Determinants of obesity in Brazil: the effects of trade liberalization and socio-

economic variables

Dr. Silvia H. G. de Miranda, ProfessorUniversity of São Paulo -ESALQ

Piracicaba, Brazil

Dr. Dragan Miljkovic, Professor (contact author)North Dakota State University

Department of Applied Economics 614A Barry Hall, NDSU Dept. 7610

Fargo, ND 58108-6050, USA Phone: 701-231-8519; FAX: 701-231-7400; E-mail: Dragan.Miljkovic@ndsu.edu

Dr. Ana L. Kassouf, Professor University of São Paulo -ESALQ

Piracicaba, Brazil

Dr. Fabíola C. R. Oliveira, ProfessorUNIMEP

Piracicaba - Brazil

Abstract

This paper aims to evaluate a possible relation between increased Brazilian trade openness and

increasing observed rates of overweight and obesity during the last 25 years. We develop an

economic model where formal trade barriers are eliminated, and resulting socio-cultural

outcomes such as the adoption of westernized lifestyle in traditional non-western countries

prevails, which could imply a health externality. In order to empirically analyze the influence of

trade flows on overweight and obesity in Brazil, a balanced fixed-effects panel model has been

estimated. Data for the 26 Brazilian states plus the Federal District are run for 1988/1989, 2002

and 2008. We found that an increase in trade openness leads to an increase in overweight and

obesity ratios in Brazil. Hence results seem to point that there is a health externality in Brazil due

to trade liberalization. However more consistent evidence may be necessary to convince

politicians and policy makers that any interference will be necessary to correct this externality.

Keywords: obesity, trade liberalization, panel data, externality, health policy

* This paper was partially sponsored by Fapesp, project number 2011/19283-7. Acknowledgements also to Ana

Laura Dias and Roselaine de Almeida, for collecting data and program assistance, respectively.

1

1. Introduction

The externalities arising as a consequence of trade liberalization have rarely been a topic

of much public debate. A possible exception would be the relationship between the international

trade and resulting environmental degradation (e.g., Markusen, 1975; Conrad, 1993; Ludema and

Wooten, 1994). Even then the mechanisms that link the volume of international trade and the

rate of environmental degradation are seldom adequately explained. Part of the problem is that

the environmental problems that are faced are extremely varied both in character and severity,

and it is fruitless to attempt to provide a single remedy for all of them.

Environmental degradation is hardly the only externality caused by trade liberalization.

There is, for example, a growing literature addressing possible adverse impact of trade

liberalization on cultural diversity within, in particular, non-western countries (e.g., Castells,

1997; Cowen, 2002; Segerstrom 2003). In addition, the relevance of the topic is underscored by

the UNESCO Universal Declaration on Cultural Diversity (2001), which emphasizes the role of

cultural goods as ”vectors of identity” and is concerned about current trade imbalances in related

sectors.

Our objective in this paper is to determine the likely effects of trade liberalization on

increasing of obesity level in Brazil. According to the World Health Organization (WHO, 2006),

approximately 1.6 billion adults were overweight and at least 400 million were obese in 2005 all

around the world. More recent statistics show that top three countries in the world with the

highest obesity rate are the United States, Egypt and Saudi Arabia (WHO, 2009). Due to this

rapid and continuous global increase of obesity, the WHO defines obesity as a global epidemic.

Obesity has not only increased in developed countries but more surprisingly in developing

countries as well. According to the Food and Agriculture Organization of the United Nations

(FAO, 2008), obesity increases rapidly in developing countries, even in those where hunger still

exists. For instance, the number of overweight people changed from less than 10 to 15 percent in

China during the last two decades. In fact, the last survey in Brazil, related to 2008-2009

reference year, showed overweight in roughly 50.1 percent of male population older than 20

years and 12.4 percent of obesity in the same population´s category, while for women these

statistics registered 48 percent and 16.9 percent, respectively (IBGE, 2010). Moreover, the

2

problem has been also identified in children over 5 years old. These figures are comparable to

those in developed countries.

In this paper, we review the literature where formal trade barriers are eliminated, foreign

direct investment promoted, and resulting socio-cultural outcomes such as the adoption of

westernized lifestyle in traditional non-western countries prevail contributing to increased

incidences of obesity, among other health and non-health externalities. We also try to establish

the extent to which health policies can become strategic trade policy instruments. We use the

theoretical models developed in these papers to empirically test in the case of Brazil, which

experienced a significant increase in obesity rates in recent decades, the hypothesis that this

process results partially from the introduction of trade liberalization policies in late 1980s, which

were even more intensified by mid of 1990s.

2. Brazilian Trade Liberalization and the Obesity Problem

2.1 Brazilian trade liberalization

In 1988, President José Sarney initiated trade liberalization and also a regional integration

initiative (Mercosur). These processes were consolidated in the following years with reforms

conducted also by his successor President Collor de Mello (1989 – 1991), and later by President

Fernando Henrique Cardoso (FHC). Besides reducing trade barriers, a privatization process has

taken place, enhancing the Foreign Direct Investments in Brazil for some years.

Before this time, Brazil had been markedly characterized by the model of imports

substitution, which prevailed in Latin American countries in general, particularly between the

1950s and the 1970s. During this period, only a few efforts were made to develop an exporting

infra-structure and international competitiveness in Brazil. Moreover, the major problem for

Brazil was to supply domestically more than generating surplus to trade, also because the share

of urban population increased due to a migration of the rural poor to the cities along those years.

Presidents Sarney and Collor de Mello both reduced drastically the tariffs, non-tariff

barriers and other import controls. According to Averbug (2002), this reduction in protectionism

to local industry lasted mostly between 1988 and 2003. One could note that at that period, there

was a surge in imports of consumption goods, including foods.

3

A relevant element to assess the Brazilian trade openness is the regional integration -

Mercosur (Argentina, Brazil, Uruguay and Paraguay1). Argentina is currently among the three

major Brazilian trade partners, and their bilateral trade flow is mostly intra-industrial. Mercosur

has launched officially through Assunción Treat, in 1991, as a free trade area and its currently

status is of a Common Market.

It is remarkable that since 2005 Brazil has also attracted increasing inflows of FDI,

although much before, since mid of 1990’s, different factors had already influenced capital

inflows. The first privatization wave occurred during Sarney’s government. Later, during the

first mandate of President FHC (1995-1998), a deeper privatization was conducted, and many

international companies came to Brazil to participate in strategic sectors such as communication,

banking, mining and energy. Again, since 2003, during Lula’s first mandate, a new boost on FDI

was registered, particularly in sectors like energy (including biofuels).

The inflow of international capital can also be highlighted in food sector, which is

important to our analysis, as the hypothesis is that international patterns of food consumption are

affecting Brazilian domestic consumption of food and lifestyle. Thus, such changes might result

in shifts of domestic preferences in food consumption and even increasing nutrition problems.

In fact, the number of fast-food chains and of their stores all over Brazil has increased

substantially in the last decades. Figure 1 shows some of the main fast-food chains present in the

country, according to data available at their websites. It is noticeable that, despite the importance

of multinational companies, some Brazilian food chains also emerged and increased their

market-share. It is also interesting to underline that stores of fast-food chains are more numerous

in the most populated and urbanized regions of Brazil.

For example, for the chain named by the code AA, 434 out of 628 stores are located in

Southeastern states; likewise, 379 stores of the brand BB out of 791 are in Southeastern states,

followed in this case by Northeastern states accounting for 134 stores (Pequenas Empresas &

Grandes Negócios, 2009).

(Figure 1 Here)

2.2. Food consumption and obesity in Brazil

1 In 2012, Venezuela has become a member.

4

According to Levy-Costa et al. (2005), the Brazilian diet in the last 30 years has trended

to reduce consumption of traditional food, like rice and beans, and to persistently increase fat

diets, excessive amount of sugar and insufficient fruits and vegetables. For the authors, all these

factors demonstrate an inappropriate pattern of food nutrition, contributing to obesity and related

chronicle diseases.

Through the Pesquisa Nacional de Saúde e Nutrição (PNSN, 1989), it is possible to

compare nutrition aspects for Brazilian population between 1975 and 1989. In this period, it had

been registered a higher prevalence of overweight (BMI > 25Kg/m2) and obesity (BMI >

30Kg/m2) among women (26.5 and 11.7%, respectively), although the biggest variation of

overweight had been observed in men (IBGE, 1990 apud Ministério da Saúde, 2009).

In 2002-2003, the National Institute of Geography and Statistics from Brazil – IBGE

conducted the Family Budget Research - POF (Pesquisa de Orçamentos Familiares) 2002-2003,

which has also collected anthropometric and nutrition data for Brazilian population. Results

showed that women (12.7%) had kept being more obese than men (8.8%), for people older than

20 years, although in terms of overweight, the male (41%) percentage had overcome female

(39.2%). Table 1 underlines a timeline of overweight and obesity frequency in Brazil.

(Table 1 Here)

The most recent official survey on family budget expenses (POF 2008-09), conducted by

IBGE, in a partnership with the Ministry of Health, points that Brazilian population has increased

in average weight along the last years. In 2009, one out of three children of 5 to 9 years old was

above the weight recommended by the Health World Organization. The share of young boys (10

to 19 years old) overweight has changed from 3.7 percent (1974-75) to 21.7 percent (2008-09),

while for young girls these numbers jumped from 7.6 percent to 19.4 percent.

According to Kac and Velasquez-Meléndez (2003), apud Ministério da Saúde (2009), the

determinants of overweight and obesity in Brazil are known. Some factors can be highlighted

such as the lifestyle increasingly sedentary and the progressive inappropriate diets.

Statistics also show that in Southern Region states the obesity problem is more serious

than in other regions and more evident for men with higher income, while for women the

problem is spread for all income categories. Moreover, the weight problem has shown to be more

5

important in urban than rural areas in Brazil, with the Southeastern region presenting the biggest

statistics in overweight among all regions. Figure 2 illustrates the regional distribution of

overweight and obesity prevalence in the country for male population, with age between 10 and

19 years old.

(Figure 2 Here)

3. Free Trade and Obesity – Literature Review

The literature that explores the link between international trade and obesity is fairly

recent. For example, Clark et al. (2012) consider the case of the impact of trade liberalization

between the United States and Mexico following the signing of the North American Free Trade

Agreement (NAFTA) on obesity and overweight. Obesity has reached epidemic proportions, in

the United States as well as among its trade partners such as Mexico. It has been established that

an “obesogenic” (obesity-causing) food environment is one influence on obesity prevalence. To

isolate the particular role of NAFTA, the North American Free Trade Agreement, in changing

Mexico's food environment, they plotted the flow of several key products between the United

States and Mexico over the 14-year NAFTA period (1994-2008) and situated them in a broader

historical context. Key sources of USDA data include the Foreign Agricultural Service’s Global

Agricultural Trade System, its official repository for current and historical data on imports,

exports and re-exports, and its Production, Supply, and Distribution online database. US export

data were queried for agricultural products linked to shifting diet patterns including: corn,

soybeans, sugar and sweeteners, consumer-oriented products, and livestock products. The

Bureau of Economic Analysis’ Balance of Payments and Direct Investment Position Data in their

web-based International Economic Accounts system also helped determine changes in US direct

investment abroad from 1982 to 2009. Directly and indirectly, the United States has exported

increasing amounts of corn, soybeans, sugar, snack foods, and meat products into Mexico over

the last two decades. Facilitated by NAFTA, these exports are one important way in which US

agriculture and trade policy influences Mexico’s food system. Clark et al. (2012) conclude that

because of significant US agribusiness investment in Mexico across the full spectrum of the

latter’s food supply chain, from production and processing to distribution and retail, the Mexican

food system increasingly looks like the industrialized food system of the United States.

6

Blouin et al. (2009) outline a conceptual framework of links between trade liberalization

and health outcomes, and review existing evidence for these by focusing on four key factors:

income, inequality, economic insecurity, and unhealthy diets. They conclude that trade

liberalization has facilitated availability of highly processed, calorie-rich, nutrient-poor food in

developing countries, but further research is needed to better understand the effects of trade on

unhealthy diets. Moreover, they suggest that policymakers and health professionals need to be

aware that the global economy affects the health of populations and understand how risks

associated with trade liberalization can be mitigated.

Thow and Hawkes (2009) the relationship between trade liberalization policies and food

imports and availability, and draws implications for diet and health, using Central America as a

case study region. The study indicates that the policies of trade liberalization in Central

American countries over the past two decades, particularly in relation to the United States, have

implications for health in the region. Specifically, they have been a factor in facilitating the

"nutrition transition", which is associated with rising rates of obesity and chronic diseases such

as cardiovascular disease and cancer. Given the significant cost of chronic disease for the health

care system, individuals and the wider community, it is critical that preventive health measures

address such upstream determinants of poor nutrition.

The book edited by Hawkes et al. (2009) provides a variety of frameworks to discuss the

potential link between trade and obesity. A different outlook is provided in Atkin’s (2013) paper

in which he introduces habit formation into an otherwise standard model of international trade.

Household tastes evolve over time to favor foods consumed as a child. The opening of trade

causes preferred goods to rise in price, as these were relatively inexpensive in autarky.

Neglecting the correlation between tastes and agro-climatic endowments overstates the short-run

nutritional gains from agricultural trade liberalization and masks potential caloric losses for

laborers. Atkin examines the predictions of this model of trade with habit formation using

household survey data from India, both by looking across Indian regions and by examining the

consumption patterns of inter-state migrants. He concludes that tastes matter for trade.

Neglecting their role overstates the caloric gains from agricultural trade liberalization, and masks

potential nutritional losses for the poorest members of society thus leading to a phenomenon

observed in less developed countries of the simultaneous presence of obesity and malnutrition.

7

Miljkovic et al. (2015) develop an economic model in which globalization factors

generate health externalities and contribute to global obesity growth. The outcome for the home

country is that, under a free-trade agreement, it has lost any influence over the level of the

externality (obesity), through relinquishing its import tariff, as well as having no means of

retaliating against the behavior of the foreign country. Thus the free-trade agreement, based on

results of this model, would likely be opposed both by protectionists and the health lobby in the

home country.

In the empirical part of the paper, the authors use the unbalanced panel data set containing

the information for 79 countries over the period 1986–2008. Fixed-effects panel data estimation

and quantile regression analysis were used to analyze the data. The fixed-effects panel model

results indicate that the impact of trade openness and the globalization social index (GSI) on

global obesity rates is positive and significant, which is consistent with prior expectations, while

surprisingly the foreign direct investments (FDI) has no impact on global obesity. Our empirical

model is based on the theoretical model developed in Miljkovic et al. (2015).

4. Empirical model and database

4.1. Database

In order to analyze the influence of trade flows over the overweight and obesity statistics

in Brazil a panel has been run. Data for the 26 Brazilian states plus the Federal District (the

capital)2 are run for 1988/1989, 2002 and 2008, the only periods for which anthropometric data

are available for all states in Brazil.

The dependent variable is number of overweight and obese people (or share in total

population). The following explanatory variables are included: trade (imports/exports flows and

trade openness), average income per capita, school years, economically active people,

consumption of sugar, soft drinks, beans and prepared food and percentage of houses with TV

sets. Besides those, several control variables are included in the model and some are highlighted:

2 States are divided in the following regions: North (Rondonia, Acre, Amazonas, Roraima, Pará, Amapá, Tocantins), Northeast (Maranhão, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, Alagoas, Sergipe, Bahia), Southeast (Minas Gerais, Espírito Santo, Rio de Janeiro, São Paulo), South (Paraná, Santa Catarina, Rio Grande do Sul) and Center-West (Mato Grosso do Sul, Mato Grosso, Goiás, Distrito Federal).

8

race, rural/urban population, gender and age of population. Some details on variables’

description follow below.

The share of overweight and obese people is obtained through the calculation of the Body

Mass Index - BMI, in which the weight of an individual is divided by his square height.

Afterwards, people who are overweight are defined as those with 25 < BMI < 30 and those

considered obese show BMI ≥ 30. Variables used to calculate BMI are explained below:

i) Inputted weight: this variable has been extracted from the PNSN 1989 and POFs

database, and the choice was to use the variable “average inputted price”, because

these surveys applied imputation procedures to deal with the no-answers problem, as

well as with the errors of answers associated to rejected values in the step of criticism.

ii) Inputted height: similarly to the previous variable, it was collected from the PNSN

1989 and POFs 2002 and 2008.

Trade statistics of total imports and exports of Brazil were taken from the official

Brazilian trade database, Sistema Alice (MDIC, 2012), which contains both volumes and values

traded internationally, per state, per year and per Harmonized System chapter and tariff item.

The total current values imported and exported, in US dollars, were taken and corrected to

inflation using a real exchange rate for R$/US$. The exchange rate was taken from IPEADATA

(www.ipeadata.gov.br) and adjusted using the Wholesale Price Index for Brazil (IPA) and for the

United States. As in Brazil, from 1988 to 2008, there were four changes in currency, it was also

necessary to make adjustments in Brazilian trade flows to guarantee consistence of all

information given in real Reais currency (R$) of 2008.

The same procedure was adopted to adjust the average income and average per capita

income, which were also converted to real Reais of 2008 using the Consumer Price Index –

INPC as deflator and collected at IBGE.

Regarding the explanatory variables, details are provided as follow:

i) States: there are 26 states and the Federal District, summing up to 27 state units in the

cross section analysis, for 2002 and 2008. The state of Goiás (Center-West) has been

divided into two in 1988, originating the states of Tocantins (North) and Goiás itself.

Because of technical reasons, this change has been incorporated to PNAD only in

1992. So, PNAD database for 1988 does not include Tocantins.

9

ii) Number of people in the family (Source: PNAD): People who are not family-

members are excluded (aggregated people, housekeepers and housekeeper´s

relatives), despite living in the house.

iii) Average age: Age information was extracted from PNAD. Information from “no-

declaration” people was excluded.

iv) Average schooling (Source: PNAD): For 2002 and 2008 the individual’s school years

could range from 0 (no instruction or less than one school year) to 14 years, and 17

for a person with 15 or more school years. Otherwise, for 1988, the survey allowed

school years ranging from 0 to 08 years, and to 09 years for people with 09 to 11

school years and 12 years (for people with more than 12 school years).

v) Household purchase of sugar, beans, and sodas (Source: POF): They refer to the food

household purchase per year per capita (in kg) for the two first and liters for the third.

vi) Household purchase of prepared food (Source: POF): It is available only for 2002

and 2008 and comprises both prepared meals as processed food such as flours, pasta,

juice, cakes and so on. It is also measured household purchase per capita (kg) per year

vii) Per capita family income: For 1988, this variable was taken from PNADs, and

calculated by dividing the total income by number of people in the family. For

PNADs 2002 and 2008, it was extracted directly from database. Similarly, the per

capita household income was also collected and run in the models.

viii) Active Economically Population (PEA): obtained directly from PNADs for 2002 and

2008. For 1988, the variable named V501 (meaning what the person was doing

during the survey week) was taken, summing up for the following categories of

answers: worked, had a job and was looking for a job.

The consumption data collected for 1988 (approximated to 1989) from POF was only

available for eight metropolitan regions and therefore it was necessary to use this information to

each of the 26 states and to the Federal District (FD). The eight metropolitan regions are: Belém

(Pará state, in the North), Fortaleza (Ceará), Recife (Pernambuco) and Salvador (Bahia state), all

the three belonging to Northeast; Belo Horizonte (Minas Gerais), Rio de Janeiro (Rio de Janeiro)

and São Paulo (São Paulo), which are in the Southeast; and Curitiba (Paraná) and Porto Alegre

(Rio Grande do Sul) located in the Southern region. POFs data for the eight regions were

10

considered a proxy for the 26 states and the FD and therefore, considering the geographical

pattern of proximity, they were extrapolated from metropolitan to states.

4.2. Empirical model

In order to verify if trade liberalization has an effect on increasing obesity and on

increasing overweight rates in Brazilian population, a model has been developed analyzing these

two variables as dependent and having trade flows and trade openness as a proxy to measure

trade liberalization and globalization in the economy. Several socio-demographic variables were

considered as controls.

Therefore, the empirical study is conducted using a balanced panel data model (Greene,

2004), i.e., the same number of observations for each cross-section unit (states), which consist on

i = 1, 2,…,27 states over time. For each year there is information related to three different

surveys conducted in 1988/1989, 2002 and 2008 (t = 1,2,3).

Let yit be the average obesity or overweight measures for each state i and time t such as,

y¿=X ¿j β+ε ¿ (1)

Where X ¿j is the value of jth explanatory variable for unit i at t and εit is the error term for the ith

unit in t.

In this model there could be unobservable variables correlated to the X variables causing

bias and inconsistency in the estimated coefficients that could be controlled by using a fixed

effect model if we assume they do not vary with time. For example, in less developed and

logistically/geographically more difficult to access areas, such as Acre State in the North of

Brazil, one would expect fewer imported goods and lower impact on obesity compared, for

instance, to São Paulo state.

The model was estimated using the STATA software, version 9.3.

5. Results and Discussion

Table 2 contains statistics collected by Brazilian agencies on consumption for Brazil and

for selected states representing each one a different macro region in Brazil. It is evident that

states differ, sometimes drastically, in their consumption levels of those foods we have

considered in the study. Moreover, one cannot affirm that all states moved to the same direction

11

in terms of reduction or increase in food consumption. For example, despite the fact that Brazil

as the country and most of the individual states are decreasing consumption of beans along the

period, Bahia State has increased and then again decreased, but it is still consuming more kg of

beans per capita per year in 2008 than in 1989. Sugar was not included in the table, but its

consumption per capita, measured by household purchase, has decreased all over Brazil between

late 1980’s and 2008.

(Table 2 Here)

It is also important to mention that, just as in major developed countries, there is also a

growing concern in Brazil about nutrition and health and therefore towards the choice of

healthier food and beverages. Therefore, the trends observed in the last two surveys for sodas

and sugar might relate also to such concerns emerging from a new lifestyle. This also might

explain why the food pattern is not the only variable to ponder when analyzing reasons behind

the increasing ratios of obese and overweight people in the Brazilian population. Hence other

factors such as the time that people spend to sports or to sedentary activities like using

computers, video-games and other modern electronic distractions should be considered in the

analysis as well.

As trade liberalization and implicitly globalization is an important issue to discussions in

this paper, Figure 3 highlights data used to model the effects of trade openness, defined as the

ratio of the sum of total imports and total exports over GDP, on obesity and overweight. It is

noteworthy that São Paulo is far ahead other states in terms of trade volume, and, in general,

Southern and Southeastern states account for the majority of Brazilian trade flows. The exception

is Amazonas state, Northern region, whose performance is most probably due to the existence of

the Free Zone in Manaus.

We are making a point again that the total imports as well as the trade openness consider

that the cultural changes that might affect overweight and obesity are not only related to food

consumption but also to behavior modifications such as spending more time using electronic

devices and driving and therefore less time doing outside exercises. Thus we believe that total

imports and trade openness are better proxies for the impact of globalization than food and

agricultural imports and trade openness alone. We also make the implicit assumption that

imports are consumed in the importing state while in reality that is not always the case, i.e., the

12

imports are not necessarily being consumed in the importing state. That could be overestimating

and underestimating the openness measure for some states, mainly those that are more inside in

the territory, but the lack of any data making direct link between imports and consumption of

imported goods at the state level forces us to make this assumption.

(Figure 3 Here)

Table 3 underscores useful data to interpret results from obesity and overweight models.

Regarding data for those states previously highlighted in Table 2, it is possible to verify how the

increasing obesity ratio differs across states and regions and, mainly, how the average household

income varies through regions. In fact, São Paulo state has a double household average per capita

income than Bahia state. In terms of school years, the range is not so large, although there is no

information about quality of school in those regions that might be considered here to support

other inferences about potential impacts of this variable in the anthropometric measures.

(Table 3 Here)

Panel models were run to test the hypothesis that trade openness might have contributed

to augmenting obesity and overweight ratios in Brazil. Table 4 shows results, analyzing three

years, 1988/1989, 2002 and 2008. Models were run in log-log form. Haussmann test confirmed

that the fixed effects model is better fit than the random effects model and thus only their results

will be presented here. Variety of models was considered, and the best fit one based on AIC and

BIC results was selected. Moreover, high correlations between variables such as per capita

income and average schooling years, or consumption of sugar and consumption of soda drinks,

led us to eliminate some of the variables to avoid the problems of multicollinearity.

(Table 4 Here)

Trade openness has shown positive and significant impact, implying that it affects

positively the ratio of obese and overweight people in the population. This result is consistent

with empirical result in Miljkovic et al. (2015) obtained for the panel of 79 different countries.

13

Indeed, the exposure and the availability of not only to processed foods, but also other goods and

services which promote more sedentary life style is likely to increase overweight and obesity. It

was shown in the literature that the exposure to new goods and services causes habit formation in

their consumption. Some of these new goods and services also contain externality, health

externality in form of obesity that represents an additional cost to society often unaccounted for

in studies about impacts of trade liberalization on an economy. This result is confirmed as we

move to more specific variables that directly reflect or proxy these changing trends in

consumption.

Regarding food consumption variables, soda and beans registered significant and positive

effects over the ratio of population with overweight. The positive impact of soda drinks on

obesity was expected and is consistent with results in some other studies (e.g., Miljkovic et al.,

2008; Lin et al., 2014). High contents of sugar in soda drinks lead to not only direct increase in

calories intake, but serve as physiological stimuli to increase an overall intake of calories

(Miljkovic et al., 2008). An increase in consumption of soda drinks per capita coincides well

with their availability made via increased imports or via foreign direct investments made in this

industry following trade liberalization. It is more difficult, however, to rationalize the positive

impact of beans consumption on obesity rate. Beans are known to be a food which is low in

saturated fat and sodium, and very low in cholesterol. Hence this positive impact on obesity rate

is somewhat unexpected. Previously was indicated in Table 2 that the consumption of these

foods varies among states as well as trends over time.

On the other hand, the income variable were not significant, although other variables

could be capturing income effects over health indicators, such as TV sets per household.

Although the correlation coefficient between the two variables is fairly high, it is in the range

where justifying the inclusion of both variables into regression equation is acceptable. It is also

true that even lower income people in Brazil, nowadays, have access to devices such as

refrigerator, TV and washing machines. Hence one could make the case that this variable could

serve as a proxy to increasing income, although we maintain the hypothesis that the number of

TV per household is more a reflection of a change in lifestyle than growth in income per se.

According to PNAD, the share of households that contains TV sets ranges from 85 percent to

more than 98 percent through the 27 states. Thus, the majority of households have access to this

technology and, therefore, we might affirm also that they are exposed to the existence of cultural

14

differences, which also includes different food habits and becomes a channel to promote changes

in people’s consumption patterns.

Interesting result is also that there is no difference in the growth of obesity between males

and females in Brazil. In many countries it has been recorded that female obesity is more

exacerbated than its male counterpart (e.g., Miljkovic et al., 2015), but our results for Brazil do

not confirm such disparity. Racially, higher rate of increase of overweight relative to the base

was estimated only for the yellow people, while there is no racial discrepancies in the growth

rate of obesity. No obvious pattern emerges about differing regional patterns and rates of

overweight and obesity growth.

6. Conclusions and Implications

This paper aimed to evaluate a possible relation between increased Brazilian trade

openness due to trade liberalization policies introduced in early 1990s and increasing observed

rates of overweight and obesity during the last 25 years.

Our empirical analysis is based on recent theoretical models developed to illustrate the

impacts of globalization, including trade liberalization, on global rise in obesity. We assume that

an individual country, Brazil in this case, imports a compound good combined of a vector of

consumption goods, capital, information, political influence, and in turn lifestyle and habit

formation in consumption, as presented in Atkin (2013). The imports result in a negative

externality, i.e., a health related externality which can be measured by an increase in obesity, for

the consumers in the importing country, though none of the firms or consumers (or their

representatives) in the exporting nations or rest of the world cares about the externality

(Miljkovic et al., 2015).

In the empirical model it was determined that the impacts of trade liberalization on

overweight and obesity in Brazil are positive and significant. Hence there is a health externality

in Brazil due to trade liberalization which has not been accounted for. Given that trade

liberalization implies lowering and/or elimination of trade barriers, the question remains what

policy is left to Brazilian policy makers to address this issue? So far, they have neither

recognized the existence of the problem nor addressed it in any way.

It would have been convenient to have an additional variable to capture directly

globalization effects, such as foreign direct investment (FDI), and particularly those coming from

food industry. Unfortunately, Brazilian database for FDI is not complete per state for the whole

15

period analyzed, compromising the use of this information. However, some of the results point in

the direction of what can be done to change things.

Given that the change in consumption of soda drinks and increased percentage of

households with TV could be considered as implicit globalization influences on traditional

Brazilian lifestyle, some related policies targeting these effects could be developed and enforced.

Just like in the case of the United States, taxes on sugar consumption including soda drinks could

be introduced (Miljkovic et al., 2008). Moreover, educational policies aiming at school age

population regarding soda consumption or availability could be developed to prevent further

escalation of obesity problem in Brazil. While conveniences of new electronics technologies are

huge, their over-use for entertainment purposes could be curbed to discourage sedentary life style

they promote. Hence extra taxes for such products that could be transferable to health sector to

deal with direct consequences of such life style would be a possible policy alternative to curb

obesity problem.

On a positive note, there are no gender and racial disparities in Brazil regarding the

presence of obesity. Hence there is no need to deal with sensitive political issues as in countries

such as the United States where obesity is more widespread among Hispanics or African

Americans than among white Americans.

International trade theory teaches us that non-intervention system guarantees most

efficient economy and maximizes social welfare. That is, if no externalities are present. In that

spirit, Brazil has recently promoted free trade and FDI openness, and these have been

contributing significantly to balance foreign currency market and guarantee better economic

growth rates. These positive outcomes make the environment sensitive to discuss potential

measures trying to restrict both trade and FDI flows.

Moreover, more consistent evidence is necessary to relate the increasing obesity and

overweight rates to globalization/trade liberalization in order to convince politicians and policy

makers that any interference will be necessary in this sense. In general, the world path is moving

towards freedom in choosing consumption patterns, which implies also cultural globalization,

and any public policy against that should be well justified to society. It is, however, necessary to

make public and policy makers aware of the existence of health externalities so that some viable

alternatives, such as the obesity tax, could be discussed in future.

16

References

Alvarenga, D. G1 - Subway chega a 583 lojas no Brasil e quer ser maior rede do país até 2015

- notícias em Negócios (2012) [online] Available at:

<http://g1.globo.com/economia/negocios/noticia/2011/03/subway-chega-583-lojas-no-brasil-

e-quer-ser-maior-rede-do-pais-ate-2015.html> (Accessed 15 March 2012).

Atkin, D. (2013). Trade, Tastes, and Nutrition in India. American Economic Review, 103(5), pp.

1629-63.

Averbug, A. (2002). Abertura e Integração Comercial Brasileira na Década de 90. [pdf].

Available at: <http://www.bndes.gov.br/SiteBNDES/export/sites/default/bndes_pt/Galerias/

Arquivos/conhecimento/livro/eco90_02.pdf >.

Blouin, C., Chopra, M. and van der Hoeven, R. (2009). Trade and social determinants of health.

The Lancet, 373(9662), pp.502-507.

Castells, M. (1997). The Power of Identity. Blackwell Publishers. Maiden and Oxford.

Clark, S.E., Hawkes, C., Murphy, S.M., Hansen-Kuhn, K.A. and Wallinga, D. (2012). Exporting

obesity: US farm and trade policy and the transformation of the Mexican consumer food

environment. International Journal of Occupational and Environmental Health, 18(1),

pp.53-64.

Conrad, K. (1993). Taxes and subsidies for pollution-intensive industries as trade policy. Journal

of Environmental Economics and Management, 25(1), pp.121-35.

Cowen, T. (2002). Creative Destruction: How Globalization is Changing the World Cultures.

Princeton University Press, Princeton.

17

Food and Agriculture Organization of the United Nations -FAO (2008). The developing world's

new burden: obesity. FAO, Rome, Italy.

Hawkes, C., Blouin, C., Henson, S., Drager, N. and Dubé, L. (2009). Trade, food, diet and

health: perspectives and policy options. John Wiley & Sons.

Instituto Brasileiro de Geografia e Estatística – IBGE (1988, 2002, 2008) [CD-ROM] Pesquisa

nacional por amostra por domicílios (PNAD): microdados. Rio de Janeiro: IBGE.

Instituto Brasileiro de Geografia e Estatística – IBGE (2002-2003, 2008-2009). [CD-ROM]

Pesquisa de orçamentos familiares (POF): microdados. Rio de Janeiro: IBGE.

Instituto Brasileiro de Geografia e Estatística – IBGE (2010). Pesquisa de Orçamentos

Familiares 2008-2009 – Despesas, Rendimentos e Condições de Vida. Rio de Janeiro: IBGE.

Instituto Nacional de Alimentação e Nutrição (INAN)/Instituto de Planejamento de Gestâo

Governamental (IPLAN)/Instituto Brasileiro de Geografia e Estatística (IBGE). (1989).

Pesquisa Nacional sobre Saúde e Nutrição (PNSN). Available at:

<http://dtr2004.saude.gov.br/nutricao/boletim_sisvan/bs_conceitpesquisa pnsn.php>.

Gilberto, K., Velásquez-Meléndez, G. (2003). A transição nutricional e a epidemiologia da

obesidade na América Latina. Cad. Saúde Pública, 19, suppl., Rio de Janeiro. Available at:

http://www.scielo.br/scielo.php?pid=S0102-311X2003000700001&script=sci_arttext

Levy-Costa, R. B.; Sichieri, R.; Pontes, N.S.; Monteiro, C.A. (2005). Disponibilidade domiciliar

de alimentos no Brasil: distribuição e evolução (1974-2003). Revista Saúde Pública, [S.l.],

39 (4), pp. 530-40.

Lin, B. H., Ver Ploeg, M., Kasteridis, P., & Yen, S. T. (2014). The roles of food prices and food

access in determining food purchases of low-income households. Journal of Policy

Modeling, 36(5), 938-952.

Ludema, R. D., Wooton, I. (1994). Cross-Border Externalities and Trade Liberalization: The

Strategic Control of Pollution. The Canadian Journal of Economics, 7(4), pp. 950-66.

Markusen, J. R. (1975). International externalities and optimal tax structures. Journal of

International Economics, 5(1), p.15-29.

Miljkovic, D., Nganje, W., and de Chastenet, H. (2008). Economic Factors Affecting the

Increase in Obesity in the United States: The Differential Response to Price. Food Policy,

33(1), p. 48-60.

18

Miljkovic, D., Shaik, S., Miranda, S., Barabanov, N. and Liogier, A. (2015). Globalisation and

Obesity. The World Economy, 38(8), pp. 1278-1294.

Ministério de Indústria, Desenvolvimento e Comércio - MDIC. Secretaria de Comércio Exterior

(2012). Sistema Alice Database. [online] Available at:<http://aliceweb2.mdic.gov.br/ >

(Accessed 20 April 2012).

Ministério da Saúde. Secretaria de Atenção à Saúde. Departamento de Atenção Básica (2009).

Indicadores de Vigilância Alimentar e Nutricional: Brasil 2006. Brasília: Ministério da

Saúde.

Ogden, C. L., Carroll, M.D., Curtin, L.R., McDowell, M.A., Tabak, C.J., Flegal, K. M. (2006).

Prevalence of Overweight and Obesity in the United States, 1999-2004. The Journal of the

American Medical Association, 295(13), pp. 1549-55.

Pequenas Empresas & Grandes Negócios (2009). Subway atinge marca de 300 lojas no Brasil.

Available at: <http://revistapegn.globo.com/Revista/Common/0,,EMI88364-17180,00-

SUBWAY+ATINGE+MARCA+DE+LOJAS+NO+BRASIL.html> Accessed 15 March

2015.

Schofield, D. (1996). The Impact of Employment and Hours of Work on Health Status and

Health Service Use. National Centre for Social and Economic Modelling, Faculty of

Management, University of Canberra, Discussion Paper No. 11. Available at:

<http://utah.natsem.canberra.edu.au/storage/dp11.pdf> Accessed 20 July 2013.

Segerstrom, P. (2003). Naomi Klein and the Anti-Globalization Movement. CEPR Discussion

Paper No. 4141.

Thow, A.M. and Hawkes, C. (2009). The implications of trade liberalization for diet and health:

a case study from Central America. Global Health, 5(5), pp.1-11.

UNESCO (2001). Universal Declaration on Cultural Diversity. 31st General Conference. Paris,

France, 2 November, 2001. UNESCO.

Wang, Y., and Beydoun, M. (2007). The Obesity Epidemic in the United States—Gender, Age,

Socioeconomic, Racial/Ethnic, and Geographic Characteristics: A Systematic Review and

Meta-Regression Analysis. Epidemiologic Reviews, 29(1), p. 6-28.

World Health Organization – WHO (2009). Data and statistics – Global database on Body Mass

Index. [Online] Available at < http://apps.who.int/bmi/index.jsp> Accessed December 2011.

19

World Health Organization – WHO (2006). Obesity and overweight. WHO Fact sheet [Online]

Available at:<http://www.who.int/mediacentre/factsheets/fs311/en/index.html > . Accessed

December, 2015.

0

100

200

300

400

500

600

1989 1994 1999 2004 2009

Num

ber o

f sto

res

AA BB CC DD

Figure 1 – Evolution of some fast-food chains in Brazil by number of stores, 1989 – 20123.

3 The source of data can be provided upon request. This procedure is to avoid identifying the companies and brands associated to each figures.

20

0 5 10 15 20 25 30

North

Northeast

Southeast

South

Center-West

Brazil

Percentage of population

Obesity

Overweight

Figure 2 – Distribution of overweight and obesity in Brazilian regions for total male population (10 to 19 years old), in percentage. Source: IBGE (2010).

21

Figure 3 – Trade openness for the 27 Brazilian states, 1989-2002-2008. In percentageSource: calculated from IBGE and MDIC databases.

22

Table 1 – Prevalence of overweight and obesity in population, older than 20 years, by gender. Brazil, ENDEF (1974-75), PNSN (1989) and POF (2002-03, 2008-09). In % of population

ENDEF (1974-1975)

PNSN (1989) POF (2002-2003)

POF (2008-2009*)

Overweight

Male 18.6 29.5 41.0 50.1Female 28.6 40.7 39.2 48.0ObesityMale 2.8 5.1 8.8 12.4Female 7.8 12.8 12.7 16.9Total - 11.7 13.1

Source: Ministério da Saúde (2009) and * IBGE (2010).

23

Table 2 – Average household purchase of some food categories, in kg or liters per capita year, Brazil and selected states. 1989, 2002 and 2008

Food State 1989 2002 2008Sodas Brazil 2.7 23.4 24

Amazonas (N) 2.2 14 20Bahia (NE) 6.0 12.1 12.1São Paulo (SE) 8.4 33.4 32.9Mato Grosso (CW) 2.2 19.7 16.8Rio Grande do Sul (S) 5 38.9 45.5

Beans Brazil 9.9 12.3 9.2Amazonas (N) 16.9 7.9 8.3Bahia (NE) 11.1 18.5 13.4São Paulo (SE) 8.1 8.4 6.3Mato Grosso (CW) 16.9 10.5 8.1Rio Grande do Sul (S) 11.6 10 5.9

Prepared food Brazil - 2.4 3.2Amazonas (N) - 2.4 2.8Bahia (NE) - 0.5 1.0São Paulo (SE) - 3.5 5.3Mato Grosso (CW) - 1.9 1.8Rio Grande do Sul (S) - 5 4.0

Note: North (N), Northeast (NE), Southeast (SE), South (S) and Center-West (CW) Regions.Source: POF/IBGE (2006, 2010).

24

Table 3 – Income, schooling and anthropometric data for selected states, Brazil. 1989, 2002 and 2008

State Overweight (% pop.)

Obesity (% pop.)

Household income p.c./year* (R$ real values 2008)

Average schooling years for people > 10 years

Brazil – 1989 25.62 8.8 4.4Amazonas (N) 27.12 8.64 366.30 4.8Bahia (NE) 20.64 5.14 229.78 3.1São Paulo (SE) 27.42 9.89 608.70 5.2Mato Grosso (CW)

24.80 7.87 392.41 4.0

Rio Grande do Sul (S)

28.42 11.86 423.75 4.9

Brazil - 2002 28.78 10.51 6.4Amazonas (N) 25.62 7.81 500.27 6.8Bahia (NE) 24.25 7.80 375.45 5.0São Paulo (SE) 30.66 12.36 924.44 7.5Mato Grosso (CW)

28.78 8.83 643.80 6.3

Rio Grande do Sul (S)

32.54 14.45 781.19 6.9

Brazil – 2008 33.97 14.64 7.3Amazonas (N) 36.19 11.01 563.43 7.2Bahia (NE) 29.78 11.14 503.38 6.1São Paulo (SE) 35.45 16.31 1002.84 8.3Mato Grosso (CW)

32.19 14.50 828.39 7.2

Rio Grande do Sul (S)

37.09 19.82 934.19 7.7

25

Table 4. Results of fixed effects panel model for Brazil (27 states). Dependent variable: log of ratio of obese and overweight people (older than 20 years) in total population. 1988/89, 2002 and 2008

Variables Overweight ObesityLog of trade openness 0.05* (0.03) 0.075* (0.04) Log of household per capita income (in real values for 2008)

0.19 (0.13) -0.10 (0.32)

Log of gender (female percentage in total population) a

-0.27 (1.09) 2.18 (2.10)

Log of “mulato” peoplea,b 0.07 (0.09) 0.24 (0.20) Log of yellow people a,b 0.03* (0.02) 0.03 (0.04) Log of black people a,b -0.00 (0.03) 0.08 (0.07) Log of white people a,b -0.16 (0.12) 0.13 (0.22) Log of percentage of households with TV 0.29** (0.11) 0.80*** (0.21) Log of beansc (kg p.c./year) 0.06 (0.06) 0.34** (0.14) log of sodasc (liters p.c./year) 0.04 (0.04) 0.21** (0.82) Binary variable for 2002 -0.11 (0.07) -0.43*** (0.15) Binary variable for 2008 -0.01 (0.09) -0.08 (0.20) Constant 2.06 (4.41) -11.92 (8.34) States fixed effects (27 units) Yes YesYear fixed effects (3 years) Yes YesNumber of observationsR2

720.924

720.931

Robust standard errors. ***Significant at 1% level, ** Significant at 5% level, * Significant at 10% level. Note: a control variables, and in the case of race, native Indian people were left out. b

Measured as share of total population; c Measured as household purchase.