SUSDIET: Implementing sustainable diets in Europe Report for Defra on SRUC work on the SUSDIET project

Cesar Revoredo-Giha1 Faical Akaichi Klaus Glenk Luiza Toma

Neil Chalmers Montserrat Costa-Font

September 2018

1 Land Economy and Environment Research Group, SRUC (Scotland’s Rural College) King’s Buildings, West Mains Road, Edinburgh EH9 3JG, UK, Phone (Cesar Revoredo-Giha): (0131)535 4344, E-mail: cesar.revoredo@sruc.ac.uk. We would like to thank the SUSDIET partners and in particular Dr. Louis-Georges Soler (INRA), coordinator of the project, for their comments and suggestions to different versions of the material. In addition, we are grateful to the UK Department for Environment, Food and Rural Affairs (Defra) for funding the research. However, the views expressed in this report are solely those of the authors.

Table of Contents

Executive Summary ....................................................................................... 4

I. Introduction .......................................................................................... 9

II. Task 2.1 – Substitution and complementarity of attributes on meat .... 11

II.1 Summary ......................................................................................... 11

II.2 Objectives ....................................................................................... 11

II.3 Methods .......................................................................................... 12

II.4 Results ............................................................................................ 14

III. Task 2.2 – Substitution and complementarity of attributes on fruit ...... 18

III.1 Summary ......................................................................................... 18

III.2 Objectives ....................................................................................... 18

III.3 Methods .......................................................................................... 18

III.4 Results ............................................................................................ 21

IV. Task 2.3 – Substitution and complementarity on food categories ........ 25

IV.1 Summary ......................................................................................... 25

IV.2 Objectives ....................................................................................... 26

IV.3 Methods .......................................................................................... 26

IV.4 Results ............................................................................................ 27

V. Task 3.4 – Impact of nutritional information on food choices ............... 31

V.1 Summary ......................................................................................... 31

V.2 Objectives ....................................................................................... 31

V.3 Methods .......................................................................................... 32

V.4 Results ............................................................................................ 33

V.4.1 Aggregated analysis ...................................................................... 33

V.4.2 Disaggregated analysis ................................................................. 36

V.4.3 Econometric evaluation of change in demand ............................... 37

V.4.4 Econometric evaluation of change in sales ................................... 39

VI. Task 4.1 – Uptake of innovative sustainable food products ................. 44

VI.1 Summary ......................................................................................... 44

VI.2 Objectives ....................................................................................... 44

VI.3 Methods .......................................................................................... 44

VI.4 Results ............................................................................................ 46

VI.4.1 Analysis of trends ......................................................................... 46

VI.4.2 Top 5 firms introducing products .................................................. 48

VI.4.3 Success on the introduction of dairy products .............................. 49

VII. Task 4.2 – Changes in prices effects on nutrition and sustainability .... 53

VII.1 Summary ......................................................................................... 53

VII.2 Objectives ....................................................................................... 53

VII.3 Methods .......................................................................................... 53

VII.4 Results ............................................................................................ 55

VIII. Conclusions ....................................................................................... 59

IX. References .......................................................................................... 63

4

Executive Summary

1. This document reports the tasks carried out by the Scotland's Rural College (SRUC) as part of the project “Implementing sustainable diets in Europe” (SUSDIET)2. The purpose of this report is to provide evidence around consumer preferences for food, and to provide an assessment of the effectiveness of policy interventions promoting sustainable food choices.

2. The SUSDIET project defines "sustainable diets" as those: a) with a high-nutritional quality and health benefit in the context of an ageing European population; b) that mitigate their environmental impacts, especially in relation to climate change; and c) are acceptable and affordable to all, including low-income groups.

3. The project aims to address three major issues through providing empirical

evidence drawing together data sets and evidence across several European countries. First, it will generate insights into the sustainability of current diets, and examine the impacts of alternative diets, from an environmental, cost and health perspective. Second, it will improve the understanding of consumers' preferences for food and the trade-offs involved in food consumption decisions and how these impact on supply chain productivity and environmentally; and third, it will inform policy on sustainable consumption.

4. Within this EU project, SRUC contributed to the following tasks (task numbers

correspond to their number in the SUSDIET project):

(i) Task 2.1 - Choice-based experiment focused on the meat sector; (ii) Task 2.2 - Choice experiment focused on the fruit sector; (iii) Task 2.3 - Estimations of a food demand system for different groups

to investigate food substitutions; (iv) Task 3.4 - Understand the impact of nutritional information on

sustainable food choices (e.g. labelling schemes); (v) Task 4.1 - Analysis of factors affecting the uptake of innovative food

products that improve sustainability; (vi) Task 4.2 - Measurement of the impact that changes in prices have

on different sustainability dimensions.

2 The final SUSDIET report can be found at: https://www6.inra.fr/sustainable diets/Media/Images/Final-report-SUSDIET-octobre-2017. The overall SUSDIET report contains an extensive literature review for each of the tasks (which provides the research base underpinning each task in the project), the questions pursued and the methodological approach followed. They have not been repeated in this report to avoid repetition. Therefore, we encourage the readers to access the entire SUSDIET report. In addition, the website of the SUSDIET project contains nine newsletters which provide summaries of each research project. Further, the project has a number of summary presentations (e.g., at the European Association of Agricultural Economists in Parma in 2017), which are available from the authors upon request.

5

5. On Task 2.1 - two large choice experiments involving 1,211 and 1,206 primary shoppers of food products in the UK and Spain were conducted to assess consumers’ preferences and willingness to pay for five meat attributes (i.e., greenhouse gas emissions (GHG) from production, type of production (organic/not organic), origin, fat content and price) as well as to analyse whether consumers consider these attributes as complements, substitutes (or overlapping) or independent.

6. The results showed consumers in the UK and Spain were willing to pay a price

premium for sustainable, healthier and local beef mince. British consumers’ willingness to pay (WTP) was found to be significantly higher than Spanish consumers’ WTP for beef mince produced with low GHG emissions, local beef mince and low fat beef mince but it was significantly lower for organic beef mince. Furthermore, the results showed that if sustainable (i.e. organic) and health (i.e. low fat) claims were present on different beef mince packages, consumers, in both countries, were found to prioritize health and, hence, buy the healthier beef mince instead of the more sustainable beef mince. Nonetheless, consumers in both countries were found to perceive sustainable (i.e. organic) and health (i.e. low fat) claims as complements if both labels were displayed on the same package of beef mince. This implies that consumers are willing to pay an additional price premium for the co-presence of these attributes on the same product.

7. On Task 2.2 - a large choice experiment involving 1,232 shoppers was

conducted only in the UK to assess consumers’ preferences and willingness to pay for five strawberries attributes (greenhouse gas emissions (GHG) from production, type of production (organic/not organic), sweetness, sweetness and price) as well as to analyse whether consumers consider these attributes as complements, substitutes (overlapping) or independent.

8. The results showed that respondents preferred environmentally-friendly

strawberries (i.e., organic strawberries or strawberries produced with low GHG emissions) over conventional strawberries (i.e., non-organic strawberries or strawberries produced with moderate or high GHG emissions). They also revealed to prefer sweet and slightly juicy strawberries. Interestingly, respondents were found to perceive the attributes environmentally-friendly and sweetness as complements and the attributes environmentally-friendly and slightly juicy strawberries as independent. The results showed that respondents were willing to pay an additional premium (on the top of their premium for individual attributes) if these attributes are bundled together.

9. On Task 2.3 - a complete system of demand for food using micro-data for six

EU countries were estimated using the EASI demand system specification across countries. The final aim was to produce a set of price and income elasticities to be used in the simulations carried out in WP 4.2. This report only presents the results for the UK. The estimation used the 2012 Kantar Worldpanel for food and drink expenditures for Scotland (which were considered representation of the average UK). Moreover, the food and drink conditional elasticities estimated with the Kantar data were transformed into

6

unconditional elasticities using data from the Office for National Statistics published in “Family Spending”; therefore, the final elasticities relate to the whole household expenditure.

10. As regards the income elasticities, all the food and drink products were found

normal and there were no inferior goods. Moreover, increases of income bring some reallocation towards meat/fish, soft drinks, alcohol and food out of home. On the price elasticities results, it was also found that the demand responds significantly and negatively to changes in their own-price. Furthermore, the demands were price inelastic for most groups, except soft drinks, snacks and food out of home. Several relatively large cross price elasticities were found in the analysis, not all of them significant, though. Relationships of complementarities were found to be at least as common as the substitution one, which made adjustments to price changes complex, with changes to the whole diet.

11. On Task 3.4 – several analyses of the impact that the introduction of the traffic

lights nutritional system (FPTL) had over consumers’ purchases were carried out. The literature on FPTL, which is mostly qualitative, found that they are easy to understand by consumers’ but there is a lack of empirical evaluation about their effectiveness. Given the data availability only since 2006, the analyses used the fact that Tesco and Morrisons, two major supermarkets only introduced FPTLs after August 2012 (all the other retailers did so in 2005 and 2006).

12. Four methods were used to test the impact of the FPTL:

Method 1, which was based on the analysis of aggregated market shares (time series), found that the introduction of traffic lights did not produced any structural change (before and after the introduction of FPTL) in any category. Method 2, which consisted of the analysis of individual purchases, did not show a strong change towards healthier products (defined based on a Kantar data attribute) except in soft drinks but that cannot be attributed to traffic lights as major manufacturers producers (e.g., were not using TFPL at that time. Method 3 was an econometric evaluation of the introduction of traffic lights and showed they had mild effects towards healthier purchases and only for some categories (breakfast cereals, cheddar cheese and potato products). Method 4 was an econometric analysis of product sales before and after the introduction of traffic lights showed only a slight effect towards reducing sales of “unhealthy” products; moreover, in most of the cases the estimated parameters were not statistically significant.

13. Potential issues are that traffic lights, whilst informative, might not be an

effective way to discriminate amongst healthy products and standard products in some categories category e.g., sweets, cheese where even healthier versions have red lights on several nutrients. For other categories (e.g., vegetables or fruits) their presence is not really important. Nevertheless, a

7

side effect of these labels that is quite important is that they encourage producers to reformulate their products towards healthier versions.

14. On Task 4.1 - the effectiveness of the policy option available for the promotion

of sustainable food choices, namely the introduction of new products with healthy and sustainable characteristics was evaluated. Mintel’s Global New Product Development together with Kantar Worldpanel data were used for the analyses. Three analyses were carried out as part of this task in order to provide an overview of the introduction of new product activities and also the acceptance of new products by consumers: (1) trends analysis on the introduction of new products with healthy and sustainable attributes; (2) an analysis of the major firms introducing new products with those attributes and (3) an analysis of the rate of success of new products considering the case of dairy products.

15. The number of launched new products with sustainable and health attributes

were shown to be increasing in most of the categories (organic products was an exception). Retailers were found to be particularly important in the introduction of products and the number of categories, and in most of the cases, they were on the top 5 product launchers.

16. On the analysis of dairy product success it was found that new products and

new varieties have less probability of success. Also, cheese and yoghurt had less probability of succeeding (maybe due to variety seeking attitudes in consumers) but this is offset if the product was a premium product. Private label products had greater rates of success. Some health and sustainability claims became significant when interacting with specific categories.

17. On Task 4.2 – the effect of a carbon consumption tax, which is one way to

reduce climate emissions associated with food consumption on greenhouse gas emissions and nutrition, was estimated. Ways to reduce GHG emissions associated with food consumption have become particularly pertinent issues given recent warnings that the planet recently has experienced its hottest year (2016). This study used consumption elasticities data, nutrient data and GHG emission data supplied by SUSDIET WP1 to estimate the impact of an ad-valorem tax and carbon consumption taxes on GHG emissions and nutrient intake. This information was combined in a model using Excel Visual Basic for Applications.

18. As regards the revenues from the simulated taxes, two further scenarios were

considered: one was that the collected revenues were kept by the Government (uncompensated) and another was that the revenues were returned to consumers in the way of subsidies to foods. These subsidies were inversely proportional the carbon emission of the food products.

19. The results suggest that the carbon consumption tax scenarios would reduce

GHG emissions by a greater quantity relative to the ad-valorem tax scenario. However, the intake of important nutrients will also decrease under these scenarios, showing a trade off between nutrition and environment. Therefore, creating an environmentally sustainable and nutritious diet through taxation is

8

challenging and requires compromise between the nutrition and environmental sustainability. The results for the compensated case were qualitatively the same as the uncompensated case, except that they implied smaller changes due to the fact that the changes in consumption were smaller.

9

Report for Defra on SRUC work on the SUSDIET project

I. Introduction Over the past 40 years, food consumption patterns have changed dramatically in industrialised countries due to increases in supply; reduction in the cost of food; changes in demographics, lifestyle; and significant increases in choice and convenience. All of these can lead to overconsumption or poor dietary choices, which in turn, raise a whole range of environmental, economic, social and health concerns. There is now broad agreement that current patterns of consumption in developed countries are fundamentally unsustainable. Associated multi-dimensional problems that result from such consumption threaten the functioning of the entire food system in the future, as well as the health prospects of a large cross section of the population in such countries. There is the need for new research and approaches to promote sustainable diets in Europe. This stems from two major shortcomings of the current response of the food system to the sustainability problem. The first relates to the lack of integration of the sustainability dimensions (economic, health and environment) when analysing the problem and seeking for solutions. Although the integration of environmental and dietary health objectives for food policy has started in some countries, the process remains in its infancy. The second shortcoming relates to the effectiveness of existing policies and measures in delivering significant changes in food consumption which is not fully understood. This is related to the complexity of the process by which consumers make decisions by weighing perceived “gains and losses” derived from the adoption of alternative diets. In the aforementioned context, the “Implementing sustainable diets in Europe” (SUSDIET) project has aimed to contribute to the existing evidence/literature by addressing the following research questions:

How do diets need to shift to enable consumers to consume healthy, nutritious food with an improved environmental impact? How different are these resulting diets from current diets?

What drivers are particularly important for achieving changes in consumption patterns?

What are the costs and benefits in terms of wellbeing (e.g., quality of life, satisfaction) for those consumers who comply with dietary, nutritional and environmental goals?

How will the project findings inform policy making across Governments, businesses and civil society

The SUSDIET project defines “sustainable diets” as those: a) with a high-nutritional quality and health benefit in the context of an ageing European population; b) that mitigate their environmental impacts, especially in relation to

10

climate change; and c) are acceptable and affordable to all, including low-income groups. The purpose of this document is to report the results of the tasks carried out by Scotland’s Rural College (SRUC) within the SUSDIET project. In turn, this report provides evidence around consumer preferences for food, and provides an assessment of the effectiveness of policy interventions promoting sustainable food choices. This will form part of an evidence base used to inform future Defra policy. These tasks contributed to two objectives: 1. To identify the major barriers preventing consumers making

sustainable dietary choices and the tasks comprise (task numbers correspond to their number in the SUSDIET project):

Task 2.1 – Choice-based experiment focused on the meat sector.

Task 2.2 – Choice experiment focused on the fruit sector.

Task 2.3 – Estimations of a food demand system for different groups to investigate food substitutions.

2. To analyse current policy approaches (Task 3.4) and interventions

(Tasks 4.1 and 4.2) addressing the uptake of sustainable food choices and gather evidence to inform policy on opportunities to encourage consumers’ dietary behaviour to be more sustainable:

Task 3.4 –Understand the impact of nutritional information on sustainable food choices (e.g. labelling schemes).

Task 4.1 – Analysis of factors affecting the uptake of innovative food products that improve sustainability.

Task 4.2 – Measurement of the impact that changes in prices have on different sustainability dimensions.

The structure of the report is as follows: for each one of the six tasks, a summary of the activities is presented followed by their objectives, methods and results. Finally, a summary of the main conclusions are presented.

11

II. Task 2.1 – Substitution and complementarity of attributes on meat II.1 Summary

II.2 Objectives The choice experiments were conducted in the UK and Spain in order to: (a) Assess consumers preferences and willingness to pay for different levels of four meat attributes, i.e. greenhouse gas emissions (low/moderate/high), type of production (organic/conventional), origin (local/national/imported) and fat content (low/moderate/high); (b) Explore whether consumers perceive the aforementioned attributes as complements, substitutes (overlapping) or independent; and (c) Assess the similarities and differences between British and Spanish consumers’ preferences and willingness to pay (WTP) for sustainable and health attributes as well as the trade-offs they may make when they are presented with these attributes at the same time (e.g., bundles of attributes).

Two large choice experiments were conducted in Spain and the UK to assess consumers’ preferences and willingness to pay for five meat attributes (greenhouse gas emissions (GHG) from production, type of production (organic/not organic), origin, fat content and price) as well as to analyse whether consumers consider these attributes as complements, substitutes (or overlapping) or independent. The results showed that consumers in the UK and Spain were willing to pay a price premium for sustainable, healthier and local beef mince. British consumers’ willingness to pay (WTP) was found to be significantly higher than Spanish consumers’ WTP for beef mince produced with low GHG emissions, local beef mince and low fat beef mince but it is significantly lower for organic beef mince. Furthermore, the results showed that if sustainable (i.e. organic) and health (i.e. low fat) claims are present on different beef mince packages, consumers in both countries were found to prioritize health and, hence, buy the healthier beef mince instead of the more sustainable beef mince. Nonetheless, consumers in both countries were found to perceive sustainable (i.e. organic) and health (i.e. low fat) claims as complements if both are displayed on the same package of beef mince. This implies that consumers are willing to pay an additional price premium for the co-presence of these attributes on the same product.

12

II.3 Methods The data were collected in the summer of 2016 through a national, web-based survey conducted in the UK and Spain. The initial design of the choice experiment and the questionnaire were developed and revised based on the input provided from a small pilot survey of 110 respondents in each country. The final versions of the surveys were administered by a market research company using its panel of meat and meat products consumers. A total of 1,211 and 1,206 primary shoppers of food products in the UK and Spain completed the survey. In both countries, the sample was required to be representative of the population in terms of gender, age, employment status and geographical area of the country. The quality of the data was checked after finishing the collection and all the ineligible observations, e.g., respondent who spent less than 10 seconds to complete each choice set, were discarded and replaced by eligible responses from new respondents. In each country, respondents were presented with a series of choice sets each including four hypothetical beef mince alternatives described in terms of five attributes: (1) level of GHG emissions: i.e., low (5.9Kg of CO2e per 500g of beef mince), moderate (19.1Kg of CO2e per 500g of beef mince), high (32.2Kg of CO2e per 500g of beef mince); (2) type of production, i.e., organic/not organic; (3) origin, i.e., local, national, imported, (4) fat content, i.e., low (3g per 100g serving of beef mince), moderate (12g per 100g serving of beef mince), high (21g per 100g serving of beef mince); and (5) price, i.e., £1.50, £3.00, £4.50, £6.00 for the UK survey and 2.30€, 3.10€, 3.90€, 4.70€ for the Spanish survey). The choice of the non-monetary attributes was based on a literature review of similar studies and the results obtained from two pilot studies conducted in the UK and Spain (interviewing 110 respondents in each country). The choice of the price levels was based on the real market prices of beef mince in each country. Given all the attributes’ levels, a full factorial design of 216 (3*2*3*3*4) profiles was generated. Since presenting participants with 216 profiles would be time consuming and cognitively challenging, we used the Ngene Software to generate a Bayesian D-optimal design with a minimum number of choice sets that allow a robust estimation of all main- and two-way interaction effects. The Bayesian D-optimal design was obtained after 25,000 iterations with 500 Halton draws per iteration, achieving a Db-error of 0.11 and 0.15 for the design of the choice experiments conducted in the UK and Spain, respectively. The obtained design consisted of 36 choice sets of four alternatives each (i.e., three beef mince alternatives plus the opt-out alternative). To make the choice task cognitively easier for respondents, the design was blocked in four blocks (i.e., 9 choice sets per respondent). In the choice task, respondents were successively shown 9 different choice sets and were repeatedly asked to choose the alternative they prefer most. In addition to collecting information on consumers’ choices, the online survey was also used to collect information on respondents’ socio-demographics, purchasing habits as well as their attitudes towards issues related with the attributes considered in the study.

13

As for the analysis of the collected data, we estimated a random parameter logit (RPL) model in WTP space. Thus, the results displayed in Table 1 represent respondents’ estimated WTP for the attributes levels as well as all the two-way interactions of the levels of the different attributes. Table 1: Estimated means and standard deviations of respondents’ willingness to pay

UK Spain

Mean Std. deviation Mean Std. deviation

Random Parameters Low GHG 0.46 *** 0.497 *** 0.23 *** -0.430 *** Moderate GHG 0.17 ** 0.584 *** 0.22 *** 0.031

High GHG (baseline) - - - - Organic 0.34 *** -0.779 *** 1.06 *** 1.198 *** Not organic (baseline) - - - - Local 0.84 *** -0.783 *** 0.45 *** 0.569 *** National 0.33 *** 0.628 *** 0.32 *** -0.511 *** Imported (baseline) - - - - Low fat 1.57 *** 1.474 *** 0.73 *** -0.933 *** Moderate fat 0.71 *** -0.633 *** 0.47 *** 0.705 *** High fat (baseline) - - - - Low GHG * Organic 0.01 -0.173 ** 0.27 *** 0.335 *** Low GHG * Local -0.05 -0.055 0.09 ** -0.018

Low GHG * National -0.25 *** 0.056 -0.15 ** -0.100

Low GHG * Low fat -0.14 0.016 0.08 ** -0.383 *** Low GHG * Moderate fat 0.23 *** -0.027 0.03 -0.137

Moderate GHG * Organic 0.04 -0.445 *** 0.19 ** 0.057

Moderate GHG * Local 0.27 *** 0.237 0.10 0.044

Moderate GHG * National 0.24 *** -0.558 *** -0.02 0.075

Moderate GHG * Low fat 0.19 -0.004 0.12 ** 0.194

Moderate GHG * Moderate Fat -0.17 -0.103 -0.04 0.198 ** Organic * Local 0.30 *** -0.332 *** 0.41 *** 0.209

Organic * National -0.06 -0.088 -0.14 -0.007

Organic * Low fat 0.13 ** -0.043 0.20 *** 0.228 ** Organic * Moderate fat -0.13 0.181 ** 0.03 -0.148

Local * Low fat 0.21 ** -0.108 0.15 *** -0.387 *** Local * Moderate fat -0.25 *** 0.118 -0.02 -0.095

National * Low fat -0.25 *** 0.229 ** 0.04 -0.398 *** National * Moderate fat 0.15 ** 0.229 ** 0.00 0.001

Non-random parameters No choice option -2.92 *** - -3.479 *** -

Likelihood (Conditional logit) -12668.78 -13290.30 Likelihood (Random parameter logit) -11039.37 -11600.19 Adjusted rho-squared 0.27 0.23

*** (**) Statistically significant at 1% (5%) level Note: GHG stand for greenhouse gas

Preference heterogeneity is revealed through the estimated standard deviations, which indicate how the valuation of the entire sample spreads around the estimated means. The results of the estimated standard deviations are displayed in Table 1. The RPL model extends the standard conditional logit model by allowing one or more of the parameters in the model to be randomly

14

distributed and the unobserved factors to be correlated over time (McFadden and Train 2000). All the parameters (i.e. all the main- and the two-way interactions effects) were assumed to be normally distributed. Since the attributes considered in this study have different units of measurement, comparing (e.g., ranking the attributes in terms of preferences) respondents’ preferences for these attribute is inappropriate. The appropriate way to compare British and Spanish respondents’ preferences for the different attributes is to calculate the marginal rate of substitution (MRS). When the price is included as the denominator in the ratio calculation, the MRS can be interpreted as marginal WTP. The estimated WTP displayed in Table 1 represents the price premium that average respondent is willing to pay for the corresponding level of attribute (e.g. organic) relative to his/her willingness to pay for the reference level (baseline) of the same attribute (e.g. not organic). II.4 Results All the estimations were conducted using the software R with 500 Halton draws to simulate the random parameters. All the estimated RPL models show significant improvement in fit when tested against the conditional logit models (see Table 1). The results show that British and Spanish respondents’ preferences for the five attribute are quite similar (i.e. the most preferred level of each attribute for respondents in both country is the same). In both countries, consumers of meat were found to prefer beef mince produced with low GHG emissions over beef mince produced with moderate or high GHG emissions. Furthermore, Spanish and British consumers were found to prefer organic beef mince over non-organic beef mince and local beef mince over national or imported beef mince. They were also found to be more likely to buy beef mince with lower fat content than beef mince with moderate or high fat content. The negative and significant sign of the “No choice option” coefficient shows that respondents preferred to buy beef mince than to opt out and choose the no-choice option. These results concur with the findings of previous studies (Onozaka and McFadden (2011); Kallas and Gil, 2012; Akaichi et al., 2017; Hung et al, 2017) that showed that there is demand for environmentally-friendly, local and more healthy food products. The results from this study and previous studies on similar topics show that the superiority of a food product in terms of sustainability, locality and healthiness could be used to differentiate them and, hence, increase their consumption, which in turn can improve the sustainability and healthiness of human diet. As mentioned, the results show that respondents’ preferences in both countries are similar in terms the most preferred level of each attribute. Nonetheless, British and Spanish respondents were found to differ in terms of the amount of the price premiums they are willing to pay for the attributes levels as well as how they perceive the interaction between these attributes levels. For instance, the results show that respondents in the UK were found to be willing to pay a price premium for the non-monetary attributes in the following decreasing order:

15

(1st) beef mince with low fat content, (2nd) local beef mince, (3rd) beef mince with low GHG emissions from production, and (4th) organic beef mince. For Spanish respondents the order of preferences is quite different and is as follows: (1st) organic beef mince, (2nd) beef mince with low fat content, (3rd) local beef mince, and (4th) beef mince with low GHG emissions from production. It is clear from these results that British consumers prioritise the healthiness of the beef mince over its sustainability; however, Spanish consumers are willing to pay more for organic beef mince than beef mince with low fat content. Furthermore, British respondents’ price premiums were found to be significantly higher than Spanish respondents’ price premiums for the attributes levels: low GHG emissions (100%), local (87%), low fat content (115%) and moderate fat content (51%). Nonetheless, Spanish respondents’ price premium for organic beef mince was found to be 212% higher than British respondents’ price premium for the same attribute. Respondents’ price premium for the other attribute levels was found to be not statistically different across countries. Similar to findings from previous studies (Onozaka and McFadden (2011); Meas et al., 2014; Akaichi et al., 2017) respondents in both countries were found to be willing to pay a price premium for beef mince with low GHG emissions. It is noteworthy that although several studies reveal that consumers are willing to pay a price premium for food products with GHG emissions, labels informing about the carbon footprint of food products are rarely used by producers, processors and retailers in both countries. The results displayed in Table 1 show that the estimated standard deviations of the main effects are all statistically significant (except the standard deviation corresponding to the attribute level “moderate GHG” in the case of Spain). This means that respondents’ preferences for the attribute levels are heterogeneous. Since we assumed that the distributions of the parameters corresponding to the non-monetary attributes are all normal, the proportion of the sample having positive or negative valuation on each attribute can also be inferred (Train 2003). For example, we found that 67% (81%) of respondents in the UK (Spain) preferred organic beef mince over non-organic beef mince. The rest of respondents preferred to buy non-organic beef mince. Most of the studies that assessed consumers’ preferences for food attributes using a choice experiment assumed that all the two-way interactions between attributes are insignificant. Thus, consumers are assumed to perceive the attributes as independent. The advantages of this approach are: big reduction in the number of choice sets that respondents has to evaluate, considerably lower sample size, and massive improvement in the estimation of the choice model in terms of complexity and estimation time. Nonetheless, this assumption can lead to significant bias of the estimated preferences and WTP, if consumers actually perceive the attributes considered in a choice experiment as complements (i.e. positive and statistically significant estimated interactions) or substitutes (i.e. negative and statistically significant estimated interactions). The resulting bias does not affect only the estimated WTP but also the results from subsequent analysis such as cost benefit analysis, where the estimated WTP is generally used as a proxy of the component “benefit”. As

16

aforementioned, in this study, we generated the choice experiment design in a way that all the two-way interactions can be estimated and how consumers perceive the interactions between the attributes can be determined. The results are also displayed in Table 1. For purposes of clarity and simplicity, only the results corresponding to the estimated two-way interactions between the most preferred levels (i.e. low GHG emissions, organic, local and low fat content) of the non-monetary attributes will be interpreted. The results show that out of 18 estimated interactions 10 and 9 are statistically significant in the case of UK and Spain, respectively. This proves that not assuming that the interactions between the non-monetary attributes are equal to zero (as assumed in most of the studies that used choice experiment) was a wise decision of this study. Spanish consumers were found to perceive the attribute levels constituting the beef mince bundles “low GHG emissions & organic”, “low GHG emissions & local”, “low GHG emissions & low fat”, “organic & local”, “organic & low fat” and “local & low fat” as complements. This implies that Spanish consumers are willing to pay an additional price premium for beef mince labelled as “organic” and “local” on the top of their price premium for the individual attributes. For example, the results in Table 1 show that Spanish consumers are willing to pay a price premium of 0.45€ for beef mince labelled as “local” and a price premium of 1.06€ for beef mince labelled as organic. The results from the estimated interactions show that if the beef mince is labelled as organic and local at the same time, consumers are willing to pay an additional (third) premium of 0.41€ for the bundle. Therefore, their total price premium for beef mince labelled as organic and local is 2.02€ (i.e. 0.45€ + 1.06€ + 0.41€). It also implies that labelling organic beef mince sold in local market as “local” is expected to increase consumers price premiums for the product from 1.06€ to 2.02€. The results also show that Spanish consumers are also willing to pay an additional price premium for beef mince that is sustainable (i.e. organic or produced with low GHG emissions) and healthy (i.e. low fat content) at the same time. Similarly, British consumers were found to consider the attribute levels constituting the beef mince bundles “organic & local”, “organic & low fat” and “local & low fat” as complements. Particularly, they are willing to pay additional premiums of 0.30€, 0.13€ and 0.21€ for these bundles, respectively. This is interesting because it shows that sellers of organic beef mince can benefit economically from labelling their product as local or healthier, obviously, if their product is actually local and healthier. In difference with Spanish consumers, British consumers were found to perceive the attributes levels constituting the bundles “low GHG emissions & organic”, “low GHG emissions & local” and “low GHG emissions & low fat” as independent. Therefore, the total price premium that British consumers are willing to pay for these bundles is simply the sum of their price premiums for the individual attribute levels (i.e., 0.80€ for the bundle “low GHG emissions & organic”, 1.30€ for the bundle “low GHG emissions & local” and 2.03€ for the bundle “low GHG emissions & low fat”).

17

Conclusions

Consumers in the UK and Spain are willing to pay a price premium for sustainable, healthier and local beef mince.

Claims presenting information on these attributes should be used to increase the demand for environmentally-friendly and healthier beef mince. Obviously, this approach is appropriate as long as it is economically viable (i.e. a cost benefit analysis in needed).

British consumers’ price premium is significantly higher than Spanish price premiums for beef mince produced with low GHG emissions, local beef mince and low fat beef mince. Spanish consumers’ price premium for organic beef mince is three times higher than British consumers’ price premium for the same product.

Therefore, if sustainable and health claims are present on different beef mince packages (i.e. not on the same product), British consumers are more likely to buy the healthier beef mince (i.e. low-fat beef mince) than the more sustainable beef mince (i.e. organic or beef mince produced with low GHG emissions). Spanish consumers are more likely to purchase healthier beef mince than beef mince produced will low GHG emissions, however, they are more likely to buy organic beef mince than healthier beef mince.

Spanish consumers perceive the attributes low GHG emissions, organic, local and low fat content as complements. As a result, they are willing to pay an additional price premium for the co-presence of any combination of these attributes.

British consumers perceive the attributes organic, local and low fat as complements. However, they were found to perceive these attributes and the attribute low GHG emissions from production as independent. Thus, they are not willing to pay a price premium (in addition to their premiums for the individual attributes) for the co-existence of the attribute low GHG emissions and any one of the attributes organic, local or low fat.

Finally, the results showed that assuming that consumers perceive food attributes as independent (as it is assumed in most of studies on similar topics) is misleading and is likely to result in biased results. Despite that designing choice experiments that allow for the estimation of all the two-way interactions between attributes is time and economically challenging, the improvement in the results validity and prediction accuracy in worth the time and effort spent to cope with that additional hurdle.

18

III. Task 2.2 – Substitution and complementarity of attributes on fruit III.1 Summary

III.2 Objectives The objectives of the study were to: (a) Assess consumers preferences and willingness to pay for different levels of four strawberries attributes i.e., greenhouse gas emissions (low/moderate/high), type of production (organic/conventional), sweetness (intensely sweet/sweet/slightly sweet) and juiciness (juicy/slightly juicy/firm), and (b) Find out whether consumers perceive these attributes as complements, substitutes (overlapping) or independent. III.3 Methods The data were collected through a national, web-based survey conducted in the UK during the summer of 2016. The initial design of the choice experiment and the questionnaire were developed and revised based on input from a small sample of 100 respondents in each country. The final version of the survey was administered by a market research company using its panel of strawberries consumers. A total of 1,232 shoppers of food products in the UK completed the survey. The sample was required to be representative of the population in terms

A large choice experiment was conducted in the UK to assess consumers’ preferences and willingness to pay for five strawberries attributes i.e. (greenhouse gas emissions (GHG) from production, type of production (organic/not organic), sweetness, sweetness and price as well as to analyse whether consumers consider these attributes as complements, substitutes (overlapping) or independent. The results showed that respondents were found to prefer environmentally-friendly strawberries (i.e., organic strawberries or strawberries produced with low GHG emissions) over conventional strawberries (i.e., non-organic strawberries or strawberries produced with moderate or high GHG emissions). They also revealed to prefer the strawberries to be sweet and slightly juicy. Interestingly, respondents were found to perceive the attributes environmentally-friendly and sweet strawberries as complements and the attributes environmentally-friendly and slightly juicy strawberries as independent. Thus, the results showed that respondents were willing to pay an additional premium (on the top of their premium for the individual attributes) when they found the attributes together.

19

of gender, age, employment status and geographical area of the country. The quality of the data was checked after finishing the collection and all the ineligible observations (e.g., respondent who spent less than 10 seconds to complete each choice set) were discarded and replaced by eligible responses from new respondents. Participants were presented with a series of choice sets each including three hypothetical strawberries alternatives described in terms of five attributes: (1) level of GHG emissions (i.e., low (120g of CO2e per 1kg of strawberries), moderate (420g of CO2e per 1kg of strawberries), high (680g of CO2e per 1kg of strawberries)), (2) type of production (i.e., organic/not organic), (3) sweetness (i.e. intensely sweet, sweet, slightly sweet), (4) juiciness (i.e. juicy, slightly juicy, firm), and (5) price (i.e., £1.50, £2.70, £3.80, £5.00). The choice of the non-monetary attributes was based on a literature review of similar studies and the results obtained from a pilot studies (interviewing 100 respondents). The choice of the price levels was based on the real market prices of strawberries in the UK. Given all the attributes’ levels, a full factorial design of 216 (3x2x3x3x4) profiles was generated. Since presenting participants with 216 profiles would be time consuming and cognitively challenging, we used the Ngene Software to generate a Bayesian D-optimal design with a minimum number of choice sets that allow a robust estimation of all main- and two-way interaction effects. The Bayesian D-optimal design was obtained after 25,000 iterations with 500 Halton draws per iteration, achieving a Db-error of 0.18. The obtained design consisted of 36 choice sets of four alternatives each (i.e., three strawberries alternatives plus the opt-out alternative). To make the choice task cognitively easier for respondents, the design was blocked in four blocks (i.e., 9 choice sets per respondent). In the choice task, respondents were successively shown 9 different choice sets and were repeatedly asked to choose the alternative they prefer most. In addition to collecting information on consumers’ choices, the online survey was also used to collect information on respondents’ socio-demographics, purchasing habits as well as their attitudes towards issues related with the attributes considered in the study. The analysis of the data was carried out using a RPL model in WTP space, as in the case for mince meat. The results displayed in Table 2 represent respondents’ estimated WTP for the attributes levels as well as all the two-way interactions of the levels of the different attributes. Preference heterogeneity is revealed through the estimated standard deviations (also showed in Table 2), which indicate how the valuation of the entire sample spreads around the estimated means. The RPL model extends the standard conditional logit model by allowing one or more of the parameters in the model to be randomly distributed and the unobserved factors to be correlated over time (McFadden and Train, 2000). All the parameters (i.e. all the main- and the two-way interactions effects) were assumed to be normally distributed.

20

Table 2: Estimated means and standard deviations of respondents’ willingness to pay

Estimated parameters Mean Standard deviation

Random Parameters Low GHG 0.42 *** 0.279 ***

Moderate GHG -0.11 0.456 ***

High GHG (baseline) - -

Organic 0.29 *** -0.835 ***

Not organic (baseline) - -

Sweet 0.22 *** 0.239 ***

Slightly sweet -0.15 *** 0.637 ***

Intensely sweet (baseline) - -

Juicy 0.19 *** 0.396 ***

Slightly juicy 0.28 *** -0.231 ***

Firm (baseline) - -

Non-random parameters

Low GHG * Organic 0.18 *** -0.086

Low GHG * Sweet 0.07 -0.039

Low GHG * Slightly sweet 0.08 ** -0.015

Low GHG * Juicy -0.16 ** 0.018

Low GHG * Slightly juicy -0.12 -0.006

Moderate GHG * Organic -0.13 -0.043

Moderate GHG * Sweet -0.07 0.002

Moderate GHG * Slightly sweet -0.16 *** 0.137 **

Moderate GHG * Juicy 0.31 *** -0.213 ***

Moderate GHG * Slightly juicy 0.36 *** -0.049

Organic * Sweet 0.09 -0.064

Organic * Slightly sweet 0.00 -0.067

Organic * Juicy 0.14 *** 0.093 **

Organic * Slightly juicy -0.08 -0.043

Sweet * Juicy 0.01 -0.069

Sweet * Slightly juicy -0.02 0.014

Slightly sweet * Juicy 0.06 -0.198 ***

Slightly sweet * Slightly juicy 0.05 -0.244 ***

No choice option -4.72 ***

Initial Likelihood -11527.94

Final Likelihood -9822.51

Adjusted Rho-squared 0.36 Note: GHG stand for greenhouse gas. *** (**) Statistically significant at 1% (5%) level

Since the attributes considered in this study have different units of measurement, comparison was made using the marginal rate of substitution (MRS). When the price is included as the denominator in the ratio calculation, the MRS is interpreted as marginal WTP. The estimated WTP displayed in Table 2 represents the price premium that average respondent is willing to pay for the corresponding level of attribute (e.g. organic) relative to his/her

21

willingness to pay for the reference level (baseline) of the same attribute (e.g. not organic). III.4 Results All the estimations were conducted using the software R with 500 Halton draws to simulate the random parameters. The estimated RPL model show significant improvement in fit when tested against the conditional logit models (see Table 2). The results show that consumers prefer strawberries produced with low GHG emissions over strawberries produced with moderate or high GHG emissions. This agrees with studies by Onozaka and McFadden (2011); Kallas and Gil, 2012; Akaichi et al., 2017; Hung et al, 2017 that showed that consumers are willing to pay a price premium for environmentally-friendly food products. They were also found to prefer organic strawberries over non-organic strawberries. Furthermore, the results show that consumers prefer sweet strawberries over intensely sweet or slightly sweet strawberries and juicy strawberries over slightly juicy and firm strawberries. The results from this study and previous studies on similar topics suggest that the superiority of a food product in terms of environmental sustainability can be used to differentiate them and, hence, increase their consumption. Organic strawberries are available in several of the major retail stores in the UK. However, to the best of the authors’ knowledge, fresh strawberries with carbon footprint claims are not yet available for consumers in the UK, despite the extensive literature evidencing the existence of a potential market for them. Consumers’ confusion in interpreting and understanding carbon labels (Gadema and Oglethorpe, 2011) and attitude-behaviour gap (Hartikainen et al., 2011) were found to be major barriers for the purchase of food products produced with low GHG emissions. Therefore, research work is still needed to minimise the effect of these barriers. Interestingly, the results show that consumers are willing to pay a higher price premium (£0.42) for strawberries with low GHG emissions than organic strawberries (£0.29). The results are similar to those found in the assessment of consumers’ preferences and WTP for beef mince attributes. We found that British consumers’ price premiums for beef mince produced with lower GHG emissions and organic beef mince are £0.39 and £0.28, respectively. Therefore, more research work is needed to find out why consumers value more the attribute low GHG emissions that organic. Is it, for example, because consumers in the UK are aware of the fact that organic food are generally produced with higher GHG emissions than non-organic foods? This seems to be a plausible explanation because 62% of respondents in this study did not agree with the statement that “the production of organic strawberries emit less GHG emissions than the production of non-organic strawberries”. There is an extensive literature on the importance of taste in the choice of food products. Most of the studies found that taste is one of the main drivers of food choice decision. Front-of-package information on the taste of food products

22

(e.g. “Thin and crispy”, “Sweet and juicy”) is becoming a common practice used by producers and retailers to help consumers choosing the fruit with their preferred taste. One of the objectives of this study was to find out whether consumers in the UK positively value the front-of-package information on taste (i.e. the level of sweetness and juiciness of strawberries). The results displayed in Table 2 show that consumers do positively value this type of information. In fact, they were found to be willing to pay £0.22 and £0.37 more for sweet strawberries than intensely sweet and slightly sweet strawberries, respectively. As you can see their price premium for sweet strawberries is higher than their price premium for organic strawberries. Consumers’ price premium for sweet strawberries (£0.37) was found to be not significantly different from their price premium for strawberries with low GHG emissions (£0.42). The juiciness of the strawberries was also found to be an important attribute of strawberries for British consumers. The results show that they are willing to pay £0.28 and £0.19 more for juicy and slightly juicy strawberries, respectively, than for firm strawberries. It was also found that the price premium for slightly juicy strawberries is not significantly different from their price premium for organic strawberries. These results imply that consumers may trade off taste and environmental attributes in real market, especially that the real market premium for organic strawberries (around £1.50) is much higher than average consumers’ price premium (£0.29). In contrast to the case of mice meat of Task 2.1 the results show that consumers like strawberries to be sweet and slightly juicy, which can be also found in the labels used in retail stores when selling strawberries. The results displayed in Table 2 show that the estimated standard deviations of the main effects are all statistically significant. This means that respondents’ preferences for the attribute levels are heterogeneous. Since it was assumed that the distributions of the parameters corresponding to the non-monetary attributes are all normal, the proportion of the sample having positive or negative valuation on each attribute can also be inferred (Train 2003). For example, it was found that 64% of consumers preferred organic strawberries over non-organic strawberries and vice versa for the rest of respondents (36%). Most of the studies that assessed consumers’ preferences for food attributes using choice experiments assumed that all the two-way interactions between attributes are insignificant. Thus, consumers are assumed to perceive the attributes as independent. The advantages of this approach are: big reduction in the number of choice sets that respondents has to evaluate, considerably lower sample size, and massive improvement in the estimation of the choice model in terms of complexity and estimation time. Nonetheless, this assumption can lead to significant bias of the estimated preferences and WTP, if consumers perceive the attributes considered in a choice experiment as complements (i.e. positive and statistically significant estimated interactions) or substitutes (i.e. negative and statistically significant estimated interactions). The resulting bias does not affect only the estimated WTP but also the results from subsequent analysis such as cost benefit analysis, where the estimated WTP is generally used as a proxy of the component “benefit”. In this study the choice experiment design was generated in a way that all the two-way interactions can be

23

estimated and how consumers perceive the interactions between the attributes can be determined. The results are also displayed in Table 2. The results show that out of 18 estimated interactions 8 are statistically significant. This proves that it was a right decision in this study not assuming the independence of the non-monetary attributes as it is assumed in most of studies that have used choice experiment. The significant and positive interaction that was found for the co-presence of low GHG emissions and organic claims implies that the use of the two types of labels generated added positive effect (£0.18) to the combined main effects (£0.42 + £0.29) of the two attributes (complementarity effect). As shown the majority of consumers think that organic strawberries emit more GHG emissions than non-organic strawberries. Therefore, it seems that the co-presence of both claims (i.e. organic and low GHG emissions) on the same product can trigger and additional price premium that consumers are willing to pay for this bundle of attributes. Regarding the interactions between the environmental and taste attributes, the results show that consumers generally perceive these attributes as independent with some exception. For instance, the results show that consumers in the UK perceive the attributes organic and juicy as complement. This implies that their total price premium for strawberries labelled as organic and juicy is £0.62 (i.e. £0.29 + £0.19 + £0.14). The significant and negative interaction that was found for the co-presence of the labels “low GHG emissions” and “juicy” suggest that consumers might perceive the values of these two labels to be overlapping when these attributes are presented simultaneously. In other words, consumers are willing to discount their total price premium by £0.16 if the strawberries are labelled as juicy and produced with low GHG emissions. The results also show that consumers perceive the attributes levels forming the bundles “moderate GHG emissions & juicy” and “moderate GHG emissions & slightly juicy” as complements. Nonetheless, they were found to perceive the attribute levels moderate GHG emissions and slightly sweet as complements. As you can see, the values of the estimated interactions are statistically and economically significant. This shows that failing to estimate the interaction effects can significantly bias the results. Conclusions

Consumers are willing to pay a price premium for sustainable strawberries. They were also found to positively value front-of-package information on the sweetness and juiciness of strawberries. Overall, consumers prefer the strawberries to be sweet and slightly juicy.

This suggests that the demand for environmentally-friendly strawberries can be increased using environmental claims. Consumers seem to find carbon footprint claims difficult to understand and interpret. Therefore, more research work is still needed to determine the best claim design that minimise this problem and, hence, increase the effectiveness of the label to improve the demand for low-carbon footprint food products.

The results also showed that front-of-package information on the juiciness and sweetness of strawberries is an effective way to provide consumers

24

with information on strawberries taste and increase the demand for strawberries by consumers who consider as the main driver of their purchasing decision of strawberries.

Consumers consider the attributes low GHG emissions and organic to be complements. Therefore, their willingness to buy organic strawberries can be increased if they are produced with low GHG emissions and labelled as so.

Consumers were found to be willing to pay a price premium for strawberries that are labelled as juicy. Therefore, the demand for organic juicy strawberries can be improved if the strawberries are clearly labelled as having a juicy taste.

Finally, the results showed that assuming that consumers perceive food attributes as independent (as it is assumed in most of studies on similar topics) is misleading and is likely to result in biased results. Despite that designing choice experiments that allow for the estimation of all the two-way interactions between attributes is time and economically challenging, the improvement in the results validity and prediction accuracy in worth the time and effort spent to cope with that additional hurdle.

25

IV. Task 2.3 – Substitution and complementarity on food categories IV.1 Summary

The purpose of the task was to estimate in six EU countries a complete system of demand for food and drink using micro-data, the EASI specification and considering a similar specification across countries. This report presents the results for the UK. The results used the 2012 Kantar Worldpanel for food and drink expenditures for Scotland and they produced elasticities conditional to the households’ food and drink expenditure. These elasticities were transformed to unconditional elasticities using information from the Office for National Statistics published in “Family Spending”. As regards the income elasticities, all the food and drink products were found normal and there were no inferior goods. Moreover, increases of income bring some reallocation towards meat/fish, soft drinks, alcohol and food out of home. On the price elasticities results, it was also found that the demand responds significantly and negatively to changes in their own-price. Furthermore, the demands were price inelastic for most groups, except soft drinks, snacks and food out of home. Several relatively large cross price elasticities were found in the analysis, not all of them significant, though. Relationships of complementarities were found to be at least as common as the substitution one, which made adjustments to price changes complex, with changes to the whole diet.

The purpose of the task was to estimate in six EU countries a complete system of demand for food and drink using micro-data, the EASI specification and considering a similar specification across countries. This report presents the results for the UK. The results used the 2012 Kantar Worldpanel for food and drink expenditures for Scotland and they produced elasticities conditional to the households’ food and drink expenditure. These elasticities were transformed to unconditional elasticities using information from the Office for National Statistics published in “Family Spending”. As regards the income elasticities, all the food and drink products were found normal and there were no inferior goods. Moreover, increases of income bring some reallocation towards meat/fish, soft drinks, alcohol and food out of home. On the price elasticities results, it was also found that the demand responds significantly and negatively to changes in their own-price. Furthermore, the demands were price inelastic for most groups, except soft drinks, snacks and food out of home. Several relatively large cross price elasticities were found in the analysis, not all of them significant, though. Relationships of complementarities were found to be at least as common as the substitution one, which made adjustments to price changes complex, with changes to the whole diet.

26

IV.2 Objectives The objective of the task was to estimate in six EU countries a complete system of demand for food using micro-data and the EASI demand system. The results presented in this section correspond to the UK estimates. The estimated elasticities were later used in Task 4.2 for the simulation of the effects of introducing a carbon tax on consumption. IV.3 Methods The data used for the analysis was a cross-section for the year 2012 coming from the Kantar Worldpanel dataset for Scotland. The choice of the year of analysis was to match the analysis in other countries. The dataset comprised 1,516 households and all of them were observed at least 40 weeks of the year. 20 food and drink categories (including a residual one) were estimated. This product classification was linked to the categories used in WP 1 in the SUSDIET project. The logic behind the category was to enable further analysis on nutrition and environment. As it is well known, changes in unit values are due to only prices but also changes in the composition of the products (changes in quality). Therefore the unit values were adjusted using Cox and Wohlgenant (1986) method. The used demand system was the EASI system (Lewbel and Pendakur, 2009). It was estimated by iterated linear 3SLS without interactions. Whilst the final estimated model was linear in real expenditure, a polynomial of fifth degree was the starting point of the estimation. The estimation requires instruments (due to the endogeneity of the shares). These were socioeconomic variables such as economic class and the Scottish Index of Multiple Deprivation. The model was estimated using the ‘easi’ routine in R by Hoareau et al. (2013). The elasticities resulting from the estimation were “transformed” to unconditional elasticities using the formulas from Carpentier and Guyomard (2001), where the first stage elasticities were computed using an Almost Ideal Demand System (Deaton and Muellbauer, 1981) and data from the Office for National Statistics publication “Family Spending”.3

3 It is important to emphasise the need to compute unconditional elasticities for simulation purposes (i.e., computed considering the total expenditure, not just food and drink). Conditional elasticities (e.g., those considering the food and drink expenditure) may overestimate the response of the demand to changes in price and income.

27

IV.4 Results As regards the income elasticities (see Table 3), all the food and drink products were found normal (i.e., the income elasticities are between 0 and 1) and there were no inferior goods (i.e., goods with negative income elasticities). Table 3: Income elasticities

Categories Income

elasticities

Grains and grain-based products 0.2982

Vegetables and vegetable products 0.3088

Starchy roots, tubers, legumes, nuts and oilseeds 0.3447

Fruit, fruit products and fruit and vegetable juices 0.2962

Beef, veal and lamb 0.4320

Pork 0.4138

Poultry, eggs, other fresh meat 0.3820

Processed and other cooked meats 0.3749

Fish and other seafood 0.3214

Milk, dairy products and milk product imitates 0.2781

Cheese 0.3271

Sugar and confectionary and prepared desserts 0.3219

Soft drinks 0.4233

Animal fats 0.3083

Plant based fats 0.3018

Tea, coffee, cocoa, and drinking water 0.2921

Alcoholic beverages 0.5208

Composite dishes (animal and vegetable composite dishes) 0.3141

Snacks and other foods 0.4137

Residual category 0.3370

Food out of home 1.1520

Non-food 2.4040

The results from Table 3 show that an increase in income (ceteris paribus) increases the purchases of non-food goods and services (e.g., housing or goods for the household) and food out of home. Within the ‘food for consumption at home’ category, increases of income bring some reallocation of purchases towards meat/fish, soft drinks and alcohol. As regards the effect of changes in prices on the purchases of food and drink, Table 4 presents the Marshallian price elasticities and Table 5 the Hicksian price elasticities. The difference between these two types of elasticities is the income effect generated by a change in prices. For instance, if some prices decrease, this creates a wealth effect on consumers (as if their income had been increased). The Marshallian elasticities represent the full effect of a change in prices and they include the income and the price effect, whilst the Hicksian only include the price effect.

28

Table 4: Marshallian price elasticities (i.e., uncompensated demand price elasticities) Grains Vegetables Starchy Fruit, Beef, Pork Poultry, Processed Fish and Milk, dairy Cheese Sugar and Soft Animal Plant Tea, Alcoholic Composite Snacks Residual

Food Non

and and roots, fruit products veal and eggs, and other other products confectionary drinks fats based coffee, beverages dishes and category out of foods

grain-based vegetable tubers, and fruit lamb other cooked seafood and milk and fats cocoa, other home

products products legumes, and fresh meats product prepared and foods

nuts and vegetable meat imitates desserts drinking

oilseeds juices water

Grains and grain-based products -0.927 -0.017 0.015 -0.036 -0.193 -0.023 -0.031 0.020 -0.038 0.020 -0.027 0.051 0.141 -0.229 -0.125 0.082 -0.062 0.096 0.091 -0.003 0.068 -0.441

Vegetables and vegetable products 0.005 -0.476 -0.104 0.009 -0.215 -0.213 -0.168 -0.014 0.014 0.041 0.125 0.014 -0.084 0.040 -0.065 0.044 -0.082 0.095 -0.073 0.175 0.071 -0.456

Starchy roots, tubers, legumes, nuts and oilseeds 0.033 -0.092 -0.800 -0.005 -0.054 0.139 -0.039 -0.036 -0.130 -0.072 0.081 0.037 0.083 0.102 -0.072 0.064 -0.035 0.006 0.177 0.010 0.079 -0.509

Fruit, fruit products and fruit and vegetable juices -0.030 -0.005 -0.040 -0.771 -0.151 -0.052 -0.098 -0.046 -0.055 0.114 0.028 0.082 -0.080 -0.104 0.000 0.005 -0.123 0.099 0.027 0.002 0.068 -0.438

Beef, veal and lamb -0.029 -0.096 -0.011 -0.021 -0.509 0.028 0.055 0.010 0.126 0.019 -0.003 0.025 -0.038 0.050 0.078 0.022 -0.036 0.018 -0.059 0.074 0.099 -0.638

Pork 0.029 -0.087 0.091 0.015 0.020 -0.919 0.029 0.022 0.033 0.014 -0.002 0.013 0.050 -0.076 -0.056 0.016 -0.039 0.041 0.056 0.052 0.095 -0.611

Poultry, eggs, other fresh meat 0.014 -0.159 -0.033 -0.039 0.077 0.040 -0.826 0.103 -0.099 0.003 -0.118 0.052 -0.003 -0.114 -0.097 0.026 -0.046 0.119 -0.020 0.075 0.088 -0.564

Processed and other cooked meats 0.046 0.002 -0.034 -0.006 -0.011 0.022 0.104 -0.639 -0.184 0.017 0.073 -0.005 -0.041 0.288 -0.027 0.041 -0.059 -0.044 -0.077 -0.102 0.086 -0.554

Fish and other seafood 0.010 0.017 -0.062 0.000 0.090 0.018 -0.048 -0.082 -0.441 0.009 -0.086 0.034 0.006 0.076 -0.071 -0.015 -0.048 0.056 -0.036 0.049 0.074 -0.475

Milk, dairy products and milk product imitates 0.018 0.035 -0.127 0.097 -0.044 -0.048 -0.037 -0.017 -0.026 -0.889 0.120 0.036 0.038 -0.050 0.028 0.044 -0.076 -0.040 0.142 0.054 0.064 -0.411

Cheese 0.012 0.087 0.056 0.031 -0.026 -0.021 -0.074 0.040 -0.111 0.078 -0.977 0.037 -0.018 -0.069 -0.029 0.023 -0.037 0.011 0.046 -0.019 0.075 -0.483

Sugar and confectionary and prepared desserts 0.068 -0.030 0.006 0.126 -0.135 -0.190 0.017 -0.122 0.015 0.052 0.022 -0.921 -0.067 0.183 0.044 -0.049 -0.027 0.006 0.046 -0.052 0.074 -0.476

Soft drinks 0.104 -0.033 0.081 -0.001 -0.052 0.070 0.009 -0.013 0.026 0.060 -0.003 0.035 -1.115 0.020 0.107 -0.056 0.019 -0.010 0.021 -0.059 0.097 -0.626

Animal fats -0.023 0.022 0.038 -0.004 0.014 -0.052 -0.032 0.081 0.049 0.006 -0.030 0.050 0.002 -0.671 -0.103 0.028 -0.027 0.013 -0.103 0.039 0.071 -0.456

Plant based fats -0.001 -0.007 -0.013 0.017 0.025 -0.035 -0.023 -0.004 -0.035 0.023 -0.008 0.033 0.034 -0.088 -0.773 0.018 -0.018 0.036 0.047 0.036 0.069 -0.446

Tea, coffee, cocoa, and drinking water 0.046 0.029 0.038 0.015 -0.004 -0.004 0.006 0.015 -0.025 0.034 0.016 0.015 -0.066 0.040 0.022 -0.805 -0.049 -0.011 -0.071 0.055 0.067 -0.432

Alcoholic beverages 0.041 -0.052 0.032 -0.053 -0.024 -0.036 0.001 -0.018 -0.059 0.005 0.005 0.104 0.151 0.003 0.063 -0.035 -0.899 -0.056 -0.009 0.026 0.119 -0.770

Composite dishes 0.118 0.156 -0.037 0.125 -0.083 0.015 0.156 -0.136 0.136 -0.064 -0.038 0.017 -0.141 -0.055 0.117 -0.092 -0.163 -0.744 -0.172 -0.329 0.072 -0.464

Snacks and other foods 0.069 -0.020 0.121 0.043 -0.067 0.058 -0.003 -0.030 -0.031 0.097 0.054 0.058 0.014 -0.189 0.109 -0.052 -0.035 -0.013 -1.124 0.002 0.095 -0.611

Residual category 0.023 0.103 0.012 0.021 0.050 0.040 0.036 -0.047 0.054 0.043 -0.013 0.024 -0.054 0.065 0.067 0.058 -0.028 -0.063 -0.006 -0.923 0.077 -0.498

Food out of home 0.006 0.004 0.006 0.006 0.006 0.005 0.008 0.008 0.002 0.004 0.003 0.013 0.007 0.001 0.001 0.002 0.028 0.009 0.005 0.003 -1.417 -0.084

Non-food 0.044 0.030 0.036 0.040 0.033 0.027 0.049 0.049 0.016 0.030 0.021 0.091 0.040 0.009 0.007 0.015 0.161 0.061 0.031 0.019 0.196 -2.085

Note: Own price elasticities are in yellow; a red cross price elasticity indicate a substitution relationship between the categories and a green complementarity. The rows of the tables are the demands and the columns indicate the price with respect which the elasticity is computed.

29

Table 5: Hicksian price elasticities (i.e., compensated price elasticities) Grains Vegetables Starchy Fruit, Beef, Pork Poultry, Processed Fish and Milk, dairy Cheese Sugar and Soft Animal Plant Tea, Alcoholic Composite Snacks Residual

Food Non

and and roots, fruit products veal and eggs, and other other products confectionary drinks fats based coffee, beverages dishes and category out of foods

grain-based vegetable tubers, and fruit lamb other cooked seafood and milk and fats cocoa, other home

products products legumes, and fresh meats product prepared and foods

nuts and vegetable meat imitates desserts drinking

oilseeds juices water

Grains and grain-based products -0.924 -0.015 0.017 -0.033 -0.192 -0.022 -0.028 0.022 -0.037 0.023 -0.025 0.057 0.143 -0.229 -0.124 0.083 -0.057 0.101 0.092 -0.002 0.230 0.232

Vegetables and vegetable products 0.009 -0.473 -0.102 0.012 -0.214 -0.212 -0.166 -0.011 0.016 0.044 0.126 0.020 -0.083 0.041 -0.065 0.045 -0.077 0.100 -0.072 0.176 0.238 0.240

Starchy roots, tubers, legumes, nuts and oilseeds 0.037 -0.089 -0.797 -0.002 -0.052 0.140 -0.036 -0.033 -0.128 -0.069 0.082 0.043 0.085 0.102 -0.072 0.066 -0.030 0.011 0.178 0.011 0.266 0.268

Fruit, fruit products and fruit and vegetable juices -0.026 -0.003 -0.038 -0.767 -0.149 -0.050 -0.095 -0.044 -0.054 0.116 0.029 0.088 -0.078 -0.104 0.000 0.007 -0.119 0.103 0.028 0.003 0.229 0.231

Beef, veal and lamb -0.024 -0.093 -0.008 -0.016 -0.507 0.030 0.059 0.013 0.128 0.023 -0.001 0.034 -0.035 0.050 0.079 0.024 -0.029 0.024 -0.057 0.075 0.334 0.336

Pork 0.033 -0.084 0.094 0.019 0.021 -0.918 0.032 0.025 0.034 0.018 0.000 0.021 0.052 -0.076 -0.056 0.018 -0.033 0.047 0.057 0.053 0.320 0.322

Poultry, eggs, other fresh meat 0.019 -0.156 -0.030 -0.034 0.079 0.042 -0.823 0.106 -0.098 0.007 -0.116 0.059 -0.001 -0.114 -0.096 0.027 -0.040 0.125 -0.018 0.077 0.295 0.297

Processed and other cooked meats 0.051 0.004 -0.032 -0.002 -0.010 0.023 0.107 -0.636 -0.183 0.020 0.075 0.002 -0.039 0.288 -0.027 0.042 -0.053 -0.039 -0.076 -0.101 0.290 0.292

Fish and other seafood 0.013 0.019 -0.060 0.003 0.092 0.019 -0.046 -0.080 -0.440 0.012 -0.085 0.040 0.007 0.077 -0.071 -0.013 -0.043 0.061 -0.035 0.051 0.248 0.250

Milk, dairy products and milk product imitates 0.021 0.037 -0.125 0.100 -0.043 -0.047 -0.035 -0.015 -0.025 -0.887 0.121 0.041 0.039 -0.049 0.028 0.045 -0.071 -0.036 0.143 0.055 0.215 0.217

Cheese 0.015 0.090 0.058 0.034 -0.024 -0.019 -0.071 0.043 -0.110 0.081 -0.976 0.043 -0.016 -0.068 -0.029 0.025 -0.032 0.015 0.047 -0.018 0.253 0.255

Sugar and confectionary and prepared desserts 0.072 -0.027 0.009 0.129 -0.133 -0.189 0.020 -0.119 0.016 0.055 0.023 -0.914 -0.065 0.184 0.045 -0.047 -0.022 0.011 0.048 -0.050 0.249 0.251

Soft drinks 0.109 -0.030 0.084 0.004 -0.050 0.072 0.012 -0.010 0.027 0.064 -0.001 0.044 -1.113 0.021 0.108 -0.054 0.025 -0.004 0.023 -0.057 0.327 0.330

Animal fats -0.020 0.024 0.040 -0.001 0.016 -0.051 -0.030 0.084 0.050 0.009 -0.028 0.056 0.003 -0.670 -0.102 0.029 -0.022 0.017 -0.102 0.040 0.238 0.240

Plant based fats 0.003 -0.005 -0.011 0.020 0.027 -0.034 -0.020 -0.001 -0.034 0.026 -0.007 0.039 0.035 -0.087 -0.772 0.019 -0.013 0.041 0.049 0.037 0.233 0.235

Tea, coffee, cocoa, and drinking water 0.050 0.031 0.040 0.018 -0.003 -0.003 0.008 0.018 -0.024 0.036 0.018 0.020 -0.064 0.040 0.023 -0.804 -0.045 -0.007 -0.070 0.056 0.226 0.227

Alcoholic beverages 0.047 -0.048 0.035 -0.047 -0.021 -0.034 0.005 -0.013 -0.057 0.010 0.008 0.114 0.154 0.004 0.064 -0.032 -0.891 -0.048 -0.007 0.028 0.402 0.405

Composite dishes 0.122 0.158 -0.035 0.129 -0.082 0.016 0.158 -0.134 0.137 -0.061 -0.036 0.023 -0.140 -0.054 0.118 -0.091 -0.158 -0.739 -0.171 -0.328 0.243 0.245

Snacks and other foods 0.073 -0.017 0.124 0.047 -0.065 0.060 0.000 -0.027 -0.029 0.100 0.056 0.066 0.016 -0.188 0.110 -0.050 -0.028 -0.007 -1.122 0.004 0.320 0.322

Residual category 0.027 0.105 0.014 0.025 0.051 0.041 0.039 -0.044 0.055 0.046 -0.011 0.031 -0.052 0.066 0.068 0.059 -0.023 -0.058 -0.005 -0.921 0.260 0.262

Food out of home 0.011 0.007 0.007 0.010 0.006 0.005 0.009 0.009 0.004 0.008 0.005 0.020 0.007 0.002 0.002 0.004 0.024 0.014 0.005 0.004 -1.319 0.120

Non-food 0.096 0.063 0.067 0.089 0.051 0.043 0.082 0.084 0.032 0.073 0.042 0.181 0.061 0.019 0.016 0.035 0.219 0.125 0.048 0.036 1.083 -0.234

Note: Own price elasticities are in yellow; a red cross price elasticity indicate a substitution relationship between the categories and a green complementarity. The rows of the tables are the demands and the columns indicate the price with respect which the elasticity is computed.

30

The Marshallian elasticities are the ones used to simulate changes in demand due to changes in prices due to the fact that they provide the full effect and Hicksian elasticities are useful to classify products according to their substitution or complementarity. Thus, if the Hicksian cross price elasticity is positive that indicate that the products are substitutes and is they are negative they are said to be complement. The yellow cells in Tables 4 and 5 are the own price elasticities, red cells indicate substitutes and green complements. It was found that the demand categories respond significantly and negatively to changes in their own-price. Furthermore, the demands were price inelastic for most groups, except soft drinks, snacks and food out of home. Several relatively large cross price elasticities were found in the analysis, not all of them significant, though. Relationships of complementarity were found to be at least as common as the substitution ones, which made demand adjustments to price changes complex, i.e., involving changes in the entire diet.

31

V. Task 3.4 – Impact of nutritional information on food choices V.1 Summary

V.2 Objectives The objective of the task was to analyse policy instruments for the promotion of sustainable food choices and formulate policy recommendations. Food labels were seen as potential tools for that task. The topic of food labels has been partly addressed at the level of the European Union (e.g., European Commission, 2012) where they found that two-thirds

The purpose of the task was to analyse the impact that the introduction of the traffic lights nutritional system (FPTL) had over consumers purchases. The literature on FPTL, which is mostly qualitative, found that they are easy to understand by consumers but there is a lack of empirical evaluation about their effectiveness. Given the availability of the data since 2006, the evaluation was focused on the fact that Tesco and Morrisons, two major supermarkets started to use the FPTL since August 2012. Four methods were used to test the impact of the FPTL. The methods and their results were: Method 1 based on aggregated market shares (time series) found that the introduction of traffic lights did not produce any structural change in the market shares of the different studied categories (before and after the introduction of FPTL). Method 2 which analysed individual purchases did not show a strong change towards healthier products except in the case of soft drink purchases but that cannot be attributed to traffic lights as major manufacturers did not used traffic lights at that time. Method 3 was an econometric evaluation of the introduction of traffic lights after August 2008. It showed only mild effects towards healthier purchases and only for some categories (i.e., breakfast cereals, cheddar cheese and potato products). Method 4 was also an econometric analysis of products sales before and after the introduction of traffic lights showed only a slight effect reducing sales of “unhealthy” products but in most of the cases they were not statistically significant. Potential issues with traffic lights are that, whilst informative, they might not be an effective way to discriminate amongst healthy products and standard products in some categories category (e.g., sweets, cheese). For other categories (e.g., vegetables or fruits) their presence is not really important.

32

(67%) of EU citizens check food purchases to see if they have quality labels indicating specific characteristics; only one-fifth (22%) of those surveyed said that they always check for these labels; just under a half (45%) say that they do this sometimes and one-third (32%) of respondents never check. However, the Eurobarometer survey focused on labels related to food quality assurance schemes (e.g., from EU’s Protected Geographical Status schemes) and not nutritional labels. In fact, the literature on nutritional labels is limited and of a qualitative character where studies have highlighted how easy it is for consumers to understand them (Malam et al., 2009; Balcombe et al., 2010; Hieke and Wilczynski, 2011). A different approach by Sacks et al. (2009) focused on the impact of the introduction of nutritional traffic lights labels on aggregated purchases of two categories (selected ready meals and sandwiches). Their study only considered the short terms behaviour of consumers focusing on the percentage change in sales four weeks before and after traffic-light labels were introduced. In 2006, the UK Food Standards Agency (FSA) recommended food retailers and manufacturers in the UK the use of front-of-pack traffic-light labels (FPTL) on products in a range of categories. The labelling format recommended by the FSA consisted of four separate colour-coded lights indicating the level of fat, saturated fat, sugar and salt in the product. The main objective of this task was to analyse the impact that the introduction of the FPTL had over consumers’ purchases. V.3 Methods The data available for the analysis consisted of the Kantar Worldpanel for Scotland for the period 2006 to 2013. Due to this it was not possible to test the effect of the introduction of the FPTL in all the supermarkets. Therefore, the work on the task used the fact that Tesco and Morrisons, two major food retailers in the UK introduced FPTL after August 2012 and concentrated the analysis on households that were customers of those retailers (at least 50 per cent of their purchases came from those supermarkets). Eight food categories were considered for the analysis (number of available households in parenthesis):

Breakfast cereals (170 households),

Soft drinks (131 households),

Sweet biscuits (128 households),

Cheddar cheese (186 households),

Total cheese (191 households),

Frozen chips (102 households),

Frozen potato products (135 households)

Ready meals (168 households). Four analyses were carried out to measure the impact of the FPTLs: 1. Method 1 - An aggregated analysis, i.e., time series were created to

compare the market shares of healthy and standard products for private labels and branded products. If the FPTL would have had a positive effect

33

some positive change in the market share of healthy private labels products would have been expected after August 2012;

2. Method 2 - An analysis comparing households’ expenditure allocation

before and after August 2012 on healthy and standard products. This analysis was carried out for each one of the household in the sample. They were classified in terms of changes in their expenditure allocation. A chi-square test was used to test the hypothesis that the market share did not change between the two periods. The results then were classified according to the change (i.e., towards healthier or standard products).

3. Method 3 - Econometric evaluation of the introduction of traffic lights on

households demand for healthy and standard products and branded and private label for all the product categories. The demand for four product groups branded, private label and healthy and standard for each of the product categories were estimated. The methodology mixed Capps and Park (1997) demand for ready meals and the average treatment effect on treated subject (Capacci and Mazzocchi, 2011). Two periods were considered (0 and 1), i.e., before (0) and after (1) the introduction of TFPL. Then, the procedure consisted of: (a) To estimate the demand in period 0 and use those parameters to predict the demand in period 1 using period 1 data; (b) to estimate the demand in period 1 and estimate the demand in period 1; (c) to compute the difference in both estimates.

4. Method 4 - Econometric estimates of the impact of nutrition traffic lights on

the product sales. This analysis used nutritional information from Kantar in order to construct traffic light indicators. Three indicators were constructed: (a) Traffic lights were introduced as nutritional information (values for fats, saturated fats, sugar and salt); (b) As a categorical variable (i.e., green=0, amber=1, red= 2) for each: fats, saturated fats, sugar and salt; (c) As individual dummies: e.g., fats_red=1, otherwise=0. Then, cross section regressions were carried out for each product category, where the endogenous variable was the change in the sales of each product in the category before and after the introduction of the FPTL and the explanatory variables were the change in price, change in promotion, whether the product was branded or private label and the aforementioned traffic light variables.

V.4 Results V.4.1 Aggregated analysis Figures 1a to 1f show the results obtained when estimated the aggregated market share for each one of the product categories (i.e., private standard, branded standard, private label healthy and branded healthy). The dotted line in the figures indicates the point when the traffic lights were introduced. Figure 1a – Market share for breakfast cereals 2006-14

34

Source: Own elaboration based on Kantar Worldpanel data

Figure 1b – Market share for soft drinks 2006-14

Source: Own elaboration based on Kantar Worldpanel data

Figure 1c – Market share for sweet biscuits 2006-14

Source: Own elaboration based on Kantar Worldpanel data

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

2006

-1

2006

-3

2006

-5

2006

-7

2006

-9

2006

-11

2006

-13

2007

-2

2007

-4

2007

-6

2007

-8

2007

-10

2007

-12

2008

-1

2008

-3

2008

-5

2008

-7

2008

-9

2008

-11

2008

-13

2009

-2

2009

-4

2009

-6

2009

-8

2009

-10

2009

-12

2010

-1

2010

-3

2010

-5

2010

-7

2010

-9

2010

-11

2010

-13

2011

-2

2011

-4

2011

-6

2011

-8

2011

-10

2011

-12

2012

-1

2012

-3

2012

-5

2012

-7

2012

-9

2012

-11

2012

-13

2013

-2

2013

-4

2013

-6

2013

-8

2013

-10

2013

-12

2014

-1

2014

-3

2014

-5

2014

-7

2014

-9

2014

-11

2014

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

0.0

10.0

20.0

30.0

40.0

50.0

60.0

20

06

-1

20

06

-3

20

06

-5

20

06

-7

20

06

-9

20

06

-11

20

06

-13

20

07

-2

20

07

-4

20

07

-6

20

07

-8

20

07

-10

20

07

-12

20

08

-1

20

08

-3

20

08

-5

20

08

-7

20

08

-9

20

08

-11

20

08

-13

20

09

-2

20

09

-4

20

09

-6

20

09

-8

20

09

-10

20

09

-12

20

10

-1

20

10

-3

20

10

-5

20

10

-7

20

10

-9

20

10

-11

20

10

-13

20

11

-2

20

11

-4

20

11

-6

20

11

-8

20

11

-10

20

11

-12

20

12

-1

20

12

-3

20

12

-5

20

12

-7

20

12

-9

20

12

-11

20

12

-13

20

13

-2

20

13

-4

20

13

-6

20

13

-8

20

13

-10

20

13

-12

20

14

-1

20

14

-3

20

14

-5

20

14

-7

20

14

-9

20

14

-11

20

14

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

100.0

20

06

-1

20

06

-3

20

06

-5

20

06

-7

20

06

-9

20

06

-11

20

06

-13

20

07

-2

20

07

-4

20

07

-6

20

07

-8

20

07

-10

20

07

-12

20

08

-1

20

08

-3

20

08

-5

20

08

-7

20

08

-9

20

08

-11

20

08

-13

20

09

-2

20

09

-4

20

09

-6

20

09

-8

20

09

-10

20

09

-12

20

10

-1

20

10

-3

20

10

-5

20

10

-7

20

10

-9

20

10

-11

20

10

-13

20

11

-2

20

11

-4

20

11

-6

20

11

-8

20

11

-10

20

11

-12

20

12

-1

20

12

-3

20

12

-5

20

12

-7

20

12

-9

20

12

-11

20

12

-13

20

13

-2

20

13

-4

20

13

-6

20

13

-8

20

13

-10

20

13

-12

20

14

-1

20

14

-3

20

14

-5

20

14

-7

20

14

-9

20

14

-11

20

14

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

35

Figure 1d – Market share for cheese 2006-14

Source: Own elaboration based on Kantar Worldpanel data

Figure 1e – Market share for processed potato products 2006-14

Source: Own elaboration based on Kantar Worldpanel data

Figure 1f – Market share for prepared foods 2006-14

Source: Own elaboration based on Kantar Worldpanel data

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

20

06

-1

20

06

-3

20

06

-5

20

06

-7

20

06

-9

20

06

-11

20

06

-13

20

07

-2

20

07

-4

20

07

-6

20

07

-8

20

07

-10

20

07

-12

20

08

-1

20

08

-3

20

08

-5

20

08

-7

20

08

-9

20

08

-11

20

08

-13

20

09

-2

20

09

-4

20

09

-6

20

09

-8

20

09

-10

20

09

-12

20

10

-1

20

10

-3

20

10

-5

20

10

-7

20

10

-9

20

10

-11

20

10

-13

20

11

-2

20

11

-4

20

11

-6

20

11

-8

20

11

-10

20

11

-12

20

12

-1

20

12

-3

20

12

-5

20

12

-7

20

12

-9

20

12

-11

20

12

-13

20

13

-2

20

13

-4

20

13

-6

20

13

-8

20

13

-10

20

13

-12

20

14

-1

20

14

-3

20

14

-5

20

14

-7

20

14

-9

20

14

-11

20

14

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

20

06

-1

20

06

-3

20

06

-5

20

06

-7

20

06

-9

20

06

-11

20

06

-13

20

07

-2

20

07

-4

20

07

-6

20

07

-8

20

07

-10

20

07

-12

20

08

-1

20

08

-3

20

08

-5

20

08

-7

20

08

-9

20

08

-11

20

08

-13

20

09

-2

20

09

-4

20

09

-6

20

09

-8

20

09

-10

20

09

-12

20

10

-1

20

10

-3

20

10

-5

20

10

-7

20

10

-9

20

10

-11

20

10

-13

20

11

-2

20

11

-4

20

11

-6

20

11

-8

20

11

-10

20

11

-12

20

12

-1

20

12

-3

20

12

-5

20

12

-7

20

12

-9

20

12

-11

20

12

-13

20

13

-2

20

13

-4

20

13

-6

20

13

-8

20

13

-10

20

13

-12

20

14

-1

20

14

-3

20

14

-5

20

14

-7

20

14

-9

20

14

-11

20

14

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

20

06

-1

20

06

-3

20

06

-5

20

06

-7

20

06

-9

20

06

-11

20

06

-13

20

07

-2

20

07

-4

20

07

-6

20

07

-8

20

07

-10

20

07

-12

20

08

-1

20

08

-3

20

08

-5

20

08

-7

20

08

-9

20

08

-11

20

08

-13

20

09

-2

20

09

-4

20

09

-6

20

09

-8

20

09

-10

20

09

-12

20

10

-1

20

10

-3

20

10

-5

20

10

-7

20

10

-9

20

10

-11

20

10

-13

20

11

-2

20

11

-4

20

11

-6

20

11

-8

20

11

-10

20

11

-12

20

12

-1

20

12

-3

20

12

-5

20

12

-7

20

12

-9

20

12

-11

20

12

-13

20

13

-2

20

13

-4

20

13

-6

20

13

-8

20

13

-10

20

13

-12

20

14

-1

20

14

-3

20

14

-5

20

14

-7

20

14

-9

20

14

-11

20

14

-13

Pe

rce

nta

ge

s

Branded-standard Private label-standard Branded-healthier Private label-healthier

36

The results show that the standard products (branded and private label) represent most of the sales. Moreover, although there are some increasing trend on the healthier category (e.g., cheese), these cannot be attributed to the introduction of FPTL as the trends started well before the introduction of FPTL (this may speak well of the information campaigns towards healthier eating) V.4.2 Disaggregated analysis Table 6 presents the results of the disaggregated analysis. They show that the FPTL may have some effect but it depends on the category. Three groups can be considered based on the results: (a) Consumers are becoming increasingly aware to healthier products (i.e., soft drinks); (b) Households spend a more on standard products but show increasing proportion of healthier products (i.e., breakfast cereal, cheese, frozen potato products and ready meals) and (c) Health message have not had much effect (i.e., frozen chips and sweet biscuits). Based on the results from Table 6, the products from group (3) could be good candidates for reformulation as consumer information seems not to be working. Table 6: Results from the disaggregated analysis Hypothesis: No change in the expenditure allocation

No rejection Rejection

Healthier Standard Remain Change Change Remain standard

healthier towards towards Improved Remained

standard healthier (1) (2) (3) (4) (5) (6) (7) Breakfast cereals 0 97 0 3 0 70 0 % in group 0 57 0 2 0 41 0 Soft drinks 61 2 47 7 3 3 8 % in group 47 2 36 5 2 2 6 Sweet biscuits 0 6 0 0 0 0 122 % in group 0 5 0 0 0 0 95 Cheddar cheese 0 83 0 0 0 1 102 % in group 0 45 0 0 0 1 55 Total cheese 0 119 0 0 0 65 7 % in group 0 62 0 0 0 34 4 Frozen chips 0 81 0 0 0 0 21 % in group 0 79 0 0 0 0 21 Frozen potato products 0 123 0 0 0 12 0 % in group 0 91 0 0 0 9 0 Chilled prepared food 0 35 0 0 0 133 0 % in group 0 21 0 0 0 79 0

Source: Own elaboration based on Kantar Worldpanel data Notes: (1) Hypothesis cannot be rejected and the share under heathier is greater than the standard. (2) Hypothesis cannot be rejected and the share under standard is greater than the healthier (3) Hypothesis was rejected but share under healthier remained greater than the standard. (4) Hypothesis was rejected and the share under standard became greater than healthier. (5) Hypothesis was rejected and the share under healthier became greater than standard. (6) Hypothesis was rejected, the share under standard is greater than healthier but the later share increased. (7) Hypothesis was rejected, the share under standard is greater than healthier.

37

V.4.3 Econometric evaluation of change in demand Table 7 presents the results for all the products categories for which it was possible to run the analysis (6 out of 8 product categories). The focus was on the demand of healthier private label categories (as some branded products were no using FPTL): Three groups were identified in the analysis: (a) The results indicated that the demand for healthier breakfast cereals, cheddar cheese and potato products had an increase after the introduction of the FPTL. (b) Ready meals, which is one of the most dynamic food categories, did not show any change in the demand for healthier varieties (c) soft drinks and total cheese show a decrease in their demand for healthier products, though it is difficult to associate this to the introduction of FPTL.

38

Table 7: Results of the econometric analysis

Variables Breakfast cereals Soft drinks Cheddar cheese Total cheese Potato products Ready meals

Diff. Sig. Diff. Sig. Diff. Sig. Diff. Sig. Diff. Sig. Diff. Sig.

All the sample 6.10 * -20.57 * 9.14 * -3.57 * 305.95 * 0.18

Region code

Accessible Rural 6.41 * -29.40 * 10.58 * -5.19 * 343.38 * 0.63

Large Urban Areas 5.89 * -6.56 9.39 * -3.28 * 307.84 * 0.02

Other Urban Areas 6.66 * -24.25 * 8.45 * -3.28 * 288.39 * 0.12

Remote Small Towns -0.82 -20.56 9.88 * -4.79 * 331.99 * 0.09

Remote Rural 5.09 * -26.33 8.37 * -1.37 320.23 * 0.50

Head of household gender

Male 6.17 * -13.36 * 9.34 * -3.98 * 308.24 * 0.30

Female 5.87 * -41.14 * 8.41 * -2.49 * 298.24 * -0.19

Social class

A .. .. .. .. .. ..

B 6.34 * -12.55 11.33 * -2.06 * 344.14 * 0.31

C1 6.84 * -29.17 * 8.95 * -3.69 * 290.41 * -0.36

C2 5.95 * -10.44 8.66 * -4.35 * 318.42 * 0.89

D 4.20 * -11.49 8.64 * -3.64 * 317.67 * 0.14

E 6.98 * -42.00 * 9.49 * -2.83 * 272.33 * 0.09

39

V.4.4 Econometric evaluation of change in sales Tables 8 and 4 provide an example (for soft drinks and cheese) of the results obtained on the econometric analysis of the change in sales due to the introduction of the FPTL. In the case of the soft drinks estimation, promotions were found important for the sales of soft drinks. The price effect showed the right sign but it was not significant. The presence of sugar shows negative effect on sales (nutritional level and dummies) but none is significant. Table 4 shows also that promotions and the price have the right sign but they are not statistically significant. Saturated fats has a negative effect on the sales but it was found only significant in the regression where the nutritional level was included (i.e., not as a traffic light).

40

Table 8: Results for the soft drinks category

Variables Nutritional levels Nutritional categories Nutritional dummies

Coeff. St. dev. Sig. Coeff. St. dev. Sig. Coeff. St. dev. Sig.

Intercept -57948.000 17190.000 * -59185.000 17660.000 * -56796.000 17180.000 *

Change in prices -9813.300 25720.000 -5934.900 24440.000 -7101.800 24470.000 Change in proportion sold under promotion 380.320 217.100 * 370.660 209.400 * 389.860 220.700 *

Calories (KC) 19609.000 12340.000 17586.000 10130.000 * 18815.000 11210.000 *

Proteins 110410.000 86510.000 118050.000 89350.000 100160.000 88620.000 Carbohydrates -75209.000 41520.000 * -78908.000 45100.000 * -80271.000 46810.000 *

Fibre 16155.000 157200.000 53783.000 166200.000 45347.000 166400.000 Levels Sugars -10048.000 17790.000 Categories Sugars 22324.000 51240.000 Dummies (base=green) Sugars – amber -17218.000 32450.000 Observations 161 161 161

Source: Own elaboration based on Kantar Worldpanel data. Note: Standard errors are from a heteroskedasticity-consistent covariance matrix.

41

Table 9: Results for the total cheese category

Variables Nutritional levels Nutritional categories Nutritional dummies

Coeff. St. dev. Sig. Coeff. St. dev. Sig. Coeff. St. dev. Sig.

Intercept -4848.300 4792.000 1887.500 4609.000 -1362.100 3847.000 Change in prices -2067.900 1907.000 -1964.300 1891.000 -2228.700 1986.000 Change in proportion sold under promotion 76.258 59.780 81.608 55.870 79.189 54.550 Calories (KC) 12.129 51.490 10.800 43.630 19.853 45.910 Proteins 112.480 258.900 22.682 193.800 -58.551 209.300 Carbohydrates -538.940 464.200 -461.150 511.100 -580.770 540.600 Fibre -924.230 1627.000 -1620.000 1578.000 -2164.500 1704.000 Levels Saturated fats -470.990 228.100 * Categories Fats -3940.700 5870.000 Saturated fats 676.260 1161.000 Sugars -496.950 4389.000 Salt -1345.200 2596.000 Dummies (base=green) Fats – amber 6069.900 4241.000 Fats – red -2058.000 9848.000 Saturated fats – amber -702.530 3294.000 Saturated fats – red -2559.600 2673.000 Sugars – amber -1100.000 4252.000 Salt – amber -1413.800 2517.000 Observations 135 135 135

Source: Own elaboration based on Kantar Worldpanel data. Note: Standard errors are from a heteroskedasticity-consistent covariance matrix.

42

Table 10: Results for the ready meals category Variables Nutritional levels Nutritional categories Nutritional dummies

Coeff. St. dev. Sig. Coeff. St. dev. Sig. Coeff. St. dev. Sig.

Intercept -367.030 4183.000 -1115.300 3835.000 -225.550 3959.000 Change in prices -431.890 475.200 -320.640 466.500 -489.420 533.100 Change in proportion sold under promotion 0.884 19.500 3.007 20.050 1.623 19.750 Calories (KC) 91.402 105.300 5.615 21.090 -3.750 30.960 Proteins -692.970 479.100 -334.370 299.000 -295.850 339.000 Carbohydrates -298.160 436.600 -4.641 130.700 -11.438 225.300 Fibre -1024.000 1168.000 -1150.200 996.700 -818.080 904.300 Levels Fats -766.660 1064.000 Saturated fats 90.900 890.200 Sugars -234.860 377.000 Salt 5331.600 5215.000 Categories Fats 4833.200 4616.000 Saturated fats -2157.400 3132.000 Sugars 1602.600 3590.000 Salt 1812.100 2951.000 Dummies (base=green) Fats - amber 6357.100 5696.000 Fats – red 10654.000 10780.000 Saturated fats - amber -5006.100 4901.000 Saturated fats - red -1000.100 5207.000 Sugars - amber 1148.500 2971.000 Sugars - red Salt - amber 1638.000 3097.000 Salt – red 2030.400 3873.000 R-square (adjusted) -0.03 -0.03 -0.03 Observations 214 214 214 Note: Standard errors are from a heteroskedasticity-consistent covariance matrix.

43

Conclusions

The literature on traffic lights, which is mostly qualitative, found that they are easy to understand by consumers’ and they also are useful to “motivate” producers to reformulate their products. As regards the results of the different methods:

(Method 1) Based on aggregated market shares (time series), the introduction of traffic lights did not show any structural change (before and after the introduction of FPTL) in any category.

(Method 2) Individual purchases did not show a strong change towards healthier products except in soft drinks but that cannot be attributed to traffic lights as major sellers, e.g., brands like Coca Cola and Pepsi did not used traffic light at that time.

(Method 3) Introduction of traffic lights seemed to have mild effects towards healthier purchases but only for some categories (breakfast cereals, cheddar cheese and potato products).

(Method 4) Based on the analysis of products, traffic lights seemed to have only a slight effect reducing sales of “unhealthy” products but in most of the cases they are not statistically significant.

Potential issues are that traffic lights, whilst informative, might not be an effective way to discriminate amongst healthy products and standard products in some categories category (e.g., sweets, cheese). For other categories (e.g., vegetables or fruits) their presence is not really important.

It is important to note that in addition of being informative for consumers, traffic light labels have the “side effect” of encouraging food producers the reformulation of the products due to the potential effect that they may have on their sales (Department of Health, 2013).

44

VI. Task 4.1 – Uptake of innovative sustainable food products VI.1 Summary

VI.2 Objectives The purpose of this task is to examine one of the policy options available for the promotion of sustainable food choices, namely the introduction of new products with healthy and sustainable characteristics. This work relies on the Mintel GNDP data base on product launches. This report presents the results for the UK. VI.3 Methods In the UK, the entire Mintel GNPD dataset for the UK for food and drink products was downloaded into an Excel workbook. A number of routines in Excels’ Visual Basic for Applications were written to compute descriptive statistics (frequency

The purpose of this task is to examine one of the policy options available for the promotion of sustainable food choices, namely the introduction of new products with healthy and sustainable characteristics. Mintel’s Global New Product Development together with Kantar Worldpanel data were used for the analyses. Three analyses were carried out as part of this task in order to provide an overview of the introduction of new product activities and also the acceptance of new products by consumers: (1) Trends analysis on the introduction of new products with healthy and sustainable attributes; (2) Analysis of the major firms introducing new products with those attributes; (3) Analysis of the rate of success of new products considering the case of dairy products. The number of launched new products with sustainable and health attributes were shown to be increasing in most of the categories (organic products was an exception). Retailers were found to be particularly important in terms of the introduction of products and the number of categories, and in most of the cases, they were on the top 5 product launchers. On the analysis of dairy product success it was found that new products and new varieties have less probability of success. Also, cheese and yoghurt had less probability of succeed (maybe due to variety seeking) but this is offset if it was a premium product. Private label products had greater rate of success. Some health and sustainability claims became significant when interacting with specific categories.

45

distributions and crosstabulations). The dataset comprised a total of 88,566 products for the period June 1996 to April 2015. The top year in terms of the introduction of products was 2013 with 10,591 introduced products. The firms with the largest number of products in the dataset were Tesco with 7,241 products, followed by M&S with 6,013. The most important manufacturer in the dataset was Nestlé with 1,402 products and was in the 7th position in terms of the total introduction of products. The ranking changes when information by category is considered. The three most important categories: Bakery (10,290 products), Meals & Meal Centres (10,060 products) Sauces & Seasonings (7,677 products). A classification of products with sustainable attributes was produced and it considers those with the following claims (based on Mintel’s classification). Three analyses were considered as part of this task: 1. Trends in the introduction of new products with sustainable attributes - We ran trend regression (y=α+β∙t) for all the categories (where y is the location quotient, t is the time trend and α and β are parameters). The use of location quotient is explained by the fact that some categories are more active than others. 2. Top 5 companies introducing new products with healthy and sustainable claims considering the period 1996-2015. 3. Rate of success on the introduction of new dairy products - Table 13 to 15 presents the results of the introduction of dairy products launched in 2011. The definition of products launched was taken broadly in the sense that it not only included new products but also cases products that were reformulated and launched again. In order to estimate the degree of success (i.e., in terms of consumers’ adoption) the dairy products selected in Mintel’s GNPD database were identified in the Kantar Worldpanel database and their total expenditure by year from 2012-2015 was computed. Then the products were classified according to 3 categories: (a) Fully failed – No sales were observed; (b) Partial success – Some sales observed but no sales in 2015; and (c) Successful - Sales were observed every year. The results were analysed using logit and multinomial logit regressions in order to explore what claims (sustainable and healthy in particular) were important to ensure success of the products.

46

VI.4 Results VI.4.1 Analysis of trends The results are presented in Table 1. Most of the cases show an increase in the share of new products with sustainable attributes within the categories. An exception was organic in most of the categories with a negative trend.

47

Table 11: Analysis of trends on the introduction of products with healthy and sustainable attributes Carbon Ethical Ethical Ethical Environmental. Environmental. Organic Low/No/ Low/No/ Low/No Low/No/

neutral animal charity human friendly friendly Reduced Reduced Reduced Reduced

production package Saturated Fat Sodium Sugar Transfat

Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff.

Bakery 0.13 (0.04) 0.05 (0.00) 0.09 (0.01) -0.04 (0.01) -0.07 (0.02) 0.16 (0.01) Meals and meal centres 0.11 (0.04) 0.05 (0.02) 0.10 (0.02) 0.14 (0.02) -0.02 (0.01) 0.04 (0.02) 0.021 (0.01) 0.21 (0.03) Sauces and seasonings 0.13 (0.03) 0.06 (0.01) 0.03 (0.01) 0.09 (0.01) -0.07 (0.02) 0.14 (0.07) 0.09 (0.01) Fish, meat and egg products 0.21 (0.08) 0.02 (0.08) 0.12 (0.01) -0.03 (0.01) 0.08 (0.03) 0.08 (0.02) 0.07 (0.26) Snacks 0.04 (0.02) 0.04 (0.01) -0.06 (0.02) 0.03 (0.02)

Dairy 0.30 (0.06) -0.13 (0.07) 0.08 (0.02) 0.07 (0.01) -0.04 (0.02) 0.03 (0.01)

Alcoholic beverages 0.09 (0.01) -0.05 (0.02)

Desserts and ice cream -0.11 (0.05) 0.05 (0.01) 0.09 (0.04) 0.05 (0.01) 0.12 (0.01) -0.02 (0.01) 0.09 (0.01) Chocolate confectionery 0.08 (0.04) 0.10 (0.04) 0.05 (0.02) 0.06 (0.01) Side dishes 0.13 (0.04) 0.02 (0.10) 0.08 (0.01) 0.07 (0.01) 0.06 (0.03)

Sugar and gum confectionery 0.04 (0.02) 0.03 (0.00) 0.017 (0.01) 0.42 (0.05) 0.05 (0.01) Fruit and vegetables 0.03 (0.01) 0.13 (0.03) 0.03 (0.01) Juice drinks 0.09 (0.02) 0.14 (0.03)

Hot beverages 0.03 (0.01) -1.08 (0.03) -0.11 (0.06) 0.04 (0.01) Breakfast cereals 0.03 (0.01) 0.32 (0.11) 0.17 (0.05) 0.08 (0.01) Soup 0.05 (0.02) 0.12 (0.02) Savoury spreads 0.10 (0.02) 0.05 (0.02) 0.11 (0.01) -0.10 (0.06) 0.07 (0.01) Sweet spreads 0.32 (0.17) 0.02 (0.01) 0.06 (0.02) 0.11 (0.01) 0.03 (0.01) 0.39 (0.15) 0.05 (0.01) Carbonated soft drinks 0.17 (0.02) 0.08 (0.03)

Baby food 0.03 (0.01) 0.08 (0.03) 0.11 (0.02) 0.53 (0.23) 0.05 (0.02) Water 1.62 (0.21) 0.13 (0.02) 0.08 (0.02)

Other beverages 0.19 (0.06) 0.11 (0.02)

Sports and energy drinks 0.09 (0.02) 0.07 (0.02) 0.11 (0.02)

Sweeteners and sugar 0.20 (0.09) 0.41 (0.01) 0.16 (0.05) 0.10 (0.01) 1.05 (0.59)

Ready to drink beverages 1.14 (0.21) 0.22 (0.09) 0.14 (0.02)

Source: Own elaboration based on Mintel GNPD data. Note: Coefficients’ standard deviations in parenthesis.

48

VI.4.2 Top 5 firms introducing products Tables 12 and 13 presents the results of the top 5 firms launching products with sustainable and healthy claims, respectively. Whilst some manufacturers appear to be introducing new products, retailers are top on the introduction of new products with sustainable attributes. Note that this does not change when the data is analysed by product category. Table 12: Top 5 firms introducing products with sustainable attributes Category/Firm Firm Total Share Category/Firm Firm Total Share

type products % type products %

Carbon Neutral

Environmentally friendly product

Kallo Foods Manufacturer 26 23.6 Waitrose Retailer 207 9.4

Alpro Manufacturer 17 15.5 Marks & Spencer Retailer 125 5.7

Ben & Jerry's Manufacturer 5 4.6 Tesco Retailer 123 5.6

Global Ethics Manufacturer 5 4.6 Sainsbury's Retailer 114 5.2

Marks & Spencer Retailer 4 3.6 Birds Eye Manufacturer 105 4.7

Ethical Animal Ethical Human

Marks & Spencer Retailer 642 27.8 Marks & Spencer Retailer 68 4.6

Waitrose Retailer 303 13.1 Sainsbury's Retailer 60 4.0

Tesco Retailer 233 10.1 The Co-operative Group Retailer 53 3.6

Morrisons Retailer 112 4.8 Tesco Retailer 50 3.4

Sainsbury's Retailer 107 4.6 Lidl Retailer 49 3.3

Ethical Charity Organic

Waitrose Retailer 62 5.4 Waitrose Retailer 187 3.8

Marks & Spencer Retailer 59 5.2 Tesco Retailer 165 3.4

Innocent Manufacturer 44 3.9 Marks & Spencer Retailer 154 3.1

Unearthed Manufacturer 40 3.5 Sainsbury's Retailer 154 3.1

Seeds of Change Manufacturer 27 2.4 Organix Brands Manufacturer 133 2.7

Environmentally friendly package

Tesco Retailer 1917 11.6 Asda Retailer 1610 9.7 Sainsbury's Retailer 1392 8.4 Morrisons Retailer 1265 7.6 Marks & Spencer Retailer 1144 6.9

Source: Own elaboration based on Mintel GNPD data.

49

Table 13: Top 5 firms introducing products with healthy attributes Category/Firm Firm Total Share Category/Firm Firm Total Share

type products % Type products %

Diabetic

High Protein

BigShot N.I Manufacturer 1 8.3 Marlow Foods Manufacturer 70 11.5

Britvic Soft Drinks Manufacturer 1 8.3 Marks & Spencer Retailer 40 6.5

Coffee Center Manufacturer 1 8.3 Asda Retailer 31 5.1

Enco Products Manufacturer 1 8.3 Sainsbury's Retailer 19 3.1

Gerber Juice Company Manufacturer 1 8.3 The Protein Drinks Manufacturer 10 1.6

Brain & Nervous System (Functional) High Satiety

H.J. Heinz Manufacturer 18 8.0 Marks & Spencer Retailer 17 15.5

GlaxoSmithKline Manufacturer 11 4.9 Nairn's Manufacturer 6 5.5

Red Bull Manufacturer 11 4.9 Oomf Manufacturer 6 5.5

Milupa Manufacturer 10 4.4 Kraft Foods Manufacturer 4 3.6

Sainsbury's Retailer 8 3.5 The Protein Drinks Manufacturer 4 3.6

Cardiovascular (Functional) Low/No/Reduced Calorie

Unilever Manufacturer 32 7.5 Tesco Retailer 292 10.6

Marks & Spencer Retailer 29 6.8 Marks & Spencer Retailer 244 8.8

Sainsbury's Retailer 26 6.1 Asda Retailer 186 6.7

Tesco Retailer 20 4.7 Somerfield Manufacturer 102 3.7

Asda Retailer 18 4.2 Coca-Cola Manufacturer 101 3.7

Gluten-Free Low/No/Reduced Saturated Fat

Sainsbury's Retailer 334 6.3 Tesco Retailer 127 11.1

Asda Retailer 181 3.4 Asda Retailer 101 8.8

Aldi Retailer 157 3.0 Sainsbury's Retailer 65 5.7

Tesco Retailer 134 2.5 Alpro Manufacturer 46 4.0

Marks & Spencer Retailer 121 2.3 Marlow Foods Manufacturer 42 3.7

Low/No/Reduced Sodium Low/No/Reduced Sugar

Sainsbury's Retailer 85 5.5 Tesco Retailer 175 4.9

H.J. Heinz Manufacturer 72 4.7 Asda Retailer 107 3.0

Ella's Kitchen Manufacturer 66 4.3 Sainsbury's Retailer 93 2.6

Tesco Retailer 66 4.3 Coca-Cola Manufacturer 88 2.5

Marks & Spencer Retailer 54 3.5 Wrigley Manufacturer 78 2.2

Low/No/Reduced Transfat

Asda Retailer 1817 44.5

Tesco Retailer 621 15.2

McVitie's Manufacturer 104 2.5

Nestlé Manufacturer 78 1.9

Marks & Spencer Retailer 72 1.8

Source: Own elaboration based on Mintel GNPD data.

VI.4.3 Success on the introduction of dairy products As shown in Table 14, overall in 2011 588 dairy product were “introduced” with a rate of success of 36.1% (i.e., they were introduced in 2011 and they still can be found in the Great Britain market in 2015).

50

Table 14: Estimation of success on the introduction of dairy products Categories Fully Partial Success Total Percentages Success

failed success Failed Partial Success Total index

Evaporated Milk 0 1 4 5 0.0 20.0 80.0 100.0 2.22

Sweetened Condensed Milk 1 0 3 4 25.0 0.0 75.0 100.0 2.08

Margarine & Other Blends 6 2 17 25 24.0 8.0 68.0 100.0 1.89

Rice/Nut/Grain & Seed Based Drinks 1 0 2 3 33.3 0.0 66.7 100.0 1.85

White Milk 15 3 29 47 31.9 6.4 61.7 100.0 1.71

Soy Based Drinks 5 2 8 15 33.3 13.3 53.3 100.0 1.48

Cream 15 3 14 32 46.9 9.4 43.8 100.0 1.21

Butter 11 0 8 19 57.9 0.0 42.1 100.0 1.17

Fresh Cheese & Cream Cheese 8 3 7 18 44.4 16.7 38.9 100.0 1.08

Processed Cheese 9 4 7 20 45.0 20.0 35.0 100.0 0.97

Flavoured Milk 13 6 10 29 44.8 20.7 34.5 100.0 0.96

Shortening & Lard 3 1 2 6 50.0 16.7 33.3 100.0 0.92

Soft Cheese & Semi-Soft Cheese 29 7 17 53 54.7 13.2 32.1 100.0 0.89

Hard Cheese & Semi-Hard Cheese 61 19 35 115 53.0 16.5 30.4 100.0 0.84

Soy Yogurt 4 1 2 7 57.1 14.3 28.6 100.0 0.79

Curd & Quark 9 2 4 15 60.0 13.3 26.7 100.0 0.74

Soft Cheese Desserts 7 4 4 15 46.7 26.7 26.7 100.0 0.74

Drinking Yoghurt & Liquid Cultured Milk 16 7 8 31 51.6 22.6 25.8 100.0 0.72

Spoonable Yoghurt 76 19 31 126 60.3 15.1 24.6 100.0 0.68

Creamers 2 0 0 2 100.0 0.0 0.0 100.0 0.00

Liquid Dairy Other 1 0 0 1 100.0 0.0 0.0 100.0 0.00

Total 292 84 212 588 49.7 14.3 36.1 100.0 1.00

Source: Based on Mintel's GNPD and Kantar Worldpanel data. Table 15 presents the results of the logit estimation of the rate of success and the rate of failure (being the base category all other cases). New products and new variety have less probability of success. In addition, cheese and yoghurt have less probability of succeed but more if they were premium products. Private label was associated with greater rate of success. The presences of organic and animal ethical claims were positively associated with success. The Britishness of the product was not found significant. In addition, several interactions were found also significant, which indicated that the claims were specific by category.

51

Table 15: Factors affecting the introduction of dairy products (logit analysis) Variable Product is successful Product failed

Coeff. St. Dev. t ratio Sig. Coeff. St. Dev. t ratio Sig.

Intercept 0.401 0.245 1.630 * -0.520 0.242 -2.140 *

New Product -1.272 0.246 -5.170 * 1.012 0.232 4.360 *

New Variety/Range Extension -0.763 0.249 -3.060 * 0.657 0.240 2.740 *

Cheese -0.851 0.265 -3.220 * 0.438 0.251 1.740 *

Yoghurt (made of milk) -1.019 0.314 -3.240 * 0.350 0.286 1.220

Dummy branded (0) and private label (1) 0.476 0.216 2.210 * -0.432 0.200 -2.160 *

Dummy product mentions it is a British product (1) 0.274 0.370 0.740 0.314 0.351 0.900

Dummy Organic(1) 0.670 0.359 1.860 * -0.691 0.352 -1.960 *

Dummy Ethical animal (1) 2.595 1.326 1.960 * -2.643 1.459 -1.810 *

Dummy Digestive (1) -1.368 0.678 -2.020 * 1.110 0.571 1.940 *

Interactions

Cheese * Children (5-12) 0.049 0.638 0.080 -1.321 0.625 -2.110 *

Cheese *Convenient Packaging 0.368 0.406 0.910 -0.739 0.385 -1.920 *

Cheese * Premium 2.259 0.835 2.710 * -2.474 1.082 -2.290 *

Yogurt * Ethical - Charity 2.300 1.389 1.660 * -0.995 1.289 -0.770

Yogurt * Ethical - Environmentally Friendly Package -1.442 0.480 -3.010 * 0.792 0.375 2.110 *

Yogurt * Gluten-Free 1.840 0.558 3.300 * -0.822 0.499 -1.650 *

Yogurt * Premium 1.313 0.670 1.960 * -1.493 0.668 -2.230 *

Liquid milk * Ethical - Environmentally Friendly Product -1.810 1.388 -1.300 2.205 1.322 1.670 *

Liquid milk * Vegetarian 0.310 0.380 0.820 -1.156 0.428 -2.700 *

Liquid milk * Gluten-Free -2.175 1.153 -1.890 * 1.437 0.887 1.620 *

Fats * Ethical - Environmentally Friendly Package 0.656 0.597 1.100 -1.047 0.643 -1.630

Fats * Seasonal -2.280 1.191 -1.910 * 0.762 0.958 0.800

Log likelihood ratio test 106.15 82.51

P-value (21 d.f.) 0.00 0.00 Pseudo R square 0.14 0.10

Note: The dependent variable take the value of 1 if it is equal to the category and 0 otherwise.

The results of the estimation of the multinomial logit are presented in Table 16. The results are close to the logit model. These are:

New products and new variety have less probability of success.

Non-milk products have greater chance of success.

Cheese and yoghurt have less probability of succeed.

Private label is associated with greater rate of success but not significant

Some health and sustainability claims become significant when interacting with category

52

Table 16: Factors affecting the introduction of dairy products (multinomial logit) Product is successful Product failed

Coeff. St. Dev. t ratio Sig. Coeff. St. Dev. t ratio Sig.

Intercept 2.719 0.505 5.390 * 2.182 0.505 4.320 *

New Product -1.048 0.359 -2.920 * 0.293 0.343 0.860

New Variety/Range Extension -0.578 0.368 -1.570 0.245 0.357 0.690

Cheese -1.585 0.515 -3.080 * -0.839 0.501 -1.680 *

Yoghurt (made of milk) -1.998 0.555 -3.600 * -1.152 0.528 -2.180 *

Dummy branded (0) and private label (1) 0.310 0.306 1.010 -0.215 0.284 -0.750

Dummy product mentions it is a British product (1) 1.396 0.692 2.020 * 1.378 0.668 2.060 *

Dummy Organic(1) 0.206 0.570 0.360 -0.574 0.563 -1.020

Dummy Ethical animal (1) 14.718 1217.667 0.010 12.065 1217.667 0.010

Dummy Digestive (1) -0.882 1.004 -0.880 0.573 0.843 0.680

Interactions

Cheese * Children (5-12) -1.099 0.711 -1.540 -1.872 0.691 -2.710 *

Cheese *Convenient Packaging -0.314 0.533 -0.590 -0.927 0.504 -1.840 *

Cheese * Premium 1.245 1.121 1.110 -1.514 1.448 -1.050 Yogurt * Ethical - Charity 16.249 1841.176 0.010 14.289 1841.176 0.010

Yogurt * Ethical - Environmentally Friendly Package -1.500 0.614 -2.440 * -0.076 0.483 -0.160

Yogurt * Gluten-Free 2.522 0.948 2.660 * 0.830 0.884 0.940

Yogurt * Premium 0.382 0.843 0.450 -1.326 0.836 -1.590

Liquid milk * Ethical - Environmentally Friendly Product 12.325 1094.138 0.010 14.561 1094.137 0.010

Liquid milk * Vegetarian -1.463 0.611 -2.400 * -2.357 0.661 -3.560 *

Liquid milk * Gluten-Free -1.963 1.327 -1.480 0.369 1.037 0.360

Fats * Ethical - Environmentally Friendly Package -0.497 0.937 -0.530 -1.446 0.994 -1.450

Fats * Seasonal -3.141 1.424 -2.210 * -1.350 1.235 -1.090

Log likelihood ratio test 145.09 *

Note: The base category is the intermediate case (i.e., product did not remain all the years).

Conclusions The results of the analysis indicated that number of launched new products with sustainable and health attributes is shown to have increased in most of the categories with the products with the organic attributes being exception. Claims are associated to product category. Retailers were found particularly important and in most of the cases they are on the top 5 product launchers. On the analysis of dairy product success it was found:

New products and new varieties have less probability of success.

Cheese and yoghurt have less probability of succeed (this might be associated to variety seeking behaviour) but offset if it is a premium product.

Private label products have greater rate of success (although not in the multinomial logit model)

Some health and sustainability claims become significant when interacting with specific categories.

53

VII. Task 4.2 – Changes in prices effects on nutrition and sustainability VII.1 Summary

VII.2 Objectives The objective of this task is to model the effects of such a tax on UK households in order to understand the likely changes in GHG emissions and the subsequent resulting nutrient intake. The contribution of the paper is to provide alternative estimation to existing estimations for the UK (Brigg et al, 2013) using unconditional elasticities instead of food conditional elasticities and also alternative data. VII.3 Methods In the UK similar analyses have been carried out by Briggs et al. (2013) and Kehlbacher et al (2016). However there are some differences with respect to those studies:

We use unconditional elasticities (based on total budget) instead of conditional (based on food expenditure only).

We use Kantar Worldpanel data for 2012 instead of UK Living Costs and Food Survey 2010 or 2011.

Source of nutritional information is National Diet and Nutrition Survey (collected in SUSDIET WP1).

Using the demand elasticities computed in Task 2.3 for the categories presented in Table 1, the estimation of the effect of the taxes on nutritional and

Greenhouse gas emissions associated with food consumption have become particularly pertinent issues given the recent warnings that the planet recently has experienced its hottest year. A way to reduce climate emissions associated with food consumption is through a carbon consumption tax. This study uses intake, nutrient and GHG emission data supplied by a European project to estimate the impact of an ad-valorem tax and carbon consumption tax on GHG emissions and nutrient intake. These estimations were performed using price elasticities computed from an Exact Affine Stone Index (EASI) demand system. The results suggest that the carbon consumption tax scenarios would reduce GHG emissions by a greater quantity relative to the ad-valorem tax scenario. However, the intake of important nutrients will also decrease in these scenarios (note that there will also be a decrease in fatty acids, sodium and free sugars). Therefore, creating an environmentally sustainable and nutritious diet through taxation is challenging and requires compromise between the nutrition and environmental sustainability.

54

GHG proceeded as follows: (a) Increase prices by the taxes (2) Compute again the demand based on elasticities (3) Estimate the changes in GHG emissions and nutrition based on factors provided by WP1. The different tax scenarios are presented in Table 17. Table 17: Tax scenarios by products Scenarios TAX 1 - Ad Valorem TAX 2: per KgCO2 eq (per kg product) tax

tax: custom Price 1: 0.0128 £/kg (current average Emission price increase Trading System (ETS) price)

Price 2: 0.0427 £/kg (mean social cost of carbon; EC medium term projection of carbon price) Price 3: 0.1709 £/kg (long term EU projection

of carbon price)

1. Beef and veal, other meats, not preserved 1.1 Beef and veal, other meats, not preserved 20% Price(i) * 40 2. All meat (by KgCO2 eq. per kg product) 2.1 Beef and veal, other meats, not preserved 20% Price(i) * 40

2.2 Pork 20% Price(i) * 7.1

2.3 Processed and other cooked meats 20% Price(i) * 7.1

2.4 Poultry, eggs, other fresh meat 20% Price(i) * 4.3 3. All animal-based products (by KgCO2 eq per kg product) 3.1 Beef and veal, other meats, not preserved 20% Price(i) * 40

3.2 Animal fats 20% Price(i) * 8.3

3.3 Cheese 20% Price(i) * 8.3

3.4 Pork 20% Price(i) * 7.1

3.5 Processed and other cooked meats 20% Price(i) * 7.1

3.6 Fish, seafood and their products 20% Price(i) * 5

3.7 Poultry, eggs, other fresh meat 20% Price(i) * 4.3

3.8 Milk, milk products 20% Price(i) * 2.3 4. All products (tax rate proportional to emissions per Kg of product, with varying rates) Rates: 30%, for products with >= 10 KgCO2 eq emissions (per kg product); 20%, for products with 5-9.9 KgCO2 eq emissions (per kg product); 10%, for products with 1-4.9 KgCO2 eq emissions (per kg product); 5%, for products with 0-0.9 KgCO2 eq emissions (per kg product); 4.1 Beef and veal, other meats, not preserved 30% Price(i) * 40

4.2 Animal fats 20% Price(i) * 8.3

4.3 Cheese 20% Price(i) * 8.3

4.4 Pork 20% Price(i) * 7.1

4.5 Processed and other cooked meats 20% Price(i) * 7.1

4.6 Fish, seafood and their products 20% Price(i) * 5

4.7 Poultry, eggs, other fresh meat 10% Price(i) * 4.3

4.8 Plant based fats 10% Price(i) * 3.38

4.9 Milk, milk products 10% Price(i) * 2.3

4.10 Vegetables, fresh and processed 10% Price(i) * 2

4.11 Sweet products and substitutes 10% Price(i) * 1.6

4.12 Alcoholic beverages 10% Price(i) * 1.43

4.13 Cereals, cereal products and substitutes 5% Price(i) * 0.98

4.14 Potatoes, tubers, nuts, legumes and their products 5% Price(i) * 0.86

4.15 Fruit, fresh and processed 5% Price(i) * 0.71

4.16 Tea, coffee, cocoa, drinking water 5% Price(i) * 0.26

All the scenarios where simulated using an uncompensated and compensated tax scheme: in both either an ad-valorem or per KgCO2 tax were applied to some/all food categories. Following Edjabou and Smed (2013) a constant share of the VAT was subtracted from all food categories: P2=P1+(KgCO2_perKg*P_CO2)-β*P1, where P1 is the pre-tax price, P2 is the post-tax price and β is the share of VAT such that the total tax revenue is unchanged. In addition, a further two scenarios were considered: an uncompensated one (i.e., taxes were collected and used by the Government) and a compensated one (i.e., collected tax revenues were redistributed to non-taxed categories in the form of price subsidies). The simulations were implemented with Excel VBA (details of the estimation and simulation can be found in Revoredo-Giha et al., 2018). This study also estimated the Mean Adequacy Ratio (MAR) and Mean Excess Ratio (MER) (Vieux et al., 2013). MAR estimates the percentage of mean daily intake of 20 beneficial nutrients with 100 percent representing a diet which would conform to all of these nutritional requirements. MAR estimates the

55

percentage of mean daily intake of 20 beneficial nutrients (i.e. proteins, fibre, retinol equivalents, thiamine, riboflavin, niacin, vitamin B-6, folates, vitamin B-12, ascorbic acid, vitamin E, vitamin D, calcium, potassium, iron, magnesium and zinc) with a value of 100 percent representing a diet which would conform to all of these nutritional requirements. Note that the components of the MAR are truncated to 100 therefore excesses of one of the nutrients cannot compensate the lack of another nutrient. The formula of the MAR is given by

the following formula, where ic is the intake of nutrient i, iR is the

recommended intake of nutrient i.

17i

i=1 i

c1MAR = 100

17 R

MER was developed as an indicator of bad nutritional quality. The MER is calculated as the mean daily percentage of maximum recommended values (

iMR ) for 3 harmful nutrients, namely saturated fatty acids, sodium and free

sugars, with a low percentage indicating a healthier diet. MER is given by the following formula.

3i

i=1 i

c1MER = 100 100

3 MR

Each ratio comprising the MER that was below 100 (i.e. below the maximum recommended value) was set as 100 in order to avoid one ratio to compensate the excesses of another. Therefore, MER values above 100 indicate a diet with excess in one or more of harmful nutrients. VII.4 Results The results are presented in Figures 2 and 4. Figure 2 presents the changes in nutritional indicators due to the taxes for the two extremes: taxing only meats versus taxing all the products (considering two taxes ad valorem and pricing the GHG). Relationship is not linear. The results indicate that carbon taxes on all the scenarios imply a contraction in the quantities purchased (and intakes) and then in nutrition content.

56

Figure 2: Results for Nutrition - Changes with respect to baseline (%)

Note: Result i-j-k indicates scenario i (i=1,..,4), j=ad valorem vs carbon tax, k (k=1,.., 3) 3 possible carbon prices.

Figure 3 shows the changes in GHG for each one of the scenarios. Although the last scenario, i.e., all the groups pay according to their contribution is the one that reduces most the GHG, it is worthwhile to note that most of the reduction is achieved when only livestock meat is taxed. Figure 3: Results of the change of GHG - Changes with respect to baseline (%)

Note: Result i-j-k indicates scenario i (i=1,..,4), j=ad valorem vs carbon tax, k (k=1,.., 3) 3 possible carbon prices. The trade-off between nutrition and GHG emissions are shown in Figure 4. Nutrition was summarised by the MAR and MER indicators. The left panels represent the relationship between the MAR and the GHG emissions whilst the right ones portray the MER and the GHG emissions. The arrows indicate the direction of improvement in terms of the MAR, MER and GHG emissions. The MAR indicated that for every tax scenario there will be a decrease in this ratio which suggests that the scenarios will result in fewer intakes of beneficial nutrients. The reduction is relatively small considering that the baseline is 84.5

Only livestock meat is taxed

Groups pay according to

their contribution

57

percent and the lowest value is 74.9 percent for carbon tax rate 3, uncompensated case considering taxation of all the products (fourth scenario) and the highest value is 84.2 percent for carbon tax price 2, compensated case with taxing only beef veal and unprocessed meats. The MER indicated a small decrease for all scenarios relative to the baseline of 107.2 percent, which indicates an improvement of the diet. The lowest MER is experienced for carbon price 3, uncompensated case for the scenario where all the products are taxed with a value of 100.13 percent, whilst the highest MER is attributed to carbon price 2, compensated case for taxation only beef veal and unprocessed meats with 106.9 percent. Note, however, that on the compensated case, the results of carbon price 2 and 3 are the best compromise between nutrition and environment because whilst carbon price 2 produces the best result in terms of carbon reduction, carbon price 3 does it in terms of nutrition. Figure 4: MAR and MER and GHG emissions by tax scenario, uncompensated and compensated cases.

Therefore, note that the overall the effect of the carbon taxes is to slightly reduce intakes of important nutrients (i.e., no carbon tax scenario improves the

Uncompensated cases

Compensated cases

74

76

78

80

82

84

86

3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1

MA

R (%

)

GHG (Kg CO2/g)

99

100

101

102

103

104

105

106

107

108

3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1

ME

R

(%)

GHG (Kg CO2/g)

75

76

77

78

79

80

81

82

83

84

85

3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1

MA

R (%

)

GHG (Kg CO2/g)

101

102

103

104

105

106

107

108

3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4 4.1

ME

R

(%)

GHG (Kg CO2/g)

58

value shown in the baseline) while also slightly decreasing intake of harmful nutrients. An interesting point is whether it would be possible to find a combination that improves nutrition (specifically the MAR) whilst still reducing the GHG emissions. Given that the MAR at the baseline has a value equal to 84.47%, any tax (even those focused on one category as meat) will make the situation worse off. Additional simulations indicated that in order to improvements the MAR’s results would require subsidizing significantly (above 50 percent) categories such as vegetables, fruits and cereals. Conclusions

The purpose of this analysis has been to estimate the potential effect that a tax on the carbon footprint on food products may have UK households as regards the GHG emissions and the nutritional quality of the diets. This type of taxes is one of the ways to encourage consumers to improve the environmental sustainability of their food choices. This is important because in the context of the discussion of sustainable diets, duties to consumption goods are often discussed as a potential policy response to fight global warming. However, normally nutritional quality is not considered.

The results show that taxing high carbon food products has the potential to reduce both GHG emissions and to some extent improve nutrient intake by reducing the quantities of harmful nutrients.

The effectiveness of this outcome is dependent upon the type of tax used. Whilst the Ad-Valorem tax would likely be the simplest to administer, it would not be as effective as an actual carbon consumption tax in terms of GHG emission reductions. The carbon consumption tax of the uncompensated scenario (price 3, considering the taxation of all the products) which used the long-term EU projection of carbon price would likely result in the largest reduction of GHG emissions by 18.7 percent relative to the baseline.

Whilst the primary purpose of a carbon consumption tax is reducing GHG emissions, the effect that they may have on nutrition is also important. Both the mean adequacy ratio (MAR) and mean excess ratio (MER), which are coefficients aiming to summarise the nutritional situation of food choices, show small change in nutrient intake. All the scenarios show a deterioration of the MAR indicating worsening of the diets in terms of recommended nutrient intake. As regards the MER it shows a small decrease in excessively consumed nutrients such as sugars. This is reflected by the decrease in the consumption of free sugars in all the tax scenarios.

The study suggests in the context of the inadequacy of the UK diet, the importance to take into consideration not just GHG emissions when considering public policy measures towards environmental sustainability that may affect consumers’ food choices since these may have effects on nutrition.

59

VIII. Conclusions The purpose of this report has been to present the results from the tasks where SRUC participate in the SUSDIET project, namely: Tasks 2.1, 2.2, 2.3, 3.4, 4.1 and 4.2. As regards Task 2.1, consumers were found to be willing to pay a price premium for sustainable, healthier and local beef mince. Therefore, claims presenting information on these attributes should be used to increase the demand for environmentally-friendly and healthier beef mince. Obviously, this approach is appropriate as long as it is economically viable to produce beef with those characteristics (i.e., a cost-benefit analysis would be needed). If sustainable and health claims are present on different beef mince packages (i.e., not on the same product), consumers are more likely to buy the healthier beef mince (i.e. low-fat beef mince) rather than the more sustainable beef mince (i.e., organic or beef mince produced with low GHG emissions). Consumers perceive the attributes organic, local and low fat as complements. However, these attributes and the attribute low GHG emissions from production were found to be independent. Thus, consumers are not willing to pay a price premium (in addition to the premiums for the individual attributes) for the co-existence of the attribute low GHG emissions and any one of the attributes organic, local or low fat. The results from Task 2.2 show that consumers are willing to pay a price premium for sustainable strawberries. They were also found to positively value front-of-package information on the sweetness and juiciness of strawberries. Overall, consumers prefer the strawberries to be sweet and slightly juicy. This suggests that the demand for environmentally-friendly strawberries can be increased using environmental claims. However, the survey showed also that consumers seem to find carbon footprint claims difficult to understand and interpret. Therefore, more research work is still needed to determine the best way to present the information to consumers. The results also showed that front-of-package information on the juiciness and sweetness of strawberries is an effective way to provide consumers with information on strawberries taste and increase the demand for strawberries who consider these attributes as the main drivers of their purchasing decision of strawberries. Consumers were found to consider the attributes low GHG emissions and organic as complements. Therefore, the premium paid for organic strawberries can be increased if they are produced with low GHG emissions and labelled as so. In addition, they are also willing to pay a price premium for strawberries that are labelled as juicy. Therefore, the demand for organic juicy strawberries can be improved if the strawberries are clearly labelled as having a juicy taste. On methodological terms, based on the results from Tasks 2.1 and 2.2, it was shown that assuming that consumers perceive food attributes as independent

60

(as it is assumed in most of studies on similar topics) is misleading and is likely to result in biased results. Despite that designing choice experiments that allow for the estimation of all the two-way interactions between attributes is time and economically challenging, the improvement in the results validity and prediction accuracy in worth the time and effort spent to cope with that additional hurdle. The results from Task 2.3 found that the demand categories respond significantly and negatively to changes in their own-price (i.e., price is an important driver of purchases). Moreover, the demands were price inelastic for most groups, except soft drinks, snacks and food out of home all of which had elastic demands. Income elasticities show that all the products were normal goods with the exception of food out of home; therefore an increase in income would substantively increase their demand. This is consistent with the evidence that during the recession households became more cautious and significantly reduced their demand for food outside home. Several relatively large cross price elasticities were found in the analysis, not all of them significant, though. Relationships of complementarity were found to be at least as common as the substitution ones, which made demand adjustments to price changes complex, i.e., involving changes in the entire diet. In other terms, changes in prices within a category generate changes on all the categories, therefore the evaluation of nutritional or GHG requires considering all the categories and not just those which have their prices changed. As regards the impact of traffic light on food purchases (Task 3.4) the different methods of evaluation show that these have at most only a mild impact on moving consumption towards healthier products. Thus, the introduction of traffic lights did not show any structural change (before and after the introduction of FPTL) in the market share of any of the four categories (standard branded and private label and healthy branded and private label). The analysis of individual purchases did not show a strong change towards healthier products except in soft drinks but that cannot be attributed to traffic lights as major sellers, e.g., brands like Coca Cola and Pepsi did not used traffic light at the time of their introduction by Tesco and Morrisons (i.e., August 2012). The econometric analysis of the introduction of traffic lights showed that they seemed to have only mild effects towards healthier purchases and were found positive only for some categories (breakfast cereals, cheddar cheese and potato products). The second econometric analysis of the traffic lights, i.e., considering the impact of the traffic lights on the sales of healthy and unhealthy products found that traffic lights seemed to have only a slight effect reducing sales of “unhealthy” products but in most of the cases their effect were not statistically significant. Potential issues are that traffic lights, whilst informative, might not be an effective way to discriminate amongst healthy products and standard products in some categories category (e.g., sweets, cheese) as several of the healthier

61

versions still show the red colour in several nutrients. For other categories (e.g., vegetables or fruits) their presence is not really important. Despite the low impact on changing consumers’ behaviour, traffic light labels have the “side effect” of encouraging food producers the reformulation of the products due to the potential effect that they may have on their sales (Department of Health, 2013). The results of the Task 4.1 indicated that number of launched new products with sustainable and health attributes is increasing in most of the categories with the products; however, the products with organic attributes were found to be an exception. As regards the major introducers in the UK of products with healthy and sustainable attributes, retailers were found particularly important as in most of the cases they are on the top 5 product launchers. On the analysis of the uptake of dairy product (i.e., their degrees of success) it was found:

New products and new varieties have less probability of success. This is challenging because it means that in order to remain in the market those products need to be relaunched.

Cheese and yoghurt have less probability of succeed (this might be associated to variety seeking behaviour) but offset if it is a premium product.

Private label products have greater rate of success (although this was found not significant in the multinomial logit model)

Some health and sustainability claims become significant when interacting with specific categories, indicating that some claims are only particular to some products.

The results from Task 4.2 indicate that a carbon tax generates a trade off between nutrition and environment. This is because carbon taxes on all the scenarios imply a contraction in the quantities purchased and in the nutrients. All the simulated cases showed reduction in GHG. As regards the type of tax, the ad valorem tax implies almost linear responses; whilst the effect of carbon prices is clear non-linear. Increasing the price shows exponential decreases in GHG. However, expanding the tax base (i.e., taxing more products) cannot be consider effective as taxing meats with a high rate produces almost the same reduction of GHG as taxing all the products. Areas for future research In terms of future work, the different tasks point out several areas of further analysis. Tasks 2.1 and 2.2 indicate the need to keep increasing the realism in the choice experiments including aspects such as flavour and also trying to incorporate additional socioeconomic information to explain some of the heterogeneity observed in the results.

62

As regards task 2.3, whilst the EASI demand system has proved to be quite flexible, the nutritional analysis of purchases requires more disaggregated models than the ones so far estimated. This is an area that is currently progressing with the SRUC estimation of demand models for each UK country considering the 259 food and drink categories of Defra’s ‘Family Food’ publication. With respect to task 3.4, given the differences in what consumers report and actually do, further work on the evaluation of labels would require a much larger sample than the one used in this study, which should start well before the introduction of traffic lights. Task 4.1 is probably the one that has been less addressed in the literature and requires further work. Whilst there is a continuous discussion of innovation as regards products with health and sustainability attributes, not much has been done in terms of the uptake by consumers of products with these attributes. The analysis performed here for the case of dairy products can be carried out for other food products. It is expected that the results of these type of evaluation would provide valuable information to firms. It is important to note that task 4.2 has focused on carbon taxes that are available in the literature; however, an extension of the work would be solve the public finance problem of finding an optimal carbon tax that reduces emission to some point, something similar to the analysis of a Laffer curve and compare the optimal tax with what is available in the literature and its implications on nutrition.

63

IX. References Abou-Zeid and Ben-Akiva, (2014). Hybrid choice models in Hess, S., & Daly, A. (Eds.).Handbook of choice modelling. Edward Elgar Publishing. Akaichi, F., Nayga, R. M., and Nalley, L. L. (2017). Are there trade-offs in valuation with respect to greenhouse gas emissions, origin and food miles attributes?. European Review of Agricultural Economics, 44(1), 3-31. Aprile, M. C., Caputo, V., and Nayga Jr, R. M. (2016). Consumers’ preferences and attitudes toward local food products. Journal of food products marketing, 22(1), 19-42.

Aschemann‐Witzel, J., and Zielke, S. (2015). Can't buy me green? A review of consumer perceptions of and behavior toward the price of organic food. Journal of Consumer Affairs. Audsley, E., Brander, M., Chatterton, J., Murphy-Bokern, Webster, D. C. and Williams, A. (2009). How low can we go? An assessment of greenhouse gas emissions from the UK food system and the scope to reduce them by 2050. WWF-UK. Balcombe, K., Fraser, I., Di Falco, S. (2010). Traffic lights and food choice: A choice experiment examining the relationship between nutritional food labels and price. Food Policy 35: 211–220. Baltussen, W. H. M., Wertheim-Heck, S. C. O., Bunte, F. H. J., Tacken, G. M. L., Van Galen, M. A., Bakker, J. H., and De Winter, M. A. (2006). Een Biologisch prijsexperiment: grenzen in zicht?. LEI. Barnett W.A. and Serletis A. (2008). Consumer preferences and demand systems. Journal of Econometrics, 147(2):210-224. Baumol, W.J. (1972). On taxation and the control of externalities. American Economic Review, 307-22. Beck, M. J., Fifer, S., and Rose, J. M. (2016). Can you ever be certain? Reducing hypothetical bias in stated choice experiments via respondent reported choice certainty. Transportation Research Part B: Methodological, 89, 149-167. Bonti-Ankomah, S. and Yiridoe, E.K. (2006). Organic and Conventional Food: A Literature Review of the Economics of Consumer Perceptions and Preferences, Organic Agriculture, Centre of Canada, Toronto. Botonaki, A., Polymeros, K., Tsakiridou, E. and Mattas, K. (2006). The role of food quality certification on consumers’ food choices, British Food Journal, 108(2-3), 77-90.

64

Briggs, A.D.M., Kehlbacher, A. Tiffin R., Garnett, T. Rayner, M. and Scarborough, P. (2013). Assessing the impact on chronic disease of incorporating the societal cost of greenhouse gases into the price of food: an econometric and comparative risk assessment modelling study. BMJ Open 3. Caillavet, F., Fadhuile, A., Nichèle, V. (2016). Taxing animal-based foods for sustainability: environmental, nutritional and social perspectives in France. European Review of Agricultural Economics 43 (4): 537-560. Capitanio, F., Coppola, A., and Pascucci, S, (2009) Indications for drivers of innovation in the food sector. British Food Journal, 111(8), 820-838. Chalak, A., Abiad, M., and Balcombe, K. (2016). Joint use of attribute importance rankings and nonattendance data in choice experiments. European Review of Agricultural Economics, jbw004.10 Chalmers, N.G., Revoredo-Giha, C., and Shackley, S. (2016). Socioeconomic Effects of Reducing Household Carbon Footprints through Meat Consumption Taxes. Journal of Food Products Marketing:1-20. Chen, M.F. (2009). Attitude toward organic foods among Taiwanese as related to health consciousness, environmental attitudes, and the mediating effects of a healthy lifestyle”, British Food Journal, 111(2), 165-78. Connor J., (1981). Food product proliferation: a market structure analysis. American Journal of Agricultural Economics,63 (4), 607-617 Costa A.I.A. and Jongen, W.M.F. (2006). New insights into consumer-led food product development. Trends In Food Science and Technology, 17, 457-465. Cox, T.L., and Wohlgenant, M.K. (1986). Prices and quality effects in cross-sectional demand analysis. American Journal of Agricultural Economics 68:908-19. Cummings, R. G., and Taylor, L. O. (1999). Unbiased value estimates for environmental goods: a cheap talk design for the contingent valuation method. The American Economic Review, 89(3), 649-665. Deaton, A., and Muellbauer, J. (1980). An Almost Ideal Demand System. American Economic Review 70, 312-26. Dijksterhuis, G. (2016). New product failure: Five potential sources discussed. Trends in Food Science and Technology, in press, 1-6. Edjabou, L. D., and Smed, S. (2013). The effect of using consumption taxes on foods to promote climate friendly diets–The case of Denmark. Food Policy, 39, 84-96.

65

European Commission (2012). Europeans’ attitudes towards food security, food quality and the countryside. Special Eurobarometer 389 Report. Available online: http://ec.europa.eu/commfrontoffice/publicopinion/archives/ ebs/ebs_389_en.pdf Feldmann, C., and Hamm, U. (2015). Consumers’ perceptions and preferences for local food: A review. Food Quality and Preference, 40, 152-164. Gatignon,J., Gotteland, D., Haon, C. (2016). Making Innovation Last. Volume 2. Sustainable Strategies for Long Term Growth. Ed. Palgrave Macmillan. London.

Gerini, F., Alfnes, F., and Schjøll, A. (2016). Organic‐and Animal Welfare‐labelled Eggs: Competing for the Same Consumers? Journal of Agricultural Economics, 67(2), 471-490. Greene, W. H. 2012. Econometric Analysis. 7th ed. Upper Saddle River, NJ: Prentice Hall. Hartikainen, H., Pulkkinen, H. (2016) Summary of the chosen methodologies and practices to produce GHGE-estimates for an average European diet. Natural resources and bioeconomy studies 58/2016. Natural Resources Institute Finland. Hieke, S. and Wilczynski, P. (2011). Colour Me In – an empirical study on consumer responses to the traffic light signposting system in nutrition labelling. Public Health Nutrition: 15(5), 773–782. Hoareau, S., G. Lacroix, M. Hoareau, and L. Tiberti (2012). Exact Affine Stone Index Demand System in R: The easi Package, http://finzi.psych.upenn.edu/library/easi/doc/easi.pdf; Huang, K.S. (1996). Nutrient elasticities in a complete food demand system. American Journal of Agricultural Economics 78 21-9. Hung, Y., Grunert, K. G., Hoefkens, C., Hieke, S., and Verbeke, W. (2017). Motivation outweighs ability in explaining European consumers' use of health claims. Food Quality and Preference. Intergovernmental Panel on Climate Change IPCC (2014). Fifth Assessment Synthesis Report: Summary for Policymakers, http://www.ipcc.ch/pdf/assess ment-report/ar5/syr /SYR_AR5_SPMcorr1.pdf. Irz, X. (2017). Demand for food and nutrients and its climate impact: A micro-econometric analysis of economic and socio-demographic drivers. Natural resources and bioeconomy studies 28/2017 (http://urn.fi/URN:ISBN:978-952-326-403-8). Johe, M. H., and Bhullar, N. (2016). To buy or not to buy: The roles of self-identity, attitudes, perceived behavioral control and norms in organic consumerism. Ecological Economics, 128, 99-105.

66

Kallas, Z. and Gil, J.M. (2012). A Dual Response Choice Experiments (DRCE) design to assess rabbit meat preference in Catalonia: A Heteroskedatistic Extreme-Value Model. British food Journal 114(10), 1394-1413 Ladenburg, J., and Olsen, S. B. (2014). Augmenting short cheap talk scripts with a repeated opt-out reminder in choice experiment surveys. Resource and Energy Economics, 37, 39-63. Lee, H. J., and Yun, Z. S. (2015). Consumers’ perceptions of organic food attributes and cognitive and affective attitudes as determinants of their purchase intentions toward organic food. Food Quality and Preference, 39, 259-267. Lewbel, A., and Pendakur, K. (2009). Tricks with Hicks: The EASI demand system. The American Economic Review, 99(3), 827-863. Lusk, J. L., Nilsson, T., and Foster, K. (2007). Public preferences and private choices: Effect of altruism and free riding on demand for environmentally certified pork. Environmental and Resource Economics, 36(4), 499-521. Lustig, R.H., L.A. Schmidt, and C.D. Brindis (2012). Public health: The toxic truth about sugar. Nature 482, 27-9. Malam, S., Clegg, S., Kirwan, S., McGinigal, S., Raats, M., Shepherd, R., Barnett, J., Senior, V., Hodgkins, C., Dean, M. (2009). Comprehension and use of UK nutrition signpost labelling schemes. Food Standards Agency, London. McFadden, D. (1973). Conditional logit analysis of qualitative choice behavior. Resource document. University of California at Berkeley. https://eml.berkeley.edu/reprints/mcfadden/zarembka.pdf McFadden, D., and Train, K. (2000). Mixed MNL models for discrete response. Journal of Applied Econometrics, 15(5), 447-470. Mazzocchi, M., Cagnone, S., Bech-Larsen, T., Niedźwiedzka, B., Saba, A., Shankar, B., and Traill, W. B. (2015). What is the public appetite for healthy eating policies? Evidence from a cross-European survey. Health Economics, Policy and Law, 10(3), 267-292. Meijer, E., and Rouwendal, J. (2006). Measuring welfare effects in models with random coefficients. Journal of Applied Econometrics. 21(2), 227-244. Mytton, O. T., Clarke, D., and Rayner, M. (2012). Taxing unhealthy food and drinks to improve health. BMJ: British Medical Journal (Online), 344. Onozaka, Y. and Mcfadden, D.T. (2011). Does local labeling complement or compete with other sustainable labels? A conjoint analysis of direct and joint values for fresh produce claim. American Journal of Agricultural Economics. 93(3), 693-706.

67

Panzone, L., Perino, G., Swanson, T., and Leung, D. (2011). Testing for the best instrument to generate sustainable food consumption. International Journal on Food System Dynamics, 2(3), 237-252. Pendakur, K. (2009). Chapter 7: EASI made easier. In: Slottje DJ, editor. Quantifying consumer preferences Bingley, UK: Emerald; p. 179. Provencher, V., and Jacob, R. (2016). Impact of Perceived Healthiness of Food on Food Choices and Intake. Current obesity reports, 5(1), 65-71. Public Health England (PHE) (2015). Sugar Reduction: The evidence for action. Annexe 2: A mixed method review of behaviour changes resulting from experimental studies that examine the effect of fiscal measures targeted at high sugar food and non-alcoholic drink. Public Health England, London, UK. Revoredo-Giha, C. (2014). Hitting the jackpot and the health agenda: the case of processed potato products. Procedia Economics and Finance, 14, 525-533. Revoredo-Giha, C., Chalmers, N., and Akaichi, F. (2018). Simulating the Impact of Carbon Taxes on Greenhouse Gas Emission and Nutrition in the UK. Sustainability, 10(1), 134. Rozenberg, S., Body, JJ., Bruyère, O. et al. (2016) Effects of Dairy Products Consumption on Health: Benefits and Beliefs—A Commentary from the Belgian Bone Club and the European Society for Clinical and Economic Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases. Calcification Tissue International, 98: 1. Sacks, G., Rayner, M., and Swinburn, B. (2009). Impact of front-of-pack ‘traffic-light’nutrition labelling on consumer food purchases in the UK. Health promotion international, 24(4), 344-352. Sarkar, S. and Costa, A.I.A (2008) Dynamics of open innovation in the food industry. Trends in Food Science and Technology, 19, 574-580. Scott, J. and Freese,J., (2014) Regression Models for Categorical Dependent Variables Using Stata. Third Edition. Stata Press Shonkwiler, J. S., Yen, S. T. (1999). Two-step estimation of a censored system of equations. American Journal of Agricultural Economics 81(4): 972-982. Shuai, C. M., Ding, L. P., Zhang, Y. K., Guo, Q., and Shuai, J. (2014). How consumers are willing to pay for low-carbon products?–Results from a carbon-labeling scenario experiment in China. Journal of Cleaner Production, 83, 366-373.

68

Slimani, N., Fahey, M., Welch, A., Wirfält, E., Stripp, C., Bergström, E. et al. (2002). Diversity of dietary patterns observed in the European Prospective Investigation into Cancer and Nutrition (EPIC) project. Public Health Nutrition 5(6b): 1311-28. Stewart-Knox, B. and Mitchell, P. (2003). What separates the winners from the losers in new food product development? Trends in food science and technology, 14, 58-64. Stobbelaar, D.J., Casimir, G., Borghuis, J., Marks, I., Meijer, L. and Zebeda, S. (2007), “Adolescents’ attitudes towards organic food: a survey of 15- to 16-year old school children”, International Journal of Consumer Studies, 31, 349-56. Talukdar, D., Sudhir, K., Ainslie, A. (2002) Investigating new product diffusion across products and countries. Marketing Science, 21, 1, 97-114. The Grocer (2015) The grocer top product survey 2015. 19 December 2015. www.thegrocer.co.uk The Grocer (2016a) The dairymen, a supplement to the grocer. 17 September 2016. www.thegrocer.co.uk The Grocer (2016b) The grocer top product survey 2016. 17 December 2016. www.thegrocer.co.uk Tiu Wright, L., Nancarrow, C., Kwok, P.M.H. (2001). Case study: Food taste preferences and cultural influences on consumption. British Food Journal 103(5): 348-357 Train, K. (2003). Discrete Choice Methods with Simulation. Cambridge, UK: Cambridge. Train, K., and Weeks, M. (2005). Discrete choice models in preference space and willingness-to-pay space (pp. 1-16). Springer Netherlands. UK Department of Health (2013). Front of pack Nutrition Labelling: Joint Response to Consultation. Available online: https://www.food.gov.uk/ sites/default /files /multimedia/pdfs /consultationresponse/frontofpacklabelling-response.pdf Van Wezemael, L., Caputo, V., Nayga, R. M., Chryssochoidis, G., and Verbeke, W. (2014). European consumer preferences for beef with nutrition and health claims: A multi-country investigation using discrete choice experiments. Food Policy, 44, 167-176. Vieux, F., Soler, L. G., Touazi, D., and Darmon, N. (2013). High nutritional quality is not associated with low greenhouse gas emissions in self-selected diets of French adults. The American journal of clinical nutrition, ajcn-035105.

69

Vranken, L., Avermaete, T., Petalios, D. and Mathijs, E. (2014). Curbing global meat consumption: Emerging evidence of a second nutrition transition. Environmental Science & Policy 39: 95-106. World Health Organization (2003) Diet, nutrition and the prevention of chronic diseases, http://www.who.int/dietphysicalactivity/publications/trs916/en/ ed2003. World Meteorological Organization (2016). Provisional WMO Statement on the Status of the Global Climate http://public.wmo.int/en/media/press-release/provisional-wmo-statement-status-of-global-climate-2016 ed2016. Zouaghi, F. and Sanchez, M. (2016) Has the global financial crisis had different effects on innovation performance in the agri-food sector by comparison to the rest of the economy? Trends in food science and technology, article in press.