environmental science & policy 39 (2014) 95106

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/envsci

Curbing global meat consumption: Emerging

evidence of a second nutrition transition

Liesbet Vranken, Tessa Avermaete, Dimitrios Petalios, Erik Mathijs *

Division of Bioeconomics, Department of Earth and Environmental Sciences, KU Leuven,
Celestijnenlaan 200 E Box 2411, 3001 Leuven, Belgium

article info abstract

Meat consumption patterns worldwide have dramatically changed over the past 50 years,
Available online 25 March 2014
putting pressure on the environment and leading especially in industrialised and emerging
countries to unbalanced diets. Given demographic projections and foresight reports, the
Keywords:
question is raised whether there are limits to the meat consumption. Based on data from 120
Meat consumption
countries, this article analyses the evolution of meat consumption in general and the
Income
relationship between meat consumption and income in particular. The study shows evi-
Nutrition transition
dence for an inverted U-shaped relationship between meat consumption and income,
meaning that at a certain level of income average meat consumption will stagnate or
even decline. The results can help policy makers to develop incentives for both environ-
mental and health policies and offers stakeholders opportunities for further research and
innovation.
# 2014 Elsevier Ltd. All rights reserved.

Also in Asia, a nutrition transition is taking place. More in

1. Introduction
Over the past 50 years, food consumption patterns in general
and meat consumption in particular has changed dramatical-
ly. Human diet and nutritional status have undergone a
sequence of major shifts, referred to as the nutrition
transition. Focusing on the recent history, the EU has
experienced an important shift after World War II. As more
food per capita became available, consumers increased their
energy intake, mainly through an increased consumption of
vegetable-based foodstuffs. This is called the expansion stage
of the nutrition transition. Another phase is referred to as the
substitution stage in which carbohydrate-rich staples such as
cereals, roots and tubers are replaced by vegetable oils, animal
products and sugar. The substitution stage is the current
nutritional phase in most industrialised and emerging
countries (like Brazil, Russia, India, China and South Africa).
particular, the Asian diet is shifting away from relatively
monotonous diets based on the indigenous staple grain or
root, local legumes, vegetables and fruits, and limited food of
animal origin. Instead, an industrialised and usually more
varied diet is adopted which includes more pre-processed
food, more food of animal origin, more added sugar and fat,
and often more alcohol (Popkin, 2006; Tey et al., 2010).
Increased meat consumption per capita and global popu-
lation growth has resulted in an overall increase of global meat
production and consumption. This tendency puts serious
pressure on the environment as livestock production, espe-
cially intensive livestock production, is one of the major
causes of greenhouse gas emissions (Garnett, 2009). Meat and
meat products account for 412% of the impact on global
warming of all consumer products (Tukker et al., 2006).
Moreover, world population is expected to increase up to
9 billion people in 2050. In addition, the share of meat

* Corresponding author. Tel.: +32 16 321450.

E-mail address: erik.mathijs@ees.kuleuven.be (E. Mathijs).
http://dx.doi.org/10.1016/j.envsci.2014.02.009
1462-9011/# 2014 Elsevier Ltd. All rights reserved.
96 environmental science & policy 39 (2014) 95106
consumption in emerging countries and developing countries
is growing rapidly because of increasing economic growth
(Delgado, 2003). According to FAO projections, meat consump-
tion in these countries may increase from an average annual
per capita consumption of 10 kg in the 1960s to 26 kg in 2000,
reaching 37 kg around the year 2030. This forecast suggests
that in a few decades, developing countries consumption of
meat will move towards that of developed countries where
meat consumption remains stagnant at a high level.
From a nutritional point of view, meat is undeniably a high
quality source of proteins. Average requirements for healthy
adults are currently estimated at 0.66 g/kg body mass per day
(EFSA). However, few data are available on the mean protein
intake on a body mass weight basis. EU statistics report daily
protein consumption between 0.8 and 1.25 g/kg body weight
for adults. Main contributors to dietary protein intake are meat
and meat products, grains and grain-based products, and milk
and dairy products. The relation between health and meat
strongly depends on the type of meat considered (Wang and
Beydoun, 2009). Basically, there are major nutritional benefits
to consuming some red white meat but almost no reason
aside from tastes to consume mostly processed meats
(Popkin, 2009). Overconsumption of meat might result in a too
high intake of saturated fatty acids, which is one of the main
reasons for high incidence of coronary heart diseases. Further,
improperly processed meat and the use of chemicals in
agriculture and animal production increases the risks of food
poisoning.
Given the already very high meat consumption rates in
industrialised countries, the question raises whether there are
limits to the increase in meat consumption. In this context,
there are promising signals. On the demand side, consumer
awareness on health, animal welfare and global environmen-
tal changes might reverse the tendency of increasing meat
consumption. A study among French households showed a
positive relation between education level of the household
head and reduced meat consumption (Allais et al., 2010).
Initiatives like the emergence of vegetarian organisations and
the introduction of meatless days (such as Meatless Monday in
the USA) illustrate increased citizens environmental and
health concerns related to meat consumption. In addition,
evidence-based guidelines on a sustainable and healthy diet
have been developed. Stehfest et al. (2009) calculated that a
healthy diet worldwide would reduce the required area of
arable land globally by 10%, and the area of grassland by 40%,
compared to the FAO projections.
An example of a healthy diet is the Willett diet (Willett, 2005),
which includes 10 g beef, 10 g pork, 47 g chicken and eggs, and
23 g fish, per person, per day on average. The associated
reduction in costs for mitigation of carbon dioxide emissions
could be as large as 50% in 2050, compared to the reference case.
An increased number of governments have integrated environ-
mental issues into national dietary guidelines. Against the lobby
of food industry, some governments have integrated sustain-
ability into their national dietary recommendations. In Finland,
experiments have even been set up to analyse the short and
long run effects of a Mandatory Vegetarian Day in Helsinki
schools (Lombardini and Lankoski, 2013).
Will meat consumption continue to raise, or is there a
limit? Will health concerns or global drivers of change
motivate consumers to decrease meat consumption? Is a
second nutrition transition in which meat is substituted back
to plant based foodstuffs emerging? This paper addresses the
evolution of meat consumption and analyses the relationship
between meat consumption and income. The key hypothesis
is that meat consumption stagnates or even decreases once a
GDP threshold is reached. Little empirical evidence is reported
on the probability of an inverted U shaped relationship
between income and meat consumption. This relationship
is similar to the Environmental Kuznets Curve (EKC) hypothe-
sis. The EKC represents graphically the hypothesis that as an
economy develops environmental pollution (e.g. SO2 concen-
tration) first increases, and then decreases after a certain
average income is attained (Grossman and Krueger, 1991,
1993). One explanation for the shape of the EKC relates to the
income elasticity of environmental quality demand. When
income increases people care more about the environment
they live in. Once a certain level of income is reached, their
willingness to pay for improved environmental quality
increases by a greater proportion than income (Pezzey,
1989; Selden and Song, 1994; Baldwin, 1995). The level of
income beyond which pollution starts to decline while income
further increases, is called the turning point. While there is a
sound theoretical background for the EKC, the empirical
evidence provides some controversy (Stern, 2005; Dinda, 2004).
Research on the EKC includes, among others, the relationship
between income and CO2 emission, sulfur emission and
deforestation (Rothman, 1998; Saboori et al., 2012; Culas, 2012;
Leitao, 2010). Empirical studies confirmed an inverted U-shape
relation with income for local pollutants with a short term
impact, but not for global pollutants with an indirect and long-
term impact such as carbon dioxide. Particularly for local
pollutants that directly affect health, the U-shape relation is
confirmed. Empirical findings also suggest that the U-shaped
relationship between income and pollution is country specific
because different national regulations result in large differ-
ences in country level per capita emissions (Dinda, 2004; Stern
et al., 1996). When it comes to meat consumption, the
hypothesis for an inverted U-shaped relationship between
income and meat consumption can be motivated because
meat consumption does not only contribute to global
pollutants but also has a direct impact on consumers health.
However, the relation might be country specific due to for
example cultural and religious differences which affect meat
consumption and result in a large difference in national levels
of per capita meat consumption.
Using panel data for 120 countries in the period 19702007,
this paper empirically analyses the link between meat
consumption and income, while controlling for fixed effects
and geographical area, culture and trade. The results of the
research indicates the need for more interdisciplinary
research and arguments for improved cooperation between
environmental and health policy in industrialised and
emerging countries as well as in developing countries.
2. Determinants of meat consumption
Food choice is determined by a complex set of socioeconomic
and demographic characteristics (Frank, 2007). This also holds
 environmental science & policy 39 (2014) 95106
for the level of meat consumption and for the type of meat
consumed. According to Kearney (2010), meat consumption is
country-specific and influenced by several factors including
cultural beliefs, religion, gender, socio-economic status (SES)
and ethnic differences. This is confirmed by Wang et al. (2010),
who analysed shifts in meat consumption across population
groups in the USA. However, there is a lack of quantitative
evidence on the determinants for national meat consumption
levels (Latvala et al., 2012).
2.1. Income and meat price
From a historical perspective, increased meat consumption
is associated with an increase of the nutritional value of
diets. Kearney (2010) illustrates how worldwide increased
incomes and lower food prices have led to increased
consumption of animal-based foods. In Europe and North
America, this explains the growth of meat consumption over
the past century. In the newly industrialised countries, such
as China and Brazil, this tendency is observed more recently.
Moreover, as the price of meat decreases, meat becomes
more easily accessible in developing countries (Tey et al.,
2010). This holds in particular for cheap, highly processed
low quality meat. In this context, it should be noted that
meat is not a homogeneous commodity with considerable
price differences between the different types of meat. The
income elasticity of meat has been studied intensely. A most
recent meta-analysis on the income elasticity of different
types of meat is carried out by Gallet (2010). For all types of
meat, price elasticity is greater than one. Park et al. (1996)
focused on the price elasticity of beef, pork and poultry. The
results confirmed that income elasticity is positive though
less than one. Furthermore, the income elasticity was higher
for people in poverty than for higher income households
(Park et al., 1996). Apparently, meat consumption increases
with income, but the rate of increase tends to flatten at
higher income levels. This can be explained by the fact that
middle and high income class consumers are more aware of
the impact of overconsumption in general, and meat
consumption in particular, on the risk for obesity and other
chronic diseases and of the environmental impact of meat
production.
Analyses of three Dutch National Food Consumption
Surveys, for example, showed that subjects with low SES
report significant higher meat consumption (Hulshof et al.,
2003). This is in line with Mullie et al. (2010), who concluded
from a cross-sectional design in 1852 military men that the
most healthy patterns were associated with a higher socio-
economic position, while lower positions were associated with
several unhealthy behaviours.
However, income growth either at the individual level or
at the national level may rather lead to a shift in the type of
meat consumed than to an overall reduction of meat
consumption. Chamorro et al. (2012), for example, conclud-
ed from a Delphi research that consumers in Spain tend to
replace beef and veal by poultry meat. Further, their results
showed an increase in the consumption of ready meals with
a meat base, whereas overall meat consumption levels in
Spain are expected to remain stable in the short run (up to
2016).
97
2.2. Geographical area
Access to open waters also has an impact on meat consump-
tion as dependence on fish is usually higher in coastal than in
inland areas. For most islands, fish rather than meat
composes the main source of proteins. The highest fish
consumption worldwide is found in Iceland, Japan and some
small island states (FAO, 2012; WHO, 2013). In this sense, one
can expect higher meat consumption in landlocked regions or
countries. The geographic term landlocked refers to a
country with no coast that is surrounded by other countries.
This status affects a countrys trade position and economic
development. Goods produced in a country without access to
open water, for example, must always be transported through
another country. As they depend on neighbouring countries
for imports and exports, they are considered as more
vulnerable as compared to non-landlocked developing coun-
tries. Furthermore, focusing on developing countries, it should
be noted that landlocked countries often have a lower GDP per
capita (Dia Kamgnia, 2011).
2.3. Food crisis
Over the last two decades, the meat sector has been struck by
several crises including the Bovine Spongiform Encephalopa-
thy (BSE) in Europe, the avian flu (H5N1) in several countries
worldwide, the dioxin crisis in Belgium and the Escherichia coli
(EHEC) outbreak in Germany. However, while these crises tend
to have considerable impact on consumer perception of meat
quality and safety (Grunert, 2006; Verbeke and Viaene, 1999),
considerations about food-safety perceptions do not seem to
affect meat consumption in the long run (Zingg et al., 2013).
Results from the European project ProSafeBeef confirm the
mismatch between citizens attitude and consumer behav-
iour. Focusing on pork consumption for example, Verbeke
et al. (2010) concluded that peoples views as citizens do not
appear to systematically influence their consumption choices.
2.4. Cultural differences and religion
Culture has an impact on food consumption behaviour in
general, and meat consumption in particular (Haverstock and
Forgays, 2012; Kearney, 2010). Even when income differences
between countries disappear, consumer behaviour will not
homogenise because consumer behaviour will be heteroge-
nous because of cultural differences (de Mooij and Hofstede,
2002). Hence, values of national cultures might have a strong
impact on consumer behaviour. Culture has been shown to
influence for example the consumption of mineral water and
soft drinks, fresh fruit, ice cream and frozen food (de Mooij,
1997, 1998, 2000). Hofstedes dimensions of national culture
can be useful for statistically analysing consumption differ-
ences across nations (de Mooij and Hofstede, 2002). The
Hofstede indices on cultural dimensions describe the effects of
a societys culture on the values of its members, and how these
values relate to behaviour. Hofstede (2001) derived these
indices and dimensions of national culture based on several
large-scale surveys among in total 88,000 IBM employees from
72 countries between 1967 and 1973. Employees were asked to
express their attitude with respect to a set of values-related
98 environmental science & policy 39 (2014) 95106
questions on a scale from 1 to 5. Initially, using factor and
cluster analysis he distinguished five dimensions of national
culture: power distance, individualism versus collectivism,
uncertainty avoidance, masculinity versus femininity and
long-term versus short-term orientation. Both the indices and
the dimensions have been validated repeatedly in later studies
(Hofstede et al., 2010), while other researchers have compared
Hofstedes approach to other approaches in the literature (see
e.g. Soares et al., 2007). The values do not change over time, as
cultural dimensions are relatively fixed, changing only very
slowly. This has also been validated by later research
(Hofstede et al., 2010).
The power distance index (PDI) measures the extent to
which the less powerful members of organisations and
institutions (like the family) accept and expect that power is
distributed unequally. This represents inequality (more versus
less), but defined from below, not from above. It suggests that a
societys level of inequality is endorsed by the followers as
much as by the leaders.
Individualism (IDV) is the degree to which individuals are
integrated into groups. Its opposite side is collectivism. On the
individualist side we find societies in which the ties between
individuals are loose: everyone is expected to look after him/
herself and his/her immediate family. On the collectivist side,
we find societies in which people from birth onwards are
integrated into strong, cohesive in-groups, often extended
families (with uncles, aunts and grandparents) which contin-
ue protecting them in exchange for unquestioning loyalty. The
word collectivism in this sense has no political meaning: it
refers to the group, not to the state. Again, the issue addressed
by this dimension is an extremely fundamental one, regarding
all societies in the world.
Uncertainty avoidance index (UAI) refers to a societys
tolerance for uncertainty and ambiguity. It ultimately
refers to mans search for Truth. It indicates to what extent
a culture programs its members to feel either uncomfort-
able or comfortable in unstructured situations. Unstruc-
tured situations are novel, unknown, surprising, and
different from usual. Uncertainty avoiding cultures try to
minimise the possibility of such situations by strict laws
and rules, safety and security measures, and on the
philosophical and religious level by a belief in absolute
Truth; there can only be one Truth and we have it. People
in uncertainty avoiding countries are also more emotional,
and motivated by inner nervous energy. The opposite type,
uncertainty accepting cultures, are more tolerant of
opinions different from what they are used to; they try to
have as few rules as possible, and on the philosophical and
religious level they are relativist and allow many currents
to flow side by side. People within these cultures are more
phlegmatic and contemplative, and not expected by their
environment to express emotions.
The masculinity index (MI) measures the distribution of
emotional roles between the genders. It refers to the
distribution of roles between the genders which is another
fundamental issue for any society to which a range of
solutions are found. Its opposite is femininity. The IBM
studies carried out by Hofstede revealed that (a) womens
values differ less among societies than mens values; (b) mens
values from one country to another contain a dimension from
very assertive and competitive and maximally different from
womens values on the one side, to modest and caring and
similar to womens values on the other. The assertive pole has
been called masculine and the modest, caring pole femi-
nine. The women in feminine countries have the same
modest, caring values as the men; in the masculine countries
they are somewhat assertive and competitive, but not as much
as the men, so that these countries show a gap between mens
values and womens values.
Despite the evidence for the impact of culture on
consumers attitudes and behaviour with respect to food
consumption, the use of the Hofstede index in food studies is
very limited. One exception is the study of Gaither and Courtin
(2007), who analysed the impact of the Danish publication on
the prophet Muhammad on the dairy company, Arla Foods.
This case study emphasises the strength of the PDI as a tool for
international public relation work and the need to apply
cultural normative tools in the development of theory of
international public relations practice.
Finally, the importance of religion on meat consumption
cannot be denied (Bonne et al., 2007). Pork is forbidden in
Judaism and Islam; pork and beef are forbidden in Hinduism
and Buddhism. Only in Christianity, there are no rules related
to meat consumption.
2.5. Trade
Overall, openness to trade in an economy has an impact on
technological development, innovation and competitive-
ness. This holds true for both developed and developing
countries (e.g. Umoh and Effiong, 2013). Focusing on food,
globalisation has drastically changed food supplies, leading
to a situation with worldwide and year-round availability of
all types of foods. Openness to trade also has its impact on
meat supplies and hence may influence meat consumption.
Limits to trade such as geographical isolation, conflicts or
trade policy on the contrary lead to reduced overall meat
supply on the local or national market and the a limited
choice in types of meats.
3. Data and methods
3.1. Model
Regression analysis is used to understand the relationship
between meat consumption and income. The model is
described by the equation
yit a bk xkit g m zmi eit (1)
where yit, the dependent variable, is the indicator of meat
consumption in country i and year t, xkit is a vector of the
time-varying explanatory variables, zmi is a vector of time-
invariant explanatory variables and eit is the idiosyncratic
error term.
First, we use a pooled ordinary least square (OLS) model in
order to analyse the relationship between the dependent and
explanatory variables, between countries and over time. To
account for heteroscedasticity and serial correlation, we use
cluster robust standard errors in all of our models.
 environmental science & policy 39 (2014) 95106
Secondly, we control for unobserved country fixed effects.
We start from the unobserved effects model which is given by
Eq. (2) as
yit ci dt bk xkit uit
where ci the unobserved country level effect, constant over
time (fixed effect), where the coefficient dt represents the time
dummies, included to capture possible time effects and where
uit represents the idiosyncratic error. Initially we estimate a
random effects model. This implies that we assume that the
unobserved effect ci is uncorrelated with each explanatory
variable in all time periods. Next, we consider the possibility
that the unobserved effect is correlated with the explanatory
variables and eliminate the unobserved effect by using a fixed
effects estimator (within estimator) and by using a first differ-
ence estimator. In this way, we control for the fact that there
are some country specific effects, that are time invariant and
correlated with our explanatory time varying variables1
(Wooldridge, 2002).
However, with this elimination all the time invariant
variables will be dropped out of our model. Consequently, we
cannot estimate the coefficient of these variables using a fixed
effects or first differencing model. For that reason we use the
fixed effect method with vector decomposition (XTFEVD),
developed by Plumper and Troeger (2007). The XTFEVD model
is based on a three steps estimation procedure. The first step is
a fixed effects estimation of the time varying variables. The
estimated country effects are a vector of omitted variables and
the effect of the time invariant variables. In the second stage,
the vector of fixed effects is decomposed into an explained
part (by the time invariant variables) and an unexplained part
(residuals), using OLS. The third stage is a pooled OLS
estimation using all the time variant explanatory variables,
the time invariant variables and the unexplained part of the
fixed effects vector (Plumper and Troeger, 2007).
3.2. Data
3.2.1. Meat consumption
Ideally, data from household-level surveys on diets should be
used as dependent variable. However, this kind of data is very
scarce and does not allow for identifying trends. Therefore, we
used the FAO Food Balance Sheets (FBS) which calculate
consumption as the residual resulting from balancing pro-
duction, stocks, feed use, exports and imports. That is:
consumption production stocks imports  exports
 feed use
Per capita meat consumption is then simply consumption
per year divided by total population (kg/capita/year). Con-
sumption calculated in this way tends to overestimate
consumption, as it assumes that all available food is actually
consumed (Kearney, 2010). However, FBS are available for a
large number of countries and for a long term period. In this
analysis, data on 120 countries over the period 19612007 are
used. We use data on total meat consumption, which equals
the sum of all categories of meat (bovine meat, mutton and
goat meat, pig meat, poultry meat, meat other, etc.).
1
Eyit jci xit ci bx0it ; or Covci xit 6 0; t 1; 2; 3; 4 . . . T.
99
3.2.2. Income
Income is measured in nominal per capita gross domestic
product (GDP/cap). GDP is the gross domestic product divided
by midyear population. GDP is the sum of gross value added by
(2)
all resident producers in the economy plus any product taxes
and minus any subsidies not included in the value of the
products. It is calculated without making deductions for
depreciation of fabricated assets or for depletion and
degradation of natural resources. Data are in current U.S.
(in thousands). To check for robustness, we check whether our
results hold when using GDP per capita based on purchasing
power parity (PPP), which is the GDP converted to international
dollars using purchasing power parity rates. An international
dollar has the same purchasing power over GDP as the US
dollar has in the United States. Data on GDP per capita
expressed in purchasing power parity are in constant 2005
international dollars (GDP_con/cap). GDP in PPP terms is a
preferable income indicator but data are only available from
1980 onwards. Income data were obtained from the World
Bank and are available for the period 19612009 for GDP in
current US dollars and for the period 19802009 for GDP in
constant 2005 international dollars (World Development
Indicators, World Bank, 2011).
3.2.3. Geographical area
We created a dummy variable for landlocked countries,
LANDLOCK, with the value 1 indicating that the country has
no coastal area and 0 if the country has a coastal area.
Information was obtained from the World Factbook of the
Central Intelligence Agency (CIA). Further, in accordance with
studies on the EKC hypothesis (e.g. Culas, 2012), a dummy for
the continent is applied.
3.2.4. Cultural differences and religion
Four variables on culture are used in the model as well as one
variable to capture religion. The selection of variables on
culture is based on Hofstede (2001). It concerns the power
distance index (PDI), the uncertainty avoidance index (UAI),
the individualism index (II), and the masculinity index (MI).
The Hofstede dimension was available for 52 countries
(Hofstede, 2001). To measure the impact of religion on meat
consumption we use data from the religion adherence
database. We used the share of the population adhering a
specific religion in 1970. By aggregating some of the adherence
fractions, we ended up with four aggregated religion catego-
ries: Christians (RELCHR) including Catholics, Protestants,
Orthodox and other Christians, Muslims (RELMUS), Eastern
Religions (RELEAS) including Hindu, Buddhist, and other
Eastern Religions, and Other Religions (RELOTH) including
Judaism, other religions and non-religion.
3.2.5. Trade
The explanatory variable TRADE is included to capture the
effect of a countrys openness. TRADE is calculated as the sum
of the exports and imports of goods and services relative to the
countrys GDP and is based on World Bank data. By this
variable we aim to capture potential effects of globalisation
and openness on meat consumption as it might affect
consumption patterns directly as more consumer goods
become available, but also indirectly, as it will affect the
100 environmental science & policy 39 (2014) 95106

Fig. 1 Evolution of meat consumption measured in kg/cap/year in Argentina, Brazil, France, Russian Federation, China and
United States, period 19802007.
Source: FAO (2012).

impact of local cultural and religious values on consumer
behaviour.
4. Results
Our analysis is based on an unbalanced data set of 120
countries. For most countries (103), we used data for a period
of 38 years (19702007). For post-USSR countries, we used data
for the period 19922007 and for Belgium and Luxembourg for
the period 20002007. However, not all observations could
always be included in the regression estimates as some
explanatory variables such as the Hofstede index, GDP/cap
expressed in 2005 constant international dollars, were only
available for a subset of the countries or a smaller time period.
Table 1 in the appendix describes the countries and years
included in the dataset that used for estimating the different
regression models.
remains below 50 kg/cap/year. Figures for African countries
suggest broad differences in meat consumption across the
continent, with 5 kg/cap/year in Burundi, around 10 kg/cap/
year in Cameroon and Senegal and almost 50 kg/cap/year in
South African (FAO, 2012).
Fig. 2 presents the relationship between meat consumption
and income, whereby income is measured in GDP/cap in
constant 2005 international dollars. For both cases, the
relationship between meat consumption and income is not
linear. Therefore, GDP in squared terms will be included as
explanatory variable into the empirical model (see further).
This is done for both GDP per capita in current US dollars and
for GDP per capita in constant 2005 international dollars,
resulting in variables (GDP/cap)2 and (GDP_con/cap).
Table A2 of the appendix presents the summary statistics
of the other variables used in the regression analysis.

4.1. Descriptive statistics

150
Meact consumption (kg/capita/year)

Fig. 1 depicts the evolution of meat consumption, measured in

kg/cap/year, in a selected number of countries in the period
100

19802007. From this graph, some global tendencies can be

derived. The United States has by far the highest meat
consumption, although the level of meat consumption tends
to stabilise. In France, there is a decreasing trend, observed
50

since the beginning of this century. Both Brazil and China are
characterised by a dramatic increase in meat consumption per
capita. In Brazil, meat consumption has increased from 40 kg/
cap/year in 1980 to over 80 kg/cap/year in 2007. In China, the
0

meat consumption almost tripled in this period, up to a 0 20 40 60 80

GDP/cap on PPP in constant 2005 (1000) US$
consumption of 53 kg/cap in 2007. Meat Consumption (kg/capita/year) Fitted values
While meat consumption in China increases rapidly, the
average annual meat consumption in a number of other Asian
countries remains extremely low. In India and Bangladesh, for Fig. 2 Relationship between meat consumption and
example, meat consumption is still below 5 kg/cap/year. In income per capita.
Japan, the average meat consumption increases slowly and Source: FAO (2012).
 environmental science & policy 39 (2014) 95106 101

4.2. Regression results well as within countries over time. We calculated the turning
point in income at which meat consumption starts to decline.
The regression analysis starts from Eq. (1). Taking into account Depending on the specification, this turning point in income lies
the aforementioned remark on the non-linear relationship at a GDP per capita between 32 thousands and 55 thousand
between meat consumption and income, meat consumption current US dollars or between 35 thousands and 53 thousands
(MC) can be rewritten as2 constant 2005 international dollars.
    As expected, religious persuasion affects meat consumption.
GDP GDP 2
MCit a b1 b2 b TRADEit Higher shares of Christians in a country corresponds to higher
cap it cap it 3
meat consumption ceteris paribus Contrarily, in countries with a
g 1 LANDLOCKi g 2 PDIi g 3 UAIi g 4 MIi higher share of the population adhering Eastern religions (Hindu,
g 5 REGCHRi g 6 REGEASi g 7 REGOTHi eit (3a) Buddhists, etc.) meat consumption is significantly lower.
Further, there is a positive correlation between the
or masculinity index (MI) and meat consumption, but the
    significance of the variable depends on the specification used.
GDP con GDP con 2
MCit a b1 b2 b3 TRADEit A possible explanation is that meat consumption is often
cap it cap it
associated with tradition, but also masculinity, strength and
g 1 LANDLOCKi g 2 PDIi g 3 UAIi g 4 MIi power (Heinz and Lee, 1998). The negative coefficient of power
g 5 REGCHRi g 6 REGEASi g 7 REGOTHi eit (3b) distance index (PDI) indicates that in societies with high
inequality among citizens, meat consumption per capita is
The share of Muslims in the population (REGMUS) was lower than in societies with low inequality. However, the
highly correlated with the share of other religions (REGOTH), coefficient of the power distance index is only significantly
while the individualism index (IDV) and the uncertainty different from zero in the specifications with income
avoidance index (UAI) were highly correlated with other expressed in current US dollars (Table 1) and not when
cultural dimensions of Hofstede. Therefore, the variables income is expressed in constant 2005 international dollars
REGMUS, IDV and UAI were not included as explanatory (Table 2). A possible explanation is that meat is still a luxury
variables in the regression analysis. In an analogous way, good with a high social status in many societies so that meat
Eq. (2) can be rewritten as consumption is less accessible to a broad category of people in
    cultures with a high degree of inequality.
GDP GDP 2
MCit ci dt b1 b2 b TRADEit uit (4a) The coefficients of the variable TRADE which measures the
cap it cap it 1
countrys openness, is only in a few specifications significantly
or different from zero. This means that while the sign of the
  coefficient is as expected, we are not able to make any strong
GDP con
MCit ci dt b1 statements on the impact of a countrys openness on its
cap it
  consumer behaviour.
GDP con 2
b2 b1 TRADEit uit (4b) Finally, the results indicate that the majority of developing
cap it countries will only reach the turning point in the very long run.
However, the establishment of a new reference framework
Tables 1 and 2 present the results of the regression analyses and policy benchmarking might lead to a situation in which
whereby income is expressed in respectively GDP in current US these countries dive through the EKC (Dinda, 2004). Cultural
dollars and in constant 2005 international dollars.3 The and religious differences together with a new reference
empirical analysis clearly confirms an inverted U-shaped frameworks e.g. driven by health concerns and policy
relationship between per capita meat consumption and per benchmarking might thus lead to situation in which these
capita income. This result holds for all of our model specifica- countries will never reach a per capita meat consumption as
tions. Our results indicate that, from a certain level of the more developed countries in our dataset.
development, per capita meat consumption starts to decrease
with income. As countries become richer they consume more 4.3. Testing for unobserved heterogeneity, model
meat, but when income continues to grow meat consumption specification and strict exogeneity
decreases. This tendency is confirmed between countries as
In presence of unobserved heterogeneity our OLS estimators
2
We do not control for the impact of meat price on meat
(model A in Tables 1 and 2) are biased because of the
consumption because only a worldwide meat price index was correlation of the error term and the explanatory variables.
available. We also do not control for food crises because health When testing for the presence of unobserved heterogeneity in
scares are only having a local and short term impact on meat the different model specifications, the null hypothesis that
consumption. there is no unobserved heterogeneity was rejected, meaning
3
We started by including only the key income variable in the that we have to control for unobserved effects which is done in
regression models (see Appendix), and continued by including
modals B, C, D and E of Tables 1 and 2.
additional control variables as explanatory variables. Due to data
After controlling for unobserved heterogeneity, we tested
availability, the number of observations included in the different
specification changes considerably. In order to arrive at compara- whether country effects are random using the Hausman test.
ble results, we re-estimated the regression specifications using an The null hypothesis was rejected which indicates that we have
identical limited dataset. to use a fixed effects estimator, a first differencing estimator or
102 environmental science & policy 39 (2014) 95106

Table 1 Regression analyses for meat consumption, with income expressed in GDP in thousand current US dollars.
Model A Model B Model C Model D Model
pooled OLS random first difference fixed effects estimator E XTFEVD
effects estimators (within estimator)
GDP/cap 1.784*** 1.104*** 0.515*** 0.461* 1.100***
(0.420) (0.259) (0.143) (0.265) (0.162)
(GDP/cap) 2 0.0185** 0.0133*** 0.00466*** 0.00737* 0.0134***
(0.00797) (0.00442) (0.00142) (0.00374) (0.00270)
TRADE 0.0713 0.121** 0.0173 0.0286 0.122***
(0.0586) (0.0529) (0.0187) (0.0559) (0.0384)
LANDLOCKED 3.491 13.62 1.348
(9.037) (10.54) (10.29)
PDI 0.168 0.248* 0.280**
(0.137) (0.140) (0.117)
MI 0.377*** 0.315*** 0.395***
(0.105) (0.118) (0.125)
RELCHR 14.75* 20.62** 16.93*
(8.688) (8.270) (10.16)
RELEAS 12.88 7.163 5.948
(8.990) (7.484) (14.44)
RELOTH 44.37*** 41.13*** 44.16***
(12.55) (14.04) (13.23)
Africa 23.68*** 23.36*** 27.40**
(7.000) (7.435) (11.97)
Asia 24.46*** 21.30** 27.38***
(8.648) (9.656) (9.739)
North America 7.074 9.829 9.692
(9.760) (10.97) (6.282)
South America 3.532 5.074 6.274
(9.222) (9.280) (6.266)
Oceania 29.87*** 31.90*** 28.99***
(4.529) (5.086) (10.72)
Constant 21.94** 26.66*** 43.55*** 29.00**
(10.74) (10.33) (3.153) (12.57)

Time dummies No No Yes Yes No

Observations 1812 1812 1761 1761 1812
R2-overall 0.716 0.692 0.057 0.167 0.935
R2 fixed effects 0.269 0.3359
R2 between 0.736 0.3049

Income turning point $48,216 $41,504 $55,258 $31,275 $41,045

t-Values in parentheses and based on robust standard errors, corrected for clustering at country level. Overall F-statistics are significant at 5%
levels for all models.
*
p < 0.1.
**
p < 0.05.
***
p < 0.01.

a fixed effects estimator with vector composition as in models
C, D and E of Tables 1 and 2.
Finally, we tested the strict exogeneity assumption on
which the first difference estimator (model C in Tables 1 and 2)
and the fixed effects estimator (model D in Tables 1 and 2) rely.
These tests revealed that all variables are strictly exogenous
when the income variable is expressed in constant 2005
international dollars (see Table 3). However, when GDP is
expressed in current US dollars, the variable TRADE does not
pass the strict exogeneity.
While an inverted U-relationship between meat consump-
tion and income is observed for all specifications, these tests
indicate that models C, D and E in Table 2 provide the best
coefficient estimates. This is in line with our assumption as
GDP in PPP terms is in general a preferable income measure
and as one might expect the country fixed effects to be
correlated with the explanatory variables.
4.4. Income elasticity of meat consumption
Since it is useful to interpret the relationship between
income and meat consumption as elasticities, we have split
the data in two groups. One group holds the observations
with income below the turning point, while the other holds
the observations with income above the turning point. On
these two groups, we have run a double log model and
provide the elasticities in Table 4. Elasticities are only
provided for the fixed effect within estimator model (i.e.
model D). Model D was chosen because our empirical
analysis indicated that a fixed model was preferred above
the pooled OLS and random effects model and because the
turning point was lowest for the fixed effects within
estimator regression. The latter was particularly important
to have sufficient observation in the group that holds
observation above the turning point.
 environmental science & policy 39 (2014) 95106 103

Table 2 Regression analyses for meat consumption, with income expressed in thousand constant 2005 international
dollars.
Model A Model B Model C Model D fixed Model
pooled OLS random effects first difference effects estimator E XTFEVD
estimators (within estimator)
GDP/cap 2.949*** 3.578*** 3.014*** 3.013*** 3.767***
(0.606) (0.653) (0.498) (0.730) (0.554)
(GDP/cap) 2 0.0280** 0.0461*** 0.0311*** 0.0430*** 0.0501***
(0.0110) (0.0116) (0.00861) (0.0118) (0.00943)
TRADE 0.0580 0.0224 0.0497** 0.0751 0.0348
(0.0427) (0.0550) (0.0616) (0.0707)
LANDLOCKED 6.757 32.59* 17.75*
(11.39) (17.75) (9.874)
PDI 0.0432 0.0920 0.0930
(0.123) (0.130) (0.138)
MI 0.320*** 0.140 0.223*
(0.111) (0.139) (0.122)
RELCHR 8.220 10.70 9.324
(8.667) (9.415) (9.695)
RELEAS 23.43* 21.41* 24.69
(13.38) (12.51) (15.07)
RELOTH 34.98*** 31.87*** 32.98***
(10.47) (10.22) (12.64)
Africa 10.63 7.731 9.298
(7.124) (7.205) (10.71)
Asia 10.89 8.128 8.840
(8.207) (7.325) (9.695)
North America 0.672 1.378 1.699
(6.834) (6.673) (5.748)
South America 6.401 4.383 3.774
(8.753) (8.282) (6.122)
Oceania 27.37*** 28.21*** 25.92**
(4.235) (5.171) (10.58)
Constant 6.424 0.888 29.707*** 3.928
(12.248) (12.686) (6.642) (13.635)

Time dummies No No Yes Yes No

Observations 1352 1352 1301 1301 1352
R2-overall 0.784 0.7490 0.092 0.5548 0.956
R2 fixed effects 0.3327 0.3646
R2 between 0.7514 0.3886

Income turning point $52,661 $38,801 $48,457 $35,035 $37,595

t-values in parentheses and based on robust standard errors, corrected for clustering at country level. Overall F-statistics are significant at 5%
levels for all models.
*
p < 0.1.
**
p < 0.05.
***
p < 0.01.

Table 3 Strict exogeneity test results ( p-values) for models C and D in Tables 1 and 2.
Model C Model D Model C Model D
first difference fixed effects first difference fixed effects
estimators estimator estimators estimator
(within estimator) (within estimator)
GDP/cap 0.858 0.858
(GDP/cap) 2 0.673 0.009
GDP_con 0.790 0.229
(GDP_con/cap) 2 0.946 0.890
TRADE 0.069 0.570 0.583 0.175
104 environmental science & policy 39 (2014) 95106

Table 4 Fixed effects within estimator regression for meat consumption, with double log specification for observations
with income above and below the income turning point.
GDP expressed in thousands current US GDP expressed in thousands 2005 in-
dollars ternational dollars

Below income Above income Below income Above income

turning point turning point turning point turning point
Ln(GDP/cap) 0.279*** 0.0231
(0.103) (0.130)
lnTRADE 0.150 0.00556
(0.122) (0.175)
Ln(GDP_con/cap) 0.512*** 1.244***
(0.138) (0.316)
lnTRADE 0.0743 0.0846
(0.100) (0.102)
Constant 2.938*** 4.572*** 2.955*** 9.299***
(0.446) (1.098) (0.435) (0.873)

Observations 1700 112 1303 49

R-squared 0.504 0.361 0.487 0.539
Number of countries 50 19 50 8
Turning point $31,275 $35,035
Robust standard errors, corrected for clustering at country level in parentheses.
*
p < 0.1.
**
p < 0.05.
***
p < 0.01.

Table 4 indicates that for countries below the turning meat
consumption increases with 0.3% if income increases with 1%
when GDP is expressed in current US dollars, while it increases
with 0.5% when GDP is expressed in PPP terms. For countries
above the income turning point, meat consumption decreases
with 1.2% if income increases with 1% when GDP is expressed
in PPP terms.
However, Allais et al. (2010) described how taxes will stimulate
the food industry to innovate and possibly undermine the
impact of the taxes. Examples of legal instruments to
stimulate reduced meat consumption can be found in Finland
where experiments have been undertaken with respect to
forced choice restriction in the format of a Mandatory
Vegetarian Day (Lombardini and Lankoski, 2013).

5. Conclusions
In this paper we studied the evolution of meat consumption
within countries and determined its relationship with income
and other variables. The results of our empirical analysis
confirm that the relationship between meat consumption and
income is non-linear and has an inverted U shape. Initially
meat consumption increases with income, but from a certain
point onwards higher levels of income lead to lower levels of
meat consumption. The results also indicate that culture and
religion can explain differences in meat consumption within
and between countries.
Since livestock production is a major cause of greenhouse
gas emissions (Garnett, 2009), this is a this promising signal.
However, many countries are still far away from the turning
point. Therefore, the findings should not demotivate govern-
ments to undertake actions in order to accelerate the
reduction of meat consumption. Decreasing meat consump-
tion can be accelerated either by using economic instruments
(e.g. meat tax), by social instruments (e.g. educational
campaigns) or by legal instruments. Nordgren (2011) argued
that developed countries and, to increasing extent, emerging
countries, should thereby take the lead. In his ethical
discourse, Nordgren (2011) outlined arguments in favour of
a meat consumption tax, rather than a meat production tax.
Appendix A. Supplementary data
Supplementary data associated with this article can be
found, in the online version, at http://dx.doi.org/10.1016/
j.envsci.2014.02.009.
references
Allais, O., Bertail, P., Nichele, V., 2010. The effects of a fat tax on
French households purchases: a nutritional approach. Am.
J. Agric. Econ. 92 (1) 228245.
Baldwin, R., 1995. Does sustainability require growth? In:
Goldin, I., Winters, L.A. (Eds.), The Economics of Sustainable
Development. Cambridge Univ. Press, Cambridge, UK, pp.
1947.
Bonne, K., Vermeir, I., Bergeaud-Blackler, F., Verbeke, W., 2007.
Determinants of halal meat consumption in France. Br. Food
J. 109 (5) 367386.
Chamorro, A., Miranda, F.J., Rubio, S., Valero, V., 2012.
Innovations and trends in meat consumption: an application
of the Delphi method in Spain. Meat Sci. 92, 816822.
Culas, R.J., 2012. REDD and forest transition: tunneling through
the environmental Kuznets curve. Ecol. Econ. 79, 4451.
 environmental science & policy 39 (2014) 95106
de Mooij, M., 1997. Mapping cultural values for global marketing
and advertising. Paper presented at ESOMAR, 50th Congress.
Edinburgh, Scotland. .
de Mooij, M., 1998. Global Marketing and Advertising:
Understanding Cultural Paradoxes. Sage, Thousand Oaks,
CA.
de Mooij, M., 2000. The future is predictable for international
marketers: converging incomes lead to diverging consumer
behaviour. Int. Market. Rev. 17, 103113.
de Mooij, M., Hofstede, G., 2002. Convergence and divergence in
consumer behavior: implications for international retailing.
J. Retailing 78, 6169.
Delgado, C.L., 2003. Rising consumption of meat and milk in
developing countries has created a new food revolution. J.
Nutr. 133 (11) 3907S3910S.
Dia Kamgnia, B., 2011. Political economy of recent global food
price shocks: gainers, losers and compensatory mechanism.
J. Afr. Econ. 20 (AERC Supplement 1) i142i210.
Dinda, S., 2004. Environmental Kuznets curve hypothesis: a
survey. Ecol. Econ. 49, 431455.
FAO, 2012. The State of World Fisheries and Aquaculture. FAO,
Rome.
Frank, J., 2007. Meat as a bad habit: a case for positive feedback
in consumption preferences leading to lock-in. Rev. Soc.
Econ. 65 (3) 319348.
Gaither, T.K., Courtin, P.A., 2007. Examining the heuristic value
of models of international public relations practice: a case
study of the Arla foods crisis. J. Publ. Relat. Res. 20 (1) 115
137.
Gallet, C.A., 2010. Meat meets meta: a quantitative review of the
price elasticity of meat. Am. J. Agric. Econ. 92 (1) 258272.
Garnett, T., 2009. Livestock-related greenhouse gas emissions:
impacts and options for policy makers. Environ. Sci. Policy
12, 491503.
Grossman, G.M., Krueger, A.B., 1991. Environmental impacts of
the North American Free Trade Agreement. NBER. Working
paper 3914.
Grossman, G.M., Krueger, A.B., 1993. In: Garber, P.D. (Ed.),
Environmental impacts of the North American Free Trade
Agreement. The U.S. Mexico Free Trade Agreement. MIT
Press, Cambridge, pp. 1356.
Grunert, K.G., 2006. Future trends and consumer lifestyles with
regard to meat consumption. Meat Sci. 74, 149160.
Haverstock, K., Forgays, D.K., 2012. To eat or not to eat. A
comparison of current and former animal product limiters.
Appetite 58 (3) 10301036.
Heinz, B., Lee, R., 1998. Getting down to the meat: the symbolic
construction of meat consumption. Commun. Stud. 49, 8699.
Hofstede, G., 2001. Cultures Consequences: Comparing Values,
Behaviors, Institutions, and Organizations Across Nations,
2nd ed. Sage, Thousand Oaks, CA.
Hofstede, G., Hofstede, G.-J., Minkov, M., 2010. Cultures and
Organizations: Software of the Mind, 3rd ed. McGraw-Hill,
USA.
Hulshof, K.F.A.M., Brussaard, J.H., Kruizinga, A.G., Telman, J.,
Lowik, M.R.H., 2003. Socio-economic status, dietary intake
and 10 y trends: the Dutch National Food Consumption
Survey. Eur. J. Clin. Nutr. 57 (1) 128137.
Kearney, J., 2010. Food consumption trends and drivers. Phil.
Trans. R. Soc. B 365, 27932807.
Kuznets, S., 1955. Economic growth and income inequality. Am.
Econ. Rev. 45, 128.
Latvala, T., Niva, M., Makela, J., Pouta, E., Heikkila, J., Kotro, J.,
Forsman-Hugg, S., 2012. Diversifying meat consumption
patterns: consumers self-reported past behaviour and
intentions for change. Meat Sci. 92 (1) 7177.
Leitao, A., 2010. Corruption and the environmental Kuznets
curve: empirical evidence for sulfur. Ecol. Econ. 69,
21912201.
105
Lombardini, C., Lankoski, L., 2013. Forced choice restriction in
promoting sustainable food consumption: intended and
unintended effects of the mandatory vegetarian day in
Helsinki schools. J. Consum. Policy 36, 159178.
Mullie, P., Clarys, P., Hulens, M., Vansant, G., 2010. Dietary
patterns and socioeconomic position. Eur. J. Clin. Nutr. 64,
231238.
Nordgren, A., 2011. Ethical issues in mitigation of climate
change: the option of reduced meat production and
consumption. J. Agric. Environ. Ethics 25, 563584.
Park, J.L., Holcomb, R.B., Raper, K.C., Capps, O., 1996. A
demand systems analysis of food commodities by U.S.
households segmented by income. Am. J. Agric. Econ. 78 (2)
290300.
Pezzey, J.C.V., 1989. Economic analysis of sustainable growth
and sustainable development. Environment Department
working paper 15. World Bank. .
Plumper, T., Troeger, V., 2007. Efficient estimation of time-
invariant and rarely changing variables in finite sample
panel analyses with unit fixed effects. Polit. Anal. 15, 124
139.
Popkin, B.M., 2006. Global nutrition dynamics: the world is
shifting rapidly toward a diet linked with noncommunicable
diseases. Am. J. Clin. Nutr. 84 (2) 289298.
Popkin, B.M., 2009. Global nutrition dynamics: the world is
shifting rapidly toward a diet linked with non-
communicable diseases. Am. J. Clin. Nutr. 84, 289298.
Rothman, D.S., 1998. Environmental Kuznets curves real
progress or passing the buck? A case for consumption-based
approaches. Ecol. Econ. 25, 177194.
Saboori, B., Sulaiman, J., Mohd, S., 2012. Economic growth and
CO2 emissions in Malaysia: a cointegration analysis of the
environmental Kuznets curve. Energy Policy 51, 184191.
Selden, T., Song, D., 1994. Environmental quality and
development: is there a Kuznets curve for air pollution
emissions? J. Environ. Econ. Manage. 27, 147162.
Soares, A.M., Farhangmehr, M., Shoham, A., 2007. Hofstedes
dimensions of culture in international marketing studies. J.
Business Res. 60, 277284.
Stehfest, E., Bouwman, L., Van Vouren, D.P., Den Elzen, M.G.J.,
Eickhout, B., Kabat, P., 2009. Climate benefits of changing
diet. Climatic Change 95, 83102.
Stern, D., 2005. The rise and fall of the environmental Kuznets
curve. World Dev. 32, 14191439 (August 2004).
Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growth
and environmental degradation: the environmental Kuznets
curve and sustainable development. World Dev. 24, 1151
1160.
Tey, Y.S., Shamsudin, M.N., Mohamed, Z., Abdullah, A.M.,
Radam, A., 2010. Demand analysis of meat in Malaysia. J.
Food Prod. Market. 16, 199211.
Tukker, A., Huppes, G., Guinee, J., Heijungs, R., de Koning, A.,
van Oers, L., Suh, S., Geerken, T., Van Holderbeke, M.,
Jansen, B., Nielsen, P., 2006. Environmental Impact of
Products (EIPRO): analysis of the life cycle environmental
impacts related to the final consumption of the EU-25 EUR
22284 EN European Commission Joint Research Centre,
Brussels. .
Umoh, O.J., Effiong, E.L., 2013. Trade openness and
manufacturing sector performance in Nigeria. J. Appl. Econ.
Res. 7 (2) 147169.
Verbeke, W., Perez-Cueto, F.J.A., de Barcellos, M.D., Krystallis,
A., Grunert, K.G., 2010. European citizen and consumer
attitudes and preferences regarding beef and pork. Meat Sci.
84 (2) 284292.
Verbeke, W., Viaene, J., 1999. Beliefs, attitude and behaviour
towards fresh meat consumption in Belgium: empirical
evidence from a consumer survey. Food Qual. Prefer. 10,
437445.
106 environmental science & policy 39 (2014) 95106
Wang, Y., Beydoun, M.A., 2009. Meat consumption is associated
with obesity and central obesity among US adults. Int. J.
Obesity 33, 621628.
Wang, Y., Beydoun, M.A., Caballero, B., Gary, T.L., Lawrence, R.,
2010. Trends and correlates in meat consumption patterns
in the US adult population. Publ. Health Nutr. 13, 13331345.
WHO, 2013. Global and Regional Food Consumption Patterns
and Trends. , http://www.who.int/nutrition/topics/
3_foodconsumption/en/index.html.
Willett, W.C., 2005. Eat, Drink, and Be Healthy: The Harvard
Medical School Guide to Healthy Eating. Free Press, New York.
Wooldridge, L.M., 2002. Econometric Analysis of Cross Section
and Panel Data. MIT Press, Massachusetts.
World Bank, 2011. World Bank Development Indicators. World
Bank, Washington, DC.
Zingg, A., Cousin, M.-E., Connor, M., Siegrist, M., 2013. Public risk
perception in the total meat supply chain. J. Risk Res. 16 (8)
10051020.