The choice options are the substitution categories and the popular RMD (Table 3). Consumers are exposed to six choice tasks for each RMD (Table 3). In total 18 choice sets (6 choice-sets per choice scenario x 3 choice scenarios) are evaluated, which seems acceptable given that consumer fatigue should be lower because different meal scenarios are surveyed.
The choice-sets include only four attributes to reduce the burden on consumers. Many attributes can cause information overload so that consumers develop decision rules based on just one or two attributes, which results in poor estimations of the remaining attributes. The four attributes are: the dish (5 levels), waiting time (3 levels), CO2-footprint (3-levels) and price (4 levels within IACBC). The choice alternatives are labelled with the dish names. Each dish occurs once per choice set (alternative specific constant). Price levels are essential to the WTP analysis and waiting time represents the value of time which matters particularly in lunch settings and hints at a convenience factor of interest to all kinds of substitution interventions. The CO2-footprint indicates just how relevant the simple naming of a footprint advantage is to the task of reducing the price penalty. The popular red meat dish is treated as no-

PLOS ONE
The price penalty for red meat substitutes in popular dishes and the diversity in substitution purchase option with constant attributes, therefore the utility derived from substitutes is anchored to the utility derived from the red meat dish. Altogether, 15 levels of 4 product attributes are evaluated ( Table 4). The design of the attribute levels in each choice set is calculated with a STATA implementation of a d-efficient design based on a modified Fedorov-algorithm [24]. The STATA code is available upon request. The design assumes an utility penalty for each substitution dish in order to allow sensitive estimations at this range of values. Further, the design allows for the alternative specific constants and a fixed red meat dish option ( Table 3). The choice-sets are followed by a set of questions on consumer characteristics.

The choice pricing mechanism-Individually adapted choice-based conjoint (IACBC)
We apply an individually adapted choice-based conjoint (IACBC) [25]. Regular choice-based conjoint often suffer from extreme response behavior, i.e. a substantial number of consumers have always or never selected the no-purchase option. Consequently, there is no information on when these consumers are willing to accept a substitute or when they would stop buying the substitute, which results in poor WTP estimations of such consumers. For this study, there is little indication of how low a price level has to be to persuade frequent meat-eaters to endorse a substitute. IACBC accommodates heterogeneity with very low and very high consumer WTP. IACBC discounts the prices of all the substitution options by a specific factor whenever the no-purchase option is selected and multiplies prices with a factor greater than 1 whenever one of the substitution options is selected. Additionally, the prices for the substitutes oscillate.
To illustrate the IACBC price algorithm, imagine a scenario of a consumer willing to buy any substitute at a price of 6 € or lower. The popular red meat dish is always offered at 8 €, which is a 2 € price premium. Such a consumer will select a substitute if the price drops below 6,01 €. In theory, the IACBC is also equipped to detect WTP differences between just the substitutes but is simplified here. The price of the dishes consists of two components. The base The substitutes are subject to base and flexible price components set in IACBC design flexible price has 4 level for each dish, see also Table 6 Total price for RMD: (1)  price level presents a stable minimum price level for each substitute, which is always added to the flexible price. The flexible price component is multiplied with a multiplier that follows the function: where n represents the number of shifts in direction between purchase and no-purchase decisions. z equals 1 if a substitute is selected and -1 if the no-purchase option is selected. The multiplier is set to 1 in the first choice-set. If a substitute is selected in choice-set 1, the multiplier is set to 2 and asymptotically converges to 1 (Table 5). For example, in choice-set 1 ( Table 5) a substitution option has been selected, so that in choice-set 2 the price for option 1 equals: base price (4 €) + flexible price (5 €) x multiplier (2) ( Table 5).
The IACBC is implemented with the methodological toolbox of dise-online (Dynamic Intelligent Survey Engine) [26]. The choice-design of the attribute levels creates unique sets and ensures regular price changes (flexible price component) between each set to capture preferences between substitution options (Table 4).

Variables
The consumption of the RMD is controlled for with a frequency scale ranging from never to at least once a week, which has previously been applied to pulses that are not a daily meal component for any consumer [27]. The RMD consumption is needed to assign less weight to consumers currently not consuming the RMD which the upcoming section explains. Other explanatory variables complement the utility function to better understand the type of consumer with preferences for a specific form of substitution. The analysis can point out consumers more or less willing to accept a specific substitute. We intend to explore the relationship with the following food related characteristics: a. Beef-eating habits, b. Experience with the substitutes, c. Attitude towards out-of-home dining, d. Convenience factor in lunch settings, e. Attitude towards the red meat dishes, f. Price sensitivity and g. Socio-demographics.
The measurement of these concepts is summarized in Table 6 and resemble previous applications. For example, attitude towards out of home dining will be operationalized with the following items, "Going out for lunch is regularly part of my eating habits" and "I do not consider it luxury to go out to have lunch in a restaurant", which have been introduced within the food related lifestyle concepts by Grunert et al. [28]. The concept allows to control for general preferences with respect to out of home consumption, the scenario consumers are exposed to.

Data analysis
The IACBC is consistent with random utility theory [33]. Consumer "n" derives utility from the choice alternative "a" in a given choice set "t". We apply a random parameter logit model, similar to Van Loo, Caputo, and Lusk [17] analysis of preferences for lab grown and plant based beef. The following utility function will be estimated: where, μ a is an alternative specific constant indicating utility derived from dish "a" relative to the red meat dish (baseline). β n are random parameters for the marginal utility derived from x nat , a vector for price, waiting time, and the CO2-footprint. This allows for consumer heterogeneity with respect to preferences, i.e. how consumers value a price penalty or a reduced waiting time. If the model cannot converge, waiting time, CO2-footprint, and if needed price, can be estimated as fixed parameters. δ a are fixed, alternative-specific coefficients on z it , a vector of case-specific variables (income, gender, experience with substitutes, etc.) that describes consumer characteristics associated with preferences for each dish. ǫ iat is a random or unobservable term that follows a type I extreme value distribution. The probability that consumer "n" The items ask for specific beef products to support participants memory of common beef product when evaluating their own consumption. The frequency for the following consumption items is addressed: Respondents were asked to "think of the out-of-home purchase situation and rate their agreement with 3 items: • I am willing to make an extra effort to find a low price for a meal • I will change what I had planned to buy in oder to take advantage of a lower price for a meal • I am sensitive to differences in price chooses alternative "a" in choice set "t", conditional on the random parameters, can be written: From the probability model, the utility for each dish will be calculated. Further, consumers are questioned on the frequency they consume the reference dish, i.e. the popular red meat dish. The response will be used for a subgroup analysis or rather to increase the weight of consumers that heavily consume the red meat dish. The weighting mechanism leads to a second model specification that better accounts for the preferences of consumers that can actually substitute. In a nutshell, the model (1) estimates the utility explained by substitution choices and consumer characteristics based on Eq 1. Model (2) repeats the estimation while weighting consumers by how often they eat the red meat dish. For both models, we will show the marginal effect sizes, p-values, not just the threshold level, and confidence intervals. A postestimation test for seemingly unrelated regressions can further indicate significant differences for coefficients between model specification 1 and 2.
In case of low discriminatory power between the utility derived from the substitutes, i.e. similar WTP for each substitution pathway, a latent class regression approach can be applied to understand the heterogeneity between consumer clusters.

Limitations
When working with stated preferences, it should be acknowledged that food choices are not necessarily conscious decisions. Stated preferences suffer from contextual biases. However, a rational endorsement in a stated preference scenario can be expected to lead to less resistance towards the substitute during food intervention scenarios. Observed and stated preferences are known to correlate. The explored choice tasks do not account for substitution behaviour between different types of cuisines. Consumers may feel like Spaghetti Bolognese, but are swayed by the price of an unrelated substitute, e.g. a salad. The full complexity of food decision making cannot be simulated. The relationship between consumer characteristics and the WTP for substitution options will be pointed out but should not be interpreted as causal inference. A causal analysis of consumer characteristics is limited through the study properties of crosssectional survey data. Prices, waiting time, and CO2-footprint are manipulated between participants and choice set to allow for causal conclusions.