The Pennsylvania State University
The Graduate School
ASSESSING THE VIABILITY OF MEAT ALTERNATIVES TO
MITIGATE THE SOCIETAL CONCERNS ASSOCIATED WITH
ANIMAL AGRICULTURE IN INDIA
A Thesis in
Energy, Environmental, and Food Economics
by
Rashmit Arora
2019 Rashmit Arora
Submitted in Partial Fulfillment of the Requirements for the Degree of
Master of Science
August 2019
ii
The thesis of Rashmit Arora was reviewed and approved* by the following:
Edward Jaenicke Professor of Agricultural Economics Graduate Program Director: Energy, Environmental, and Food Economics Thesis Co-Advisor
Daniel Brent Assistant Professor of Environmental Economics Thesis Co-Advisor
Amit Sharma Professor of Hospitality Management/Finance Director, Food Decisions Research Laboratory
Robert Chiles Assistant Professor of Rural Sociology
*Signatures are on file in the Graduate School
iii
Abstract
Meat alternatives such as plant-based and cell-based meat offer a demand-side solution to the environmental, nutritional, and other societal concerns associated with animal-intensive agriculture. However, little is known about the consumer preferences of meat alternatives, which will ultimately dictate their effectiveness in shifting demand away from conventional animal-based meat products. This thesis attempts to address this gap by assessing consumer preferences for four sources of protein – conventional meat, plant-based meat, cell-based meat, and chickpeas – in India, a rapidly developing country that has been consistently witnessing an increase in demand for animal-based protein. The sheer size of India’s population makes its existing and future consumption trends of global import. Using a discrete choice experiment (n = 394) that was conducted via a face-to-face survey in the city of Mumbai and analyzed by a latent class model, four heterogeneous segments in the market are identified. On average, respondents are willing to pay a premium of ₹138 ($1.97) per kg for plant-based meat and ₹57 ($0.81) per kg for cell-based meat over the price of conventional meat, with estimates ranging from -₹65 (-$0.93) to ₹261 ($3.73) for plant-based meat and -₹68 (-$0.97) to ₹320 ($4.57) for cell-based meat between segments.
Participants consistently rated plant-based meats higher than cell-based and conventional meat across several metrics, indicating that plant-forward substitutes have a higher potential for substitutability than cell-based alternatives. Furthermore, we conclude that vegetarians in India are not the target market for meat alternatives. The findings suggest that public policy and business strategy will be most effective when tailored to specific market segments based on their respective preferences and demographic makeups.
iv
Table of Contents
List of Tables ...... vi
List of Figures ...... vii
Acknowledgments...... viii
CHAPTER 1 ...... 1
Introduction ...... 1
CHAPTER 2 ...... 2
Background Information and Literature Review ...... 2 2.1. India and Meat ...... 2 2.2. The Current State of the Food System ...... 4 2.3. How Meat Alternatives Factor into the Equation ...... 7 2.4. Simulated Meat Preference ...... 10
CHAPTER 3 ...... 15
Survey Methodology ...... 15 3.1. Overview ...... 15 3.2. Pre-Testing ...... 17 3.3. Missing Data ...... 18
CHAPTER 4 ...... 19
Analytical Methodology ...... 19 4.1. The Choice Experiment ...... 19 4.2. Model Selection...... 23 4.3. Methodology ...... 25
CHAPTER 5 ...... 31
Results ...... 31 5.1. Descriptive Statistics ...... 31 5.2. The Latent Class Model ...... 37 5.3. Willingness-to-Pay ...... 40 5.4. Class-Membership Model ...... 43 5.5. Even Price Scenario ...... 48 5.6. Model Fit ...... 48 5.7. Robustness Checks ...... 50 5.8. Peripheral Results ...... 51
CHAPTER 6 ...... 57
Discussion ...... 57
v
6.1. Discussion of Results ...... 57 6.2. Policy Simulation & Implementation ...... 64 6.3. Shortcomings of the Survey ...... 68 6.4. Potential for Future Research ...... 69
CHAPTER 7 ...... 71
Conclusion ...... 71
Appendix A ...... 77
Appendix B ...... 78
References ...... 79
vi
List of Tables
Table 1: Price attribute levels in INR (₹)...... 22
Table 2: BIC, CAIC, and AIC results with different numbers of classes...... 26
Table 3: Sample descriptive statistics and relevant 2011 Census statistics...... 32
Table 4: Latent class model and class-membership model results...... 39
Table 5: Willingness-to-Pay estimates and confidence intervals with chana as the reference category...... 40
Table 6: WTP estimates with respect to conventional meat in INR and USD...... 41
Table 7: Logit regression of unaided awareness of plant-based and clean meat on demographic factors...... 52
Table 8: Logit regression of willingness-to-buy plant-based and clean meat on demographic factors...... 54
Table 9: Average consumer favorability ratings for conventional, plant-based, and clean meat ...... 56
Table 10: D-optimal fractional factorial design of attribute levels...... 77
Table 11: Results from conditional logit and mixed logit specifications...... 78
vii
List of Figures
Figure 1: Ward map of Mumbai (left); Survey interview locations (right)...... 16
Figure 2: Sample choice set...... 21
Figure 3: Income distribution comparison between the online Bryant Survey and our face-to-face survey ...... 33
Figure 4: Distribution of Household Income in India (Source: Patnaik (2019) and CMEI Consumer Pyramids Database) ...... 34
Figure 5: Age distribution of full sample and age distribution of vegetarians in sample...... 35
Figure 6: Weekly frequency of animal-based meat consumption...... 36
Figure 7: Vegetarianism by religion...... 45
Figure 8: Kernel density plots of the distribution of individual-level coefficients for clean meat (top left), plant-based meat (top right), and conventional meat (bottom)...... 49
Figure 9: Unaided awareness of plant-based meat and clean meat...... 51
Figure 10: Percentage of respondents willing to buy plant-based meat and clean meat prior to choice experiment...... 53
Figure 11: Choice card to answer consumer favorability questions...... 55
Figure 12: Consumers’ choice outcomes in a party scenario where they do not have to pay for food...... 56
Figure 13: Meat consumption of males and females in India by state...... 66
viii
Acknowledgments
I would like to thank my advisors Edward Jaenicke and Daniel Brent for their constant support and mentorship throughout the course of this study; Robert Chiles and Amit Sharma for their valuable advice and insight; Reenu Sharma, Sandhya Yadav, and Gerush Bahal for painstakingly collecting survey data with me in the nooks and corners of Mumbai; Ganesh Padwal and Subhash Kumar
Gautam for assisting in the execution of the survey; Chris Bryant, Eric Plutzer, and Kate Zipp for
their important insight; Penn State’s Department of International Agriculture and the College of
Agricultural Sciences Competitive Grant program for funding the data collection; James Shortle
for guiding me into this field and planting the seeds of this project in my head; and my family for
always believing in me and supporting me in my (often ridiculous) endeavors.
ix
To my grandfather, Professor Keshar Singh aka Darjee: For your wisdom, your compassion, and your inspiration, thank you.
CHAPTER 1
Introduction
The role of domesticated animals in agriculture can be traced as far back as the Neolithic revolution
over 10,000 years ago (Chiles & Fitzgerald, 2018). While animals have always been considered an
essential part of the human food supply, increasing environmental, ethical, and health-related issues
with intensive animal husbandry are raising concerns about their role in our agricultural systems.
For instance, one of the major environmental problems associated with animal agriculture is
livestock’s contribution to anthropogenic climate change. Most reliable estimates place livestock’s
share of global emissions between 12%-18%, while food systems in their entirety are estimated to
contribute between 19%-29% of global emissions (Gerber et al., 2013; Bellarby et al., 2013;
Vermeulen, Campbell, & Ingram, 2012; Van der Werf et al., 2009; Steinfeld et al., 2006). The
Intergovernmental Panel on Climate Change (IPCC) estimates that human populations have until
2030 to prevent global temperatures rising by 1.5°C, at which level of warming we can expect the
worst impacts of climate change, such as severe droughts, floods, and consequent surges in the flow
of climate refugees (Allen et al., 2018). The 1.5°C projection is seen as modest and conservative;
accounting for the potential of Earth’s positive feedback mechanisms—such as the recent earlier-
than-expected melting of Canadian permafrost (Reuters, 2019)—paints an even bleaker image of
our future, with temperatures rising between 3.5-4°C by 2050 and a dramatic intensification of the
aforementioned impacts, manifesting into an existential threat to human civilization (Spratt &
Dunlop, 2018; Steffen et al., 2018). Demand-side policies aimed at promoting behavioral change have been identified as key to mitigating the impacts of such environmental problems (Roglej et
1
al., 2018; Poore & Nemecek, 2018). Meat alternatives fall under this category of market-based, consumer-side solutions.
Meat alternatives – also known as meat analogues, meat substitutes, faux meat, mock meat, or
imitation meat – emulate the textural, aesthetic, and chemical characteristics of meat products
(Joshi & Kumar, 2016). They may be made from plants (plant-based), or artificially replicated in a lab (cell-based). Plant-based meat uses plant ingredients to imitate the taste, texture, and nutrition of conventional meat. Cell-based meat is produced through the culturing of animal cells and has an identical sensory and nutritional profile to conventional meat (Post, 2012). There has been some research on consumer acceptance and preferences towards meat alternatives, particularly cell-based
meat; however, next to none of this research is in the context of developing countries. This gap in
the literature is problematic since growing demand for animal-based proteins from populous
developing countries poses a substantial threat to global environmental resources (Bryant et al.,
2019; Taufik, 2018). India and China, specifically, have been recognized as hotspots where little is
known regarding consumer preferences towards meat alternatives (Bryant & Barnett, 2018).
The focus of this study is India, the second most populous – projected to be the most populous
within the next decade – country in the world (Samir et al., 2018). Home to approximately 1.3
billion people, 80% of who follow Hinduism, India is often misperceived to be majority vegetarian;
in fact, upwards of 70% of Indians consume meat at least occasionally (International Institute for
Population Sciences (IIPS, 2017). Although India’s per-capita meat consumption is low due to
nearly a third of the country not consuming meat, the total consumption and total livestock counts
are some of the highest in the world (OECD, 2019; Ritchie & Roser, 2019). India is observing one
of the highest rates of growth in chicken, mutton (goat/sheep meat), and fish consumption in the
world (OECD, 2018; Robinson & Pozzi, 2011). Keeping with the issue of climate change,
agriculture contributes to 19.6% of India’s total GHG emissions, with over 70% of those emissions
2
coming from livestock products such as mutton, poultry, eggs, and milk (Vetter et al., 2017). Vetter et al. (2017) and Green et al. (2018) warn that higher consumption of animal-based products such
as mutton and poultry are likely to drastically increase India’s GHG emissions and worsen water scarcity, pointing to the important role of consumption choices in meeting environmental goals.
Aside from the environmental impacts, there are substantial health-related, and food security concerns associated with India’s growing meat consumption (Cao and Li, 2013; Steinfeld et al.,
2006a; Chao et al., 2005).
Several studies suggest that plant- and cell-based meat alternatives provide considerable environmental and nutritional savings compared to animal-based meat (Ritchie, Reay, & Hiigins,
2018; Heller & Keoleian, 2018; Tuomisto & Teixeira de Mattos, 2011; Hoek et al., 2011).
Regardless of the suggested benefits, it is not clear whether consumers are willing to adopt meat alternatives, which will ultimately determine the effectiveness of any government policy or
business strategy aiming to divert animal-meat consumption towards them. This brings us to the
primary objective of this study: to assesses the viability of meat alternatives as a solution to the societal concerns associated with animal agriculture in the context of a developing country with rising demand for animal-based protein. The primary research question can be articulated as:
Are meat alternatives a viable demand side solution to mitigating the societal concerns
associated with animal agriculture in India?
A hypothetical discrete choice experiment (DCE) is employed to analyze consumer preferences for four alternative sources of protein: conventional animal-based meat, plant-based meat, cell-based meat, and chana. Chana, the fourth alternative, is the Hindi word for chickpeas, a popular legume
3
that contains 20g of protein per 100g (USDA, 2019). Meat is defined as any animal flesh; this includes red meat items such as beef, mutton, and pork, as well as white meat items such as chicken and fish. Primary data was collected using a multi-language face-to-face survey (n = 394) in
Mumbai, India, in December 2018. A latent class model is used to identify how preferences towards protein alternatives vary within the market. “Viability” is gauged based on respondents’ receptiveness to meat alternatives, which is indicated by their willingness-to-pay (WTP) estimates.
This study contributes to the literature by being the first to estimate and report willingness-to-pay for, both, plant- and cell-based meat alternatives. It is also one of the first studies to assess consumer preferences for simulated meat in the context of a developing country. The plant-based meat market in India is small, while the market for clean meat does not exist at all. As interest in this area of food technology ramps up, our results will prove to be beneficial for entrepreneurs and policy makers trying to encourage meat substitution in the subcontinent.
This section serves as an introduction to the paper’s context. The next section will provide background information and a literature review on several aspects of meat alternatives, focusing on important relevant studies in the space. The following two sections provide an overview of the survey and analytical methodologies. The concluding sections delineate our results, relevant discussion, and conclusions.
4
Note: The Good Food Institute conducted a survey exploring the receptiveness of consumers to
different names for cell-based meat (Szejda, 2018). They found that consumers responded most positively to the prefixes of “clean” and “safe”. In keeping with the findings of their survey and with the nomenclature used in recent literature, we will use the term “clean meat” to describe cell-
based meat. Additionally, in following the nomenclature set in Slade (2018), hereon, we will refer
to plant-based and clean meat alternatives collectively as “simulated meat”. It is also important to
note that there are no commonly used Hindi counterparts for these terms. Conventional animal-
based meat was translated to “saadharan maans”, plant-based meat “paudhon pe aadharit maans”,
and clean meat “saaf maans”, which are near-literal translations. Lastly, the contents of this thesis
use grouped personal pronouns such as “we” and “us” to allow for this transcript to be modified
for publication purposes with ease. While the final publication will include co-authors, the contents
of this thesis are solely the listed author’s work.
1
CHAPTER 2
Background Information and Literature Review
2.1. India and Meat
Dining at a restaurant in India, you will likely be offered a menu with dedicated ‘Veg’ and ‘Non-
Veg’ sections. These two colloquial short names for ‘Vegetarian’ and ‘Non-Vegetarian’ are
symbolic of the broad relevance of vegetarianism in the subcontinent. Vegetarianism in India often
takes the form of a lacto-vegetarian diet, i.e. most vegetarians consume some form of dairy products
(Agrawal et al., 2014). A predominantly Hindu country, India is often inaccurately assumed to be
majority vegetarian—an estimated 64-74% of the population consume some form of animal-based meat at least occasionally (IIPS, 2017).
The cultural and traditional value of meat has been explored in Western countries (Chiles, 2017;
Marshall & Anderson 2002). Extending findings from Western studies on meat consumption to
India is problematic considering the significantly different role meat has played in the subcontinent’s history (Sathyamala, 2018). A handful of academic articles document the dynamics of meat consumption in India (e.g., Sathyamala, 2018; Staples, 2018; DeLessio-Parson, 2016; Devi
et al., 2014). Since 2014, Narendra Modi has held the post of the Prime Minister of India. Modi
belongs to the Bhartiya Janta Party (BJP), a right-wing political party that has strong ties with fundamentalist Hindu organizations such as the Rashtriya Swayamsevak Sangh (RSS). While Modi has been in power, there has been an increase in communal violence against non-Hindus.
Specifically, beef-consuming minorities, which largely tend to be Muslims and Christians, have come under attack. Reports of violence against these minorities perpetrated by extremist gau
2
rakshaks, which literally translates to “cow protectors”, have become commonplace. Over 97% of such attacks have occurred since the BJP assumed power in 2014, and 86% of the attacks have been against Muslims (Abraham & Rao, 2017). Sathyamala (2018) and Staples (2018) provide deeper insight into the politics of eating meat in India.
While beef consumption is under attack, demand for animal-based meat, in general, is rising. The three most consumed types of meat in India are chicken, mutton, and fish (Devi et al., 2014). In a recent systematic review of studies projecting future dietary trends in India, Alae-Carew et al.
(2019) find that all eleven included studies project an increase in meat consumption in India, regardless of projection methodology. A 2011 analysis by the FAO aimed at projecting global consumption trends to 2030 finds that India is observing one of the highest growth rates in the consumption of chicken, mutton, and fish in the world (Robinson & Pozzi, 2011). Surges in the demand for animal-based proteins are already being experienced within Indian markets. Recent poultry industry reports, for example, indicate that rising prices for poultry in India – by as much as 50% in some cities – are largely due to rapidly increasing demand and an inability of suppliers to meet growing demand (Jha, 2019).
The increasing influence of Westernization as well as urbanization and spread of retail supermarkets have been documented to be drivers of rising consumption of meat products in India
(Pingali et al., 2019). The increasing demand for animal-based protein in developing countries is often attributed rising incomes (Steinfeld, Wassenaar, & Jutzi, 2006; Rosegrant, Agcaoili-
Sombilla, & Perez, 1995). In the above-mentioned systematic review, Alae-Carew et al. (2019) confirm that future increases in meat consumption in India will be directly proportional to increases in income. However, they also, interestingly, conclude that increases meat consumption will be relatively moderate, indicating that per-capita consumption of meat in India is expected to remain low. In a recent analysis on protein trends in India, Minocha et al. (2019) draw a similar conclusion
3
based on per-capita consumption of meat being low. This brings to light the importance of the
dichotomy between per-capita statistics and total statistics in this context.
In a country where approximately a third of the population does not consume meat, assessing meat
consumption through the lens of per-capita statistics paints a misleading image of reality.
Considering the size of India’s population, total consumption of all goods and services in the
economy is expected to be high. For instance, in terms of total consumption, India is the largest consumer of mutton (703,000 tons) and the fifth largest consumer of poultry meat (3,257,000 tons) in the world (OECD, 2019). On the other hand, in terms of per-capita values, India’s mutton consumption (0.5 kg per person) and poultry consumption (2.1 kg per person) are among the lowest
in the world (OECD, 2019). As we can see, both measures tell very different stories. It may be
argued that per-capita values are more useful for food security and nutritional purposes, whereas
total values are more telling of the environmental and ethical impacts of a country’s meat
consumption. Both statistics have their role in discourse; however, the focus on per-capita statistics by recent studies downplays the significance of India’s growing meat consumption, which is expected to place a significant strain on limited global resources.
2.2. The Current State of the Food System
Livestock production has, both, positive and negative externalities associated with it. The positive externalities have been documented and may include environmental benefits such as sustainable rangeland management and the improvement of soil fertility and nutrient cycling (Mearns, 1996).
More widely documented, however, are the negative environmental externalities. Recent calls for urgent action on climate change emphasize the existential threat it poses to human society (Spratt
& Dunlop, 2018; Steffen et al., 2018). Accounting for both indirect and direct emissions, food systems contribute between 19%-29% of global anthropogenic GHG emissions (Van der Werf et
4
al., 2009). Animal agriculture and its processing industries contribute between 12%-18% of global
GHG emissions (Gerber et al., 2013). India makes up 6.55% of global GHG emissions, making it the third largest contributor of anthropogenic GHG emissions (World Resources Institute, 2019).
The Indian livestock sector, specifically, is considered a significant emitter of GHG emissions globally (Naqvi & Sejian, 2011). Agriculture accounts for over 19.6% of India’s total GHG emissions (INCCA, 2010). In one of the few directly relevant studies, Vetter et al. (2019) conduct a rigorous analysis on India’s agricultural GHG emissions. They find that mutton is associated with the highest GHG emissions per kilogram, followed by other livestock products such as poultry, eggs, and milk. The order reverses in terms of contributions to total GHG emissions from India’s agricultural sector. Livestock products – primarily poultry and eggs – contribute to 50.5% of total agricultural emissions in India, with ruminant meat products such as mutton being the second highest contributors, making up 21.7% of total agricultural emissions. Combined, the livestock sector makes up over three fourths of India’s total agricultural emissions.
Green et al. (2018) echo these findings in their study exploring the GHG contributions and water footprints of different dietary patterns in India. They find that diets with some form of animal-based meat in them produce the highest GHG emissions. In terms of water footprints, they find that poultry is one of the most water intensive animal-based meats in India, requiring more than 20 liters of water per gram. India, Pakistan, China, and the United States use 72% of all the irrigation water used globally, with India alone accounting for 36% of that total (West et al., 2014). Agriculture accounts for 90% of India’s freshwater usage and freshwater scarcity is a severe problem faced by
Indians (FAO, 2016). A recent exhibition of this scarcity took place in Chennai, one of the largest cities in India, which ran out of drinking water in June of 2019 (Pathak, 2019).
There are other environmental concerns associated with livestock production, too. Growing meat consumption in Asian countries is seen as a major concern in terms of biodiversity loss (Machovina,
5
Feeley, & Ripple, 2015). Biodiversity loss in our oceans is being accelerated at unprecedented rates
due to overfishing (Montoya, Donohue, & Pimm, 2018). Nutrient pollution from agricultural
activities on land and from aquaculture operations in our oceans is leading to negative impacts on
water quality and widespread loss in aquatic life (Howarth et al., 2000). In general, food systems
that rely heavily on animals require more water, land, and energy and take a greater toll on
environmental resources than plant-forward systems (Pimentel & Pimentel, 2003).
Food security is a significant problem for India, where 195.9 million people are undernourished, which is approximately 15% of the population (FAO & UNICEF, 2018). Food insecurity is associated with losses in cognitive ability and productivity, which tend to have ripple effects across economies. Positive externalities from livestock are notable in this context. Animal products are an important source of micronutrients for some of the poorest people in the world (Whaley et al.,
2003). The Indian diet is a perfect example of a scenario in which the introduction of meat could improve dietary quality, since micronutrient-poor cereals are the primary source of protein and per- capita consumption of meat is low (Swaminathan et al., 2012). While recent studies call for an increase in animal foods consumption in India to tackle this issue (Minocha et al., 2019), they neither account for the environmental consequences of this solution nor for the potential of meat alternatives to address these concerns.
In terms of health benefits, animal-based meat is a good source of protein, vitamins B12 and B6, as well as iron, zinc, and phosphorous (Williams, 2007). However, there are significant health risks associated with overconsumption of meat products. Although consumers perceive health benefits to be associated with meat products (Verbeke et al., 2010), their consumption has been shown to
be connected to cardiovascular diseases and some types of cancer (Bovalino, Charleson, & Szoeke,
2016; Chao et al., 2005). India is experiencing an alarming increase in the number of individuals
6
with diabetes, and Indian diets containing meat products have been associated with higher
likelihoods of diabetes compared to plant-based diets in the subcontinent (Agrawal et al., 2014).
This paper is premised on our judgment that the negative externalities from livestock production
likely outweigh the positive externalities, and the inefficiencies within the system may, potentially,
be addressed by the introduction of plant- and cell-based meat alternatives. The following section
provides an overview of how meat alternatives can fulfil this requirement.
2.3. How Meat Alternatives Factor into the Equation
Several studies have focused on the environmental and food security benefits of consuming less meat or completely adopting plant-based diets (Springmann et al., 2018b; Aleksandrowicz et al.,
2016; Sau Chun Yip, Crane, & Karnon, 2013; De Boer & Aiking, 2011; Stehfest et al., 2009, etc.).
A meta-analysis of LCAs of agricultural products by Nijdam, Rood, and Westhoek (2012) finds that animal-based proteins produce higher emissions and require more land than vegetal sources of protein. Reducing meat consumption is also seen as the most important step to ensuring biodiversity conservation (Machovina et al., 2015). The health benefits of plant-based diets have been widely documented, too. Appleby et al. (1999) and Tuso et al. (2013) showed that individuals on primarily plant-based diets are generally healthier than those on other diets. Whole foods, plant-based diets have the potential of lowering cholesterol levels, body mass index, and blood pressure, while being cost-effective (Tuso et al, 2013). Several papers show that higher consumption of plant-based foods is associated with significantly reduced risks of cardiovascular disease (Mishra et al, 2013;
Ferdowsian & Barnard, 2010; Hu, 2003). Furthermore, low fat vegetarian and vegan diets have proven to be more effective in combating diabetes than non-vegetarian diets (Barnard et al., 2009).
Specific to India, lower likelihoods of diabetes are associated with lacto- lacto-ovo and semi- vegetarian diets in India (Agrawal et al., 2014). Lastly, from a food security perspective, animal-
7
based proteins are an inefficient converter of energy, as 1 kg of requires roughly 6 kg of plant-
protein (Aiking, 2011; Pimentel & Pimentel, 2003). Transitioning to a plant-forward food system
can be expected produce more food for more people using out existing agricultural land. Clean
meat, particularly, boasts an identical nutritional profile to conventional meat while eliminating the risk of animal-borne diseases, and can therefore be seen as a more sustainable, safer direct replacement for conventional meat in populations where meat is a key source of micronutrients.
A handful of studies focus specifically on the scope of meat substitutes across the societal metrics discussed above (Ritchie et al., 2018b; Apostolidis & McLeay, 2016; Nijdam, Rood, & Westhoek,
2012; Hoek et al., 2011). Compared to animal-based proteins, meat alternatives provide significant
benefits in terms of the environment. Ritchie et al. (2018b) showed that meat substitutes have a large potential for reducing GHG emissions in high-income nations (up to 583 MtCO2e per year).
Several life cycle analyses (LCAs) of existing plant-based meat products have been done.
California-based manufacturer Beyond Meat’s popular Beyond Burger produces 90% less GHG
emissions, requires 93% less land, 46% less energy, and has a 99% lower impact on water security
than a quarter-pound U.S. beef burger (Heller & Keoleian, 2018). Similarly, The Kellogg
Company’s subsidiary Morningstar Farms’ meatless products produce an average of 40% in environmental savings across several impact metrics; meatless meals using their products required
between 64%-84% less water and produced between 58%-77% less GHG emissions compared to
meals containing animal-based proteins, with dinner being the meal-of-the-day producing greatest
savings (Dettling et al., 2016). A 2011 LCA of clean meat technology found that it would require
7%-45% lower energy use, 99% lower land use, 82%-96% lower water use, and produce 78%-96%
lower GHG emissions (Tuomisto & Teixeira de Mattos, 2011). However, more recent analyses
suggest complex tradeoffs and consequent uncertainty in the environmental impacts of in-vitro
technology (Lynch & Pierrehumbert, 2019; Mattick et al., 2015).
8
Meat alternatives have directly been associated with lower cholesterol, reduced mortality due to
cancer, improved bone health, and eye health (Sadler, 2014). While there is limited research on the health impacts of clean meat, it is generally perceived to be nutritionally equivalent to conventional meat but significantly safer since it reduces the risk of animal-borne illnesses almost entirely (Post,
2012). In terms of addressing food security problems, meat alternatives show strong potential.
Shepon et al. (2018) show that plant-based equivalents of animal-based foods in the US alone can produce twenty-fold more nutritionally equivalent food per unit of cropland compared to conventional meat. They further find that a complete replacement of conventional meat by plant- based options would add enough food to feed an additional 350 million people to eliminate food- loss in the supply chain. Ritchie et al. (2018b) estimate that meat substitutes have a significant potential for improving nutritional outcomes, reporting the potential of preventing up to 52,700 premature deaths per year. Simulated meat products a resource efficient means of meeting food security needs in India – and this hypothesis can be carried over to other underdeveloped and developing countries with food security issues. Meat substitutes are nutritionally equivalent or superior to conventional meat counterparts, and therefore have the ability of providing high-quality nutrition to lower income populations, while minimizing the environmental damage from increased food production (Sadler, 2014). Clean meat alternatives have identical profiles to conventional meat (Van der Weele et al., 2019). An often-cited critique of plant-based meat alternatives is that they tend to be highly processed, which is an important challenge for the space to overcome in order to improve healthfulness (Patel, 2017).
9
2.4. Simulated Meat Preference
There has been considerable interest in consumer acceptance of clean meat since the technology’s
proof-of-concept in 2012 (Post, 2012). Verbeke, Sans, and Van Loo (2015) first analyzed the
consumer acceptance of clean meat using a survey in Belgium, conducted in April 2013. They
found low levels of unaided awareness of such alternatives (about 13% were aware); however,
when informed, there were high levels of acceptance (only 9% rejected the idea). Since then, several
papers exploring consumer acceptance towards clean meat have been published. For a systematic
review of this space, please see Bryant and Barnett (2018).
Plant-based meat alternatives such as tofu and seitan have been around for centuries – their roots can be traced back to early followers of Buddhism in China. Two popular meat substitutes in India are soya chaap and nutrela soy chunks. However, longstanding meat replications of the sort, at best, offer a partial solution to meat-attached populations due to existing negative perceptions towards them (Graça, Oliveira, and Calheiros, 2015). While there is limited research specifically analyzing consumer acceptance of plant-based meat alternatives, several studies have explored consumers’ attachment to meat and consequent unwillingness to adopt plant-based diets (Graça, Calheiros, &
Oliviera, 2015; Schösler, de Boer, & Boersema 2012; Lea, Crawford, & Worsley, 2006). The unwillingness is attributed to factors such as a lack of cooking skill and familiarity with options. A lack of sensory appeal has been determined to be a barrier to the adoption of plant-based meat substitutes (Hoek et al., 2011). Plant-based meat alternatives, however, are becoming more and more meat-like due to the innovation of texturized vegetable protein (TVP) and mycroprotein (Wild
et al., 2014). Replications produced by GoodDot India, Impossible Foods, and Beyond Meat are
examples of this new generation of plant-based meat alternatives. There is limited to no research
on the consumer acceptance of this new generation of “realistic” plant-based alternatives.
10
We identify and delineate the findings from four studies in the simulated meat literature that are
directly relevant to our work. The following papers will be frequently referenced throughout the
thesis, and we will compare our findings to those presented in these papers in the discussion of
results.
i. Apostolidis and McLeay (2016)
Apostolidis and McLeay (2016) is the closest study to ours in terms of subject and
methodology. They use a DCE to assess consumer preferences for meat and meat substitutes
in the UK using a latent class approach. Based on a set of attributes – fat content, carbon
emissions, type of mince, brand, method of production, region of origin, and price – they
identify six unique classes of consumers with differing preferences. The “type of mince”
included the following alternatives: beef, turkey, lamb, pork, and meat-free, with the last
category being defined with as meat products derived “from a non-animal source (meat
substitute) such as soya, tofu, Quorn etc”. They find low levels of utility associated with meat
substitutes, with only the ‘vegetarian’ class indicating positive preferences towards meat
substitutes; subsequently, they warn stakeholders of the challenges associated with getting
consumers to switch to meat alternatives from meat. They highlight the benefits to
manufacturers of meat substitutes to adopt different strategies for different segments of the
market, proceeding to provide segment-specific strategy recommendations.
Our study differs from their work in three primary capacities: first, we include clean meat in
our analysis and provide respondents with a traditionally vegetarian alternative; second, we
report willingness-to-pay estimates; and third, our work is based in a developing country with
considerably different preferences to the UK.
11
ii. Slade (2018)
To our knowledge, Slade (2018) is the only other study attempting to experimentally estimate
preferences for clean meat alternatives. The study explores consumer preferences towards
conventional, plant-based, and clean beef burgers in Canada.1 Employing a mixed-logit model,
they find that individuals exhibit strong preferences for conventional beef burgers as opposed
to plant-based and clean meat burgers. While they don’t report willingness-to-pay estimates in
their paper, they can be calculated using the coefficients reported in the paper. Using their
results, we extrapolate a WTP of -$3.03 for plant-based burgers and -$3.72 for clean meat
burgers with respect to conventional meat burgers, indicating that consumers are not willing to
pay premiums for simulated meat over the price of conventional meat. Our study differentiates
itself from their work by conducting latent-class market segmentation in addition to reporting
mixed-logit and conditional-logit specifications. Furthermore, our study explores preferences
in the context of a developing country, and also includes a traditional vegetarian alternative in
the DCE. iii. Bryant et al. (2019)
While nearly all the research assessing consumer acceptance of meat alternatives is in the
context of developed countries, Bryant et al. (2019) is the only study that directly explores
current consumer perceptions of plant-based and clean meat alternatives in developing
countries. The study collected primary data via an online survey in USA, China, and India. In
the Indian survey, they find that 10.7% were not at all likely to purchase clean meat; 37.7%
were somewhat or moderately likely; and 48.7% were likely or extremely likely. The respective
1 Since the survey does not specify a location, we presume a location based on the author’s employer, the survey-company’s office location, and the fact that monetary amounts were listed in dollars.
12
statistics for likelihood of purchasing plant-based meat were 5.5%, 31.7%, and 62.8%. Using
linear regression models, they further explore how demographic factors affect likelihood of
purchasing plant-based and clean meat in India. As per their findings, non-vegetarians; those
in higher income brackets; those who are politically liberal; and those who are more aware of
the concept of clean meat are more likely to purchase it. For plant-based meat, they find the
same groups, with the addition of those individuals that are more educated, to have higher
likelihood of purchase.
Our study builds on this project by using more advanced econometric modeling to understand
the potential future meat alternative market in India. Most importantly, we experimentally
estimate willingness-to-pay using an indirect survey mechanism. Our questionnaire was partly
based on the survey instrument used in this study, enabling us to make some direct comparisons
with their results. In direct comparisons, hereon, we will refer to their survey as the “Bryant
Survey”.
iv. Anderson and Bryant (2018)
Anderson and Bryant (2018) attempt to estimate the WTP of clean meat. They estimate WTP
by asking customers what monetary amount they would be willing to pay for clean meat relative
to varying price levels of conventional meat. While they do not provide monetary estimates of
WTP, they conclude that consumers, when prompted that conventional meat is unnatural, are
willing to pay premiums for clean fish, chicken, and beef. However, Breidert, Hahsler, and
Reutterer (2006) find that direct survey techniques such as these have proven to be ineffective
means of eliciting WTP, whereas indirect surveying techniques such as DCEs and conjoint
13
analysis provide consistent and reliable estimates of WTP. For a systematic overview of the various methods used to estimate WTP, please see Breidert et al. (2006).
14
CHAPTER 3
Survey Methodology
3.1. Overview
Primary data was collected via a face-to-face survey in December 2018 in Mumbai, India. The
survey was coded in Qualtrics software and implemented using tablets. The survey was conducted
on the ground by the primary author, alongside a team of three surveyors. Participants were
monetarily incentivized for taking part in the survey. The survey instrument was written in English
and translated to Hindi by the author—the survey tablets displayed the instrument in both English
and Devanagari scripts. Graber et al. (2018) conducted a discrete choice experiment (DCE) in rural
India to estimate the willingness-to-pay for solar microgrids – their published Hindi survey
instrument was used as a guide for designing our Hindi questionnaire. The questionnaire contained
the following sections: an introduction; a section on dietary behavior; three consecutive sections
assessing consumer perceptions towards conventional meat, plant-based meat, and clean meat; a
section containing the choice experiment; and a concluding section on demographic information.
Respondents took an average of 15 minutes to complete the survey in both languages.
The city of Mumbai is divided by a zonal ward system. Six zones containing three to five wards
each make up the urban conglomerate referred to as Greater Mumbai. There are a total of 24 wards
in district of Greater Mumbai. Rather than stratify and select wards using a probability proportional
to size (PPS) method, interviews were conducted in all 24 wards of Mumbai to get a more
representative sample. A total of 424 interviews were conducted across Mumbai City districts’ 24
15
wards (see image), averaging 18 interviews per ward. Figure 1 contains a map of Mumbai by ward,
and a map of the locations in which interviews were conducted.
Figure 1: Ward map of Mumbai (left); Survey interview locations (right).
In each ward, two general categories of locations were used to obtain a diverse sample of
respondents in terms of income and background. The first location in each ward was the area
surrounding the local railway station. In Mumbai, most railway stations are complemented by open-
air vegetable or fish/meat markets (known as sabzi mandis and macchi/meat markets). The second location within each ward was a slightly more affluent establishment like a mall or a supermarket in order to sample individuals that are more likely to be the first buyers of simulated meat (Bryant et al., 2019).
We used a systematic sampling technique in which surveyors randomly selected participants using a pre-determined rule in order to minimize selection bias. The procedure was as follows: the surveyor would first determine a corner or a space that has some kind of privacy to conduct the interview. The surveyor would then turn to a randomization app (Randomizer+) on their tablets and
16
obtain a random start between 1 and 9. Let’s say the random start provided by the app is 7. In the
supermarket/mall setting, the respondent would wait for the 7th adult to walk out of the mall exit;
at the open-air market, they would wait for the 7th person to pass by from their right side to their left side. After the first interview, the surveyor would approach every 3rd person to pass them by to
request their response. If denied, the surveyor would proceed to approach the 3rd person after each
rejection. Other random sampling methods such as random digit dialing and voter-registration based sampling were considered, but ultimately rejected due to feasibility concerns, lack of access to data, and financial restrictions. Further commentary on why a face-to-face survey was employed as opposed to an online survey is provided in the forthcoming sections.
3.2. Pre-Testing
Two rounds of pre-testing were conducted to evaluate the survey instrument. The first round took place in State College, Pennsylvania in USA. A total of 8 individuals were approached and interviewed in the HUB building on The Pennsylvania State University’s main campus. In addition, two experts in survey design and implementation at Penn State were approached to assess the survey transcript at this stage. A handful of revisions were made based on this round of pre-testing: the survey instrument was shortened and some of the questions were restructured to be more succinct.
The Hindi version of the survey instrument could not be pre-tested in the US. Therefore, a second round of pre-testing was conducted in Mumbai after the surveyors’ training - this served as field practice for the surveyors, too. This round of pre-testing recorded 10 interviews outside K-Star
Mall in Chembur, Mumbai. Further revisions were made to the Hindi instrument: several questions were reworded to make them more colloquially relevant to the style of Hindi spoken in Mumbai, and the instrument was further shortened to keep it under 15 minutes in both languages.
17
3.3. Missing Data
Of the 424 participants, 21 were dropped due to incomplete responses. Missing data patterns were
analyzed and some additional observations that were missing across several key demographic
variables were dropped in order to obtain a consistent n for the choice experiment analysis. Income data was determined to be Missing Not at Random (MNAR). ‘Don’t Know’ and ‘Refused’ responses were recoded as missing during estimation. The working n for the analysis is 394.
18
CHAPTER 4
Analytical Methodology
4.1. The Choice Experiment
Discrete choice experiments (DCE) are a popular tool among economists and marketers to
understand consumer behavior and have been used widely in the food marketing literature (e.g.
Sonoda et al. 2018; Mugera, Burton, & Downsborough, 2017; Camarena & Sanjuán. 2008). A DCE
elicits preferences by requiring respondents to state their choice (a “stated preference”) over a set
of alternatives in a hypothetical choice situation (Mangham, Hanson, & McPake, 2009). In order to assess consumer preferences for protein alternatives, this study employs a labeled DCE. In most cases, researchers have the option of using “revealed preferences” to understand consumer behavior; this involves analyzing how consumers have actually behaved in markets, instead of analyzing how they would behave in a hypothetical scenario. However, using a stated preference method is the only option in this case because the market for clean meat does not exist at all, while the market for plant-based meat is rather limited.
The DCE provided respondents with four alternatives (labels): conventional meat, plant-based
meat, clean meat, and chana. As a reminder: chana is the Hindi word for chickpeas, which are a
popular vegetarian source of protein. The purpose of including chana was to test whether
vegetarians would opt for a meat alternative over an existing vegetarian option that they are
accustomed to. There is widespread use of “opt-out” or “no choice” options in DCE literature (e.g.
Slade 2018; Apostolidis & McLeay, 2016; Lancsar & Louviere, 2008; Ryan & Farrar 2000). While
the inclusion of an opt-out option of this sort offers respondents a more realistic choice situation,
19
some recent papers call for a re-evaluation of its effectiveness (Campdell & Erdem, 2019; Veldwijk et al., 2014). Extreme response behavior is a common trend in DCEs with opt-out options, with respondents always or never using a opt-out option (Schlereth & Skiera, 2016; Gensler et al., 2012).
Additionally, opt-out options tend to be riddled with context effects (Schlereth & Skiera, 2016).
For instance, if alternatives are too similar, a respondent might choose the opt-out option so as to not make a decision, known as choice deferral. Veldwikj et al. (2014) find that the opt-out options are selected significantly more often by individuals with lower levels of education, with there being minimal differences in parameter estimates between models with and without the opt-out option.
They further suggest that, since respondents are forced to make a choice, datasets not containing a opt-out option might be of higher quality. Considering that this was a face-to-face survey in a developing country with diversity in literacy levels, we elected to not include a opt-out option.
The DCE is considered labeled because the alternatives have specific names attached to them rather than unlabeled titles with attributes defining all relevant characteristics of the alternative. For example, in this DCE, respondents would see an option in their choice sets as “Conventional Meat” as opposed to “Alternative 1” with “Conventional” described as an attribute pertaining to the type of meat. Jin et al. (2017) explore the differences of labeled and unlabeled naming strategies on statistical estimation. They conclude that there are not statistically significant differences in estimates of coefficient parameters between either strategy, but they do find that a labeled strategy yields slightly higher WTP estimates. Prioritizing approachability and computational ease for participants, we acknowledged this tradeoff and elected to continue with the labeled approach.
This choice experiment is unconventional in the sense that only one attribute - the price of the good
- was varied. Apostolidis and McLeay (2016), on the other hand, assess preferences for meat substitutes across several attributes in their DCE. Veldwijk et al. (2014) point out to the importance of keeping DCEs simple and accessible for participants. Accounting for factors such as the
20
respondents’ expected unfamiliarity with choice experiments; the novel nature of the subject
matter; diverse levels of literacy in the population affecting the survey’s comprehensibility; our
desire to collect information on peripheral subjects of interest; and our interest in keeping the survey
time under 15 minutes, we limited the attributes to price only.
Each respondent received four hypothetical choice occasions; they had to make a choice between
the four alternatives on four different choice occasions, with no two choice occasions having the
same price levels for all alternatives. The order in which respondents saw selection-options was
randomized to reduce primacy or recency bias. Figure 2 contains a sample choice occasion.
Figure 2: Sample choice set.
In order to obtain effective estimates of WTP, a pricing structure must meet a few criteria. First,
we need a range of price levels, since commodity prices in the marketplace fluctuate due to a myriad
of factors. Second, we need the prices to reflect reality to some extent so that they seem plausible
to respondents. Price spreads need not match reality exactly, but they must be believable enough
that they do not lead to protests from respondents. Bearing this in mind, we designed the pricing structure displayed in Table 1. These prices were approximated using current market prices for
21
chana and conventional chicken. The reason chicken was chosen as the type of meat is that it is the
most consumed form of animal-based protein in India.
Table 1: Price attribute levels in INR.
Alternative Low Price ( /kg) Medium Price ( /kg) High Price ( /kg) Chana 90 120 150 ₹ ₹ ₹ Conventional Meat 120 150 180 Plant-Based Meat 150 180 210 Clean Meat 180 210 240
While in the current state of the market clean meat is the most expensive, scale analyses indicate that clean meat has the potential of attaining price parity with conventional meat (Specht, 2019).
Over the past decade, the price of clean meat has decreased substantially and is expected to continue
to do so (Stephens et al., 2018). For example, Israeli startup Future Meat Technologies projects
being able to be price competitive with conventional meat by 2020 (Nelson, 2018). Plant-based meat is a lot more established than clean meat. It is currently priced higher than conventional meat in most markets (Ritchie et al., 2018); however, plant-based alternatives are rapidly gaining price
parity with conventional meat and costs are expected to keep going down as further economies of
scale set in (Grover, 2019). Since we’re exploring the viability of these alternatives to solve societal
problems, it is important to allow the pricing structure to accommodate scenarios where plant-based and clean meat are priced competitively with conventional meat.
As a balance between the current state of the market and what we foresee as a future state, our pricing structure features an overlapping design which prices chana, conventional meat, plant-based meat, and clean meat from least expensive to most expensive. There is a consistent price difference of 30 between any two price levels for all alternatives. However, given the overlapping structure, there₹ exist choice occasions where up to three of the four alternatives are at the same price. This allows for an “even price scenario” where, if conventional meat was at its highest price level, plant- 22
based meat was at its medium price level, and clean meat was at its lowest price level, all three
meat alternatives would cost the same – 180 per kg.
₹ Given the price levels in Table 2, a full factorial design would require total of 34 = 81 choice sets.
Recognizing the impracticality of providing respondents with 81 choice occasions to respond to, a blocked fractional factorial design was applied to obtain 20 choice sets divided into 5 blocks of 4 choice sets each. The choice sets were determined using a D-optimality criterion that maximizes the determinant of the X’X matrix; i.e. the information matrix of all possible price-level combinations. PASS16 software was used to run 50,000 iterations of the full-factorial design, narrowing the full factorial design down to 5 blocks of 4 choice sets (occasions) each. Respondents were randomly assigned one block to respond to using Qualtrics’ built-in randomization tool; each block was answered approximately the same number of times. The choice blocks and levels can be found in Appendix A.
4.2. Model Selection
Heterogeneity in respondent preferences is seen as a desirable feature of discrete choice modeling
since preferences in the real world are not homogenous (Boxall & Adamowicz, 2002; Wedel &
Kamakura, 2002). Accounting for preference heterogeneity is one way of relaxing the
Independence of Irrelevant Alternatives (IIA) assumption required for running conditional logit
models. The IIA assumption states that an individual’s decision between options A and B would be unaffected by the inclusion of an option C, the “irrelevant” alternative. Whether the IIA
assumption holds can be tested using the Hausman specification test. This tests whether a
statistically significant difference exists in the estimation of the conditional logit model with and
without the irrelevant alternative. If the IIA assumption holds, the conditional logit model (CLM)
is a viable model.
23
In our dataset, the Hausman specification test indicates that the IIA assumption does hold, which
validates the use of the CLM. Despite this, we still see value in accounting for preference
heterogeneity between individuals. Considering how little is known about the market for simulated
meat in developing countries, understanding how preferences vary within a population is useful.
The mixed logit model (MLM) and latent class model (LCM) are two recognized methods of
accounting for preference heterogeneity (Greene & Hensher, 2003).
The MLM relaxes the IIA by allowing for a continuous distribution of coefficients across
individuals, whereas the LCM does the same by allowing a discrete distribution of coefficients
across a specified number of latent classes. In other words: the MLM allows observed variables to
vary across individuals within a specified population, while the LCM divides populations into a
discrete number of latent “classes” or “segments”, allowing for preferences to vary between
segments – however, individuals within a segment are assumed to have homogenous preferences.
Shen (2009) conducts a detailed comparison between the MLM and the LCM, using two transport
choice data sets. They find that the LCM performs better in both datasets across a number of
different statistical criteria, supporting the claim in Greene & Hensher (2003) of the LCM’s
superior statistical properties compared to the MLM. Regardless of the statistical benefits, segmentation of the market is seen beneficial from a perspective of government policy or business strategy. For these reasons, our model of choice is the LCM. The results from the MLM and the
CLM are reported as robustness checks.
24
4.3. Methodology
DCEs rest on the random utility theory (RUT), which proposes a probability-oriented approach to
understanding consumers’ decision-making in discrete choice situations (McFadden, 2001). The
individual’s true utility, , consists of two components, the observable (deterministic)
𝑖𝑖𝑖𝑖𝑖𝑖 component, , and the 𝑈𝑈unobservable component , assumed to have a Type 1 extreme
𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖 distribution. 𝑉𝑉 𝜀𝜀
= + (1)
𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑈𝑈 𝑉𝑉 𝜀𝜀 ∀ 𝐽𝐽 ∈ 𝐶𝐶𝐶𝐶
This generalized setup produces the utility observed by consumer for alternative on choice occasion , from choice set . Assuming rational behavior, individual𝑖𝑖 will select such𝑗𝑗 that their utility is maximized.𝑡𝑡 𝐶𝐶𝐶𝐶 𝑖𝑖 𝑗𝑗
Latent-class models (LCM) are used to capture heterogeneity across respondents by segmenting the sample into discrete classes. Such models are used when we suspect that there exists heterogeneity within the respondents’ preferences towards the dependent variable, but there is no
single, obvious element establishing the dissection of the respondents on the basis of this perceived
heterogeneity (Greene & Hensher, 2003). A consumer’s inclination to purchase meat alternatives
is a complex construct that could depend on a variety of factors. The LCM allows us to understand
how these factors interact towards the outcome of the dependent variable – the purchase decision
between four protein alternatives – and accordingly segments the sample into a specified number
of classes.2
2 Hong Il Yoo’s lclogit2 program in STATA 15.1 was used to estimate the latent class model. Arne Risa Hole’s wtp command was used to obtain willingness-to-pay estimates.
25
The number of “classes” or “groups” to divide respondents into is an important decision in latent
class analyses. Nylund, Asparouhov, and Muthen (2007) assess the effectiveness of several fit criterion for determining the number of classes in the model. As per their findings, the Bayesian
Information Criterion (BIC) can be used to determine the number of classes that provide the best model fit. Kim (2014) finds the Consistent Akaike Information Criterion (CAIC) to be a reliable measure as well. Table 2 shows the results for the BIC and CAIC obtained from the latent class model. For reference, we report the AIC as well.
Table 2: BIC, CAIC, and AIC results with different numbers of classes. 3
Number of Classes BIC CAIC AIC 2 3801.85 3810.85 3766.06 3 3546.45 3560.45 3490.78 4 3166.04 3185.04 3090.49 5 3176.35 3200.35 3080.92 6 3185.243 3214.24 3069.93 7 3187.14 3221.14 3051.94 8 3187.20 3226.20 3032.12
= 2 log L + p log(n), = 2 log L + p(log(n) + 1), = 2 log L + 2p
𝐁𝐁𝐁𝐁𝐁𝐁 − 𝐂𝐂𝐂𝐂𝐂𝐂𝐂𝐂 − 𝐀𝐀𝐀𝐀𝐀𝐀 − where L is the likelihood, p is the number of parameters to be estimated, and n is the sample size.
As we can see in the formulation, the AIC only penalizes the model for its number of parameters, while the BIC and the CAIC include an additional penalty for sample size. The BIC and CAIC are minimized at 4 classes, while the AIC continues to decrease. We use C = 4 classes in our analysis.
3 The listed BIC, CAIC, and AIC were estimated without class-membership variables. The model with class-membership variables did not converge after 7 classes, but we still observed BIC and CAIC minimizing at 4 classes. 26
Establishing the number of segments, the utility relationship in (1) for can be reinterpreted for an
LCM as:
| = | + | (2)
𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 𝑈𝑈 𝑉𝑉 𝜀𝜀
where | = + |
𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 𝑐𝑐 𝑖𝑖𝑖𝑖𝑖𝑖 𝑗𝑗 𝑐𝑐 𝑖𝑖𝑖𝑖 𝑉𝑉 𝛽𝛽′ 𝑋𝑋 γ ∀ 𝐽𝐽 ∈ 𝐶𝐶𝐶𝐶
The deterministic (observed) portion | is defined as follows. is a vector of parameters
𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 c specific to class , is a vector of attributes𝑉𝑉 of choice on choice𝛽𝛽 occasion for individual ,
𝑖𝑖𝑖𝑖𝑖𝑖 | is the alternative𝑐𝑐 𝑋𝑋-specific constant (ASC) for alternative𝑗𝑗 specific to class 𝑡𝑡, and is the𝑖𝑖
𝛾𝛾𝑗𝑗 𝑐𝑐 𝑗𝑗 𝑐𝑐 𝐶𝐶𝐶𝐶 𝑖𝑖𝑖𝑖 choice set faced by individual on choice occasion , and. The error term | is distributed with
𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 a Type I extreme value (IID). T𝑖𝑖herefore, the overall 𝑡𝑡utility function can be𝜀𝜀 specified as:
| = + | + | + | + (3)
𝑈𝑈𝑖𝑖𝑖𝑖𝑖𝑖 𝑐𝑐 𝛽𝛽𝑐𝑐𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖𝑖𝑖 𝛾𝛾1 𝑐𝑐𝐶𝐶ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑖𝑖𝑖𝑖 𝛾𝛾2 𝑐𝑐𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 𝛾𝛾3 𝑐𝑐𝑃𝑃𝑃𝑃𝑖𝑖𝑖𝑖 | + |
4 𝑐𝑐 𝑖𝑖𝑖𝑖 𝑖𝑖𝑖𝑖 𝑐𝑐 𝛾𝛾 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶 𝜀𝜀
In this specification, is the class-specific vector of price coefficients, and - are the
coefficients on the dummy𝛽𝛽𝑐𝑐 variables that for chana, conventional meat, plant-based meat,γ1 γ 4and clean meat. The purpose of these dummy variables is to identify an alternative. The coefficients on these identifiers are called the above-mentioned alternative-specific constants (ASCs). The ASCs capture the average effect of all the factors that were not included in the model on utility (Train, 2009).
When ASCs are included, the expected mean of the error term is zero. Since we are concerned with differences in utility between choice scenarios and not absolute utility, we are consequently
27
concerned with differences in ASCs and not their absolute values. Considering that these are
constants, there exist an infinite number of combinations of constants that will yield the exact same difference. To prevent this from happening, we normalize the coefficient for one of the alternatives, chana, to zero. Therefore, only coefficients for conventional meat, plant-based meat, and clean meat are reported. The magnitudes of these coefficients should be interpreted as the average effect of omitted factors on an alternative relative to chana, the normalized base alternative.
The conditional probability of individual in class choosing alternative from a set of = 4 alternatives across = 4 choice occasions𝑖𝑖 is represented𝑐𝑐 by: 𝑘𝑘 𝐽𝐽
𝑇𝑇
( ) (4) | = 𝑇𝑇 ( ) 𝑒𝑒𝑒𝑒𝑒𝑒 𝛽𝛽′𝑐𝑐𝑋𝑋𝑖𝑖𝑖𝑖𝑖𝑖 𝑃𝑃𝑖𝑖𝑖𝑖 𝑐𝑐 � 𝐽𝐽 𝑡𝑡=1 𝑗𝑗=1 𝑐𝑐 𝑖𝑖𝑖𝑖𝑖𝑖 ∑ 𝑒𝑒𝑒𝑒𝑒𝑒 𝛽𝛽′ 𝑋𝑋
The class share - also known as the prior probability of being in a particular class - is driven by the class-membership model and is calculated by:
( ) (5) = 1 + ′ ( ) 𝑒𝑒𝑒𝑒𝑒𝑒 𝜃𝜃𝑐𝑐𝑍𝑍𝑖𝑖 𝑐𝑐 𝐶𝐶−1 ′ 𝜋𝜋 𝑐𝑐 𝑖𝑖 ∑𝑐𝑐=1 𝑒𝑒𝑒𝑒𝑒𝑒 𝜃𝜃 𝑍𝑍
where is a vector of socio-economic indicators that help us understand the composition of the
𝑖𝑖 classes,𝑍𝑍 and is a vector of class-specific coefficient parameters for the membership variables in
𝑐𝑐 , with the last𝜃𝜃 class normalized to zero as the reference segment. This is a fractional multinomial
𝑍𝑍logit formulation where the coefficients tell us whether the given membership variable increases
θ
28
of decreases the probability of an individual belonging to a segment; is normalized to zero to allow for a reference segment. θ𝐶𝐶
The unconditional choice probability, i.e. the probability of an individual choosing an alternative regardless of class membership can be given by:
(6) = 𝐶𝐶 𝑇𝑇 |
𝑃𝑃𝑖𝑖𝑖𝑖 � 𝜋𝜋𝑐𝑐 � 𝑃𝑃𝑖𝑖𝑖𝑖 𝑐𝑐 𝑐𝑐=1 𝑡𝑡=1
The posterior probability of class membership is the probability of an individual being in a
particular class based on their revealed choice patterns across different choice occasions and can
be calculated using the following formula:
(7) | 𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑇𝑇 𝐽𝐽 𝑦𝑦 𝑐𝑐 𝑡𝑡=1 𝑗𝑗=1 𝑃𝑃𝑖𝑖𝑖𝑖 𝑐𝑐 | 𝜋𝜋 ∏ ∏ � � 𝑖𝑖𝑖𝑖𝑖𝑖 𝐺𝐺𝑛𝑛𝑛𝑛 𝐶𝐶 𝑇𝑇 𝐽𝐽 𝑦𝑦 𝑐𝑐=1 𝑐𝑐 𝑡𝑡=1 𝑗𝑗=1 𝑃𝑃𝑖𝑖𝑖𝑖 𝑐𝑐 ∑ 𝜋𝜋 ∏ ∏ � �
Willingness-to-Pay (WTP) is defined as the price at which the consumer is indifferent between
purchasing and not purchasing a product and can be estimated using stated preferences. We
calculate WTP using the following formula:
, , (8) , = = 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑒𝑒𝑘𝑘 𝑐𝑐 𝛽𝛽𝑘𝑘 𝑐𝑐 𝑊𝑊𝑊𝑊𝑃𝑃𝑘𝑘 𝑐𝑐 − − 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑒𝑒𝑐𝑐 𝛾𝛾𝑐𝑐
29
Equation (8) produces class-specific estimates of WTP. To calculate the WTP for the entire sample, i.e. the mean WTP across classes, we can use the formula below.
(9) , , = 𝐶𝐶 = 𝐶𝐶 𝑘𝑘 𝑐𝑐 𝑘𝑘 𝑐𝑐 𝑘𝑘 𝑐𝑐 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑒𝑒 𝑐𝑐 𝛽𝛽 𝑀𝑀𝑀𝑀𝑀𝑀𝑀𝑀 𝑊𝑊𝑊𝑊𝑃𝑃 � 𝜋𝜋 �− 𝑐𝑐 � � 𝜋𝜋 �− 𝑐𝑐 � 𝑐𝑐=1 𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑒𝑒 𝑐𝑐=1 𝛾𝛾
30
CHAPTER 5
Results
5.1. Descriptive Statistics
The survey collected demographic information on religion, gender, household income, and
education. Respondents were also asked if they identified as vegetarians or non-vegetarians. The descriptive statistics from the survey, as well as relevant statistics from the 2011 Census of India are reported in Table 3.
Gender is reported as binary, but respondents were offered an option to specify non-binary gender identity; no respondents selected this option. The 2011 Census of India reports a sex distribution of 51.5% males, 48.4% females, and 0.1% transgender. In comparison, the gender distribution in our sample is highly skewed towards males. The distribution of gender in Mumbai is 42.6% female to 57.4% male as per the 2011 Census. Mumbai is a melting pot of sorts, with workers from all over the country moving to the city to find better-paying opportunities (AFP, 2011). Often times, these workers tend to be male members of the family, who send money back to their families living in their rural hometowns to avoid the financial costs of sustaining a family in an urban environment.
However, despite Mumbai having a higher percentage of males compared to the overall population, our gender distribution remains significantly skewed towards males. Considering the random, systematic nature of our sampling procedure, we suspect the skew is due to female respondents being more cautious of engaging in conversations with strangers. The unwillingness of women to partake in face-to-face surveys in India has been experienced by other researchers (Graber et al.,
2018). It could be argued that a potential benefit of this skew in the context of our research is that
31
more men than women consume meat in India; approximately 22% of men are vegetarian, whereas
30% women identify as vegetarian as per the fourth round of the National Family Health Survey
(IIPS, 2017).
Table 3: Sample descriptive statistics and relevant 2011 Census statistics.4
Variable Categories Sample (%) 2011 Census (%) Religion Hinduism 65.74 79.80 Islam 10.66 14.23 Christianity 3.3 2.30 Sikhism 3.3 1.72 Buddhism 3.3 <1 Jainism 4.82 <1 Not Religious 3.55 <1 Other Religion 2.03 <1 Gender Male 72.34 51.5 Female 26.9 48.4 Income < ₹ 5,000 1.15 ₹ 5,001 - ₹ 7,500 2.88 ₹ 7,501 - ₹ 10,000 6.63 ₹ 10,001 - ₹ 20,000 23.63 ₹ 20,001 - ₹ 50,000 36.31 > ₹ 50,000 29.39 Education Less than 10th 7.11 10th Pass 11.93 12th Pass 20.56 Technical Qualification 2.03 Bachelor's Degree or Higher 57.61 Diet Vegetarian 35.28 Non - Vegetarian 64.72
4 The table reports statistics from the 2011 Census of India for those variables that could directly be compared. Income and literacy levels are measured differently in the Census and are therefore not reported. The extent of vegetarianism is not measured in the Census, but compared to the NFHS-4, our proportions are very close to the expected number of vegetarians in the population. 32
A comparison of the spread of household income between the Bryant Survey (conducted online) and our survey can be found in Figure 3. Online surveys in developing countries tend to attract
individuals in higher income brackets. While it may be argued that these higher income individuals
will be the first consumers of meat alternatives, it is important to keep in mind that this paper argues
that these newer technologies should be viewed as tools to correct societal problems; in order to
occupy this role, meat alternatives must . Considering the potential for these technologies to be
competitively priced with conventional meat (Specht, 2019), we see benefit in not restricting the
sample to higher income brackets. As is evident, our face-to-face survey was able to attain a more
diverse, representative sample in terms of income. This income distribution is also closer to the national distribution of income in India, displayed in Figure 4.
> ₹ 50,000
₹ 20,001 - ₹ 50,000
₹ 10,001 - ₹ 20,000
₹ 7,501 - ₹ 10,000
₹ 5,001 - ₹ 7,500
< ₹ 5,000
0 10 20 30 40 50 60
Bryant Survey Our Survey
Figure 3: Income distribution comparison between the online Bryant Survey and our face-to-face survey
33
> ₹ 33,001
₹ 16,601 - ₹ 33,000
₹ 8,301 - ₹ 16,600
₹ 6,001 - ₹ 8,300
< ₹ 6,000
0% 10% 20% 30% 40% 50%
Figure 4: Distribution of Household Income in India (Source: Patnaik (2019) and CMEI Consumer Pyramids Database)
Between the 2001 and 2011 Census of India, the number of native English speakers increased by
15% from 226,449 to 260,416. The number of Hindi speakers, however, increased by 25% from
422,048,642 to 528,347,193 over the same period. The 2011 Census reports the number of second- and third-language speakers of English and Hindi as well. Including these secondary and tertiary speakers, about 10.6% of the population speaks English and 57.1% of the population speaks Hindi.
Additionally, as our data corroborates, the ability to speak in English in developing countries is often correlated with higher levels of education and higher incomes. In order to keep our survey inclusive and demographically diverse, our survey presented respondents the option of responding in Hindi or English. Approximately 61% of the participants chose to take the survey in Hindi. This is not an indication of the number of people that speak Hindi in Mumbai, but a representation of the number of people in our sample that were more comfortable taking the survey in Hindi than
English. We see this as an advantage over the Bryant Survey, which was conducted in English only.
Future surveys in developing countries should include local language options to improve accessibility.
34
The mean age of respondents in our sample was 34, with the youngest respondent being 18 and the
oldest respondent being 78. The average age of vegetarians in our sample was 36, with a minimum
of 18 and a maximum of 76, indicating that we saw vegetarians across the spectrum. This is
indicative of the age-diverse nature of vegetarianism in India, which is expected, considering its
cultural and historical significance in the region. Figure 5 contains a representation of the age
distribution of the full sample, as well as the age distribution of vegetarians in the sample.
20
15
10 Frequency 5
0 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 59 61 63 65 70 72 78 Age in Years
Full Sample Vegetarians in Sample
Figure 5: Age distribution of full sample and age distribution of vegetarians in sample.
Of the 394 respondents included in the sample, 35% perceived themselves to be vegetarian. This is
close to the figures estimated by several surveys looking at vegetarianism in Indian, with most
estimating somewhere between 31%-45% of the population identifying as vegetarian (IIPS, 2017).
Studies in the United States have observed that self-identified vegetarians tend to consume some form of meat occasionally (Juan, Yamini, & Britten, 2015). For instance, using data from the 2009-
2010 NHANES, Juan, Yamini, and Britten (2015) find that 48% of self-perceived vegetarians in the US ate some form of red meat, poultry, or seafood. To assess the prevalence of this pattern in
35
our data, we asked respondents to select the types of foods they had consumed within the last 30 days from a similar food list. Of the self-identified vegetarians in our sample, only 3.5% of respondents indicated consuming poultry, fish, mutton, pork, or beef at least occasionally. Within this set of self-perceived vegetarians that ate some type of animal meat, 57% ate fish. This is indicative of a stricter adherence to vegetarianism in India compared to the United States, which can be attributed to the historic relevance of vegetarianism in South Asia.
Self-identifying non-vegetarian respondents were asked to report their weekly frequency of eating meat. Figure 6 indicates that most non-vegetarians in our sample eat meat 1-2 days a week. This is reflective of the cultural perceptions of meat in India. For example, several groups of Hindus choose to not eat meat on certain days of the week (Tuesdays are dedicated to Lord Hanuman,
Wednesdays to Lord Ganesha, etc.). However, despite the infrequency with which Indians consume meat relative to Western populations, the aforementioned environmental, nutritional, and food security concerns persist—these problems are attached to future expectations of more frequent meat consumption and in larger quantities due to rising incomes.
60
50
40 Vegetarians - 30
20 % of Non of %
10
0 6-7 days 3-5 days 1-2 days Less than once a week
Figure 6: Weekly frequency of animal-based meat consumption.
36
5.2. The Latent Class Model
The latent class model (LCM) identified four distinct latent classes of consumers in our sample.
Table 4 show the results of the latent class logit regression, as well as the results of the class- membership model. Since the LCM is an extension of a multinomial logit, we need a reference alternative to compare our results to; therefore, we estimate K-1 sets of coefficients, where K is the number of alternatives. The coefficients for chana are normalized to zero, making it the reference
alternative (see discussion of ASCs in Chapter 4). Based on results from the BIC and CAIC, we
will divide our sample into C = 4 latent classes (see Table 2). The reported values are the
coefficient, the robust standard error clustered by respondent ID, and the class shares. Statistical
significance levels of 1%, 5%, 10%, and 25% are indicated, and the phrases “highly significant”,
“significant”, “moderately significant”, and “weakly significant” will be used to describe these,
respectively.
The reported coefficients stem from an extension of a logit regression, so the magnitude of the
coefficient has little direct interpretive value (Greene & Hensher, 2003). However, the sign of the
coefficient is very informative: it represents whether a given class has positive or negative
preferences for a particular attribute. All four classes indicate significant or highly significant
coefficients for price, validating the pricing structure and choice set design. Class shares, also
referred to as the prior probabilities of class membership, were relatively even and are listed in the
top row of Table 4. A class share of 19.6% for Class 4, for instance, indicates that the prior
probability of a given individual being in that Class 4 is 0.196.
Class 1 reports negative coefficients for conventional, plant-based, and clean meat that are highly
significant. This indicates a strong, positive preference for chana relative to these alternatives.
Comparing the absolute magnitude between alternatives, the coefficient for conventional meat is the largest and most significant, indicating that this class has the strongest negative preference for
37
conventional meat compared to chana. This suggests that this class is likely made up of vegetarians,
who, as a group, would be expected to display strong negative preferences towards conventional
meat. Class 2 exhibits opposite preferences to Class 1. We observe highly significant, positive coefficients for all three meat alternatives with respect to chana, with the coefficient for conventional meat being the largest in magnitude relatively and most statistically significant. This indicates that this class has a strong preference for meat products, particularly conventional meat, compared to chana, suggesting that this class comprises of non-vegetarians.
Classes 3 and 4 show less straightforward patterns of preferences. Class 3 has a positive, highly significant coefficient for plant-based meat, and a positive, significant coefficient for clean meat.
The class indicates a positive but weakly significant positive preference for conventional meat. In comparing absolute magnitudes, the coefficient for plant-based meat is the strongest. This, along with the high statistical significance of the coefficient, suggests that this class has a strong preference for plant-based meat. Class 4 exhibits, to some extent, the opposite story. The members of this class indicated highly significant and strong positive preferences for clean meat, and comparatively weaker but still significant positive preferences for plant-based meat. Similar to
Class 3, this class had a positive but weakly significant coefficient for conventional meat. These choice outcomes indicate that individuals in Class 4 have a strong preference for clean meat.
Classes 3 and 4 collectively make up 51.6% of our sample, indicating that over half of our sample exhibited strong, positive preferences towards simulated meat compared to chana.
38
Table 4: Latent class model and class-membership model results.
Class 1 (21.0%) Class 2 (27.5%) Class 3 (32%) Class 4 (19.6%) Attributes Coef. Robust SE Coef. Robust SE Coef. Robust SE Coef. Robust SE Price -0.010** 0.005 -0.019*** 0.005 -0.011** 0.004 -0.010** 0.004 Conventional Meat -3.518*** 0.514 3.609*** 0.485 0.965** 0.478 0.673^ 0.443 Plant-Based Meat -1.295*** 0.358 2.338*** 0.529 3.732*** 0.513 2.022** 0.806 Clean Meat -3.166*** 0.787 2.283*** 0.641 1.149^ 0.974 3.976*** 0.685
Class Membership Variables Vegetarian 2.466** 0.455 -1.396** 0.548 0.424 0.417 Environmentally Conscious -0.331 0.390 -0.476^ 0.362 -0.353 0.358 College Educated -0.014 0.428 0.685* 0.378 0.584^ 0.384 Hindu -0.258 0.435 0.349 0.438 0.626^ 0.462 Muslim 0.129 0.957 1.378** 0.665 0.938^ 0.706 Age 0.014 0.019 0.002 0.017 -0.005 0.020 Female -0.245 0.412 -0.535^ 0.380 -0.548^ 0.388 Monthly Income > ₹ 50,000 0.394 0.524 0.446 0.509 0.716^ 0.502 Monthly Income between ₹ 20,000 and ₹ 50,000 -0.570^ 0.454 -0.025 0.399 0.176 0.398 Conservative 0.001 0.425 0.037 0.409 0.467^ 0.396 Liberal -0.548 0.697 0.167 0.545 0.307 0.556
*** significant at 1%, described as “highly significant” ** significant at 5%, described as “significant” * significant at 10%, described as “moderately significant” ^ significant at 25%, described as “weakly significant”
39
5.3. Willingness-to-Pay
We can make these conclusions more tangible by estimating consumers’ willingness-to-pay (WTP)
for conventional, plant-based, and clean meat with respect to chana. The WTP estimates are
reported in Table 5, along with 95% confidence intervals (CIs). The CIs were calculated using the
Krinsky-Robb method, which is also known as the parametric bootstrap method for calculating CIs for WTP. This method relies on taking a large number of draws from a multivariate distribution to calculate percentiles of the specified model based on the specified level of confidence (Hole, 2007).
Table 5: Willingness-to-Pay estimates and confidence intervals with chana as the reference category.
Class Conventional Meat Plant-Based Meat Clean Meat 1 -352.49 -129.79 -317.18 [-2074.72, -111.23] [-867.85, -16.37] [-2011.56, -60.20] 2 185.29 120.01 117.18 [140.14, 318.86] [85.41, 188.91] [80.57, 166.12] 3 91.14 352.34 108.44 [-1.90, 337.70] [215.91, 1290.39] [-149.55, 303.84] 4 65.29 196.08 385.48 [-37.46, 254.80] [85.05, 463.30] [250.85, 1339.42] Mean WTP 18.89 156.93 75.87
The WTP estimates take chana as the base alternative, so the magnitudes should be interpreted with
respect to chana. These estimates reflect the preferences of the classes established from the
regression coefficients. It is also important to keep in mind that the following interpretations are
based on our pricing structures—percentages may be calculated to extend these results to other prices.
Given our pricing structure, individuals in Class 2 are willing to pay a premium of 185.29 for conventional meat relative to the price of chana. The same class is willing to pay a premium₹ of
40
120.01 for plant-based meat with respect to chana. Alternatively, we could see this as a WTP of
₹120.01 - 185.29 = -65.28 for plant-based meat relative to conventional meat for the individuals
in₹ Class 2.₹ This is to₹ say that, while members of this class have a positive WTP for plant-based meat with respect to chana, they don’t see it as a worthy alternative to conventional meat and would not pay a premium on top of the price of conventional meat for an equivalent quantity of the product. Arguably, this is a more useful interpretation of these estimates since plant-based and clean meat will compete more directly with conventional meat than with chana. Based on this re-
interpretation, we report WTP estimates for plant-based meat, clean meat, and chana with respect
to conventional meat in Table 6. Values are reported in Indian rupees (INR) and US dollars (USD).
Table 6: WTP estimates with respect to conventional meat in INR and USD.5
Class Plant-Based Meat Clean Meat Chana (₹/kg) ($/kg) (₹/kg) ($/kg) (₹/kg) ($/kg) 1 222.70 3.18 35.30 0.50 352.49 5.04 2 -65.28 -0.93 -68.11 -0.97 -185.29 -2.65 3 261.20 3.73 17.30 0.25 -91.14 -1.30 4 130.79 1.87 320.19 4.57 -65.29 -0.93 Mean WTP 138.04 1.97 56.98 0.81 -18.89 -0.27
This reinterpretation provides useful insight. Even though Class 1 has negative preferences for
plant-based and clean meat relative to chana, they are willing to pay a significant premium of 223
for plant-based meat and a premium of 35 for clean meat over the price of conventional meat.₹
This affirms a positive attitude towards₹ plant-based and conventional meat compared to conventional meat; however, individuals in Class 1 still need plant-based meat to cost 130 per kg
₹
5 Using a conversion rate of 1 USD = 70 INR. 41
less than chana, and clean meat to cost 317 per kg less than chana to substitute towards them. At
the very least, even if we look at the upper₹ bounds of confidence intervals, they need plant-based
meat to cost 16 per kg less than chana, and clean meat to cost 60 per kg less than chana. This
confirms our ₹earlier hypothesis that Class 1 has strong, negative preferences₹ for all types of meat.
These individuals are content with chana as a source of protein and are least likely to substitute
away from it unless simulated meat products are priced significantly lower than chana. This further strengthens our expectation that this class comprises of vegetarians. Plant-based meat fares substantially better than clean meat with this class, indicating that they are far more likely to be allured by plant-based alternatives as opposed to cell-based or animal-based meats.
The results in Table 6 provide an important insight into Class 2 as well. While this class associated positive preferences for all types of meat compared to chana, they exhibit negative preferences when conventional meat is the base alternative. These participants do not associate premiums with simulated meat products relative to conventional meat. Relative to the price of conventional meat, plant-based meat will have to cost 65 and clean meat 68 less per kg to induce substitution. This reinforces our expectation of non-₹vegetarians being in₹ this class. In the most positive scenario
(taking the difference between the lower bound of conventional meat and the upper bounds of plant based and clean meat), some individuals in the class would associate a premium of 49 per kg for plant-based meat and 26 per kg for clean meat over conventional meat. This is₹ the absolute maximum an individual₹ in this class might be willing to pay for a simulated meat alternative.
Classes 3 and 4 exhibit preferences that are more in agreement with the current state of the market, which prices simulated meat products over conventional meat and chana. They associate premiums with plant-based, clean, and conventional meats with respect to chana, with Class 3 associating higher value with plant-based meat and Class 4 with clean meat. The mean WTP is useful to understand the preferences of the entire sample. Weighting by class shares, the sample – as a whole
42
– associates premiums with plant-based and clean meat over conventional meat, and also associates a premium with conventional meat over chana. The sample is collectively willing to pay the highest premium for plant-based meat.
5.4. Class-Membership Model
The class-membership model allows us to take a deeper look into the make-up of these classes. The results are presented in the lower half of Table 4. All parameters – except age – were coded as binary dummy variables. This model takes Class 4 as the base class; Class 4 is normalized to zero during estimation, so the parameter estimates for the other classes should be interpreted relative to
Class 4. The parameters were estimated via a fractional multinomial logit model; therefore, the coefficients indicate the log-odds that an individual is in a given class relative to Class 4.
Respondents had six tiers of income to choose from. As of March 2019, the average per-capita income in Mumbai is approximately 11,900 per month or $2,042 per annum (CEIC Data, 2019).
Bearing this in mind, we determined ₹those earning more than 50,000 per month as high income.
Individuals earning in the second highest tier, between 20,001₹ and 50,000 were seen as the middle-income group. Anyone earning less than 20,001₹ was in the₹ low-income, reference category. Environmental consciousness was assessed₹ by asking respondents whether they favored
New Delhi Chief Minister, Arvind Kejriwal’s odd-even rule aimed at curbing air pollution in New
Delhi, a policy that was divisive in terms of public opinion when implemented in 2017. Those who favored the policy were coded as environmentally conscious (1) and those who did not fell in the reference category (0). Political conservatism and liberalism were assessed by asking respondents which political party they voted for in the 2014 Indian general election. The reason respondents weren’t directly asked whether they identified as “conservative” or “liberal” is that these words are largely not used to describe ideology within the Indian political discourse, and consequently, there
43
do not exist sensible counter-parts for these words in Hindi that would be understood by the lay
person. Therefore, those who voted for the historically right-wing Bharatiya Janta Party (BJP) in
the previous election were labeled conservative, and those who voted for the historically
center/center-left Indian National Congress (INC) were labeled as liberal.
Class 1 has a positive, significant coefficient for vegetarianism. This should be interpreted as a
higher likelihood of an individual being vegetarian in Class 1 compared to Class 4. This supports
our initial hypothesis that this class is made up of vegetarians that have strong negative preferences
towards conventional meat. The class also reports a negative, significant coefficient for the middle-
income category indicating that those earning a monthly income between 20,000 and 50,000 are
less likely to belong to this class. Although this parameter is only weakly significan₹ t, it still₹ provides
valuable insight regarding the demographic composition of this segment. This raises an important
point for the forthcoming discussion: even though several of the estimates aren’t statistically
significant, their signs can still be interpreted to understand the make-up of the identified latent
segments (Liao, 1994). This is a pilot study, and any inferences that can be made about the market
are valuable.
Considering our expectation of Class 1 comprising of vegetarians, the negative coefficient for
Hindu in Class 1 might be surprising to some, but it isn’t all that shocking if we consider the distribution of vegetarianism within religious identities (see Figure 7). In our sample, 94% of individuals who identified with Jainism also self-identified as vegetarian. This same statistic was
70% for Sikhs. However, only 37.8 % of Hindus identified as vegetarian. Jainism and Sikhism, both majority vegetarian religions, fall in the reference category of our Hindu and Muslim dummy
44
variables. This offers a possible explanation for the negative coefficient for Hindu in the vegetarian class.6
100.00
80.00
60.00
40.00
20.00
0.00
Figure 7: Vegetarianism by religion.
Class 2 exhibits a negative, significant coefficient for vegetarianism. This, too, bolsters our initial hypothesis based on alternative-specific preferences that this class is comprises of non-vegetarians.
A positive, significant coefficient is also reported for Muslim. This is understandable since most
Muslims tend to consume meat; in our sample, 93% of Muslim-identifying participants also identified as non-vegetarians.
Being college-educated was a positive and moderately significant indicator of membership in this
Class 2, and positive and weakly significant for Class 3 relative to Class 4. This can be interpreted as college-educated individuals having the strongest preferences for conventional meat, followed by plant-based meat, clean meat, and chana, in that order. All three classes exhibit negative, albeit
6 To further understand how vegetarianism relates to the other demographic variables, we logistically regressed vegetarianism on all the other membership variables. The odds ratio for being Hindu is actually insignificant and <1, indicating that being Hindu is not an indicator of vegetarianism. Being Muslim, however, is consistently a significant indicator of not identifying as vegetarian. 45
mostly insignificant coefficients for the environmental consciousness dummy variable. This can be
inferred as Class 4 containing respondents that most likely to comprise of individuals that are in
favor of environmental policies. This supports our earlier hypothesis that this class contains
relatively more progressive-thinking individuals that are excited about food innovations that have
environmental benefits associated with them. Furthermore, all classes reported negative coefficients for the female regressors, indicating that Class 4 is more likely to contain female- identifying individuals relative to the other classes. This is important since, in most Indian households, women tend to be the member of the household responsible for buying groceries and their receptiveness towards food innovation is a good sign for the clean meat manufacturers. The lack of significance in the female dummy suggests that the skew towards males in our sample is not as problematic as it could have been.
The middle-income category reports negative coefficients for Classes 1 & 2. This implies that those individuals earning between 20,000 and 50,000 per month are less likely to be in Classes 1 & 2
relative to Class 4. These individuals₹ are more₹ likely to be in Class 3 relative to Class 4. In terms
of product-preference, this has an important inference: middle-income earners are more likely to
be in Classed 3 and 4, indicating that middle-income earners are more likely to exhibit negative
preferences towards conventional meat and chana than high- or low-income earners. The coefficient of membership in Class 3 is positive for high- and middle-income categories, but it is only (weakly) significant for the high-income category. This suggests that that those individuals earning more than 50,000 per month are more likely to exhibit positive preferences for plant-based meat compared to ₹clean meat. Middle-income earners still prefer plant-based meat to clean meat, but the preference isn’t as strong as that observed for high-income earners.
The only continuous dependent variable in our class-membership model is age. Classes 1 & 2 report positive coefficients for age, while Class 3 exhibits a negative relationship. Albeit insignificant,
46
this supports findings from other studies concluding younger individuals are more receptive to
change. Those that are more set in their ways are less likely to be open to newer alternatives. As per this, we can deduce that younger individuals are more likely to exhibit positive preferences towards simulated meat, with stronger preferences towards plant-based meat than clean meat.
All three classes reported positive coefficients for the conservative dummy, leading us to suspect that Class 4 is more likely to contain individuals with non-conservative political stances. However, the coefficients for the liberal dummy vary from being negative for Class 1 to being positive for
Classes 2 and 3. This could be interpreted as liberals being less likely to positive preferences for
clean meat over chana, but negative preferences for clean meat over conventional and plant-based meat.
As a socio-economic indicator, vegetarianism was consistently a strong and significant indicator of membership in Classes 1 and 2. Additionally, members of Class 1 exhibited positive, highly significant preferences for chana compared to the other alternatives, while those in Class 2 exhibited positive, highly significant preferences for conventional meat relative to the other alternatives. Classes 3 and 4 did not have a socio-economic indicator that was as rigid, but they consistently expressed positive, highly significant preferences for one of the two simulated meat alternatives.
Based on the above discussion, we are naming the classes: the veggie lovers, the meat lovers, the plant-based meat enthusiasts, and the clean meat enthusiasts. While it may seem appealing and intuitive to refer to the veggie lovers as “the vegetarians” and the meat lovers as “the non-
vegetarians”, such a classification would be marginally misleading because it isn’t necessary that
everyone in Class 1 is a vegetarian or everyone in Class 2 is a non-vegetarian, but it is highly likely
that vegetarians will be in Class 1 and non-vegetarians will be in Class 2.
47
5.5. Even Price Scenario
Due to the random block design, 84 of the 394 respondents received the even-price scenario
described in the analytical methodology. In this choice occasion, the price of plant-based meat,
clean meat, and conventional meat was the same at 180 per kg, and chana cost 90 per kg. Of the
84 respondents, 42% said they would purchase plant₹ -based meat, 16.6% would₹ purchase clean
meat, and 15.5% would purchase conventional meat. This is indicative of the importance of
competitive pricing in the simulated meat space. Our expectation is that the remaining 25% who
selected chana were in the vegetarian latent class and highly price sensitive.
5.6. Model Fit
We do not compute a regular goodness-of-fit (G2) statistic for our likelihood functions due to sparseness in response patterns such as those who have positive preferences for chana also having high positive preferences for conventional meat. In terms of latent class separation, we observed four classes with distinctly different preferences, which indicates that the number of classes was sufficient and accurate. To see how well our model fit, we estimated the average of the highest posterior probabilities of class membership across individuals (Yoo, 2019). The posterior probabilities indicate the probability of an individual belonging to a particular segment based on their choice behavior. The segment that an individual has the highest probability of belonging to is of interest to us because it can tell us how well the model is segmenting the sample. The average of these “highest” class-membership probabilities across individuals and choice situations is
94.5%, which tells us that our model is doing a good job at predicting how well, on average, an individual belongs to specific segment given their revealed preferences (Yoo, 2019). The model is effective in identifying heterogeneity within the sample and differentiating multiple classes of preferences.
48
We can also use the prior probabilities of class-membership to estimate the distribution of individual-level coefficients in the sample. We can view this distribution graphically using Kernel density estimation (see Figure 6). As we can see, there is quite a bit of variability in the individual- level coefficients; but the highest concentration of the coefficients is towards the positive end, reinforcing the positive overall preferences observed in the sample using the mean WTP.
Figure 8: Kernel density plots of the distribution of individual-level coefficients for clean meat (top left), plant-based meat (top right), and conventional meat (bottom).
49
5.7. Robustness Checks
In order to check whether our results hold in alternate specifications testing the same hypothesis, we ran a conditional logit model (CLM) and a mixed logit model (MLM) including interaction terms for conventional meat, plant-based meat, and clean meat with all the membership variables
in the latent class model. In both specifications, the dependent variable and the interactions were treated as fixed while price, conventional meat, plant-based meat, and clean meat were treated as random. The results can be found in Appendix B, Table 11.
Both models report positive coefficients for all three types of meat with respect to chana (although significance varies), which is reflective of the mean WTP observed in the LCM presented in Table
5. In the main model, the sample collectively associates premiums with all three types of meat when chana is the base alternative, indicating the positive preferences. This is corroborated by the positive coefficients for the three types of meat observed in the CLM and MLM. The interpretation of the coefficients on the interaction terms can be compared to the relationships determined by our main model. Both the CLM and MLM indicate significant, negative coefficients for the interactions between vegetarianism and the three types of meat. This is consistent with the negative WTP estimates for all three types of meat with respect to chana for the vegetarian class (class 1). These interaction terms do provide some additional interpretive value for those interested. We can see that those earning an income between 20,000 and 50,000 exhibit positive and significant preferences for plant-based and clean meat.₹ This is an important₹ inference since it indicates that individuals in the middle-income category are most likely to purchase these alternatives. Overall, both robustness checks validate our main model, while providing some additional insight.
50
5.8. Peripheral Results
This sub-section of results contains the outcomes of some of the survey’s peripheral questions. i. Consumers’ Unaided Awareness
The survey assessed participants’ unaided awareness of the concepts of plant-based and clean meat.
These results are particularly useful to understand the current state of the market in terms of
knowledge. In our sample, 15.78% of respondents were aware of the concept of plant-based meat;
1.27% were unsure; and 82.95% were unaware. For clean meat, 9.92% were aware; 1.02% were
unsure; and 89.06% were unaware. These statistics are depicted in Figure 9.
Plant-Based Meat Clean Meat 89.06 82.95 15.78 9.92 1.27 1.02
AWARE UNAWARE UNSURE
Figure 9: Unaided awareness of plant-based meat and clean meat.
Binary dummy variables for awareness of plant-based and clean meat were generated. Those that
responded “Unsure” were coded as missing since they made up <5% of the sample. The coefficients and average marginal effects from a logit regression of awareness on demographic variables of interest are reported in Table 7. Since most of our covariates were coded as dummy variables, the
51
interpretation for the reported marginal effects would be the expected percentage change in the
probability of a respondent being willing to buy a simulated meat alternative (Prob(y = 1)), given
a true (x = 1) outcome for the independent variable.
Table 7: Logit regression of unaided awareness of plant-based and clean meat on demographic factors.
Plant-Based Meat Clean Meat Variable Coef. Avg. Marginal Effects Coef. Avg. Marginal Effects Vegetarian 0.0906253 0.0113536 -0.0909777 -0.006721 Environmentally Conscious -0.079971 -0.0100188 -0.4360493 -0.0322132 College Educated -0.0445839 -0.0055855 2.186*** 0.162*** Hindu -0.585** -0.073*** -1.345*** -0.010*** Muslim -0.1265464 -0.0158538 -0.1653225 -0.0122132 Age 0.0041919 0.0005252 0.0259512 0.0019171 Female 0.363216 0.0455038 -0.4787774 -0.0353698 Monthly Income > ₹ 50,000 0.916*** 0.115*** 0.2330243 0.0172147 Monthly Income between ₹ 20,000 and ₹ 50,000 -0.2533863 -0.0317443 -0.1007308 -0.0074415 Conservative -0.4432768 -0.0555339 -1.413*** -0.104*** Liberal -0.5760012 -0.0721616 -0.7902276 -0.0583783
Hindus are significantly less likely to be aware of plant-based and clean meat; an individual that identifies as Hindu is 7.3% less likely to be aware of plant-based meat and 1% less likely for clean meat. High-income earners are significantly more likely to be aware of plant-based meats. If someone earns more than 50,000, they are 11.5% more likely to be aware of plant-based meat.
College-educated individuals₹ are 16.2% more likely to be aware of clean meat than those individuals that have less than a college-degree. Those who adhere to more conservative, right- wing thinking are 10.4% less likely to be aware of clean meat than liberals and individuals with other political views.
52
ii. Consumers’ Willingness to Buy
Prior to the choice experiment, respondents were directly asked if they would be willing to buy
plant-based and clean-meat alternatives based on the information provided to them. While the DCE
“infers” a respondents’ willingness to buy based on their purchase decisions on different choice
occasions, this question aims to directly elicit the preference. Figure 10 depicts the results; 79% of
respondents indicated they would be willing to buy plant-based meat, while 57% of the respondents
indicated that they would be willing to buy clean meat.
Clean Meat
Plant-Based Meat
0 10 20 30 40 50 60 70 80 90
Figure 10: Percentage of respondents willing to buy plant-based meat and clean meat prior to choice experiment.
Averaging the willingness-to-buy across both alternatives, we could say that 68% of the sample is
receptive to simulated meat alternatives. In the LCM, classes 3 & 4 valued simulated meat items
over, both, chana and conventional meat. These classes collectively make up 51% of our sample.
As expected, when prices are attached and preferences indirectly inferred, there are fewer
consumers of simulated meat products due to their higher prices compared to conventional meat.
53
We use a binary logit to further understand the factors influencing willingness to buy. Table 8
reports the coefficients and the average marginal effects from a logit regression of willingness to
buy on the same demographic variables that were included in our class-membership model. The
marginal effects can be interpreted in the same fashion as coefficients from a simple linear
regression. For example, the expected probability of an individual being willing to buy plant-based meat decreases by 18% if they identify as vegetarian.
Table 8: Logit regression of willingness-to-buy plant-based and clean meat on demographic factors.
Plant-Based Meat Clean Meat Variable Coef. Avg. Marginal Effects Coef. Avg. Marginal Effects Vegetarian -1.450*** -0.180** -2.044*** -0.380*** Environmentally Conscious 0.610** 0.076** 0.641** 0.119*** College Educated 0.100 0.012 -0.209 -0.039 Hindu 0.554 0.069 0.194 0.036 Muslim -0.647 -0.081 -0.205 -0.038 Age -0.009 -0.001 -0.008 -0.001 Female -0.182 -0.023 0.229 0.042 Monthly Income > ₹ 50,000 0.148 0.018 -0.373 -0.069 Monthly Income between ₹ 20,000 and ₹ 50,000 0.911** 0.113** 0.126 0.024 Conservative 0.768** 0.095** 0.492* 0.0914* Liberal -0.230 -0.029 0.106 0.020
Vegetarianism, environmental consciousness, income, and political stances are significant
indicators of willingness to buy. Those who self-identified as non-vegetarians; those who showed
support for a policy aimed at curbing air pollution; those who earned an income above 50,000;
those who earned an income between 20,000 and 50,000; and those who voted for the BJP₹ in the
2016 elections are more likely to be willing₹ to buy plant₹ -based meat. These same categories, except
those who earned an income greater than 50,000, are more likely to buy clean meat as well.
₹
54
iii. Consumer Favorability Ratings After receiving standardized information on conventional meat, plant-based meat, and clean meat,
respondents were asked to rate each type of meat on four characteristics. The Likert-style rating for
the variables is displayed in the figure below.
Figure 11: Choice card to answer consumer favorability questions.
The mean values of these ratings allow us to gather valuable insight into consumer attitudes towards
meat alternatives and are reported in Table 9. Participants consistently rated plant-based meat most
favorably in terms of health, environmental friendliness, and animal welfare impact. While
consumers expect plant-based meat to cost roughly the same as conventional meat, clean meat is
perceived to be more expensive than clean meat. In terms of animal welfare, consumers perceive
the benefits of simulated meat over conventional meat, with both plant-based and clean meat rating
higher than conventional meat.
55
Table 9: Average consumer favorability ratings for conventional, plant-based, and clean meat
Characteristic Scale Conventional Meat Plant-Based Meat Clean Meat
Health 1 to 5 3.35 4.03 3.49 Environment 1 to 5 2.59 3.78 3.69 Affordability 1 to 5 3.27 3.26 2.89 Animal Welfare 1 to 5 1.89 4.12 3.8
iv. Consumers’ Choice in No-Cost Scenario
Lastly, respondents were asked to imagine themselves at a party where they did not have to pay for
food and select which type of meat they would prefer. The results are presented in Figure 12.
Surprisingly, approximately 32% of respondents indicated that they would try clean meat. Another
26% reported a preference for plant-based meat; 21% for conventional meat; and another 18% said
they would only eat other dishes at the party.
None
Clean Meat
Plant-Based Meat
Conventional Meat
0 5 10 15 20 25 30 35
Figure 12: Consumers’ choice outcomes in a party scenario where they do not have to pay for food.
56
57
CHAPTER 6
Discussion
6.1. Discussion of Results
The goal of this study was to assess the viability of meat alternatives as a solution to the negative
impacts of animal-intensive agriculture in India. To reiterate, we articulated the primary research question as follows:
Are meat alternatives a viable demand side solution to mitigating the societal concerns
associated with animal agriculture in India?
To answer this question, we identified four unique segments in the Indian market for protein, namely: the veggie lovers, the meat lovers, the plant-based meat enthusiasts, and the clean meat enthusiasts. The nomenclature of the segments immediately provides stakeholders with an overview of how preferences vary in the market. We observed that respondents, as a collective, associated premiums of ₹138 ($2) per kg with plant-based meat and ₹57 ($0.8) per kg with clean meat over the price of conventional meat—respondents, as a collective, were receptive to plant- and clean meat alternatives. However, preferences varied substantially between segments.
The veggie lovers consistently displayed highly significant, negative preferences for all three types of meat – conventional meat, plant-based meat, and clean meat – compared to chana (chickpeas).
The veggie lovers also had a significant indication of vegetarianism. This paper views meat
57
58
substitutes as a tool for guiding the rising demand for meat away from animal-based proteins.
Bearing that in mind, those who are likely vegetarian and/or prefer chana over conventional meat are already choosing a relatively more sustainable source of protein. Put differently, if the goal is sustainability, spending effort to try and get individuals who are already interested in a relatively low-impact alternative to substitute towards simulated meat is not in line with the desired dietary
shift. Interventions aimed at reducing dairy and egg consumption are more relevant to vegetarians
(Vetter et al., 2017).
This brings us to the first key finding of this study: vegetarians are not the target market for meat
alternatives in India. In their latent class analysis of meat substitution, Apostolidis and McLeay
(2016) identify six classes of consumers: price conscious, green, taste-driven, healthy, organic, and
vegetarian. The vegetarian class in their analysis observes strong preferences towards plant-based
meat alternatives compared to conventional beef, turkey, lamb, pork, and an opt-out option. Slade
(2018), too, concludes that vegetarians exhibit stronger preferences for plant-based and cultured
meat than meat-eaters. In this study, while the veggie lovers – consistent with Apostolidis and
McLeay (2016) and Slade (2018) – expressed positive preferences for simulated meat alternatives
against conventional meat, they displayed negative preferences for simulated meat against the non-
meat option: chana. Our conclusion of negative preferences towards simulated meat is supported
by the conclusion in Bryant et al. (2018) that non-vegetarians in India are more likely to purchase
simulated meat than vegetarians. Further support comes from Verbeke et al. (2015), who find that
respondents eating mostly vegetarian meals did not perceive clean meat to be healthy and were
therefore not the ideal target market.
One possible explanation for this disparity in results is that Slade (2018) and Apostolidis and
McLeay (2016) primarily observed newer or recently turned vegetarians, while Bryant et al. (2018),
Verbeke et al. (2015), and our study observed long-term vegetarians. Individuals who grew up
58
59
eating meat are naturally more likely to want alternatives that taste like meat, compared to those individuals that never developed a taste for meat. Therefore, recent vegetarians are more likely to have positive preferences for meat-free options that resemble the taste and texture of animal-based meat, whereas lifelong vegetarians are likely to have no desire for food items that fulfill the functionality of animal-based meat. In other words, long-term vegetarians are likely to have found ways to meet their nutritional needs and satisfy their taste buds using existing vegetarian options.
Corroborating this hypothesis, we find that nearly 70% of the vegetarians in our sample were lifelong vegetarians.
Allowing for a widely-available, non-meat vegetarian option like chana is an important decision point because “meat” – no matter the type – does not appeal to most vegetarians. It is important to understand that, in a real-life choice situation, individuals choose between meat and non-meat options to meet their requirements. In the case of vegetarians, they actively and consistently choose to not purchase the meat option. Apostolidis and McLeay (2016) argue that an opt-out option serves the same purpose as chana does in our experiment. However, we assert that a well-defined vegetarian, non-meat option might yield more accurate and defined preferences than an opt-out option, since opt-out options tend to be ambiguous for respondents and researchers. As has been discussed in the analytical methodology, opt-out options may lead to extreme choice behavior and context effects within responses, and are most often selected by lower educated groups (Schlereth
& Skiera, 2016; Veldwikj et al., 2014). Therefore, the difference between an opt-out strategy and a defined fourth alternative could be feeding into the disparity in conclusions, as well. It is not clear from the results in Apostolidis and McLeay (2016) and Slade (2018) how often the opt-out option was selected by vegetarians, making it difficult for us to comment further on this. Another possible explanation for the disparity in conclusions could be the cultural differences between India and
59
60
Western nations such as the UK and Canada. However, at odds with this reasoning is the fact the
Verbeke et al. (2015) survey took place in Belgium.
Excluding the veggie lovers, we are still left with 79% of the sample that do prefer meat products to chana: the meat lovers, the plant-based meat enthusiasts, and the clean meat enthusiasts. The plant-based meat enthusiasts and the clean meat enthusiasts, as the names suggest, are excited about simulated meat alternatives. The two segments make up 32% and 19.6% of the sample, respectively. These segments accept the current pricing structure of the market where clean meat is the most expensive, followed by plant-based meat, conventional meat, and chana. Both the plant- based meat enthusiasts and the clean meat enthusiasts exhibit positive preferences for conventional meat, but they associate stronger preferences for simulated meat comparatively. Based on this, we hypothesize that these are individuals who, regardless of their diet choice, associate higher monetary value with conventional meat as a source of protein than chana. However, the key take- away about these classes is their “enthusiasm” for simulated meat products. Collectively, these classes make up over 50% of our sample, indicating that more than half of the sample exhibits strong, positive preferences for simulated meat products over conventional meat and chana. The individuals in these segments are likely the first adopters of simulated meat products in India. This is a strong sign of positive prospects for a simulated meat industry in India.
On the other hand, the meat lovers comprise 27.5% of the total sample. Individuals in this class exhibit strong preferences for conventional meat, as is indicated by the negative WTP estimates of
- 65 and - 68 per kg for plant-based and clean meat relative to conventional meat. This is an important₹ result.₹ It indicates that, in order entice those with strong preferences for conventional meat to substitute towards meat alternatives, the pricing of simulated meat must be considerably lower than that of conventional meat. These are the meat-attached consumers of this market. If meat substitution is the goal, then those individuals with the strongest preferences towards
60
61
conventional meat will require the most effort to induce substitution. The plant-based meat enthusiasts and the clean meat enthusiasts accept the pricing structure of the market as is and indicate preferences for simulated meat products. The meat lovers, however, opt to continue eating conventional meat if the market follows a pricing structure where simulated meat alternatives are more expensive than their conventional counterparts. This the most challenging set of consumers, and also the most necessary to account for when pricing simulated meat products because of their strong preference for an unsustainable source of protein.
Having identified the nature of the classes, the class-membership model in Table 4 can be used by policymakers and businesses to identify the demographic and socioeconomic characteristics that are associated with the classes so that consumers can be targeted more effectively. For instance, if a business is trying to enter the plant-based meat market and wants to know who their first-buyers might be, the results of the class-membership model for Class 3, the plant-based meat enthusiasts, will provide the demographic characteristics of those individuals that are most likely to purchase plant-based meat relative to Class 4. The coefficient for vegetarianism, for example, suggests that vegetarians are more likely to be in the plant-based enthusiasts class compared to the clean meat enthusiasts and the meat lovers, but less likely to be in the plant-based enthusiasts compared to the veggie lovers.
Conclusions regarding the impact of socio-economic and demographic factors on meat substitution vary between studies. Older and lower educated groups have found to be less willing to reduce meat intake (de Boer & Aiking 2011). Within higher educated groups, individuals have been observed to be wary of clean meat (Hocquette et al. 2015), suggesting stronger potential for plant- based options. Slade (2018) finds that higher educated Canadians are more likely to purchase simulated meat. In our study, college education was a mostly insignificant indicator, across the latent class logit, mixed logit, and conditional logit specifications. The highest significance
61
62
observed was for class-membership in the meat eaters, suggesting that those who attend college are likely to have the strongest preferences towards conventional meat. College educated individuals are also more likely to be plant-based enthusiasts than clean meat enthusiasts or veggie lovers.
De Boer, Schosler, and Aiking (2017) find that females are less likely to be heavy meat eaters than males. In our latent class model, we find that women are most likely to be clean meat enthusiasts.
The mixed logit specification also reports a significant coefficient for the female and clean meat interaction. This suggests, intriguingly, that females are more likely to be appealed by clean meat.
A similar affinity for clean meat was noticed amongst those who favored environmentally conscious public policy. Educational and informational campaigns are an effective tool for encouraging meat substitution. Slade (2018) finds that environmental consciousness affects desirability of meat substitutes significantly, suggesting policy surrounding information campaigns regarding the adverse impacts of animal agriculture. Based on our results, we believe that such educational campaigning will be more beneficial to manufacturers of clean meat products than plant-based meat products.
Bryant et al. (2018) find that willingness to purchase clean meat increases with income, political liberalism, and familiarity with the concept. Similarly, they find that willingness to buy plant-based meat increases with non-vegetarianism, higher incomes, and political liberalism, with the addition of higher education. To further understand how these conclusions compare to ours, we can turn to the binary logistic regressions of willingness-to-buy in Table 8, as well as the WTP estimates and the class-membership model. Consistent with their results, we find that non-vegetarians are more receptive to simulated meat products than vegetarians. The rest of our conclusions are not as consistent. We find that political conservatism increases the likelihood of being willing to buy both plant-based meat and clean meat, whereas liberalism decreases the likelihood for plant-based meat
62
63
and increases the likelihood for clean meat. This can, perhaps, be explained by the religious, conservative history of vegetarianism in India. We also find that middle-income groups are significantly more likely to purchase plant-based meat; the coefficients on clean meat were positive
but statistically insignificant. We did not observe significant coefficients for education and high
income for both simulated meat products.
Unfamiliarity with meat substitutes has been identified as a key barrier to the adoption of meat
substitutes (Hoek et al., 2011). Consumers’ unaided awareness was low for both plant-based meat
and clean meat; there is a significant educational hurdle for policymakers and businesses to
overcome. A fascinating result was observed when an alternate specification (not reported) of the
latent class model was run with plant-based meat awareness and clean meat awareness as class-
membership variables. Classes 1-3 observed significant, negative coefficients for plant-based and
clean-meat awareness. This implies that the individuals that were identified as veggie lovers, meat
lovers, and plant-based meat enthusiasts were less likely to purchase plant-based meat alternatives
and clean-meat alternatives if they were aware of them, relative to the clean meat enthusiasts. In the mixed logit specification containing the awareness variables, individuals who were aware of plant-based meat alternatives were less likely to purchase plant-based meat; however, those who were aware of clean meat were more likely to purchase clean meat. This is a good sign for the clean meat industry but a point of concern for the plant-based meat industry. The negative relationship between awareness and likelihood of purchase of plant-based options, as indicated by the mixed logit specification, suggests that existing perceptions of plant-based meat are at play. We suppose that respondents who claimed to be aware of plant-based meat connected it to existing meat replicates such as soya chaap and nutrela, but not to the aforementioned “newer generation” of increasingly meat-like plant-based substitutes. Stakeholders in the plant-based meat alternative space must employ information campaigns to consciously try and reverse this existing negative
63
64
relationship between awareness and purchasing behavior, in order to encourage adoption of their products. Considering how few people were aware of these alternatives, the awareness variables were not included in our main model – but the lesson is important, nonetheless.
However, in general, the idea of plant-based alternatives seems to fare better with respondents. The consumer favorability ratings in Table 9 are very telling. We see that consumers perceive simulated meat products to have health-related, environmental, and animal-welfare benefits compared to conventional meat. We find largely positive attitudes towards meat substitutes within our sample, which, alongside the results of the market segmentation, suggests that the Indian market will be receptive to meat substitutes. However, between the two simulated meat alternatives, plant-based meat rated consistently higher than clean meat. Even in the choice experiment, plant-based meat had a significantly higher mean WTP than clean meat. An enhancement to this receptiveness is that the plant-based meat enthusiasts were the largest of the four identified segments, making up 32%
of our sample. All these factors lead us to our second important conclusion: of the two simulated meat alternatives, plant-based meat alternatives have stronger prospects in India compared to clean meat.
6.2. Policy Simulation & Implementation
We simulate two policy options for regulators. The variability in preferences between segments indicates that not one approach will fit all. However, adopting a generalized modeling framework provides a direction to draft policy around. The EAT-Lancet commission predicts that increased consumptions of plant-based diets could reduce GHG emissions by up to 80% (Willet et al., 2018).
In order to bring about The Great Food Transformation, the commission suggests a 50% reduction in the consumption of unhealthy foods – such as red meat and sugar – to simultaneously meet global environmental and nutritional goals. Research has shown that food price changes are the most
64
65
impactful in low-income countries (Cornelsen et al., 2014). Two possible price-oriented policy instruments to influence the demand for conventional meat are a tax on conventional meat, and a subsidy on meat alternatives.
Based on the unconditional choice probabilities and the marginal effects, we estimate what price changes would be needed to bring about the suggested 50% reduction in demand for conventional meat. The initial average prices for conventional meat, plant-based meat, and clean meat were ₹149,
₹180, and ₹210 respectively. If a meat-tax were to be implemented, our results suggest that the
price of conventional meat will need to increase by 63% or ₹93 in order to decrease its market share
by 50%. Alternatively, we could explore the potential of subsidies towards meat substitutes to
increase their consumption by 50%. We must keep in mind, however, that a 50% increase in the
consumption of a meat alternative does not imply a 50% reduction in the demand of conventional
meat; demand could be reduced from any one of the remaining three options. Bearing that in mind,
given our hypothetical pricing structure, to increase the market share of plant-based meat by 50%,
its price needs to decrease by ₹117 or 65%. To increase the market share of clean meat by 50%, its price needs to decrease by ₹200 or 95%. Considering the environmental and societal benefits explored earlier in this paper, there is ample reason for governmental support towards these sunrise industries. It is also important to remember that the price of conventional meat is artificially low.
If subsidies for existing animal-agriculture operations are redirected towards the meat alternative sector, the food system is likely to alter towards a more sustainable iteration much faster.
Targeted and segment-specific policy approaches to encourage meat substitution are effective
(Apostolidis & McLeay, 2016; Spiller & Nitzko, 2015). The above discussions establish that veggie
lovers need the least attention from a policy standpoint, while the meat lovers need the most.
Vegetarianism was consistently a significant indicator for membership in veggie lovers and meat lovers. Therefore, how meat consumption varies in India is a good proxy for targeting these two
65
66
classes of consumers. Figure 13 shows the percentage of males and females that consume meat across states. States such as Kerala and Assam have extremely high levels of meat consumption, while states such as Punjab and Rajasthan have very low levels of meat consumption. This is just
an example of one possible strategy for targeting; stakeholders can engage in more specific
targeting strategies based on the availability and reliability of data. Another policy example could
be the use of financial incentives for meat lovers, similar to how Apostolidis and McLeay (2016)
suggest financial incentives for their ‘price conscious’ segment.
Figure 13: Meat consumption of males and females in India by state.
Source: NFHS-4, graphic by Yadavar (2018).
Further important lessons regarding strategies to encourage meat substitution can be learned from
the existing literature. Focusing on the “warm glow” associated with eating less meat has been
deemed an effective strategy for reducing meat consumption, but not necessarily an effective
66
67
strategy for increasing the consumption of meat substitutes in China (Taufik, 2018). This message could be extended to the Indian context, too, considering the bevy of existing vegetarian options which could yield lower meat consumption but not necessarily higher adoption of meat alternatives.
Specific to increasing the consumption of meat alternatives, studies have found that clean meat has a “yuck” factor associated with it (van der Weele & Driessen, 2013). In their systematic review of consumer acceptance of clean meat, Bryant and Barnett (2018) conclude that rejection of clean
meat rests on three primary factors: taste, price, and safety. Macdonald and Vivalt (2017) find that
embracing the unnaturalness of cell-based meat is the only successful strategy to overcome such negative social perceptions of in-vitro meat technology. Anderson and Bryant (2018) find that reversing the “yuck” factor and associating unnaturalness with conventional meat yields better preferences towards clean meat. This “yuck” factor is likely to prevail in India as well, and is something proponents of clean meat must consciously and strategically try to overcome. More
recently, Graça et al. (2019) highlight the importance of accepting consumers’ orientations and
perceptions towards plant-based diets while designing strategies to encourage meat substitution.
Innovative thinking in policy solutions is needed to target the environmental problems associated
with agriculture (Smith et al., 2007). Allen and Hof (2019) propose biodiversity offsetting to reduce
meat consumption, a policy that would involve producers requiring to quantify and correct
ecological impacts of their activities by investing in biodiversity gains elsewhere. Other studies
emphasize that it is vegetarians that lead the transition towards meat substitutes by communicating
with their peers regarding their benefits, and thereby creating “hype” (Apostolidis & Mcleay,
2016). Seeing how vegetarians perceive meat substitutes to be better than conventional meat but
worse than chana, we are uncertain if vegetarians will aid the adoption of meat substitutes in India.
67
68
6.3. Shortcomings of the Survey
The biggest shortcoming of this survey is its lack of specification of the form of meat in question, i.e. chicken, mutton, or fish. The survey could have benefitted from having labeled the options more directly. For instance, naming the labels plant-based chicken, conventional chicken, and clean chicken would have provided respondents with more context for how realistic the price levels were.
Surprisingly, we faced little protest from respondents regarding this lapse. In describing the alternatives, participants were told that simulated meat alternatives could include chicken, fish, mutton, or any other form of conventional animal-based meat. We find some recluse for this shortcoming in a recent interview by plant-based meat manufacturer Hungry Planet’s CEO, Todd
Boyman. He argues that that making plant-based beef, chicken, and crab, all cost, for example,
$3.50 per pound, due to relatively consistent variable costs. This pricing, he claims, makes the plant-based alternative highly competitive with conventional crab, moderately competitive with conventional beef, and poorly competitive with conventional chicken (Zacharias, 2019). This,
along with the prospect of clean meat to attain price parity with all forms of conventional meat
(Specht, 2019), warrants our decision to keep the label ambiguous to some degree.
Snowball and Willis (2011) compare the effectiveness of DCEs conducted using face-to-face
methods against those conducted using self-completion methods such as online surveys. They find
that self-completion methods consistently generate lower attribute coefficients than face-to-face surveys. This can possibly be due to interviewer bias. While we tried to minimize interviewer bias, this is certainly a limitation of the data used in this study. The primary trade-off between the two methods was the possibility of obtaining a respondent set that diverse in income and levels of
education, which is something we valued. Additionally, a face-to-face situation allowed for a conversational-style interview, which was helpful because the subject matter dealt with technological innovations that most respondents were expected to be unfamiliar with.
68
69
Lastly, the believability of the standardized prompts in our survey could be scrutinized.
Respondents were told that these plant-based meat and clean meat had similar taste and texture to
conventional meat product. However, Slade (2018) finds that most respondents do not, in fact,
believe a prompt of the sort. Considering recent improvements in technology, imitation is getting
closer and it is our hope that such a prompt will experience more believability in future surveys.
Additionally, there were possibly slight discrepancies in the Hindi and English versions of the
survey due to difference in rhetoric between the languages.
6.4. Potential for Future Research
Vetter et al. (2019) indicate high contributions to agricultural GHG emissions in India from milk
and eggs. Cultured milk technology exists and is currently being further developed by a company
called Perfect Day Foods. Plant-based milks already exist and are gaining popularity in Asia (Sethi,
Tyagi, and Anurag, 2016). Another startup, Clara Foods is developing a cell-based alternative to
egg whites using yeast, while a long-term player in the space, JUST, has already launched a plant-
based egg alternative made from mung beans (popularly known in India as moong, which is used
to make cheelas, a vegetarian pancake similar to an omelet) that mimics the functionality of animal-
based eggs. While these technologies exist, they are not a focus of this paper. Future work exploring
the economic prospects of plant-based and cell-based dairy and egg products is needed. Even
though our conclusions regarding preferences towards meat alternatives might be extended to the
potential preferences towards egg alternatives, they cannot be extended to the dairy sector. Dairy
presents a completely different set of challenges in India due to religious attachment. Considering
the widespread consumption of dairy and eggs and their reported unsustainability, future research
is key to help policymakers and manufacturers navigate demand within this market. Similarly, there
is a lack of research on the economic prospects of simulated fish alternatives and their potential for
preventing ecosystem collapse in our aquatic systems. Lastly, in terms of environmental research,
69
70
our literature review suggests that additional work is needed to explore impacts of animal-based agriculture on metrics other than GHG emissions and water footprints. These could include nutrient pollution, biodiversity loss, soil health, etc. Specific analyses quantifying the merits of meat substitution on food security in developing countries is needed, too.
As of 2016, India was one of the largest exporters of beef in the world. While this supply caters mostly to foreign populations, we think there is potential for research on supply side interventions to shift Indian production from animal-based proteins towards the production of simulated meat alternatives. For instance, one such policy could be the introduction of retraining programs enabling farmers to switch away from animal-based farm products towards producing sustainable crops that could be used in manufacturing of simulated meat. Research exploring such strategies will catalyze change on the supply side. Along the same lines, a transition away from animal-based meats is bound to have an impact on smallholders in the Indian agricultural sector. The poorer half of the world’s population only produces 10% of global emissions while bearing over 75% of the consequences. Is it then, fair, that smallholders who depend on livestock for their livelihoods be punished because of coal-powered industrialization and legacy emissions in Western countries?
Further research is needed on how the introduction of meat alternatives can incorporate and empower these smallholders during the transitioning of our global food systems away from the intensive use of animals.
70
71
CHAPTER 7
Conclusion
Considering the broad impacts of our food systems, dietary trends in populous countries like India
are of global import. A recent headline in Forbes echoes this message: “Demand for Meat Is
Growing Rapidly in India. This Could Impact All of Us” (Rowland, 2018). The composition of the
Indian diet is moving from pulses and cereals towards highly processed foods and animal-based
fats (Aleksandrowicz et al., 2019). While Indians still don’t consume nearly as much meat as
Western populations per capita (Ritchie & Roser, 2019), maneuvering rising demand away from
animal-based proteins towards more sustainable, healthier meat alternatives at an early stage has societal merit. For instance, solely in terms of environmental impact, if the entire Indian population
were to adopt affluent high-meat diets, it would result in increases of 19–36% across GHG
emissions, water footprints, and land use (Aleksandrowicz et al., 2019).
Even though the dietary guidelines issued by the National Institute for Nutrition (NIN) of India are
one of the most sustainable in the world (Ritchie, Reay, & Higgins, 2018c), they call for an increase
in meat and fish intake. This is primarily to address nutrient deficiencies since small inclusions of
meat in diets have shown to have nutritional benefits in reducing malnourishment (Rivera et al.,
2003). In their recently published work, Minocha et al. (2019), too, call for governmental support
to increase demand for fish and meat in an effort to alleviate nutrient deficiencies. While we agree
with the aim of enhancing dietary quality, we would assert that doing so by increasing animal-
protein intake is counterproductive with environmental goals and ethical implications. Following
NIN’s guidelines, for instance, increases India’s GHG emissions, worsens nutrient pollution, and
71
72
land use (Behrens et al., 2017). This evident unsustainability of meat products causes a conflict in environmental and global health objectives. Springmann et al. (2018) discuss the importance embracing the interconnectedness of public policies aiming to meet health objectives and policies
aiming to mitigate environmental impacts. The EAT-Lancet report is a recent example of how
frameworks for prioritizing multiple goals can be established.
While this thesis has largely focused on the environmental, nutritional, and food security benefits
associated with meat alternatives, the ethical implications of meat consumption cannot be ignored.
The onset of large-scale factory farming of animals has exacerbated the slaughter and mistreatment
of animals. The economic and ethical arguments for the humane treatment of animals are extensive
and complex (see Wang & Chan, 2017; and Miele, 2016). Humans slaughter over 100 billion
animals every year, which translates to nearly 2 billion animals per week, which is about a third of
the human population in animals slaughtered on a weekly basis (Miele, 2016). Therefore, the ethical
implications of a switch to meat alternatives are monumental. Meat alternatives lay down a path
for a food system that would not rely upon the domestication and slaughter of non-human animals,
but still provide consumers with the nutrition, taste, and texture of animal-based meat products.
As we have established throughout this paper, the list of societal concerns associated with meat
consumption is long; therefore, at some level, we need to prioritize. This thesis places a stronger
emphasis on environmental issues than nutritional and societal concerns due to the urgency of the
environmental problems we face. In a world where climate change threatens human and non-human
existence like never before, the environment needs to take priority; for, without an environment
that can sustain us, the likelihood of future generations of humans having a society to have societal
concerns about is minimal.
Poore and Nemecek (2018) report than dietary change will outweigh the impact of any supply-side
policies aimed at mitigating the harmful environmental impacts of our food systems. As a
72
73
foundational tenant of microeconomic theory, suppliers produce what markets demand—if the composition of the demand changes to demand more sustainable alternatives, supply will follow.
While it is important to fix and improve supply systems, manipulating the demand allows for a complete revamping of the entire system. That is the promise and prospect of meat alternatives.
The potential for simulated meat products to battle environmental problems needs to be considered.
Plant-based meat alternatives are becoming increasingly more realistic, something that is valued by consumers (Hoek et al., 2011; Elzerman, Hoek, van Boekel, and Luning, 2011). A handful of companies in this space have re-defined what meat alternatives look like and how meaty they can get. With improved imitation and increasing information regarding their environmental and nutritional benefits, they are rapidly gaining popularity across the world. Recent industry reports indicate that the global market for meat substitutes is expected to be worth $9.25 billion by 2023, reporting a compound annual growth rate (CAGR) of 4.0%, one of the highest among other storage categories (Market Research Future, 2019). At its IPO, Beyond Meat was priced at $25 per share.
As of the time of writing this paper, a little over a month after its IPO, the stock is priced has increased by 528%. In India, the main company in the plant-based market is GoodDot, who have created plant-based imitations of mutton and chicken. They recently announced a fried-chicken imitation of KFC’s popular Zinger.
Clean meat is still not commercially available for sale in most of the world. Leaders in the technology include the Netherlands, Israel, and the United States. The technology comes with the promise of creating 100% real animal meat without the need to slaughter an animal. Clean meat manufacturers such as Memphis Meats are generating a lot of excitement—conventional meat giant
Tyson acquired a stake in Memphis Meats, alongside notable billionaires Bill Gates and Richard
Branson (Sorvino, 2018).
73
74
As regulatory frameworks for a cell-based meat market in the US and EU are coming together, the
Indian government is not far behind. In 2018, the Humane Society of India and the Centre for
Cellular and Molecular Biology in Hyderabad, India joined forces to announce an initiative to promote and pursue research in the clean meat space (Humane Society International, 2018). The state of Maharashtra announced a cellular agriculture research center at the Institute of Chemical
Technology in Mumbai in February 2019 (Kulkarni, 2019). Most recently, in April 2019, the Indian government announced $600,000 in research funding for Center of Molecular Biology in
Hyderabad to develop cellular meat technology (Ramamurthy, 2019). Chiles (2013) discusses the importance of focusing on scalability of cultured meat technology and approaching scalability with caution. Dagevos, Tolonen, and Quist (2019) echo this sentiment by discussing strategies to create markets for simulated meat by practicing “cautious optimism”, while simultaneously creating the necessary hype surrounding this space. However, while clean meat is getting its fair share of attention from Indian regulators, plant-based meat innovation is not being tapped into as much. In a country with an unparalleled history and expertise in vegetarian cuisine, this seems like a lost opportunity. Seeing how, based on our findings, plant-based alternatives fare much better with
Indian consumers than clean meat, innovation in this space needs to become a focus.
Ritchie et al. (2018b) suggest that sustainable food security in India will not stem for self- sufficiency but trade with other countries to ensure that macronutrient needs are met. This signals an opportunity for international and domestic, established and hopeful simulated meat manufacturers to tap into the market. The WTP estimates presented in this paper have important implications for such companies, as well as policymakers trying to steer demand away from conventional meat. They allow an understanding of what price-points meat alternatives need to
attain in order to be competitive with conventional meat. Furthermore, the class-membership model
allows these stakeholders to further understand the makeup of these segments and strategize
74
75
accordingly. We strongly encourage a segment-specific approach to be adopted in the introduction of meat alternatives into the Indian market. Accommodating for differences in the above-identified segments takes us a step closer to accommodating for the reality that the same strategy does not
and will not work on everyone. For businesses and changemakers in this space, we specifically
suggest a focus on simulated alternatives of chicken, mutton, and fish based on their widespread
consumption and environmental impacts (Vetter et al. 2017). Dairy and eggs also must be focused
on from an ecological standpoint; however, we cannot confidently extend our findings to
alternatives of these food items.
India presents a unique case. One cannot fully understand meat consumption in the Indian
subcontinent without acknowledging the complex socio-cultural dynamics surrounding it. In 2017,
the BJP banned the sale and purchase of beef under an animal welfare act. The anti-beef movement
has gone on to use popular environmentalist and actor, Leonardo Di Caprio, to denounce the
support their beef-ban on the basis of environmental reasoning (Blackburn-Dwyer, 2016). While a
ban of this sort may achieve the goal of reducing beef consumption in the subcontinent, it does so
under the guise of “food fascism” (Staples, 2018). The driving factors for the BJP’s actions have
historically been religious, and they do not shy away from admitting that the driving forces behind
this ban were religious (Sathyamala, 2018). Therefore, we would like to make a clarification: while
this paper targets rising meat consumption in India, it does not do so in support of or in animosity
towards any religious group.
So, are meat alternatives a viable solution for mitigating the negative impacts associated with
animal-based proteins in India? Provided that production and supply-chain challenges can be resolved, the evidence would suggest so. We hope these results aid policymakers and entrepreneurs to bring about the necessary changes to our food systems. The optimism offered by
meat alternatives in the Indian context rests on a desire to navigate tastes before the Indian
75
76
population finds an appetite for meat that is as hard to maneuver as that of Western populations.
Think of it as needing India to skip a step in the elaborate food chain of affluence. Instead of going from meat to more meat, we propose that Indians go from meat to “meat”.
76
77
Appendix A
Table 10: D-optimal fractional factorial design of attribute levels.
Block Conventional Meat Plant-Based Meat Clean Meat Chana 1 150 210 180 90 1 120 180 240 90 1 180 150 180 120 1 120 150 210 150 2 150 150 210 90 2 120 180 180 120 2 180 210 180 150 2 120 210 240 150 3 120 150 180 90 3 180 210 240 90 3 120 150 240 120 3 150 180 210 150 4 180 180 180 90 4 150 210 240 120 4 120 210 180 150 4 180 150 240 150 5 120 210 210 90 5 180 150 240 90 5 150 150 180 150 5 180 210 240 150
77
78
Appendix B
Table 11: Results from conditional logit and mixed logit specifications.
Conditional Logit Mixed Logit Covariates Coefficients Coefficients Price -0.007*** -0.014*** Conventional Meat 1.342** 1.585^ Plant-Based Meat 1.566*** 1.896^ Clean Meat 1.567** 0.928
Interaction Variables Conventional Meat Plant-Based Meat Clean Meat Conventional Meat Plant-Based Meat Clean Meat Vegetarian -3.140*** -1.536*** -2.089*** -5.569*** -2.001*** -3.163*** Environmentally Conscious -0.306 -0.190 0.112 -0.788^ -0.008 0.447 College Educated 0.484^ 0.307 0.097 0.436 0.541 -0.025 Hindu 0.464^ 0.704** 0.150 0.975^ 1.382** 0.877 Muslim 1.214* 0.730^ -0.026 2.585** 1.058 0.478 Age -0.010 -0.018^ -0.011 -0.011 -0.037* -0.037^ Female -0.376^ -0.226 0.114 -0.515 -0.411 1.280* Monthly Income > ₹ 50,000 0.086 0.138 -0.254 0.302 0.195 -0.393 Monthly Income between ₹ 20,000 and ₹ 50,000 0.508^ 0.611** 0.631* 0.850^ 1.145** 1.582** Conservative -0.172 0.147 -0.160 0.018 0.197 -0.612 Liberal 0.188 0.304 0.205 -0.020 0.594 0.012
*** significant at 1% ** significant at 5 % * significant at 10% ^ significant at 25%
78
79
References
Abraham, D., & Rao, O. (2017, December 08). 84% Dead In Cow-Related Violence Since 2010 Are
Muslim; 97% Attacks After 2014. Retrieved from https://archive.indiaspend.com/cover-story/86-
dead-in-cow-related-violence-since-2010-are-muslim-97-attacks-after-2014-2014
AFP. (2011, September 18). Mumbai: India's megacity melting-pot. Retrieved from
https://www.independent.co.uk/travel/news-and-advice/mumbai-indias-megacity-melting-pot-
2303332.html
Agrawal, S., Millett, C. J., Dhillon, P. K., Subramanian, S. V., & Ebrahim, S. (2014). Type of
vegetarian diet, obesity and diabetes in adult Indian population. Nutrition journal, 13(1), 89.
Aiking, H. (2011). Future protein supply. Trends in Food Science & Technology, 22(2-3), 112-120.
Alae-Carew, C., Bird, F. A., Choudhury, S., Harris, F., Aleksandrowicz, L., Milner, J., ... & Green, R.
(2019). Future diets in India: A systematic review of food consumption projection studies. Global
Food Security, 23, 182-190.
Aleksandrowicz, L., Green, R., Joy, E. J., Smith, P., & Haines, A. (2016). The impacts of dietary
change on greenhouse gas emissions, land use, water use, and health: a systematic review. PloS
one, 11(11), e0165797.
Aleksandrowicz, L., Green, R., Joy, E. J., Harris, F., Hillier, J., Vetter, S. H., ... & Haines, A. (2019).
Environmental impacts of dietary shifts in India: A modelling study using nationally-
representative data. Environment international, 126, 207-215.
79
80
Allen, A. M., & Hof, A. R. (2019). Paying the price for the meat we eat. Environmental Science &
Policy, 97, 90-94.
Allen, M. R., Dube, O. P., Solecki, W., Aragón–Durand, F., Cramer, W., Humphreys, S., & Zickfeld,
K. (2018). Framing and context. Global warming of, 1, 47-91.
Anderson, J., & Bryant, C. (2018). Messages to Overcome Naturalness Concerns in Clean Meat
Acceptance: Primary Findings.
Apostolidis, C., & McLeay, F. (2016). Should we stop meating like this? Reducing meat
consumption through substitution. Food Policy, 65, 74-89.
Appleby, P. N., Thorogood, M., Mann, J. I., & Key, T. J. (1999). The Oxford vegetarian study: an
overview. The American journal of clinical nutrition, 70(3), 525s-531s.
Barnard, N. D., Cohen, J., Jenkins, D. J., Turner-McGrievy, G., Gloede, L., Green, A., & Ferdowsian,
H. (2009). A low-fat vegan diet and a conventional diabetes diet in the treatment of type 2
diabetes: a randomized, controlled, 74-wk clinical trial. The American journal of clinical
nutrition, 89(5), 1588S-1596S.
Behrens, P., Kiefte-de Jong, J. C., Bosker, T., Rodrigues, J. F., De Koning, A., & Tukker, A. (2017).
Evaluating the environmental impacts of dietary recommendations. Proceedings of the National
Academy of Sciences, 114(51), 13412-13417.
Bellarby, J., Tirado, R., Leip, A., Weiss, F., Lesschen, J. P., & Smith, P. (2013). Livestock
greenhouse gas emissions and mitigation potential in Europe. Global change biology, 19(1), 3-18.
Blackburn-Dwyer, B. (2016, June 23). Leonardo DiCaprio joins the anti-beef movement. Retrieved
from https://www.globalcitizen.org/en/content/leonardo-dicaprio-anti-beef/
80
81
Bovalino, S., Charleson, G., & Szoeke, C. (2016). The impact of red and processed meat
consumption on cardiovascular disease risk in women. Nutrition, 32(3), 349-354.
Boxall, P. C., & Adamowicz, W. L. (2002). Understanding heterogeneous preferences in random
utility models: a latent class approach. Environmental and resource economics, 23(4), 421-446.
Breidert, C., Hahsler, M., & Reutterer, T. (2006). A review of methods for measuring willingness-to-
pay. Innovative Marketing, 2(4), 8-32.
Bryant, C., & Barnett, J. (2018). Consumer acceptance of cultured meat: A systematic review. Meat
Science, 143, 8-17.
Bryant, C. J., Szejda, K., Deshpande, V., Parekh, N., & Tse, B. (2019). A Survey of Consumer
Perceptions of Plant-Based and Clean Meat in the USA, India, and China. Frontiers in
Sustainable Food Systems, 3, 11.
Camarena, D. M., & Sanjuán, A. I. (2008). Ethnic food preferences in the Spanish market.
In presentation at the 12th EAAE Congress ‘People, Food and Environments: Global Trends and
European Strategies’, Gent (Belgium).
Campbell, D., & Erdem, S. (2019). Including opt-out options in discrete choice experiments: issues to
consider. The Patient-Patient-Centered Outcomes Research, 12(1), 1-14.
Cao, Y., & Li, D. (2013). Impact of increased demand for animal protein products in Asian countries:
Implications on global food security. Animal frontiers, 3(3), 48-55..
CEIC Data. (2019). India GDP per Capita [1958 - 2019] [Data & Charts]. Retrieved from
https://www.ceicdata.com/en/indicator/india/gdp-per-capita
Chao, A., Thun, M. J., Connell, C. J., McCullough, M. L., Jacobs, E. J., Flanders, W. D., ... & Calle,
E. E. (2005). Meat consumption and risk of colorectal cancer. Jama, 293(2), 172-182.
81
82
Chiles, R. M. (2013). If they come, we will build it: in vitro meat and the discursive struggle over
future agrofood expectations. Agriculture and Human Values, 30(4), 511-523.
Chiles, R. M., & Fitzgerald, A. J. (2018). Why is meat so important in Western history and culture?
A genealogical critique of biophysical and political-economic explanations. Agriculture and
human values, 35(1), 1-17.
Chiles, Robert M., et al. “Greenhouse Gas Emissions from Agriculture: Reconciling the
Epistemological, Ethical, Political, and Practical Challenges.” Journal of Agricultural and
Environmental Ethics, vol. 31, no. 3, 2018, pp. 341–348., doi:10.1007/s10806-018-9728-5.
Cornelsen, L., Green, R., Turner, R., Dangour, A. D., Shankar, B., Mazzocchi, M., & Smith, R. D.
(2015). What happens to patterns of food consumption when food prices change? Evidence from
a systematic review and meta‐analysis of food price elasticities globally. Health
economics, 24(12), 1548-1559.
D'Silva, J., & Webster, J. (Eds.). (2017). The Meat Crisis: Developing More Sustainable and Ethical
Production and Consumption. Routledge.
Dagevos, H., Tolonen, E., & Quist, J. (2019). Building a Market for New Meat Alternatives:
Business Activity and Consumer Appetite in the Netherlands. In Environmental, Health, and
Business Opportunities in the New Meat Alternatives Market(pp. 183-201). IGI Global.
De Boer, J., & Aiking, H. (2011). On the merits of plant-based proteins for global food security:
Marrying macro and micro perspectives. Ecological Economics, 70(7), 1259-1265.
De Boer, J., Schösler, H., & Aiking, H. (2017). Towards a reduced meat diet: Mindset and motivation
of young vegetarians, low, medium and high meat-eaters. Appetite, 113, 387-397.
DeLessio-Parson, A. (2016). The Protective Effect of Plant-Based Diets in Urbanizing India.
82
83
Dettling, J., Tu, Q., Faist, M., DelDuce, A., & Mandlebaum, S. (2016). A comparative life cycle
assessment of plant-based foods and meat foods. Morning Star Farms.
Devi, S. M., Balachandar, V., Lee, S. I., & Kim, I. H. (2014). An outline of meat consumption in the
Indian population-A pilot review. Korean journal for food science of animal resources, 34(4),
507.
FAO (Food and Agriculture Organization of the United Nations). (2016). FAOSTAT food balance
sheets.
FAO, I., & UNICEF. (2018). WFP and WHO (2017) The State of Food Security and Nutrition in the
World 2017: Building Resilience for Peace and Food Security. Rome: Food and Agriculture
Organization of the United Nations (FAO).
Ferdowsian, H. R., & Barnard, N. D. (2009). Effects of plant-based diets on plasma lipids. The
American journal of cardiology, 104(7), 947-956.
Gensler, S., Hinz, O., Skiera, B., & Theysohn, S. (2012). Willingness-to-pay estimation with choice-
based conjoint analysis: Addressing extreme response behavior with individually adapted
designs. European Journal of Operational Research, 219(2), 368-378.
Gerber, P. J., Steinfeld, H., Henderson, B., Mottet, A., Opio, C., Dijkman, J., ... & Tempio, G.
(2013). Tackling climate change through livestock: a global assessment of emissions and
mitigation opportunities. Food and Agriculture Organization of the United Nations (FAO).
Graça, J., Calheiros, M. M., & Oliveira, A. (2015). Attached to meat?(Un) Willingness and intentions
to adopt a more plant-based diet. Appetite, 95, 113-125.
Graça, J., Oliveira, A., & Calheiros, M. M. (2015). Meat, beyond the plate. Data-driven hypotheses
for understanding consumer willingness to adopt a more plant-based diet. Appetite, 90, 80-90.
83
84
Graça, J., Truninger, M., Junqueira, L., & Schmidt, L. (2019). Consumption orientations may support
(or hinder) transitions to more plant-based diets. Appetite, 140, 19-26.
Graber, S., Narayanan, T., Alfaro, J., & Palit, D. (2018). Solar microgrids in rural India: Consumers'
willingness to pay for attributes of electricity. Energy for Sustainable Development, 42, 32-43.
Green, R. F., Joy, E. J., Harris, F., Agrawal, S., Aleksandrowicz, L., Hillier, J., ... & Haines, A.
(2018). Greenhouse gas emissions and water footprints of typical dietary patterns in
India. Science of the total environment, 643, 1411-1418.
Greene, W. H., & Hensher, D. A. (2003). A latent class model for discrete choice analysis: contrasts
with mixed logit. Transportation Research Part B: Methodological, 37(8), 681-698.
Grover, S. (2019, January 09). What happens when plant-based 'meat' is cheaper than the real thing?
Retrieved from https://www.treehugger.com/green-food/what-happens-when-plant-based-meat-
cheaper-real-thing.html
Heller, M. C., & Keoleian, G. A. (2018). Beyond Meat’s Beyond Burger Life Cycle Assessment: A
detailed comparison between.
Hocquette, A., Lambert, C., Sinquin, C., Peterolff, L., Wagner, Z., Bonny, S. P., ... & Hocquette, J. F.
(2015). Educated consumers don't believe artificial meat is the solution to the problems with the
meat industry. Journal of Integrative Agriculture, 14(2), 273-284.
Hoek, A. C., Luning, P. A., Weijzen, P., Engels, W., Kok, F. J., & De Graaf, C. (2011). Replacement
of meat by meat substitutes. A survey on person-and product-related factors in consumer
acceptance. Appetite, 56(3), 662-673.
Hole, A. R. (2007). A comparison of approaches to estimating confidence intervals for willingness to
pay measures. Health economics, 16(8), 827-840.
84
85
Howarth, R. W., Anderson, D. B., Cloern, J. E., Elfring, C., Hopkinson, C. S., Lapointe, B., ... &
Walker, D. (2000). Issues in ecology: Nutrient pollution of coastal rivers, bays, and seas.
Hu, F. B. (2003). Plant-based foods and prevention of cardiovascular disease: an overview. The
American journal of clinical nutrition, 78(3), 544S-551S.
Humane Society International. (2018, March 28). First collaboration to promote Clean Meat in India
takes off. Retrieved from https://www.hsi.org/news-media/laboratory-grown-meat-india-032818/
INCCA. (2010). India: greenhouse gas emissions, 2007.
IIPS, I. (2017). National Family Health Survey (NFHS-4), 2015–16: India. Mumbai: International
Institute for Population Sciences.
Jin, W., Jiang, H., Liu, Y., & Klampfl, E. (2017). Do labeled versus unlabeled treatments of
alternatives’ names influence stated choice outputs? Results from a mode choice study. PloS
one, 12(8), e0178826.
Joshi, V., & Kumar, S. (2016). Meat Analogues: Plant based alternatives to meat products- A review.
International Journal of Food Fermentation and Technology. 5.
Jha, D. K. (2019, March 20). Poultry prices surge by up to 50% on rising demand, lower production.
Retrieved from https://www.business-standard.com/article/economy-policy/poultry-prices-surge-
by-up-to-50-on-rising-demand-lower-production-119032000967_1.html
Jose, A. P., Shridhar, K., & Dorairaj, P. (2018). Diet, Nutrition and Cardiovascular Disease: The Role
of Social Determinants. doi:10.16943/ptinsa/2018/49451
Juan, W., Yamini, S., & Britten, P. (2015). Food intake patterns of self-identified vegetarians among
the US population, 2007-2010. Procedia Food Science, 4, 86-93.
85
86
Kulkarni, P. (2019, February 18). You could get to eat 'clean' meat in India by next year. Retrieved
from https://economictimes.indiatimes.com/news/politics-and-nation/you-could-get-to-eat-green-
meat-in-india-by-next-year/articleshow/68043360.cms
Lancsar, E., & Louviere, J. (2008). Conducting discrete choice experiments to inform healthcare
decision making. Pharmacoeconomics, 26(8), 661-677.
Lea, E. J., Crawford, D., & Worsley, A. (2006). Public views of the benefits and barriers to the
consumption of a plant-based diet. European journal of clinical nutrition, 60(7), 828.
Liao, T. F. (1994). Interpreting probability models: Logit, probit, and other generalized linear
models (No. 101). Sage.
Lin, D. (2018, July 06). Is Laboratory-Grown Meat Vegan? Retrieved from
https://www.thoughtco.com/laboratory-grown-meat-is-not-vegan-127673
Lynch, J., & Pierrehumbert, R. (2019). Climate impacts of cultured meat and beef cattle. Frontiers in
Sustainable Food Systems, 3, 5.
Macdonald, B., & Vivalt, E. (2017). Effective strategies for overcoming the naturalistic heuristic:
Experimental evidence on consumer acceptance of “clean” meat.
Machovina, B., Feeley, K. J., & Ripple, W. J. (2015). Biodiversity conservation: The key is reducing
meat consumption. Science of the Total Environment, 536, 419-431.
Mangham, L. J., Hanson, K., & McPake, B. (2009). How to do (or not to do)… Designing a discrete
choice experiment for application in a low-income country. Health policy and planning, 24(2),
151-158.
Market Research Future. (2019, April 09). Meat Substitutes Market Size Worth USD 9.25 Billion by
2023 at 9.4% CAGR: Concern for Animal Welfare to Impact Meat Substitutes Market. Retrieved
86
87
from https://www.globenewswire.com/news-release/2019/04/09/1799883/0/en/Meat-Substitutes-
Market-Size-Worth-USD-9-25-Billion-by-2023-at-9-4-CAGR-Concern-for-Animal-Welfare-to-
Impact-Meat-Substitutes-Market.html
Mattick, C. S., Landis, A. E., Allenby, B. R., & Genovese, N. J. (2015). Anticipatory life cycle
analysis of in vitro biomass cultivation for cultured meat production in the United
States. Environmental science & technology, 49(19), 11941-11949.
McFadden, D. (2001). Economic choices. American economic review, 91(3), 351-378.
Mearns, R. (1996). When livestock are good for the environment: Benefit-sharing of environmental
goods and services.
Miele, M. (2016). Killing animals for food: How science, religion and technologies affect the public
debate about religious slaughter. Food Ethics, 1(1), 47-60.
Minocha, S., Makkar, S., Swaminathan, S., Thomas, T., Webb, P., & Kurpad, A. V. (2019). Supply
and demand of high quality protein foods in India: Trends and opportunities. Global Food
Security, 23, 139-148.
Mishra, S., Xu, J., Agarwal, U., Gonzales, J., Levin, S., & Barnard, N. D. (2013). A multicenter
randomized controlled trial of a plant-based nutrition program to reduce body weight and
cardiovascular risk in the corporate setting: the GEICO study. European journal of clinical
nutrition, 67(7), 718.
Montoya, J. M., Donohue, I., & Pimm, S. L. (2018). Planetary boundaries for biodiversity:
implausible science, pernicious policies. Trends in ecology & evolution, 33(2), 71-73.
87
88
Mugera, A., Burton, M., & Downsborough, E. (2017). Consumer preference and willingness to pay
for a local label attribute in Western Australian fresh and processed food products. Journal of
Food Products Marketing, 23(4), 452-472.
Naqvi, S. M. K., & Sejian, V. (2011). Global climate change: role of livestock. Asian Journal of
Agricultural Sciences, 3(1), 19-25.
Nelson, D. (2018, September 27). $5 lab-grown burger could be ready by 2021. Retrieved from
https://geneticliteracyproject.org/2018/09/28/5-lab-grown-burger-could-be-ready-by-2021/
Nijdam, D., Rood, T., & Westhoek, H. (2012). The price of protein: Review of land use and carbon
footprints from life cycle assessments of animal food products and their substitutes. Food
policy, 37(6), 760-770.
Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes in latent
class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural
equation modeling: A multidisciplinary Journal, 14(4), 535-569.
OECD (2018), "Fish and seafood", in OECD-FAO Agricultural Outlook 2018-2027, OECD
Publishing, Paris, https://doi.org/10.1787/agr_outlook-2018-11-en.
OECD (2019), Meat consumption (indicator). doi: 10.1787/fa290fd0-en (Accessed on 02 July 2019)
Patel, A. (2017, November 13). Vegetarian and vegan 'meats' are more popular than ever, but are they
good for you? Retrieved from https://globalnews.ca/news/3850562/vegetarian-meat-healthy/
Pathak, S. (2019, June 25). No Drips, No Drops: A City Of 10 Million Is Running Out Of Water.
Retrieved from https://www.npr.org/sections/goatsandsoda/2019/06/25/734534821/no-drips-no-
drops-a-city-of-10-million-is-running-out-of-water
88
89
Pimentel, D., & Pimentel, M. (2003). Sustainability of meat-based and plant-based diets and the
environment. The American journal of clinical nutrition, 78(3), 660S-663S.
Patnaik, I. (2019, April 06). The Congress's NYAY plan is just a palliative. Retrieved from
https://www.hindustantimes.com/analysis/the-congress-s-nyay-plan-is-just-a-palliative/story-
URG0UcQ9KQHONp9V7oCwOM.html. The data in this article was sourced from the CMEI
Consumer Pyramids database.
Pingali, P. L., Aiyar, A., Abraham, C. M., & Rahman, A. (2019). Transforming food systems for a
rising India. Cham, Switzerland: Palgrave Macmillan.
Poore, J., & Nemecek, T. (2018). Reducing food’s environmental impacts through producers and
consumers. Science, 360(6392), 987-992.
Post, M. J. (2012). Cultured meat from stem cells: Challenges and prospects. Meat science, 92(3),
297-301.
Ramamurthy, R. (2019, April 26). Indian Government Grants Over $600,000 to Cell-Based Meat
Research. Retrieved from https://www.gfi.org/2019-04-26
Rosegrant, M. W., Agcaoili-Sombilla, M. C., & Perez, N. D. (1995). Global food projections to
2020: Implications for investment (Vol. 5). Diane Publishing.
Rowland, M. P. (2018, April 02). Demand For Meat Is Growing Rapidly in India. This Could Impact
All Of Us. Retrieved from
https://www.forbes.com/sites/michaelpellmanrowland/2017/12/17/india-meat-
increase/#748d926233b1
89
90
Reuters. (2019, June 18). Scientists shocked by Arctic permafrost thawing 70 years sooner than
predicted. Retrieved from https://www.theguardian.com/environment/2019/jun/18/arctic-
permafrost-canada-science-climate-crisis
Ritchie, H., Reay, D., & Higgins, P. (2018a). Sustainable food security in India—Domestic
production and macronutrient availability. PloS one, 13(3), e0193766.
Ritchie, H., Reay, D. S., & Higgins, P. (2018b). Potential of Meat Substitutes for Climate Change
Mitigation and Improved Human Health in High-Income Markets. Frontiers in Sustainable Food
Systems,2. doi:10.3389/fsufs.2018.00016
Ritchie, H., Reay, D. S., & Higgins, P. (2018c). The impact of global dietary guidelines on climate
change. Global environmental change, 49, 46-55.
Ritchie, H., & Roser, M. (2019). Meat and Seafood Production & Consumption. Retrieved from
https://ourworldindata.org/meat-and-seafood-production-consumption
Robinson, T. P., & Pozzi, F. (2011). Mapping supply and demand for animal-source foods to
2030. Animal Production and Health Working Paper, 2, 1-154.
Rogelj, J., Shindell, D., Jiang, K., Fifita, S., Forster, P., Ginzburg, V., ... & Mundaca, L. (2018).
Mitigation pathways compatible with 1.5 C in the context of sustainable development.
Ryan, M., & Farrar, S. (2000). Using conjoint analysis to elicit preferences for health
care. Bmj, 320(7248), 1530-1533.
Sadler, M. J. (2004). Meat alternatives—market developments and health benefits. Trends in Food
Science & Technology, 15(5), 250-260.
Salter, A. M. (2013). Impact of consumption of animal products on cardiovascular disease, diabetes,
and cancer in developed countries. Animal Frontiers, 3(1), 20-27.
90
91
Sathyamala, C. (2018). Meat-eating in India: Whose food, whose politics, and whose rights?. Policy
Futures in Education, 1478210318780553.
Schlereth, C., & Skiera, B. (2016). Two new features in discrete choice experiments to improve
willingness-to-pay estimation that result in SDR and SADR: Separated (adaptive) dual
response. Management Science, 63(3), 829-842.
Schösler, H., De Boer, J., & Boersema, J. J. (2012). Can we cut out the meat of the dish?
Constructing consumer-oriented pathways towards meat substitution. Appetite, 58(1), 39-47.
Sethi, S., Tyagi, S. K., & Anurag, R. K. (2016). Plant-based milk alternatives an emerging segment
of functional beverages: a review. Journal of food science and technology, 53(9), 3408-3423.
Shen, J. (2009). Latent class model or mixed logit model? A comparison by transport mode choice
data. Applied Economics, 41(22), 2915-2924.
Shepon, A., Eshel, G., Noor, E., & Milo, R. (2018). The opportunity cost of animal based diets
exceeds all food losses. Proceedings of the National Academy of Sciences, 115(15), 3804-3809.
Slade, P. (2018). If you build it, will they eat it? Consumer preferences for plant-based and cultured
meat burgers. Appetite, 125, 428-437.
Smith, P., Martino, D., Cai, Z., Gwary, D., Janzen, H., Kumar, P., ... & Scholes, B. (2007). Policy
and technological constraints to implementation of greenhouse gas mitigation options in
agriculture. Agriculture, Ecosystems & Environment, 118(1-4), 6-28.
Snowball, J. D., & Willis, K. G. (2011). Interview versus self-completion questionnaires in discrete
choice experiments. Applied Economics Letters, 18(16), 1521-1525.
91
92
Sonoda, Y., Oishi, K., Chomei, Y., & Hirooka, H. (2018). How do human values influence the beef
preferences of consumer segments regarding animal welfare and environmentally friendly
production?. Meat science, 146, 75-86.
Sorvino, C. (2018, January 30). Tyson Invests In Lab-Grown Protein Startup Memphis Meats,
Joining Bill Gates And Richard Branson. Retrieved from
https://www.forbes.com/sites/chloesorvino/2018/01/29/exclusive-interview-tyson-invests-in-lab-
grown-protein-startup-memphis-meats-joining-bill-gates-and-richard-branson/#6d555b3f3351
Specht, L. (2019, February 13). An analysis of culture medium costs and production volumes for
cell-based meat. Retrieved from https://www.gfi.org/files/sci-tech/clean-meat-production-
volume-and-medium-cost.pdf
Spiller, A., & Nitzko, S. (2015). 12. Peak meat: the role of meat in sustainable
consumption. Handbook of research on sustainable consumption, 192.
Spratt, D., Dunlop, I., & Schellnhuber, F. H. J. (2018). What lies beneath: the understatement of
existential climate risk. Breakthrough (Nafional Centre for Climate Restorafion).
Springmann, M., Clark, M., Mason-D’Croz, D., Wiebe, K., Bodirsky, B. L., Lassaletta, L., ... &
Jonell, M. (2018a). Options for keeping the food system within environmental
limits. Nature, 562(7728), 519.
Springmann, M., Wiebe, K., Mason-D'Croz, D., Sulser, T. B., Rayner, M., & Scarborough, P.
(2018b). Health and nutritional aspects of sustainable diet strategies and their association with
environmental impacts: a global modelling analysis with country-level detail. The Lancet
Planetary Health, 2(10), e451-e461.
Staples, J. (2018). Appropriating the cow: beef and identity politics in contemporary India.
92
93
Stehfest, E., Bouwman, L., Van Vuuren, D. P., Den Elzen, M. G., Eickhout, B., & Kabat, P. (2009).
Climate benefits of changing diet. Climatic change, 95(1-2), 83-102.
Steinfeld, H., Gerber, P., Wassenaar, T. D., Castel, V., Rosales, M., Rosales, M., & de Haan, C.
(2006a). Livestock's long shadow: environmental issues and options. Food & Agriculture Org..
Steinfeld, H., Wassenaar, T., Jutzi, S. (2006b). Livestock production systems in developing countries:
status, drivers, trends. Rev. Sci. Tech. Off. Int. Epiz. 25, 505–516
Steffen, W., Rockström, J., Richardson, K., Lenton, T. M., Folke, C., Liverman, D., ... & Donges, J.
F. (2018). Trajectories of the Earth System in the Anthropocene. Proceedings of the National
Academy of Sciences, 115(33), 8252-8259.
Stephens, N., Di Silvio, L., Dunsford, I., Ellis, M., Glencross, A., & Sexton, A. (2018). Bringing
cultured meat to market: Technical, socio-political, and regulatory challenges in cellular
agriculture. Trends in food science & technology, 78, 155-166.
Swaminathan, S., Vaz, M., & Kurpad, A. V. (2012). Protein intakes in India. British Journal of
Nutrition, 108(S2), S50-S58.
Taufik, D. (2018). Prospective “warm-glow” of reducing meat consumption in China: Emotional
associations with intentions for meat consumption curtailment and consumption of meat
substitutes. Journal of Environmental Psychology, 60, 48-54.
Train, K. E. (2009). Discrete choice methods with simulation. Cambridge university press.
Tuomisto, H. L., and Teixeira de Mattos, M. J. (2011). Environmental impacts of cultured meat
production. Environ. Sci. Technol. 45, 6117–6123. doi: 10.1021/es200130u
Tuso, P. J., Ismail, M. H., Ha, B. P., & Bartolotto, C. (2013). Nutritional update for physicians: plant-
based diets. The Permanente Journal, 17(2), 61.
93
94
USDA. (2019, January 4). FoodData Central Search Results. Retrieved from
https://fdc.nal.usda.gov/fdc-app.html#/food-details/173756/nutrients
Van der Weele, C., & Driessen, C. (2013). Emerging profiles for cultured meat; ethics through and as
design. Animals, 3(3), 647-662.
Van der Werf, G. R., Morton, D. C., DeFries, R. S., Olivier, J. G., Kasibhatla, P. S., Jackson, R. B., ...
& Randerson, J. T. (2009). CO 2 emissions from forest loss. Nature geoscience, 2(11), 737.
Veldwijk, J., Lambooij, M. S., de Bekker-Grob, E. W., Smit, H. A., & de Wit, G. A. (2014). The
effect of including an opt-out option in discrete choice experiments. PloS one, 9(11), e111805.
Verbeke, W., Pérez-Cueto, F. J., de Barcellos, M. D., Krystallis, A., & Grunert, K. G. (2010).
European citizen and consumer attitudes and preferences regarding beef and pork. Meat
science, 84(2), 284-292.
Verbeke, W., Sans, P., & Van Loo, E. J. (2015). Challenges and prospects for consumer acceptance
of cultured meat. Journal of Integrative Agriculture, 14(2), 285-294.
Vermeulen, S. J., Campbell, B. M., & Ingram, J. S. (2012). Climate change and food systems. Annual
review of environment and resources, 37.
Vetter, S. H., Sapkota, T. B., Hillier, J., Stirling, C. M., Macdiarmid, J. I., Aleksandrowicz, L., ... &
Smith, P. (2017). Greenhouse gas emissions from agricultural food production to supply Indian
diets: Implications for climate change mitigation. Agriculture, ecosystems & environment, 237,
234-241.
Wang, Y., & Chan, P. (2017). Animal Mistreatment in Business: Ethical Challenges and
Solutions. International Business Research, 10(5), 159-168.
94
95
Wedel, M., & Kamakura, W. A. (2002). Introduction to the special issue on market
segmentation. Intern. J. of Research in Marketing, 19, 181-183.
Wellesley, L., Froggatt, A., & Happer, C. (2015). Changing climate, changing diets: Pathways to
lower meat consumption.
West, P. C., Gerber, J. S., Engstrom, P. M., Mueller, N. D., Brauman, K. A., Carlson, K. M., ... &
Siebert, S. (2014). Leverage points for improving global food security and the
environment. Science, 345(6194), 325-328.
Whaley, S. E., Sigman, M., Neumann, C., Bwibo, N., Guthrie, D., Weiss, R. E., ... & Murphy, S. P.
(2003). The impact of dietary intervention on the cognitive development of Kenyan school
children. The Journal of nutrition, 133(11), 3965S-3971S.
Wild, F., Czerny, M., Janssen, A. M., Kole, A. P., Zunabovic, M., & Domig, K. J. (2014). The
evolution of a plant-based alternative to meat. From niche markets to widely accepted meat
alternatives. Agro Food Industry Hi-Tech, 25(1), 45-49.
Willett, W., Rockström, J., Loken, B., Springmann, M., Lang, T., Vermeulen, S., ... & Jonell, M.
(2019). Food in the Anthropocene: the EAT–Lancet Commission on healthy diets from
sustainable food systems. The Lancet, 393(10170), 447-492.
Williams, P. (2007). Nutritional composition of red meat. Nutrition & Dietetics, 64, S113-S119.
World Resources Institute, Greenhouse Gas Emissions Over 165 Years. (2019, March 22). Retrieved
from https://www.wri.org/resources/data-visualizations/greenhouse-gas-emissions-over-165-
years
95
96
Yadavar, S. (2018, September 11). Most Indians Non Vegetarian; Men More Than Women, South &
East Most. Retrieved from https://www.indiaspend.com/most-indians-non-vegetarian-men-more-
than-women-south-east-most-66469/
Yip, C. S. C., Crane, G., & Karnon, J. (2013). Systematic review of reducing population meat
consumption to reduce greenhouse gas emissions and obtain health benefits: effectiveness and
models assessments. International journal of public health, 58(5), 683-693.
Yoo, H. I. (2019). LCLOGIT2: Stata module to estimate latent class conditional logit models.
Zacharias, N. (2019). Hungry Planet. Retrieved from https://eftp.co/hungryplanet
96