ScienceDirect
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.
consumption in emerging countries and developing countries motivate consumers to decrease meat consumption? Is a
is growing rapidly because of increasing economic growth second nutrition transition in which meat is substituted back
(Delgado, 2003). According to FAO projections, meat consump- to plant based foodstuffs emerging? This paper addresses the
tion in these countries may increase from an average annual evolution of meat consumption and analyses the relationship
per capita consumption of 10 kg in the 1960s to 26 kg in 2000, between meat consumption and income. The key hypothesis
reaching 37 kg around the year 2030. This forecast suggests is that meat consumption stagnates or even decreases once a
that in a few decades, developing countries consumption of GDP threshold is reached. Little empirical evidence is reported
meat will move towards that of developed countries where on the probability of an inverted U shaped relationship
meat consumption remains stagnant at a high level. between income and meat consumption. This relationship
From a nutritional point of view, meat is undeniably a high is similar to the Environmental Kuznets Curve (EKC) hypothe-
quality source of proteins. Average requirements for healthy sis. The EKC represents graphically the hypothesis that as an
adults are currently estimated at 0.66 g/kg body mass per day economy develops environmental pollution (e.g. SO2 concen-
(EFSA). However, few data are available on the mean protein tration) first increases, and then decreases after a certain
intake on a body mass weight basis. EU statistics report daily average income is attained (Grossman and Krueger, 1991,
protein consumption between 0.8 and 1.25 g/kg body weight 1993). One explanation for the shape of the EKC relates to the
for adults. Main contributors to dietary protein intake are meat income elasticity of environmental quality demand. When
and meat products, grains and grain-based products, and milk income increases people care more about the environment
and dairy products. The relation between health and meat they live in. Once a certain level of income is reached, their
strongly depends on the type of meat considered (Wang and willingness to pay for improved environmental quality
Beydoun, 2009). Basically, there are major nutritional benefits increases by a greater proportion than income (Pezzey,
to consuming some red white meat but almost no reason 1989; Selden and Song, 1994; Baldwin, 1995). The level of
aside from tastes to consume mostly processed meats income beyond which pollution starts to decline while income
(Popkin, 2009). Overconsumption of meat might result in a too further increases, is called the turning point. While there is a
high intake of saturated fatty acids, which is one of the main sound theoretical background for the EKC, the empirical
reasons for high incidence of coronary heart diseases. Further, evidence provides some controversy (Stern, 2005; Dinda, 2004).
improperly processed meat and the use of chemicals in Research on the EKC includes, among others, the relationship
agriculture and animal production increases the risks of food between income and CO2 emission, sulfur emission and
poisoning. deforestation (Rothman, 1998; Saboori et al., 2012; Culas, 2012;
Given the already very high meat consumption rates in Leitao, 2010). Empirical studies confirmed an inverted U-shape
industrialised countries, the question raises whether there are relation with income for local pollutants with a short term
limits to the increase in meat consumption. In this context, impact, but not for global pollutants with an indirect and long-
there are promising signals. On the demand side, consumer term impact such as carbon dioxide. Particularly for local
awareness on health, animal welfare and global environmen- pollutants that directly affect health, the U-shape relation is
tal changes might reverse the tendency of increasing meat confirmed. Empirical findings also suggest that the U-shaped
consumption. A study among French households showed a relationship between income and pollution is country specific
positive relation between education level of the household because different national regulations result in large differ-
head and reduced meat consumption (Allais et al., 2010). ences in country level per capita emissions (Dinda, 2004; Stern
Initiatives like the emergence of vegetarian organisations and et al., 1996). When it comes to meat consumption, the
the introduction of meatless days (such as Meatless Monday in hypothesis for an inverted U-shaped relationship between
the USA) illustrate increased citizens environmental and income and meat consumption can be motivated because
health concerns related to meat consumption. In addition, meat consumption does not only contribute to global
evidence-based guidelines on a sustainable and healthy diet pollutants but also has a direct impact on consumers health.
have been developed. Stehfest et al. (2009) calculated that a However, the relation might be country specific due to for
healthy diet worldwide would reduce the required area of example cultural and religious differences which affect meat
arable land globally by 10%, and the area of grassland by 40%, consumption and result in a large difference in national levels
compared to the FAO projections. of per capita meat consumption.
An example of a healthy diet is the Willett diet (Willett, 2005), Using panel data for 120 countries in the period 19702007,
which includes 10 g beef, 10 g pork, 47 g chicken and eggs, and this paper empirically analyses the link between meat
23 g fish, per person, per day on average. The associated consumption and income, while controlling for fixed effects
reduction in costs for mitigation of carbon dioxide emissions and geographical area, culture and trade. The results of the
could be as large as 50% in 2050, compared to the reference case. research indicates the need for more interdisciplinary
An increased number of governments have integrated environ- research and arguments for improved cooperation between
mental issues into national dietary guidelines. Against the lobby environmental and health policy in industrialised and
of food industry, some governments have integrated sustain- emerging countries as well as in developing countries.
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 2. Determinants of meat consumption
schools (Lombardini and Lankoski, 2013).
Will meat consumption continue to raise, or is there a Food choice is determined by a complex set of socioeconomic
limit? Will health concerns or global drivers of change and demographic characteristics (Frank, 2007). This also holds
environmental science & policy 39 (2014) 95106 97
for the level of meat consumption and for the type of meat 2.2. Geographical area
consumed. According to Kearney (2010), meat consumption is
country-specific and influenced by several factors including Access to open waters also has an impact on meat consump-
cultural beliefs, religion, gender, socio-economic status (SES) tion as dependence on fish is usually higher in coastal than in
and ethnic differences. This is confirmed by Wang et al. (2010), inland areas. For most islands, fish rather than meat
who analysed shifts in meat consumption across population composes the main source of proteins. The highest fish
groups in the USA. However, there is a lack of quantitative consumption worldwide is found in Iceland, Japan and some
evidence on the determinants for national meat consumption small island states (FAO, 2012; WHO, 2013). In this sense, one
levels (Latvala et al., 2012). can expect higher meat consumption in landlocked regions or
countries. The geographic term landlocked refers to a
2.1. Income and meat price country with no coast that is surrounded by other countries.
This status affects a countrys trade position and economic
From a historical perspective, increased meat consumption development. Goods produced in a country without access to
is associated with an increase of the nutritional value of open water, for example, must always be transported through
diets. Kearney (2010) illustrates how worldwide increased another country. As they depend on neighbouring countries
incomes and lower food prices have led to increased for imports and exports, they are considered as more
consumption of animal-based foods. In Europe and North vulnerable as compared to non-landlocked developing coun-
America, this explains the growth of meat consumption over tries. Furthermore, focusing on developing countries, it should
the past century. In the newly industrialised countries, such be noted that landlocked countries often have a lower GDP per
as China and Brazil, this tendency is observed more recently. capita (Dia Kamgnia, 2011).
Moreover, as the price of meat decreases, meat becomes
more easily accessible in developing countries (Tey et al., 2.3. Food crisis
2010). This holds in particular for cheap, highly processed
low quality meat. In this context, it should be noted that Over the last two decades, the meat sector has been struck by
meat is not a homogeneous commodity with considerable several crises including the Bovine Spongiform Encephalopa-
price differences between the different types of meat. The thy (BSE) in Europe, the avian flu (H5N1) in several countries
income elasticity of meat has been studied intensely. A most worldwide, the dioxin crisis in Belgium and the Escherichia coli
recent meta-analysis on the income elasticity of different (EHEC) outbreak in Germany. However, while these crises tend
types of meat is carried out by Gallet (2010). For all types of to have considerable impact on consumer perception of meat
meat, price elasticity is greater than one. Park et al. (1996) quality and safety (Grunert, 2006; Verbeke and Viaene, 1999),
focused on the price elasticity of beef, pork and poultry. The considerations about food-safety perceptions do not seem to
results confirmed that income elasticity is positive though affect meat consumption in the long run (Zingg et al., 2013).
less than one. Furthermore, the income elasticity was higher Results from the European project ProSafeBeef confirm the
for people in poverty than for higher income households mismatch between citizens attitude and consumer behav-
(Park et al., 1996). Apparently, meat consumption increases iour. Focusing on pork consumption for example, Verbeke
with income, but the rate of increase tends to flatten at et al. (2010) concluded that peoples views as citizens do not
higher income levels. This can be explained by the fact that appear to systematically influence their consumption choices.
middle and high income class consumers are more aware of
the impact of overconsumption in general, and meat 2.4. Cultural differences and religion
consumption in particular, on the risk for obesity and other
chronic diseases and of the environmental impact of meat Culture has an impact on food consumption behaviour in
production. general, and meat consumption in particular (Haverstock and
Analyses of three Dutch National Food Consumption Forgays, 2012; Kearney, 2010). Even when income differences
Surveys, for example, showed that subjects with low SES between countries disappear, consumer behaviour will not
report significant higher meat consumption (Hulshof et al., homogenise because consumer behaviour will be heteroge-
2003). This is in line with Mullie et al. (2010), who concluded nous because of cultural differences (de Mooij and Hofstede,
from a cross-sectional design in 1852 military men that the 2002). Hence, values of national cultures might have a strong
most healthy patterns were associated with a higher socio- impact on consumer behaviour. Culture has been shown to
economic position, while lower positions were associated with influence for example the consumption of mineral water and
several unhealthy behaviours. soft drinks, fresh fruit, ice cream and frozen food (de Mooij,
However, income growth either at the individual level or 1997, 1998, 2000). Hofstedes dimensions of national culture
at the national level may rather lead to a shift in the type of can be useful for statistically analysing consumption differ-
meat consumed than to an overall reduction of meat ences across nations (de Mooij and Hofstede, 2002). The
consumption. Chamorro et al. (2012), for example, conclud- Hofstede indices on cultural dimensions describe the effects of
ed from a Delphi research that consumers in Spain tend to a societys culture on the values of its members, and how these
replace beef and veal by poultry meat. Further, their results values relate to behaviour. Hofstede (2001) derived these
showed an increase in the consumption of ready meals with indices and dimensions of national culture based on several
a meat base, whereas overall meat consumption levels in large-scale surveys among in total 88,000 IBM employees from
Spain are expected to remain stable in the short run (up to 72 countries between 1967 and 1973. Employees were asked to
2016). 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 very assertive and competitive and maximally different from
cluster analysis he distinguished five dimensions of national womens values on the one side, to modest and caring and
culture: power distance, individualism versus collectivism, similar to womens values on the other. The assertive pole has
uncertainty avoidance, masculinity versus femininity and been called masculine and the modest, caring pole femi-
long-term versus short-term orientation. Both the indices and nine. The women in feminine countries have the same
the dimensions have been validated repeatedly in later studies modest, caring values as the men; in the masculine countries
(Hofstede et al., 2010), while other researchers have compared they are somewhat assertive and competitive, but not as much
Hofstedes approach to other approaches in the literature (see as the men, so that these countries show a gap between mens
e.g. Soares et al., 2007). The values do not change over time, as values and womens values.
cultural dimensions are relatively fixed, changing only very Despite the evidence for the impact of culture on
slowly. This has also been validated by later research consumers attitudes and behaviour with respect to food
(Hofstede et al., 2010). consumption, the use of the Hofstede index in food studies is
The power distance index (PDI) measures the extent to very limited. One exception is the study of Gaither and Courtin
which the less powerful members of organisations and (2007), who analysed the impact of the Danish publication on
institutions (like the family) accept and expect that power is the prophet Muhammad on the dairy company, Arla Foods.
distributed unequally. This represents inequality (more versus This case study emphasises the strength of the PDI as a tool for
less), but defined from below, not from above. It suggests that a international public relation work and the need to apply
societys level of inequality is endorsed by the followers as cultural normative tools in the development of theory of
much as by the leaders. international public relations practice.
Individualism (IDV) is the degree to which individuals are Finally, the importance of religion on meat consumption
integrated into groups. Its opposite side is collectivism. On the cannot be denied (Bonne et al., 2007). Pork is forbidden in
individualist side we find societies in which the ties between Judaism and Islam; pork and beef are forbidden in Hinduism
individuals are loose: everyone is expected to look after him/ and Buddhism. Only in Christianity, there are no rules related
herself and his/her immediate family. On the collectivist side, to meat consumption.
we find societies in which people from birth onwards are
integrated into strong, cohesive in-groups, often extended 2.5. Trade
families (with uncles, aunts and grandparents) which contin-
ue protecting them in exchange for unquestioning loyalty. The Overall, openness to trade in an economy has an impact on
word collectivism in this sense has no political meaning: it technological development, innovation and competitive-
refers to the group, not to the state. Again, the issue addressed ness. This holds true for both developed and developing
by this dimension is an extremely fundamental one, regarding countries (e.g. Umoh and Effiong, 2013). Focusing on food,
all societies in the world. globalisation has drastically changed food supplies, leading
Uncertainty avoidance index (UAI) refers to a societys to a situation with worldwide and year-round availability of
tolerance for uncertainty and ambiguity. It ultimately all types of foods. Openness to trade also has its impact on
refers to mans search for Truth. It indicates to what extent meat supplies and hence may influence meat consumption.
a culture programs its members to feel either uncomfort- Limits to trade such as geographical isolation, conflicts or
able or comfortable in unstructured situations. Unstruc- trade policy on the contrary lead to reduced overall meat
tured situations are novel, unknown, surprising, and supply on the local or national market and the a limited
different from usual. Uncertainty avoiding cultures try to choice in types of meats.
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 3. Data and methods
Truth; there can only be one Truth and we have it. People
in uncertainty avoiding countries are also more emotional, 3.1. Model
and motivated by inner nervous energy. The opposite type,
uncertainty accepting cultures, are more tolerant of Regression analysis is used to understand the relationship
opinions different from what they are used to; they try to between meat consumption and income. The model is
have as few rules as possible, and on the philosophical and described by the equation
religious level they are relativist and allow many currents
yit a bk xkit g m zmi eit (1)
to flow side by side. People within these cultures are more
phlegmatic and contemplative, and not expected by their where yit, the dependent variable, is the indicator of meat
environment to express emotions. consumption in country i and year t, xkit is a vector of the
The masculinity index (MI) measures the distribution of time-varying explanatory variables, zmi is a vector of time-
emotional roles between the genders. It refers to the invariant explanatory variables and eit is the idiosyncratic
distribution of roles between the genders which is another error term.
fundamental issue for any society to which a range of First, we use a pooled ordinary least square (OLS) model in
solutions are found. Its opposite is femininity. The IBM order to analyse the relationship between the dependent and
studies carried out by Hofstede revealed that (a) womens explanatory variables, between countries and over time. To
values differ less among societies than mens values; (b) mens account for heteroscedasticity and serial correlation, we use
values from one country to another contain a dimension from cluster robust standard errors in all of our models.
environmental science & policy 39 (2014) 95106 99
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 remains below 50 kg/cap/year. Figures for African countries
behaviour. 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
4. Results South African (FAO, 2012).
Fig. 2 presents the relationship between meat consumption
Our analysis is based on an unbalanced data set of 120 and income, whereby income is measured in GDP/cap in
countries. For most countries (103), we used data for a period constant 2005 international dollars. For both cases, the
of 38 years (19702007). For post-USSR countries, we used data relationship between meat consumption and income is not
for the period 19922007 and for Belgium and Luxembourg for linear. Therefore, GDP in squared terms will be included as
the period 20002007. However, not all observations could explanatory variable into the empirical model (see further).
always be included in the regression estimates as some This is done for both GDP per capita in current US dollars and
explanatory variables such as the Hofstede index, GDP/cap for GDP per capita in constant 2005 international dollars,
expressed in 2005 constant international dollars, were only resulting in variables (GDP/cap)2 and (GDP_con/cap).
available for a subset of the countries or a smaller time period. Table A2 of the appendix presents the summary statistics
Table 1 in the appendix describes the countries and years of the other variables used in the regression analysis.
included in the dataset that used for estimating the different
regression models.
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
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)
a fixed effects estimator with vector composition as in models 4.4. Income elasticity of meat consumption
C, D and E of Tables 1 and 2.
Finally, we tested the strict exogeneity assumption on Since it is useful to interpret the relationship between
which the first difference estimator (model C in Tables 1 and 2) income and meat consumption as elasticities, we have split
and the fixed effects estimator (model D in Tables 1 and 2) rely. the data in two groups. One group holds the observations
These tests revealed that all variables are strictly exogenous with income below the turning point, while the other holds
when the income variable is expressed in constant 2005 the observations with income above the turning point. On
international dollars (see Table 3). However, when GDP is these two groups, we have run a double log model and
expressed in current US dollars, the variable TRADE does not provide the elasticities in Table 4. Elasticities are only
pass the strict exogeneity. provided for the fixed effect within estimator model (i.e.
While an inverted U-relationship between meat consump- model D). Model D was chosen because our empirical
tion and income is observed for all specifications, these tests analysis indicated that a fixed model was preferred above
indicate that models C, D and E in Table 2 provide the best the pooled OLS and random effects model and because the
coefficient estimates. This is in line with our assumption as turning point was lowest for the fixed effects within
GDP in PPP terms is in general a preferable income measure estimator regression. The latter was particularly important
and as one might expect the country fixed effects to be to have sufficient observation in the group that holds
correlated with the explanatory variables. 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)
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
Table 4 indicates that for countries below the turning meat However, Allais et al. (2010) described how taxes will stimulate
consumption increases with 0.3% if income increases with 1% the food industry to innovate and possibly undermine the
when GDP is expressed in current US dollars, while it increases impact of the taxes. Examples of legal instruments to
with 0.5% when GDP is expressed in PPP terms. For countries stimulate reduced meat consumption can be found in Finland
above the income turning point, meat consumption decreases where experiments have been undertaken with respect to
with 1.2% if income increases with 1% when GDP is expressed forced choice restriction in the format of a Mandatory
in PPP terms. Vegetarian Day (Lombardini and Lankoski, 2013).
5. Conclusions
Appendix A. Supplementary data
In this paper we studied the evolution of meat consumption
within countries and determined its relationship with income Supplementary data associated with this article can be
and other variables. The results of our empirical analysis found, in the online version, at http://dx.doi.org/10.1016/
confirm that the relationship between meat consumption and j.envsci.2014.02.009.
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
references
meat consumption. The results also indicate that culture and
religion can explain differences in meat consumption within
and between countries.
Allais, O., Bertail, P., Nichele, V., 2010. The effects of a fat tax on
Since livestock production is a major cause of greenhouse
French households purchases: a nutritional approach. Am.
gas emissions (Garnett, 2009), this is a this promising signal.
J. Agric. Econ. 92 (1) 228245.
However, many countries are still far away from the turning Baldwin, R., 1995. Does sustainability require growth? In:
point. Therefore, the findings should not demotivate govern- Goldin, I., Winters, L.A. (Eds.), The Economics of Sustainable
ments to undertake actions in order to accelerate the Development. Cambridge Univ. Press, Cambridge, UK, pp.
reduction of meat consumption. Decreasing meat consump- 1947.
tion can be accelerated either by using economic instruments Bonne, K., Vermeir, I., Bergeaud-Blackler, F., Verbeke, W., 2007.
Determinants of halal meat consumption in France. Br. Food
(e.g. meat tax), by social instruments (e.g. educational
J. 109 (5) 367386.
campaigns) or by legal instruments. Nordgren (2011) argued Chamorro, A., Miranda, F.J., Rubio, S., Valero, V., 2012.
that developed countries and, to increasing extent, emerging Innovations and trends in meat consumption: an application
countries, should thereby take the lead. In his ethical of the Delphi method in Spain. Meat Sci. 92, 816822.
discourse, Nordgren (2011) outlined arguments in favour of Culas, R.J., 2012. REDD and forest transition: tunneling through
a meat consumption tax, rather than a meat production tax. the environmental Kuznets curve. Ecol. Econ. 79, 4451.
environmental science & policy 39 (2014) 95106 105
de Mooij, M., 1997. Mapping cultural values for global marketing Lombardini, C., Lankoski, L., 2013. Forced choice restriction in
and advertising. Paper presented at ESOMAR, 50th Congress. promoting sustainable food consumption: intended and
Edinburgh, Scotland. . unintended effects of the mandatory vegetarian day in
de Mooij, M., 1998. Global Marketing and Advertising: Helsinki schools. J. Consum. Policy 36, 159178.
Understanding Cultural Paradoxes. Sage, Thousand Oaks, Mullie, P., Clarys, P., Hulens, M., Vansant, G., 2010. Dietary
CA. patterns and socioeconomic position. Eur. J. Clin. Nutr. 64,
de Mooij, M., 2000. The future is predictable for international 231238.
marketers: converging incomes lead to diverging consumer Nordgren, A., 2011. Ethical issues in mitigation of climate
behaviour. Int. Market. Rev. 17, 103113. change: the option of reduced meat production and
de Mooij, M., Hofstede, G., 2002. Convergence and divergence in consumption. J. Agric. Environ. Ethics 25, 563584.
consumer behavior: implications for international retailing. Park, J.L., Holcomb, R.B., Raper, K.C., Capps, O., 1996. A
J. Retailing 78, 6169. demand systems analysis of food commodities by U.S.
Delgado, C.L., 2003. Rising consumption of meat and milk in households segmented by income. Am. J. Agric. Econ. 78 (2)
developing countries has created a new food revolution. J. 290300.
Nutr. 133 (11) 3907S3910S. Pezzey, J.C.V., 1989. Economic analysis of sustainable growth
Dia Kamgnia, B., 2011. Political economy of recent global food and sustainable development. Environment Department
price shocks: gainers, losers and compensatory mechanism. working paper 15. World Bank. .
J. Afr. Econ. 20 (AERC Supplement 1) i142i210. Plumper, T., Troeger, V., 2007. Efficient estimation of time-
Dinda, S., 2004. Environmental Kuznets curve hypothesis: a invariant and rarely changing variables in finite sample
survey. Ecol. Econ. 49, 431455. panel analyses with unit fixed effects. Polit. Anal. 15, 124
FAO, 2012. The State of World Fisheries and Aquaculture. FAO, 139.
Rome. Popkin, B.M., 2006. Global nutrition dynamics: the world is
Frank, J., 2007. Meat as a bad habit: a case for positive feedback shifting rapidly toward a diet linked with noncommunicable
in consumption preferences leading to lock-in. Rev. Soc. diseases. Am. J. Clin. Nutr. 84 (2) 289298.
Econ. 65 (3) 319348. Popkin, B.M., 2009. Global nutrition dynamics: the world is
Gaither, T.K., Courtin, P.A., 2007. Examining the heuristic value shifting rapidly toward a diet linked with non-
of models of international public relations practice: a case communicable diseases. Am. J. Clin. Nutr. 84, 289298.
study of the Arla foods crisis. J. Publ. Relat. Res. 20 (1) 115 Rothman, D.S., 1998. Environmental Kuznets curves real
137. progress or passing the buck? A case for consumption-based
Gallet, C.A., 2010. Meat meets meta: a quantitative review of the approaches. Ecol. Econ. 25, 177194.
price elasticity of meat. Am. J. Agric. Econ. 92 (1) 258272. Saboori, B., Sulaiman, J., Mohd, S., 2012. Economic growth and
Garnett, T., 2009. Livestock-related greenhouse gas emissions: CO2 emissions in Malaysia: a cointegration analysis of the
impacts and options for policy makers. Environ. Sci. Policy environmental Kuznets curve. Energy Policy 51, 184191.
12, 491503. Selden, T., Song, D., 1994. Environmental quality and
Grossman, G.M., Krueger, A.B., 1991. Environmental impacts of development: is there a Kuznets curve for air pollution
the North American Free Trade Agreement. NBER. Working emissions? J. Environ. Econ. Manage. 27, 147162.
paper 3914. Soares, A.M., Farhangmehr, M., Shoham, A., 2007. Hofstedes
Grossman, G.M., Krueger, A.B., 1993. In: Garber, P.D. (Ed.), dimensions of culture in international marketing studies. J.
Environmental impacts of the North American Free Trade Business Res. 60, 277284.
Agreement. The U.S. Mexico Free Trade Agreement. MIT Stehfest, E., Bouwman, L., Van Vouren, D.P., Den Elzen, M.G.J.,
Press, Cambridge, pp. 1356. Eickhout, B., Kabat, P., 2009. Climate benefits of changing
Grunert, K.G., 2006. Future trends and consumer lifestyles with diet. Climatic Change 95, 83102.
regard to meat consumption. Meat Sci. 74, 149160. Stern, D., 2005. The rise and fall of the environmental Kuznets
Haverstock, K., Forgays, D.K., 2012. To eat or not to eat. A curve. World Dev. 32, 14191439 (August 2004).
comparison of current and former animal product limiters. Stern, D.I., Common, M.S., Barbier, E.B., 1996. Economic growth
Appetite 58 (3) 10301036. and environmental degradation: the environmental Kuznets
Heinz, B., Lee, R., 1998. Getting down to the meat: the symbolic curve and sustainable development. World Dev. 24, 1151
construction of meat consumption. Commun. Stud. 49, 8699. 1160.
Hofstede, G., 2001. Cultures Consequences: Comparing Values, Tey, Y.S., Shamsudin, M.N., Mohamed, Z., Abdullah, A.M.,
Behaviors, Institutions, and Organizations Across Nations, Radam, A., 2010. Demand analysis of meat in Malaysia. J.
2nd ed. Sage, Thousand Oaks, CA. Food Prod. Market. 16, 199211.
Hofstede, G., Hofstede, G.-J., Minkov, M., 2010. Cultures and Tukker, A., Huppes, G., Guinee, J., Heijungs, R., de Koning, A.,
Organizations: Software of the Mind, 3rd ed. McGraw-Hill, van Oers, L., Suh, S., Geerken, T., Van Holderbeke, M.,
USA. Jansen, B., Nielsen, P., 2006. Environmental Impact of
Hulshof, K.F.A.M., Brussaard, J.H., Kruizinga, A.G., Telman, J., Products (EIPRO): analysis of the life cycle environmental
Lowik, M.R.H., 2003. Socio-economic status, dietary intake impacts related to the final consumption of the EU-25 EUR
and 10 y trends: the Dutch National Food Consumption 22284 EN European Commission Joint Research Centre,
Survey. Eur. J. Clin. Nutr. 57 (1) 128137. Brussels. .
Kearney, J., 2010. Food consumption trends and drivers. Phil. Umoh, O.J., Effiong, E.L., 2013. Trade openness and
Trans. R. Soc. B 365, 27932807. manufacturing sector performance in Nigeria. J. Appl. Econ.
Kuznets, S., 1955. Economic growth and income inequality. Am. Res. 7 (2) 147169.
Econ. Rev. 45, 128. Verbeke, W., Perez-Cueto, F.J.A., de Barcellos, M.D., Krystallis,
Latvala, T., Niva, M., Makela, J., Pouta, E., Heikkila, J., Kotro, J., A., Grunert, K.G., 2010. European citizen and consumer
Forsman-Hugg, S., 2012. Diversifying meat consumption attitudes and preferences regarding beef and pork. Meat Sci.
patterns: consumers self-reported past behaviour and 84 (2) 284292.
intentions for change. Meat Sci. 92 (1) 7177. Verbeke, W., Viaene, J., 1999. Beliefs, attitude and behaviour
Leitao, A., 2010. Corruption and the environmental Kuznets towards fresh meat consumption in Belgium: empirical
curve: empirical evidence for sulfur. Ecol. Econ. 69, evidence from a consumer survey. Food Qual. Prefer. 10,
21912201. 437445.
106 environmental science & policy 39 (2014) 95106
Wang, Y., Beydoun, M.A., 2009. Meat consumption is associated Willett, W.C., 2005. Eat, Drink, and Be Healthy: The Harvard
with obesity and central obesity among US adults. Int. J. Medical School Guide to Healthy Eating. Free Press, New York.
Obesity 33, 621628. Wooldridge, L.M., 2002. Econometric Analysis of Cross Section
Wang, Y., Beydoun, M.A., Caballero, B., Gary, T.L., Lawrence, R., and Panel Data. MIT Press, Massachusetts.
2010. Trends and correlates in meat consumption patterns World Bank, 2011. World Bank Development Indicators. World
in the US adult population. Publ. Health Nutr. 13, 13331345. Bank, Washington, DC.
WHO, 2013. Global and Regional Food Consumption Patterns Zingg, A., Cousin, M.-E., Connor, M., Siegrist, M., 2013. Public risk
and Trends. , http://www.who.int/nutrition/topics/ perception in the total meat supply chain. J. Risk Res. 16 (8)
3_foodconsumption/en/index.html. 10051020.