ISSN 2042-2695

CEP Discussion Paper No 1098
November 2011
Anorexia, Body Image and Peer Effects:
Evidence from a Sample of European Women
Joan Costa-Font

and Mireia Jofre-Bonet



Abstract
Excessive preoccupation with self-image (or identity) is regarded as a factor contributing to
the proliferation of food disorders, especially among young women. This paper models how
self-image and peer effects influence health-related behaviours, specifically food disorders.
We empirically test our claims using data from the European survey. Our findings suggest
that the larger the peers’ body-mass, the lower the likelihood of being anorexic. Self-image is
correlated with body weight. We use several definitions of peers’ body mass and we find that
all are negatively associated with the likelihood of women being thin or extremely thin.

Keywords: self-image, identity, body image, eating disorders, anorexia, European women
JEL Classifications: I12, Q18



This paper was produced as part of the Centre’s Wellbeing Programme. The Centre for
Economic Performance is financed by the Economic and Social Research Council.



Acknowledgements
The authors express their gratitude to the editor professor Frank Cowell and the two
anonymous journal referees for their extremely helpful and precise comments, to the
participants at the London Health Economics Group seminar at LSE, FEDEA-Health
workshop in Barcelona, and the internal seminar at the Economics Department of City
University for helpful comments. Also, support from CESifo Munich Germany is gratefully
acknowledged.
Joan Costa-Font

is a Research Associate at the Centre for Economic Performance,
London School of Economics. He is also a Senior Lecturer in Political Economy – a joint
appointment at the Department of Social Policy and the European Institute at LSE. Mireia
Jofre-Bonet is a Senior Associated Researcher at LSE Health and Social Policy and Director
of Health Economics at City University, London.




Published by
Centre for Economic Performance
London School of Economics and Political Science
Houghton Street
London WC2A 2AE

All rights reserved. No part of this publication may be reproduced, stored in a retrieval
system or transmitted in any form or by any means without the prior permission in writing of
the publisher nor be issued to the public or circulated in any form other than that in which it
is published.

Requests for permission to reproduce any article or part of the Working Paper should be sent
to the editor at the above address.

 J. Costa-Font and M. Jofre-Bonet, submitted 2011
1

1. Introduction

It is becoming increasingly apparent that standards of physical appearance are important and
powerful motivators of human behaviour, especially regarding health and food. However, the
content and formation of these ideal-body standards have yet to be explored in the economics
literature. Case studies of eating disorders constitute a natural example to investigate how
changes in social attitudes towards physical appearance explain irregular health behaviour
such as eating disorders among women. Anorexia together with other food disorders such as
bulimia nervosa can be characterized by a distorted body image accompanied by an eating
obsession. Eating disorders can have damaging, and even devastating and life-threatening
effects (APA, 2000). About 6% of those who suffer from anorexia nervosa die from it
(Birmingham et al, 2005). Given that the relatively young females are more at risk, it
becomes especially important to understand how food disorders are engendered, identifying
the motivation underpinning them.

The reasons for the increasing trend towards food disorders are yet to be fully understood.
Some critics have argued that the phenomenon of anorexia is a paradigmatic example where
economics needs to examine preference formation given that anorexia and food disorders
more generally are recognised as “an extreme response to the pressures (for example, to
thinness) that are experienced by most if not all women” (Fine, 1995). This paper is the first
attempt to model anorexia and test some of these claims empirically. Only one other paper by
Goldfarb et al. (2009) proposes a model explaining anorexic disorders (low calorie intakes,
purging behaviours) that is based on taste variations and on an implicit rational choice to be
underweight. However, Goldfarb’s model does not attempt to include or explain the
formation of self-image, which, as we argue, is essential to understand food disorders.
Similarly, Ham (2009) proposes a model of addiction to explain bulimia nervosa, but does
not look at preference formation either. In the social science literature the formation of social
identity is seen as a key factor, and it is thought that food disorders are probably the result of
some ‘socially transmitted’ standard of ‘ideal’ body image affecting food intake and exercise.
Mainstream literature in social psychology regards social image as being continually under
construction and essential in determining physical, psychological and social equilibrium
(Schilder, 1958), hence anorexia results from a tension between aesthetic and utilitarian
reactions to food ingestion (Orbach, 1993). When applied to food disorders, this could
explain some extreme forms of weight aversion. This is the first claim of the paper.

Secondly, in explaining the formation of self-image, we claim that network phenomena
appear to be relevant to the biological

and behavioural trait of obesity (Christakis and Fowler,
2007) as reflecting the tensions between the imperatives to eat and not to eat (Fine, 1995).
Although the correlation between network effects and obesity is contested in Cohen-Cole and
Fletcher (2008), Trogdon et al. (2008) using a sample of adolescents found that mean peer-
weight is correlated with individual weight, suggesting that early health behaviour is
significantly determined by social influences. However, the specific mechanisms behind
peer-pressure on food disorders are unknown and require careful examination; the fact that
members of the peer group have a similar self-identity is a necessary, but not sufficient,
condition for the presence of social-multiplier effects (Costa-Font and Gil, 2004). To measure
the strength of such socially transmitted influences on individual behaviour it seems
appropriate to use the concept of peer or social-multiplier effect, as applied in Glaeser et al.
(1996) and in Sacerdote (2000). This concept arises not only when women have similar
behaviour or representations (self-identity) due to sharing a common environment, but also
when they belong to certain unobservable social groups (see Manski, 1993). Economic
2

explanations of health prevention require a better understanding of the effects of social
identity and self-image on health, and the development of empirical strategies to measure
these effects.

Recent contributions to the economics literature enable baseline modelling. Akerlof and
Kranton (2000) wrote the seminal paper in this area and included an application to gender
attitudes. Bodenhorn and Ruebeck (2003) created models for the influence of identity on
ethnic preferences. However, there is not much in the literature on the role of social identity
as a determinant of health. Blanchflower et al 2008 used Eurobarometer data for 29 countries
to show that overweight perceptions and dieting were influenced by the individual relative
body mass index (BMI). Lakdawalla and Philipson (2002) referred to an ‘ideal weight’, and
Etile (2007) examined the role of social norms on obesity and concluded that social norms
have an effect on ideal body-weight (for women). Gardner (1996) discussed the role of body-
image in behavioural reactions in cases where individuals perceived a large gap between their
desired image and the one they actually had, suggesting that this gap gave rise to permanently
distorted self-perceptions of the body. Altogether, the power exerted by media stereotypes of
beauty and the social norms that individuals are immersed in – especially the association
between thinness, aesthetic ideals and success (Hill, 1993) – is widely accepted. Further, it
has been suggested that the consequent fear of rejection based on physical appearance is
behind the increase in the number of persons suffering from eating disorders. Hence, one can
hypothesize that eating disorders are ‘socially formed’ rather than a biological pathology
(Bordo, 2003). Hutchinson (1982) points out that ‘body image’ refers not only to the
description of the body but to the place ‘where body, mind and culture meet’. Accordingly,
different cultural backgrounds are likely to exert idiosyncratic influences on the prevalence of
food disorders, and these need to be controlled for.

The aim of this paper is to build an economic model of eating disorders which relates social
and environmental factors to ‘self-image’ and objective weight. Some of the implications of
this model are taken to the data and the effect of underlying determinants is estimated. We
use a women-only sample from a representative European data-set (Eurobarometer 59.0). We
restrict attention to women since according to the APA (2000) women account for 90% of all
anorexia nervosa. Extreme thinness is different from thinness in that it is influenced by some
deviance form a “fine rule of healthiness”. This paper focuses particularly on the effect of
‘peer weight’ (which is likely to influence self-image or social identity) on the likelihood of
anorexia, and the influence of self-image on individual weight. In a joint-modelling exercise,
the paper then estimates the determinants of the probability of a woman being extremely thin
and, at the same time, seeing herself as fine or too fat. It then takes the two processes apart
and estimates a recursive probit model of being extremely thin and perceiving oneself as
being fine or too fat, finding that the unobserved factors explaining both processes are
correlated. This paper supports the hypothesis that social pressure through peer-shape is
determinant in explaining anorexia nervosa and distorted self-perception of one’s own body.
To the authors’ knowledge, there is no previous study examining anorexia that uses an
economic decision-model perspective combining self-image – or self-identity – formation
and individual health production. The Oswald and Powdthavee (2007) study on wellbeing an
obesity finds that anorexics typically exhibit a convex utility function with respect to weight.
A contribution of our study is to introduce identity or self-image to capture the tradeoffs
women face between following social pressures to attain a certain body shape and their
health.

The structure of the paper is as follows: Section 2 provides some background on the issue of
3

self-image and healthy eating among women. Section 3 proposes an economic model for
eating disorders. Section 4 sets out the empirical strategy used, describes the data-set and
estimates a reduced-form equation derived from the model. Section 5 presents the estimation
results and Section 6 contains a discussion and conclusions.


2. Background

Standard health production models fail to explain irregular behaviours such as those of
anorexics. Different factors have been suggested as possible determinants of anorexia, and
generally evidence does not suggest a clear consistency with the inclusion of social effects.
Some of these are related to ‘nature’, i.e. gender, genes and predisposition. Other factors are
more closely related to ‘nurture’ i.e. parental values and socio-cultural influences. In the
main, these determinants seem to make individuals susceptible to having their food and
exercise intake shaped by the strong socio-environmental pressures, which in this paper we
define as influencing what an ideal body looks like.

Puberty and anorexia. Girls who achieve sexual maturity ahead of their peers, with the
associated development of breasts, hips, and other physical signs of womanhood, are at
increased risk of becoming eating-disordered (Bordo, 2003). These girls often wrongly
interpret their new curves as signs of fatness and feel uncomfortable because they no longer
look like their peers, who still have childish bodies. A young woman in this group may
‘tackle’ her body, partly because she wants3 to take control and ‘fix’ her insecurity and
importantly because they are under the influence of a culture that equates success and
happiness with thinness. For this group of young women, dieting, bingeing, purging,
exercising, and other strange forms of behaviour are not random, but the result of a conscious
decision process.

Family and anorexia. There is some evidence indicating that eating disorders may run in
families. Parents influence their off-springs’ values and priorities, including those towards
food. Some people with eating disorders report having felt smothered in overprotective
families. Others have felt abandoned, misunderstood and alone. Parents who overvalue
physical appearance can unwittingly contribute to an eating disorder, as can parents who
make critical comments, even in jest, about their children's bodies. Furthermore, families that
include a person with an eating disorder tend to be rigid and ineffective at resolving conflicts.
In some such cases mothers are emotionally cool while fathers are physically and/or
emotionally absent. At the same time, there are high expectations of achievement and
success. Children in this type of family learn not to disclose doubts, fears, anxieties, and
imperfections. Instead they try to solve their problems by manipulating weight and food, in
an attempt to achieve the appearance of success, even if they do not feel successful (Bordo,
1993). Generally, anorexics reveal great fear of the criticism and rejection that would occur if
their perceived flaws and shortcomings should become known (Bachar et al, 2001).

Genetic Factors. Some studies suggest that there may be a genetic component in anorexia.
According to recent research (Fairburn et al 2005) genetic factors account for more than half
(56%) of the risk of developing anorexia nervosa and work on the genetics of bulimia and
binge-eating is under way. There are suggestions that women who develop anorexia nervosa
have excess activity in the brain's dopamine receptors, which regulate pleasure. This may
explain why they feel driven to lose weight but receive no pleasure from shedding pounds
(Frank, et al 2005).
4

Network effects: the media. Many people believe media stereotyping helps explain why about
90% of people with eating disorders are women and only 10% are men (Thompson and
Heinberg, 2002). In westernised countries, characterized by competitive striving for success,
women often experience unrealistic cultural demands for thinness. According to Health
magazine (April 2002), in the United States (US) 32% of female TV-network characters are
underweight, while only 5% of the female audience is underweight. Similarly, only 3% of
female TV-network characters are obese, while 25% of US women fall into that category.
The differences between media images of happy, successful men and women are interesting.
While women appear young, beautiful and thin, men are young or old, but strong and
powerful in all the areas that matter – physically, in business, and socially. Thin is not
desirable in men; power, strength and firmness are.

Despite TV being a dominant media type, some studies have found magazine-reading to be a
more consistent predictor than television-viewing (Harrison and Cantor, 2006). Studies of
undergraduate women have associated reading fashion magazines with having higher
preference for lower weight, having lower confidence on their own body image, feeling
frustrated for this reason etc (Turner et al., 1997). The ‘ideal’ body image portrayed by the
media influences social interaction and this may in turn make it more dominant. This
circularity only makes the power of social interactions in shaping people’s self-identity more
extreme.

To sum up, females of similar age, education and background are likely to have been exposed
to similar media and social environments and, accordingly, to have similar ideal self-
identities.


3. An Economic Decision Model for Eating Disorders

Current empirical evidence makes modelling eating disorders complex as one of the
assumptions of consumer-choice theory is the principle of non-satiation. According to extant
literature, food seems to need to be modelled as an economic ‘good’ up to a certain caloric
intake – which is idiosyncratic due to socially influenced self-perception – and as an
economic ‘bad’ thereafter.

In order to model anorexia, the self-identity model of Akerlof and Kranton (2000) was found
to be particularly useful and was adapted to the subject of interest. We assume that
individuals choose both food and exercise-related ‘actions’ - namely their net caloric intake -
in order to maximize an implicit utility function that depends not only on their net caloric
intake, but also on their self-image (or self-identity) and health. Besides these individual
factors, the utility function of individuals is conditioned by their peers’ net caloric intake -
and also their appearance and their characteristics - and by socio-cultural environmental
factors. Thus, the utility function can be modelled as:

) 1 ( ) , ; , , , , (
j j j j j j j j j
Z z H SI c a a U U
÷
=


where a
j
is j’s net caloric intake; a
-j
is the appearance of the j’s group of reference; c
j
reflects
j’s other actions – not related to caloric intake; SI
j
is j’s self-image; H
j
is j’s health-production
function; z
j
are j’s characteristics; and Z
j
the environmental factors in which j is immersed. To
simplify our model, we assume that all others’ actions, c
j
are embedded in the environmental
factors, Zj. Further, we assume that utility depends on the rather abridged concept of ‘net
5

caloric intake’ because food and exercise are a source of satisfaction beyond the body weight
they achieve.

Similarly to Akerlof and Kranton (2000), self-image SI
j
depends not only on j’s net caloric
intake, a
j
, but also on others’ body-weight-related actions or appearance, a
-j
; and is
conditioned by j’s individual characteristics and environmental factors, z
j
and Z
j
; and by j’s
status’, s
j
- as a person with higher status may have a better self-image than an identical one
with lower status. Thus, the equation for self-image is written as:

) 2 ( ) , , ; , (
j j j j j j j
Z z s a a I SI
÷
=

Finally, a health-production function H
j
is added.

This depends on j’s net caloric intake, a
j
; j’s
status’, s
j
; and any other individual and environmental factors, z
j
and Z
j
. The health-
production equation is written as follows:

( ; , , ) (3)
j j j j
H H a s z Z =



Standard utility maximization subject to a budget constraint under the usual regularity
assumptions would lead to an associated first-order condition as follows:

0
j j j j j
a
j j j j j
U
SI H
U U SI U H
P
a SI a H a
ì
c c c c c
+ + ÷ =
c c c c c
(4)

where ì is the usual income-multiplier and P
a
the monetary price of net caloric intake or the
combination of food price and exercise monetary cost including the opportunity cost of the
time invested in it.

A person without any eating disorder and in a range of net-caloric-intake would be expected
to receive a positive marginal utility from net-caloric-intake, from health and also from an
improved self-image. Also, it is assumed that a normal net-caloric-intake has a positive
marginal impact on health, since nutrition is necessary for survival. Thus, the first two
summands in equation (4) are expected to be positive.

In contrast, one can expect a negative marginal impact of net caloric intake on self-image
after a certain level of net-caloric-intake, which would make the sign of the second term in
equation (4) negative. The net-caloric-intake chosen to optimise overall utility will vary
depending on the relative magnitude of the positive and negative signs in equation (4) above,
bearing in mind that both anorexic and non-anorexic women will eventually confront the
economic principle of non-satiation. The difference lies in satiation among anorexic women
taking place at lower levels of consumption. In other words, the ‘bliss point’ of food
consumption for anorexics is lower, because the negative effect of eating on self-image is
greater for them. The opposite would apply for extremely overweight individuals.

Given the empirical evidence, a person with anorexia will have an extraordinarily large
negative term associated with the effect of net-caloric-intake on self-image. In this special
case the utility of net-caloric-intake would achieve a maximum at a much lower level than for
a non-anorexic person (see Figure 1). Note that the sign of the self-image term is
idiosyncratic insofar as it depends on the impact on each individual of the societal ideal-
6

body-shape that is in fashion.


Figure 1. Optimal equilibrium with and without anorexia















Thus, an anorexic individual chooses a net-caloric-intake a
j
that is under the optimal net-
caloric-intake â
j
associated with his/her characteristics had that individual not been anorexic.

This minimum-necessary net-caloric-intake threshold can be thought of as the one that would
keep individual j on a body mass index (BMI) considered ‘healthy’. Hence, whilst obesity
refers to the negative slope, the problem in this paper refers to the positive slope instead.

From equation (4), we can infer an implicit reduced form of net-caloric-intake that depends
on individual status, individual characteristics and the social environment (peer or network
effects), which includes the appearance/net-caloric-intake of others. In particular, under
standard normality and linearity assumptions, the likelihood of being anorexic, e.g. the
probability that the net-caloric-intake of an individual j is below his/her minimal healthy level
â
j
can be expressed as:

ˆ ( ) ( , , , , )
j j j j j j j
P a a s z Z a c ¢
÷ ÷
< = (5)

The next section describes how equation (5) is taken to the data, where observed variables
are included and others can be controlled for as country specific effects.


4. Data and Methods

4.1 Data

We combine two types of variables, namely individual-level variables and socio-
environmental variables. The former are taken directly from the answers to the
Eurobarometer 59.0 questionnaire, study number 3903. Eurobarometer 59.0 is one of the
Eurobarometer Surveys that have been conducted each spring and autumn since autumn
1973, adding countries as the European Union has expanded. The usual sample in standard
Eurobarometer Surveys is 1,000 people per country, with the exception of Luxembourg (600)
and the United Kingdom (1,000 in Great Britain and 300 in Northern Ireland). Also, since
Eurobarometer 34, 2,000 people have been sampled in Germany (1,000 in East Germany and
a
j
â
j
Utility

Net caloric intake

7

1,000 in West Germany) in order to monitor the integration of the five new Länder into
unified Germany and the European Union. In each of the 15 member states, the survey is
carried out by national institutes associated with the European Opinion Research Group. A
special issue, Eurobarometer 59.0, was carried out in all European Union countries between
15th January and 19th February 2003 on behalf of the European Opinion Research Group.
The questions from this Special Eurobarometer centred around attitudes towards life-long
learning, health issues, dietary habits and alcohol consumption, safety issues, partnership,
household tasks, childcare and family planning. It focussed particularly on the incidence of
chronic illness, on long-term treatment, on dental health and, in more depth, on health
maintenance (by discussing doctor's visits and various screening tests), on women's health
and medical tests relating specifically to women's health, and on general and children’s
safety.

Given that the mechanisms that give rise to anorexia and bulimia particularly affect women
(Hill, 1993) this paper focuses on women’s behaviour and thus only evidence on women was
selected. This gave a sample of 8,740 valid observations on women above 15 years of age.
Furthermore, given that anorexia is especially prevalent among younger women, we use a
sample of younger women between 15 and 34, which altogether make a sample of 2871
women. The percentage of extremely thin women of all ages ranges from 11% in France to
2% in Germany. The prevalence of Anorexia reaches a peak of 2% in some countries when
all women’s sample is examined, but when we restrict the analysis to a sample of younger
women the maximum prevalence rises to almost 5%.

We scrutinise a set of individual variables ranging from socio-demographic characteristics to
biometric measures and behavioural attitudes. This set of variables includes: (self-reported)
weight, height, own-body perception, healthiness of eating habits, age, gender, being married,
educational level, professional category, political attitudes, and residence in an urban or rural
area. Furthermore, to reflect the freedom and quality of the answers, in some of the
specifications the number of people present during the interview and the level of cooperation
is included.

Women are categorised as anorexic if they are extremely thin (BMI<17.5) but at the same
time perceive themselves as being ‘just fine’ or ‘too fat’. For that purpose, an indicator
variable called ‘anorexia’ was created, which took a value of 1 if a woman had a BMI of less
than 17.5 and at the same time saw herself as being ‘fine’ or ‘too fat’. A second anorexia
indicator variable labelled as ‘severe anorexia’ restricted the previous definition by
accounting for the fact that the respondent declared herself to be eating adequately (self
reported eating). Finally, self image was classified to identify seeing oneself in the ‘right
weight range or above it’, a variable called ‘fine or too fat’ was created, which took a value
of 1 if the individual declared she saw herself as fine or too fat, and of 0 otherwise.

A variable was also created that measured health consciousness through the number of
declared gynaecological check-ups received during the previous six months of the interview.
This variable was used as a control in the regression. A larger number of check ups is
interpreted to proxy women’s concerns with their own health, which can be hypothesized to
explain some of the unobserved heterogeneity in the model.

8

Figure 2: Prevalence of extreme thinness and anorexia among different age groups


0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
15-24 25-34 35-44 45-54 55-64 65+
Extreme Thinness Anorexia Severe Anorexia


Figure 2 reports the prevalence of the three variables examined in different age groups.
Extreme thinness in terms of very low BMI was highest during early youth (age group 15-24)
and progressively decreased until 55-64 years of age, increasing slowly again in the late years
of life. This could be due to attrition alongside later life metabolic conditions. Anorexia, as
defined here, had a prevalence of 3% for women aged between 15 and 24, just slightly higher
than severe anorexia. Both conditions followed a decreasing pattern till the age of 35, after
which they remained relatively constant at about 1%. We found that the prevalence of
anorexia is just below 4% for younger age groups and just below 2% among women aged 25-
34. Therefore, women 34 or below are expected to present a different pattern from women
over 34.

The first panel in Table 1 provides some overall statistics for the data (N=8,740). The average
age of the women in the sample was 45. Of these, 57% were married, 37.5% were heads of
household and 27% lived in a small town or rural area. Roughly 26% had completed primary
education, 41% had finished compulsory secondary education, 24% had studied up to 18
years of age, and 9.4% held a university degree. The average value of the variable ‘being
health conscious’ for the full sample was 1.25, indicating the average number of
gynaecological screenings received over the previous 6 months.

The second panel in Table 1 shows the descriptive statistics for young women, who had a
higher incidence of food disorders. This group included the women between 15 and 34 years
of age, with an average age of 25 years. Only 47% were married (in part due to increasing
cohabitation), 28% were heads of household and 30% lived in a small town or rural area.
Eight per cent had completed primary education, 41% compulsory secondary education, 23%
had studied until the age of 18, and 27% held a university degree (indicating some cohort
differences with previous generations). For this younger group (N=2,871), the average
number of gynaecological check-ups during the last half-year was 1. Other variables such and
individual’s income were initially included but were never significant, and were finally not
included in the reported specifications below. Our analysis includes country fixed effects to
explore the variability within countries. Finally, we report a set of robustness checks, which
account for the inclusion of different controls.
9

Table 1. Descriptive statistics

Variable ALL WOMEN WOMEN 15-34 years

Mean
(Std. Dev)
Mean
(Std. Dev)
Age
45.07
(17.9)
25.40
(5.59)
Married 0.568 0.469
Being head of household 0.375 0.276
Living in rural area 0.268 0.301
Having completed primary education 0.260 0.083
Having completed secondary education 0.407 0.411
Having received education up to 18
years 0.239 0.234
Holding a university degree 0.094 0.273
Being health conscious (using number
of gynaecological check-ups)
1.3
(1.5)
1.0
(1.3)
Peer effects: group BMI
25.40
(5.59)
24.7
(1.9)
Observations 8,740 2,871
Source: own elaboration using data from Eurobarometer 59.0 study number 3903

In order to reflect the woman’s peer effect (pressure felt in terms of acceptable body-shape in
her social environment/setting) different variables were created to represent the appearance of
others around her. These measures were: the average BMI of women with the same education
level, same reference age group (in ten-year groups), living in a similar environment (rural or
urban) and in the same immediate region of residence – all estimated using a recursive
system. Since individual BMI and BMI of the group of reference may be affected by
common unobserved factors, an instrument for peer BMI was developed using the average
BMI of women meeting the above criteria but from a different generation – i.e. five to ten
years older. All these effects would be expected to account for contemporaneous
geographical influences and, arguably could pick up to some reverse causality form younger
generations to older ones. This strategy can be found in other studies such as Grodner and
Kniesner (2006) and Etilé (2007) where ideal BMI is used. In an attempt to reflect social
norms and image patterns, a variable called ‘women’s magazines per capita’ was included
(measured by subscriptions), referring to the number of magazines categorized as ‘for
women.’ that were made available.

Table 2 provides a breakdown of average BMI by country and age-group, the percentage of
women with a BMI below 17.5; the percentage of women ‘seeing themselves as fine or too
fat’, the percentage of women who saw themselves as fine or too fat while having a BMI
below 17.5 (defined here as anorexic); the percentage of women who believed they ate
adequately; and finally, the percentage of women with a BMI below 17.5, who saw
themselves as fine or too fat and believed they were eating adequately (defined as severe
anorexic). The last column reports the circulation of women's magazines per 1,000 persons.
At the bottom of the table, we report the aggregate average BMI and the standard deviation
for these groups of women as categorized above.

The average BMI by country is about 25 for all ages and 23 for those between 15 and 34
years of age. The country with highest average BMI for all women was Greece (25.85) and
10

the lowest average BMI was that of Austria (23.67); for young women the highest average
BMI was that of Belgium (23.41) and the lowest average BMI was that of Italy (21.40). The
country with the highest prevalence of female anorexia (column 4) was Austria (1.55%),
followed by France and Ireland. The lowest prevalence was in Northern Ireland. For the
younger group of women, the highest prevalence of anorexia was in Austria (4.60%) and the
lowest in West Germany. Almost all countries contained a population that was generally
worried about its body weight, ranging from 56.29 in Northern Ireland to 35.97% in Ireland.
In the younger group of women, Luxembourg and Netherland had the highest percentage
with 49.42%, while Spain had only 22%. Consistently, the highest percentages of women
who declared they ate adequately were found in Denmark (95.53%) and Finland (90.59%)
and the lowest were found in Greece (64.07%) and Austria (78.86%). And, consistently with
all the above, the highest prevalence of anorexia was found in Austria, Italy and Ireland
among both young women and the full sample. Substantial differences were found between
the sample of younger women and the sample of women of all ages, suggesting that anorexia
is very much a recent phenomenon. Significant differences between younger women and
women as a whole were found in Austria, Ireland, Italy and Luxembourg, where a significant
high prevalence of anorexia was found among younger women. Interestingly, with hardly any
exceptions it was found that self-reported perception of eating adequately was higher among
older women that among younger ones. Finally, aggregate circulation of women’s magazines
was particularly high in Northern Ireland and Austria, though it was also high in West
Germany and Luxembourg.

Empirical Strategy

Given the dual health production and social dimension of the question examined, we
followed a two-step approach. First we estimated the impact of network effects and several
covariates on the likelihood of just being anorexic, i.e. equation (5) above. Then, in a second
step we used a bivariate recursive (seemingly unrelated) probit specification to separate the
two processes involved the definition of anorexia, namely being extremely thinness (an
outcome of the health production function in equation 3 above) and ‘seeing oneself as fine or
too fat’ (self-image formation in equation 2 above). The bivariate recursive probit allows a
better empirical representation of the theoretical model as it accounts for common
unobserved heterogeneity in these two processes. Also, it allowed us to disentangle the effect
of different variables on the two separate processes. For instance, variables such as magazine
circulation are expected to correlate with self-image but not with health production, unless
channelled through self-image. That is, although printed health stories influence extreme
thinness, we assume they do so through self-image rather than other routes.

In the light of both the empirical evidence and the model specification in section 2, we model
the individual’s propensity to be anorexic using a latent variable, A
j
*
, which depends on
individual and socio-environmental characteristics:


A
j
*
= |w
j
+ ¸Z
j
+oa
÷ j
+c
j
(6)

where w
j
are individual-specific controls and determinants of j’s status such as gender, age,
professional status, political affiliation and education, labelled as in equation (5); Z
j
refers to
the socio-environmental factors that individual j faces - including prevalence of women’s


11

Table 2: Country-specific BMI* average and other measures of thinness and self-image

Country obs BMI BMI < 17.5 Fine or Too Fat BMI<17.5 and
Fine or
Too Fat
Eating fine BMI<17.5,
Fine or
Too Fat at
and Eating
Fine
Magazines
All Young All Young All Young All Young All Young All Young
Belgium 519 24.67 23.41 1.35% 1.72% 48.31% 46.81% 0.39% 0.57% 85.08% 78.19% 0.39% 0.57% 0.053
Denmark 494 24.19 22.90 1.21% 2.22% 44.44% 40.29% 0.81% 1.48% 92.53% 82.01% 0.61% 0.74% 0.042
Germany East 508 24.03 22.19 0.62% 2.90% 41.56% 34.19% 0.21% 1.45% 82.77% 82.58% 0.00% 1.45% 0.037
Germany West 487 25.45 22.96 0.98% 1.56% 45.78% 37.06% 0.39% 0.00% 82.79% 76.22% 0.39% 0.00% 0.214
Greece 467 25.85 23.37 0.43% 1.20% 51.30% 38.15% 0.21% 0.60% 64.07% 54.91% 0.00% 0.00% 0.022
Spain 500 24.86 21.91 0.62% 0.54% 37.79% 22.40% 0.21% 0.54% 88.95% 90.10% 0.21% 0.54% 0.025
France 484 23.34 21.67 3.66% 5.14% 48.37% 40.22% 1.42% 1.14% 80.42% 75.54% 1.02% 0.00% 0.044
United Kingdom 492 25.98 25.05 1.18% 2.44% 54.73% 49.17% 0.51% 0.98% 82.84% 76.45% 0.34% 0.49% 0.037
Ireland 454 24.06 24.42 1.76% 3.19% 35.97% 30.48% 1.32% 2.66% 87.94% 87.62% 1.32% 2.66% 0.027
Italy 167 23.66 21.40 1.80% 3.90% 41.21% 33.94% 0.80% 2.60% 79.96% 75.15% 0.80% 2.60% 1.32
Luxembourg 309 24.00 22.03 3.56% 4.65% 48.10% 39.77% 1.29% 2.33% 84.49% 75.00% 1.29% 2.33% 1.159
The Netherlands 480 25.63 24.23 0.63% 0.00% 51.94% 49.42% 0.00% 0.00% 90.12% 86.63% 0.00% 0.00% 0.069
Northern Ireland 486 25.26 22.80 0.00% 0.00% 56.29% 49.21% 0.00% 0.00% 82.04% 76.19% 0.00% 0.00% 0.019
Austria 592 23.67 21.94 1.74% 5.17% 38.30% 29.63% 1.55% 4.60% 78.86% 76.72% 1.35% 4.02% 0.037
Portugal 557 25.41 23.01 1.23% 2.84% 42.12% 28.57% 0.62% 1.14% 83.88% 85.71% 0.62% 1.14% 0.097
Finland 499 25.31 23.16 1.08% 1.62% 49.75% 37.93% 0.54% 0.54% 90.59% 85.71% 0.54% 0.54% 0.043
Sweden 517 24.28 22.57 1.60% 2.50% 49.91% 38.07% 0.60% 1.25% 84.77% 82.95% 0.40% 0.63% 0.052
Average BMI
(standard
dev)

24.69
(4.88)
22.88
(4.28)
16.60
(0.85)
16.64
(0.82)
27.03
(4.84)
26.06
(4.56)
16.66
(1.01)
16.66
(0.95)
24.48
(4.72)
22.68
(4.12)
16.62
(1.08)
16.62
(1.04)


*Body mass index
Note: Eurobaometer and World Magazine Trends FIPP/ ZenithOptimedia World Magazine Trends§
12

magazines, country’s access to the internet, trust in the press, etc. and, for simplicity we
assume that Z
-j
contains c-
j
all others’ actions not related to a-
j
. which stands for peers’
appearance; and, as usual, c
j
represents j‘s unobserved idiosyncratic characteristics. Based
on our own definition of anorexia (see above to address the DSM-IV
1
criteria), we created a
dichotomous variable that takes value 1 if the person can be considered anorexic and 0
otherwise:

¹
´
¦ >
= =
>
otherwise
A if
A
j
A
j
0
0 1
1
*
) 0 (
*


Assuming normality of the error term in equation (6), it is possible to estimate the likelihood
of being anorexic in the form of the probit model:


P(A
j
=1| w
j
,Z
j
,a
÷ j
) =u(c
j
s |w
j
+ ¸Z
j
+oa
÷ j
) (7)

where

is the normal-distribution cumulative-probability function.

b) Joint estimation of own-body self image and health-production function
This second empirical exercise investigates how w
j
, Z
j
and a
-j
affect the two different
processes involved in anorexia according to the paper’s own definition: 1. having a self-
image of being in the right weight range or simply too fat; and, 2. being extremely thin.

We assumed that seeing oneself on the right weight range or too fat is an outcome reflecting
one’s latent (body) ‘self-identity’ as in equation (2). In order to simplify the analysis, we
assume that SI
*
j
depends linearly on individual characteristics and status (summarized in
variable w
j
,) and on peers’ appearance (a-
j
.):

*
j j j j
SI w a e , 0
÷
= + + (8)

Once again, in the survey we do not observe the ‘propensity’ to see oneself as fine or too fat,
so we create a dichotomous variable, SI
j
, that takes a value of 1 if the individual declares she
sees herself to do so and 0 otherwise. Under the usual linearity and normality assumptions,
this dichotomous variable allows us to obtain the probit model below:

) 8 ( ) ( ) , | 1 ( b a w e a w SI P
j j j j j j ÷ ÷
+ s u = = 0 ,

We test different specifications of our health production function outcome, we define another
measure of the health production function being extremely thin (BMI<17.5) may be thought
of as a partial representation of the individual’s latent health-production function H
j
,. As
before, what is being observed is the dichotomous variable, associated with this process. The
variable H
j
, takes value 1 when the individual is extremely thin and 0 otherwise.

We assume that the propensity to being extremely thin, labelled UW
j
*
, depends linearly on
individual characteristics and status, environmental variables, peer appearance, and also –
and very importantly – on the individual’s own-body perception or self-image, SI
*
. Being
extremely thin may also depend on additional factors and it is not necessary caused by a

1
The DSM-IV stands for the Diagnostic and Statistical Manual of Mental Disorders, which is a coding system
developed by the American Psychiatric Association to identify mental health disorders.
13

personal propensity to having a distorted self-image, x
j
.


j j j j
j
j
x Z w SI UW µ µ ì v ç + + + + =
* *
(9)

As before, in the survey we do not observe the ‘propensity’ to be extremely thin, so we create
a dichotomous variable, UW
j
, that takes a value of 1 if the individual declares is extremely
thin and 0 otherwise.

) 9 ( ) ( ) , , , , | 1 (
*
b x Z w SI a x Z w SI UW P
j j j
j
j j j j j j j
µ ì v ç µ + + + s u = =
÷


The system formed by equations (8b) and (9b) above is estimated on a recursive probit model
by assuming that the idiosyncratic terms
j
µ

and e
j
are jointly normally distributed. The
identification of parameters in the recursive probit model defined by equations (8b) and (9b)
is satisfied by the inclusion of variables x
j
in equation (9b) that do not appear in equation (8b)
and the triangularity created by the fact that a propensity to being underweight is influenced
by a distorted own-body self image but not the other way around.

By estimating the recursive probit model above, it was possible to investigate how individual
and environmental factors influence these two processes, while allowing the unobserved
factors affecting self-image and extreme thinness to be correlated. Furthermore, cross-
country genetic variations in BMI can be controlled for by the inclusion of country dummies
and clustering standard errors were always clustered. A country fixed effect allows exploring
the individual’s variability within each country, which can pick up some of the unobserved
cultural influenced on self-image and desired weight.


5. Results and Discussion

This section contains the results of estimating the probit model for being anorexic according
to the definition outlined above in equation (7). It also displays the results of estimating a
recursive probit model formed by equations (8) and (9), which made it possible to correlate
not only error terms but also self-image and extreme body-weight (thinness) in line with the
theoretical prediction displayed in Figure 1.

Table 3 displays the results of the preliminary strategy, which consisted of estimating a probit
model to determine which observable factors could cause a woman to see herself as fine or
too fat whilst being extremely thin (BMI<17.5) - which we call Anorexia. These are
displayed in column 1 or, if the woman also thought she was eating adequately (Extreme
Anorexia), in column 2. One probit model was estimated for the full sample and one for
women between 15 and 34 years of age. There were several explanatory variables or potential
determinants of such behaviour: marital status; living in a rural setting; being the head of the
household age (being the excluded group girls aged 18 to 24); education; and a proxy of
health-consciousness based on the declared number of gynaecological screenings taken in the
last 6 months. Some potential socio-environmental factors were also included: the circulation
of women’s magazines per capita in the country of residence, and peer BMI – based on the
BMI of women in the same age-bracket living in the same region.

For the full sample of women, the estimated marginal effects (see Table 3) showed, as
expected, that the BMI of the group of reference in terms of age, gender and location was
14

very significant and negative -0.0015 for the full sample, -0.0026 for the sample of younger
women. Thus, the higher the BMI of the peer group, the lower the probability of suffering
from anorexia. For the younger women, the effect of peer BMI was even more marked, in
terms of decreasing the probability of being anorexic. Taking into account that the peers’
BMI is 24.11 with a standard deviation of about 2, we can interpret the magnitude of this
coefficient in the right context: A change in one standard deviation in the peers’ BMI reduces
the probability of anorexia by 0.5%, i.e. 0.0026 times 2 is 0.0052. Being married, as opposed
to not, was only significant for the full sample. Ceteris paribus, the probability of anorexia is
higher for the excluded group (women aged 15 to 24 years of age) followed by women aged
45-55, and in turn followed by women aged 25 to 34 and 35-44. These results suggest the
presence of cohort effects. However, in the case of severe anorexia, cohort effects are less
marked, suggesting that the younger the women, the more likely it is to suffer anorexia.
Results show that having secondary education or having been to university all decreased the
likelihood of being anorexic or severely anorexic as defined above. These values reflect
cohort effects as explained latter in the results for the recursive probit specification. The signs
of the results for being married are influenced by cohort effects which alongside respondent’s
age and education are the expected ones. Network of peer effects are in line with expectation
and the literature on social-multiplier effects (Glaeser et al, 1996; Sacerdote, 2000).
Nevertheless, this result should be interpreted with caution because only a crude measure of
‘peer effect’ was used. Yet, we will come back to this point when dealing with robustness
checks.

Surprisingly, living in a rural setting, being the head of the family, and the measure of
women’s magazine circulation were not significant covariates explaining the probability of
anorexia for either group. Being in a rural environment was found not to be significant,
although urban women were expected to be subject to more social pressure with regards to
their appearance than those living in rural settings. However, this might have had to do with
other household-related variables such as quality and type of parenthood (Fairburn et al,
1999), which remained unobservable due to lack of data. The non-significance of being the
head of the household was sensitive to the inclusion of education, which may be picking up
part of the ‘being head of the household’ variation effect. However, given that what is being
studied is a combination of self-image and thinness, it might well be that the effects cancel
each other out, and this calls for a separate estimation strategy. The result of non-significance
for the women’s magazine circulation per capita was quite puzzling as it was not consistent
with some specific studies on the subject (Turner et al., 1997). This may be due to the
crudeness of the country measure and the possibility that the categories are not comparable
across countries; perhaps better quality data was required to measure the effect of
environmental or media-related variables.

The effects of these variables on the probability of being severely anorexic were qualitatively
very similar but slightly less marked than the ones commented on above.

15

Table 3. Probit model of the likelihood of suffering from anorexia

Anorexia § Severe Anorexia §§
VARIABLES
*
All women Younger women 15-34 All women Younger women 15-34
Peer effects: group BMI -0.0015*** -0.0026* -0.0014*** -0.0023**
(0.00457) (0.01) (0.0042) (0.01)
Being married -0.0381* -0.0370 -0.00459* -0.00681
(0.002) (0.006) -(0.002) (0.006)
Being health conscious† 0.00076** 0.00265 0.000493 0.00162
(0.001) (0.00214) (0.001) (0.002)
Between 25 and 34 years
old
-0.0327*** -0.0134*** -0.0221** -0.0752*
(0.001) (0.005) (0.001) (0.004)
Between 35 and 44 years
old
-0.0487*** -0.0368**
(0.001) (0.002)
More than 45 years of age -0.0101*** -0.0682***
(0.003) (0.002)
Having completed
secondary education
-0.0256** -0.0539 -0.0182** -0.0278
(0.001) (0.004) (0.001) (0.004)
Having a university degree -0.0247** -0.0535 -0.0210* -0.0404
(0.001) (0.006) (0.001) (0.005)
Controlled by country of
origin
Yes Yes Yes Yes
Number of observations 8012 2654 8012 2654
Psudo R squared 0.0859 0.0484 0.0788 0.0217
LogLikelihood -275.1 -180.8 -242.5 -145
Number of clusters
(countries)
17 17 17 17

Robust standard errors in brackets *** p<0.01, ** p<0.05, * p<0.1
Women’s Magazine circulation per capita, living in a rural environment, being the head of the household, and having education up to 18 years of age, were included but were
not significant.
§ Anorexia is defined as when the subject is below 17.5 BMI and sees herself as a fine or too fat person.
§§ Severe Anorexia is defined as when the subject is below 17.5 BMI, considers herself fine or too fat and also thinks she is eating appropriately.
† Being health conscious is defined as the sum of all gynaecological screenings received in the last 6 months.
16

Robustness checks for the probability of being anorexic
To test the robustness of the specifications in Table 3, their impact on the peer marginal-
effect coefficient was estimated by adding the controls incrementally (see Table 4 below).
The results obtained suggested that the peer marginal-effect was robust to these changes,
although the introduction of additional covariates progressively decreased the coefficient
from -0.02 to -0.014 for the less restrictive definition of anorexia, and from -0.016 to -0.014
for the strictest definition. This coefficient was barely more stable for the younger sample.
Even when additional controls were introduced; the coefficient decreased from -0.014 to -
0.010 for both definitions of anorexia. Overall, the peer (network) effects coefficients are
fairly stable, and change less than twice a standard deviation.

Joint estimation of fat self-image and low weight
Table 5 shows the marginal effects of a bivariate recursive probit model (Greene, 1998) of
being extremely thin and seeing oneself as fine or too fat separately, but allows the
unexplained variation in both processes to be related. Once again, the bivariate model was
estimated for the full sample of women first, and then for those in the 15-34 year age range
which allows to pick up the presence of cohort effects. Results turn out to suggest that
unobservable factors influencing the two processes are only significant for the total sample of
all women, but not for the subsample of younger women only. The identification restrictions
were that peer BMI and women’s magazine circulation were presumably related to body self-
perception but not to own-weight, while seeing oneself as fine or too fat (self-identity) was
likely to influence the probability of being extremely thin.

The estimates in the recursive bivariate probit model, where the two processes involved in the
paper’s simplistic definition of anorexia were disaggregated, gave rise to some interesting
findings. Peer BMI had a positive effect on the probability of seeing oneself as fine or too fat
(0.079 for the full sample) but as expected it revealed a negative effect -0.0377 for the
younger sample). Among younger women, being married had a positive effect on the
probability of being extremely thin (0.12) and on seeing oneself as fine or too fat (0.20). Age
had a non-linear effect on the probability of being extremely thin for the full sample but not
on the younger sample, probably because of cohort effects and the limited age-variation in
that group. Extreme thinness decreases with age at a decreasing rate, and increases again after
the age of 63. Living in a rural area had a positive effect on the likelihood of seeing oneself
as fine or too fat (0.056) but this was not significant for the younger sample. Having been to
university had a significant negative effect on seeing oneself as fine or too fat for both
samples (-0.345 and -0.214) but only in the full sample did it negatively affect the probability
of being extremely thin (-0.169). Surprisingly, neither being head of the household nor being
health-conscious was statistically significant. Being married did not explain extreme thinness

Seeing oneself as fine or too fat (self-image) had a very negative and significant effect on the
probability of being extremely thin (-0.424 and -0.451 for the full and younger sample,
respectively), as one would expect following the model outlined in Figure 1. This result is
particularly important as it provides evidence consistent with the idea that self-image and
identity do exert an influence on health-production, at least in the case of anorexia. Finally,
the coefficient representing the correlation of the error terms of both processes, Atrho, is
positive and highly significant for both samples, corroborating the fact that there are some
unobserved factors influencing both women’s body identity and extreme thinness that are
positively correlated.


17

Table 4. Robustness checks using alternative probit model specifications

Sample Anorexia § Severe Anorexia § §
Anorexia (All
women)

Peer Effects:
group BMI
-0.002 -
0.0017
-
0.0017
-
0.0015
-
0.0014
-
0.0018
-
0.0016
-
0.0015
-
0.0015
-
0.0014
(0.001) (0.001) (0.001) (0.003) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Personal
characteristics
Yes Yes Yes Yes Yes Yes Yes Yes
Age variables Yes Yes Yes Yes Yes Yes
Education Yes Yes Yes Yes
Health
consciousness
Yes Yes
Controlled by
country of origin
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Anorexia (Young
Women)

Peer effects:
group BMI
-
0.0014
-
0.0011
-
0.0011
-
0.0010
-
0.0010
-
0.0014
-
0.0011
-
0.0011
-
0.0011
-
0.0010
(0.001) (0.001) (0.001) (0.003) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001)
Personal
Characteristics
Yes Yes Yes Yes Yes Yes Yes Yes
Health
consciousness
Yes Yes Yes Yes Yes Yes
Age variables Yes Yes Yes Yes
Education Yes Yes
Controlled by
country of origin
Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes


18

Table 5. Recursive-probit models for being extremely thin and seeing oneself as 'fine’ or
‘too fat' (marginal effects)

All women Younger
women 15-34

VARIABLES
*
Extreme
thinness ‡
Seeing oneself as
fine or too fat
Extreme
thinness ‡
Seeing oneself as
fine or too fat
Peer effects: group
BMI

0.079
(0.07)

-0.0377**
(0.012)
Being married -0.0034 0.176*** 0.12** 0.20***
-(0.049) (0.042) (0.07) (0.07)
Between 25 and 34
years old
0.180*** 0.13***
(0.036) (0.039)
Between 35 and 44
years old
0.309***
(0.045)
More than 45 years of
age
0.338***
(0.051)
Age -0.02* 0.005
-(0.01) (0.03)
Age squared 0.0001* -0.0002
(0.000) -(0.001)
Living in a rural area 0.046 0.056** 0.0320 0.0089
(0.031) (0.023) (0.052) (0.036)
Having been to
university
-0.169* -0.345*** -0.162 -0.214**
-(0.095) -(0.089) -(0.110) -(0.098)
Seeing oneself as fine
or too fat
-0.424*** -0.451***
-(0.134) -(0.17)
Constant 0.289* -3.01*** -0.0381 -1.960***
(0.146) -(0.50) -(0.424) -(0.569)
Atrho 1.583*** 7.564
(0.431) (19.280)

Controlled by country
of origin
Yes Yes
Number of
observations
8740 2871
Chi-Square for rho=0 15.1 1.18
Reject null rho=0 Yes No
Degrees of freedom 14 14
Loglikelihood -3765.45 -1710.34
Number of clusters
(countries)
17 17

Robust standard errors in brackets
*** p<0.01, ** p<0.05, * p<0.1
* Women’s Magazine circulation per capita, living in a rural environment, being the head of the household,
being health conscious, and having compulsory education and up to 18 years of age, were included but were not
significant.
† Being health conscious is defined as the sum of all gynaecological screenings had in the last 6 months.
‡ Thinness is defined as having a BMI of below 17.5
19

Robustness checks for the joint estimation of fat self-image and low weight
Table 6 shows that the impact of self-image on being extremely thin remained almost
constant when additional controls were introduced for both the ‘all women’ and ‘younger
women’ samples. Results show that show that as expected, the coefficient of fat self image
was negative and robust, revealing that women that see themselves as ‘fine’ or ‘too fat’ are
44%-46% less likely to be extremely thin. The instruments employed performed well
following traditional Hausman test methods and were theoretically relevant. However,
additional factors might still be present. For instance, unobservable variables affecting
women of different ages may bias upward the relationship between BMI and peer-average
BMI (and thus bias downward the peer effect coefficient).

Table 6. Robustness checks using alternative specifications

All women Dependent variable: Extreme thinness
Seeing oneself as fine or too
fat
-0.42 -0.44 -0.44 -0.44
(standard error) (0.014) (0.130) (0.129) (0.130)
Personal characteristics Yes Yes Yes Yes
Age variables No Yes Yes Yes
Education No No Yes Yes
Health consciousness No No No Yes
Controlled by country of
origin
Yes Yes Yes Yes
Young Women 15-34
Seeing oneself as fine or too
fat
-0.45 -0.44 -0.44 -0.44
(standard error) (0.200) (0.210) (0.210) (0.210)
Personal characteristics Yes Yes Yes Yes
Age variables No Yes Yes Yes
Education No No Yes Yes
Health consciousness No No No Yes
Controlled by country of
origin
Yes Yes Yes Yes


6. Conclusions

This paper presents a simple theoretical framework to explain the influence of self image and
other people’s body shapes on female caloric intake, in line with the identity model of
Akerlof and Kranton (2000). It then uses an empirical strategy to identify the determinants of
food disorders (anorexia) and self-image (seeing oneself as fine or too fat) following first a
joint process and secondly a recursive one. We find that the larger the peers’ body-mass, the
lower the likelihood of being anorexic. Additionally, self-image correlates with body weight
even when unobservable factors explaining both processes are controlled for.

Our results were consistent with the assumption that individuals trade off health against self-
image. Also, in agreement with the epidemiological literature, we found that weight-related
food disorders happen mostly at younger ages and require attention before they extend to
older age groups. Note that the findings showed that anorexia primarily affected women aged
between 15 and 34, and that it was primarily socially induced. These results have serious
20

policy implications. They call for urgent action on individual identity, probably while it is
still being formed, so as to prevent severe damage to women’s health and in order to improve
their well-being and that of their families and friends.

The influence of a crude measure of peer effects is significant and robust throughout the
samples, indicating that socio-environmental factors play an important role. This result
should be corroborated using longitudinal data, but these are not available in Europe at the
moment. The paper’s findings were the best that could be done with the existing cross-
sectional data on Europe. They provide some important results that can act as a basis for
future literature. In addition the paper contributes to behavioural economics by using a model
for eating disorders that allows for net-caloric-intake being a ‘bad’ instead of a ‘good’ in the
consumer utility function above a certain intake.

Our findings are in line with the Clark and Oswald (1998) model of comparison utility in that
deviant behaviour - such as anorexia - may occur when an individual attempts to deviate from
some social norm using her own-BMI as an instrument. Hence, it is important to understand
how individuals come to value what they do. Consistently, Etile (2007) funds that social
norms regarding body shape have significant effects on perceptions of ideal BMI only for
those women who want to loose weight. Anorexics, by definition, want to lose weight. In the
health policy arena, this implies understanding how preferences for smoking, eating
unhealthy food or avoiding physical activity - with their costs in terms of health and well-
being - are incorporated in people’s utility maximization. Underlying this debate is the
question of time-discount rates and the formation of preferences; which has important
consequences for health-policy given that preferences for health-related activities are likely
both to be influenced by and to influence health outcomes. These results go a step beyond
Christakis and Fowler (2007) by exploiting self image as an intermediary source of social
pressure.

In the light of this study, government intervention to adjust individual biases in self-image
would be justified to curb or at least prevent the spread of a potential epidemic of food
disorders. The distorted self-perception of women with food disorders and the importance or
the peer effects may prompt governments to take action to influence role models and
compensate for social pressure on women driving the trade-off between ideal weight and
health. However, given the nature of the data and the absence of natural experiments we can’t
prove our results as being causal and should be taken with caution.
21

References

Akerlof, G. A. and Kranton, R. E. (2000), ‘Economics and Identity’, Quarterly Journal of
Economics 115:3 pp. 715-753.

APA (2000), American Psychiatric Association, Diagnostic on Statistical Manual of Mental
Disorders. DSM-IV-TR., American Psychiatric Association: Arlington, VA.

Bachar, E., Latzer, Y., Canetti, L., Gur, E., Berry, E. M. and Bonne, O. (2001), ‘Rejection of
Life in Anorexic and Bulimic Patients’, International Journal of Eating Disorders,
31(1): 43–48.

Blanchflower, D., Oswald, A. and Van Landeghem, B. (2008), ‘Imitative Obesity and
Relative Utility’, NBER Working Paper No. 14337.

Birmingham, C. L., Su, J., Hlynsky, J. A.; Goldner, E. M. and Gao, M. (2005), ‘The
Mortality Rate from Anorexia Nervosa’, International Journal of Eating Disorders,
38: 143-146.

Bodenhorn, H. and Ruebeck, C. S. (2003), ‘The Economics of Identity and the Endogeneity
of Race’, NBER Working Paper No. 9962.

Bordo, S. (2003), Unbearable Weight: Feminism, Western Culture, and the Body. University
of California Press; 2nd revised edition.

Christakis, N. A. and Fowler, J. H. (2007), ‘The Spread of Obesity in a Large Social Network
over 32 Years’, N Engl J Med 2007; 357:1866-1868.

Clark, A. E. and Oswald, A. J. (1998), ‘Comparison-Concave Utility and Following
Behaviour in Social and Economic Settings’, Journal of Public Economics 70 (1998),
133–155.

Cohen-Cole, E. and Fletcher, J. M. (2008), ‘Is Obesity Contagious? Social Networks Vs.
Environmental Factors in the Obesity Epidemic’, Journal of Health Economics,
Elsevier, Vol. 27(5), 1382-1387.

Connor Gorber S., Shields, M., Tremblay, M. S. and McDowell, I. (2008), ‘The Feasibility of
Establishing Correction Factors to Adjust Self-Reported Estimates of Obesity’,
Health Rep. Sep;19(3):71-82. Health Information and Research Division, Statistics
Canada, Ottawa.

Costa-Font, J. and Gil Transfi, J. (2004), ‘Social Interactions and the Contemporaneous
Determinants of Individuals’ Weight’, Applied Economics, 36 (20), pp. 2253-2263.

Etile, F. (2007), ‘Social Norms, Ideal Body Weight and Food Attitudes’, Health Economics
(in press).

Fairburn, C. G., Cooper, Z., Doll, H. A. and Welch, S. L. (1999), ‘Risk Factors for Anorexia
Nervosa’, Arch Gen Psychiatry, 56:468-476.

22

Fine, B. (1995), ‘Towards a Political Economy of Anorexia’, Appetite, 24(3), 231-42.

Frank, G., Bailer, U. F., Henry, S. E., Drevets, W., Meltzer, C. C., Price, J. C., Mathis, C. A.,
Wagner, A., Hoge, J., Ziolko, S., Barbarich-Marsteller, N., Weissfeld, L. and Kaye,
W. H. (2005), ‘Increased Dopamine D2/D3 Receptor Binding After Recovery from
Anorexia Nervosa Measured by Positron Emission Tomography and [
11
C]raclopride’,
Biol. Psychiatry, 58 11, 908–912

Gardner, R. M. (1996), ‘Methodological Issues in Assessment of the Perceptual Component
of Body Image Disturbance’, British Journal of Psychology, 87, 327-337.

Glaeser, E., Sacerdote, B. and Scheinkman, J. (1996), ‘Crime and Social Interactions’,
Quarterly Journal of Economics, 91: 507-548.

Goldfarb, R. B., Leonard, T. C., Markowitz, S. and Suranovic, S. (2009), ‘Can A Rational
Choice Framework Make Sense of Anorexia Nervosa?’, NBER Working Paper
No.14838.

Greene, W. (1998), ‘Gender Economics Courses in Liberal Arts Colleges: Further Results’,
Journal of Economic Education, Fall 1998, p. 291-301.

Goeree, M. S., Ham, J. C. and Iorio, D. (2009), ‘Caught in the Bulimic Trap? On the
Incidence, Persistence and State Dependence of Bulimia Among Blacks and Whites’,
Mimeo, University of Zurich.

Harrison K. and Cantor, J. (2006), ‘The Relationship Between Media Consumption and
Eating Disorders’, Journal of Communication, Volume 47 Issue 1, Pages 40-67,

Hill, A. (1993), ‘Causes and Consequences of Dieting and Anorexia’, Proceedings of the
Nutrition Society 52, 211-218.

Hudson, J. I., Hiripi, E., Pope, H. G. and Kessler, R. C. (2007), ‘The Prevalence and
Correlates of Eating Disorders in the National Co morbidity Survey Replication’, Biol
Psychiatry, 61(3): 348-358.

Hutchinson, M. G. (1982), ‘Transforming Body-Image: Your Body, Friend or Foe?’, Women
and Therapy, 1, 59-67.

Lakdawalla, D. and Philipson. T. J. (2002), ‘The growth of obesity and technological change:
a theoretical and empirical examination’, NBER Working Paper No. 8946.

Manacorda, M. and Moretti, E. (2006), ‘Why do Most Italian Youths Live with Their
Parents? Intergenerational Transfers and Household Structure’, Journal of the
European Economic Association, 4(4), 800-829.

Manski, C. (1993), ‘Identification of Endogenous Social Effects: The Reflection Problem’,
Review of Economic Studies, 60, 531-542.

Orbach, S. (1993), Hunger Strike: The Anorectic's Struggle As A Metaphor For Our Age,
Harmondsworth: Penguin.
23


Oswald, A. J. and Powdthavee, N. (2007), ‘Obesity, Unhappiness, and The Challenge of
Affluence: Theory and Evidence’, Economic Journal, Vol. 117(521), 441-454.

Sacerdote, B. (2000), ‘Peer Effects with Random Assignment: Results for Dartmouth
Roommates’, NBER Working Paper No.7469.

Schilder, P. (1958), The Image and Appearance of the Human Body. New York: International
Universities Press.

Slade, P. P. (1988), ‘Body Image in Anorexia Nervosa’, British Journal of Psychiatry, 153
(2), 20-22.

Thompson J. K. and Heinberg, L. J. (2002), ‘The Media's Influence on Body Image
Disturbance and Eating Disorders: We've Reviled Them, Now Can We Rehabilitate
Them?’, Journal of Social Issues, Volume 55 Issue 2, Pages 339 – 353, Published
Online: 17 Dec 2002.

Trogdon, J., Nonnemaker, G. and Pais, J. (2008), ‘Peer Effects in Adolescent Overweight’,
Journal of Health Economics 27, 5, 1388-1399.

Turner S. L., Hamilton, H., Jacobs, M., Angood, L. M. and Hovde Dwyer, D. (1997), ‘The
Influence of Fashion Magazines on the Body Image Satisfaction of College Women:
An Exploratory Analysis’, Adolescence, Vol. 32, n127: p 603-615.

World Magazine Trends FIPP / ZenithOptimedia World Magazine Trends.
CENTRE FOR ECONOMIC PERFORMANCE
Recent Discussion Papers
1097 Michal White
Alex Bryson
HRM and Workplace Motivation:
Incremental and Threshold Effects
1096 Dominique Goux
Eric Maurin
Barbara Petrongolo
Worktime Regulations and Spousal Labor
Supply
1095 Petri Böckerman
Alex Bryson
Pekka Ilmakunnas
Does High Involvement Management
Improve Worker Wellbeing?
1094 Olivier Marie
Judit Vall Castello
Measuring the (Income) Effect of Disability
Insurance Generosity on Labour Market
Participation
1093 Claudia Olivetti
Barbara Petrongolo
Gender Gaps Across Countries and Skills:
Supply, Demand and the Industry Structure
1092 Guy Mayraz Wishful Thinking
1091 Francesco Caselli
Andrea Tesei
Resource Windfalls, Political Regimes, and
Political Stability
1090 Keyu Jin
Nan Li
Factor Proportions and International Business
Cycles
1089 Yu-Hsiang Lei
Guy Michaels
Do Giant Oilfield Discoveries Fuel Internal
Armed Conflicts?
1088 Brian Bell
John Van Reenen
Firm Performance and Wages: Evidence from
Across the Corporate Hierarchy
1087 Amparo Castelló-Climent
Ana Hidalgo-Cabrillana
The Role of Educational Quality and Quantity
in the Process of Economic Development
1086 Amparo Castelló-Climent
Abhiroop Mukhopadhyay
Mass Education or a Minority Well Educated
Elite in the Process of Development: the Case
of India
1085 Holger Breinlich Heterogeneous Firm-Level Responses to
Trade Liberalisation: A Test Using Stock
Price Reactions
1084 Andrew B. Bernard
J. Bradford Jensen
Stephen J. Redding
Peter K. Schott
The Empirics of Firm Heterogeneity and
International Trade
1083 Elisa Faraglia
Albert Marcet
Andrew Scott
In Search of a Theory of Debt Management
1082 Holger Breinlich
Alejandro Cuñat
A Many-Country Model of Industrialization
1081 Francesca Cornaglia
Naomi E. Feldman
Productivity, Wages and Marriage: The Case
of Major League Baseball
1080 Nicholas Oulton The Wealth and Poverty of Nations: True
PPPs for 141 Countries
1079 Gary S. Becker
Yona Rubinstein
Fear and the Response to Terrorism: An
Economic Analysis
1078 Camille Landais
Pascal Michaillat
Emmanuel Saez
Optimal Unemployment Insurance over the
Business Cycle
1077 Klaus Adam
Albert Marcet
Juan Pablo Nicolini
Stock Market Volatility and Learning
1076 Zsófia L. Bárány The Minimum Wage and Inequality - The
Effects of Education and Technology
1075 Joanne Lindley
Stephen Machin

Rising Wage Inequality and Postgraduate
Education

1074 Stephen Hansen
Michael McMahon
First Impressions Matter: Signalling as a
Source of Policy Dynamics
1073 Ferdinand Rauch Advertising Expenditure and Consumer
Prices
1072 Alberto Galasso
Mark Schankerman
Carlos J. Serrano
Trading and Enforcing Patent Rights
1071 Daniel Hale
John Coleman
Richard Layard
A Model for the Delivery of Evidence-Based
PSHE (Personal Wellbeing) in Secondary
Schools
1070 David Marsden Individual Voice in Employment
Relationships: A Comparison Under Different
Forms of Workplace Representation
1069 Ross Levine
Alexey Levkov
Yona Rubinstein
Racial Discrimination and Competition
1068 Klaus Adam
Albert Marcet
Internal Rationality, Imperfect Market
Knowledge and Asset Prices
The Centre for Economic Performance Publications Unit
Tel 020 7955 7673 Fax 020 7955 7595
Email info@cep.lse.ac.uk Web site http://cep.lse.ac.uk
CENTRE FOR ECONOMIC PERFORMANCE
Recent Discussion Papers
1098 Joan Costa-Font
Mireia Jofre-Bonet
Anorexia, Body Image and Peer Effects:
Evidence from a Sample of European Women
1097 Michal White
Alex Bryson
HRM and Workplace Motivation:
Incremental and Threshold Effects
1096 Dominique Goux
Eric Maurin
Barbara Petrongolo
Worktime Regulations and Spousal Labor
Supply
1095 Petri Böckerman
Alex Bryson
Pekka Ilmakunnas
Does High Involvement Management
Improve Worker Wellbeing?
1094 Olivier Marie
Judit Vall Castello
Measuring the (Income) Effect of Disability
Insurance Generosity on Labour Market
Participation
1093 Claudia Olivetti
Barbara Petrongolo
Gender Gaps Across Countries and Skills:
Supply, Demand and the Industry Structure
1092 Guy Mayraz Wishful Thinking
1091 Francesco Caselli
Andrea Tesei
Resource Windfalls, Political Regimes, and
Political Stability
1090 Keyu Jin
Nan Li
Factor Proportions and International Business
Cycles
1089 Yu-Hsiang Lei
Guy Michaels
Do Giant Oilfield Discoveries Fuel Internal
Armed Conflicts?
1088 Brian Bell
John Van Reenen
Firm Performance and Wages: Evidence from
Across the Corporate Hierarchy
1087 Amparo Castelló-Climent
Ana Hidalgo-Cabrillana
The Role of Educational Quality and Quantity
in the Process of Economic Development
1086 Amparo Castelló-Climent
Abhiroop Mukhopadhyay
Mass Education or a Minority Well Educated
Elite in the Process of Development: the Case
of India
1085 Holger Breinlich Heterogeneous Firm-Level Responses to
Trade Liberalisation: A Test Using Stock
Price Reactions
1084 Andrew B. Bernard
J. Bradford Jensen
Stephen J. Redding
Peter K. Schott
The Empirics of Firm Heterogeneity and
International Trade
1083 Elisa Faraglia
Albert Marcet
Andrew Scott
In Search of a Theory of Debt Management
1082 Holger Breinlich
Alejandro Cuñat
A Many-Country Model of Industrialization
1081 Francesca Cornaglia
Naomi E. Feldman
Productivity, Wages and Marriage: The Case
of Major League Baseball
1080 Nicholas Oulton The Wealth and Poverty of Nations: True
PPPs for 141 Countries
1079 Gary S. Becker
Yona Rubinstein
Fear and the Response to Terrorism: An
Economic Analysis
1078 Camille Landais
Pascal Michaillat
Emmanuel Saez
Optimal Unemployment Insurance over the
Business Cycle
1077 Klaus Adam
Albert Marcet
Juan Pablo Nicolini
Stock Market Volatility and Learning
1076 Zsófia L. Bárány The Minimum Wage and Inequality - The
Effects of Education and Technology
1075 Joanne Lindley
Stephen Machin

Rising Wage Inequality and Postgraduate
Education

1074 Stephen Hansen
Michael McMahon
First Impressions Matter: Signalling as a
Source of Policy Dynamics
1073 Ferdinand Rauch Advertising Expenditure and Consumer
Prices
1072 Alberto Galasso
Mark Schankerman
Carlos J. Serrano
Trading and Enforcing Patent Rights
1071 Daniel Hale
John Coleman
Richard Layard
A Model for the Delivery of Evidence-Based
PSHE (Personal Wellbeing) in Secondary
Schools
1070 David Marsden Individual Voice in Employment
Relationships: A Comparison Under Different
Forms of Workplace Representation
1069 Ross Levine
Alexey Levkov
Yona Rubinstein
Racial Discrimination and Competition
The Centre for Economic Performance Publications Unit
Tel 020 7955 7673 Fax 020 7955 7595
Email info@cep.lse.ac.uk Web site http://cep.lse.ac.uk

Abstract Excessive preoccupation with self-image (or identity) is regarded as a factor contributing to the proliferation of food disorders, especially among young women. This paper models how self-image and peer effects influence health-related behaviours, specifically food disorders. We empirically test our claims using data from the European survey. Our findings suggest that the larger the peers’ body-mass, the lower the likelihood of being anorexic. Self-image is correlated with body weight. We use several definitions of peers’ body mass and we find that all are negatively associated with the likelihood of women being thin or extremely thin. Keywords: self-image, identity, body image, eating disorders, anorexia, European women JEL Classifications: I12, Q18

This paper was produced as part of the Centre’s Wellbeing Programme. The Centre for Economic Performance is financed by the Economic and Social Research Council.

Acknowledgements The authors express their gratitude to the editor professor Frank Cowell and the two anonymous journal referees for their extremely helpful and precise comments, to the participants at the London Health Economics Group seminar at LSE, FEDEA-Health workshop in Barcelona, and the internal seminar at the Economics Department of City University for helpful comments. Also, support from CESifo Munich Germany is gratefully acknowledged. Joan Costa-Font is a Research Associate at the Centre for Economic Performance, London School of Economics. He is also a Senior Lecturer in Political Economy – a joint appointment at the Department of Social Policy and the European Institute at LSE. Mireia Jofre-Bonet is a Senior Associated Researcher at LSE Health and Social Policy and Director of Health Economics at City University, London.

Published by Centre for Economic Performance London School of Economics and Political Science Houghton Street London WC2A 2AE All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means without the prior permission in writing of the publisher nor be issued to the public or circulated in any form other than that in which it is published. Requests for permission to reproduce any article or part of the Working Paper should be sent to the editor at the above address.  J. Costa-Font and M. Jofre-Bonet, submitted 2011

but does not look at preference formation either. 1993). is essential to understand food disorders. the fact that members of the peer group have a similar self-identity is a necessary. Economic 1 . Introduction It is becoming increasingly apparent that standards of physical appearance are important and powerful motivators of human behaviour. Secondly. (1996) and in Sacerdote (2000). However. The reasons for the increasing trend towards food disorders are yet to be fully understood. Trogdon et al. it becomes especially important to understand how food disorders are engendered. psychological and social equilibrium (Schilder. This is the first claim of the paper. and even devastating and life-threatening effects (APA. Goldfarb’s model does not attempt to include or explain the formation of self-image. Some critics have argued that the phenomenon of anorexia is a paradigmatic example where economics needs to examine preference formation given that anorexia and food disorders more generally are recognised as “an extreme response to the pressures (for example. but not sufficient. in explaining the formation of self-image. When applied to food disorders. as we argue. condition for the presence of social-multiplier effects (Costa-Font and Gil. suggesting that early health behaviour is significantly determined by social influences. and it is thought that food disorders are probably the result of some ‘socially transmitted’ standard of ‘ideal’ body image affecting food intake and exercise. 2004). This paper is the first attempt to model anorexia and test some of these claims empirically. the specific mechanisms behind peer-pressure on food disorders are unknown and require careful examination. 2005). About 6% of those who suffer from anorexia nervosa die from it (Birmingham et al. This concept arises not only when women have similar behaviour or representations (self-identity) due to sharing a common environment. 1995). Case studies of eating disorders constitute a natural example to investigate how changes in social attitudes towards physical appearance explain irregular health behaviour such as eating disorders among women. 1995). the content and formation of these ideal-body standards have yet to be explored in the economics literature. Mainstream literature in social psychology regards social image as being continually under construction and essential in determining physical. we claim that network phenomena appear to be relevant to the biological and behavioural trait of obesity (Christakis and Fowler. 1993). 2000). this could explain some extreme forms of weight aversion. but also when they belong to certain unobservable social groups (see Manski. especially regarding health and food. However. Only one other paper by Goldfarb et al. Eating disorders can have damaging. (2009) proposes a model explaining anorexic disorders (low calorie intakes. Anorexia together with other food disorders such as bulimia nervosa can be characterized by a distorted body image accompanied by an eating obsession. to thinness) that are experienced by most if not all women” (Fine. Ham (2009) proposes a model of addiction to explain bulimia nervosa. as applied in Glaeser et al. To measure the strength of such socially transmitted influences on individual behaviour it seems appropriate to use the concept of peer or social-multiplier effect. purging behaviours) that is based on taste variations and on an implicit rational choice to be underweight. (2008) using a sample of adolescents found that mean peerweight is correlated with individual weight. Given that the relatively young females are more at risk. which. In the social science literature the formation of social identity is seen as a key factor. 1958). 2007) as reflecting the tensions between the imperatives to eat and not to eat (Fine. identifying the motivation underpinning them. Although the correlation between network effects and obesity is contested in Cohen-Cole and Fletcher (2008). However. hence anorexia results from a tension between aesthetic and utilitarian reactions to food ingestion (Orbach. Similarly.1.

2003). We use a women-only sample from a representative European data-set (Eurobarometer 59. This paper supports the hypothesis that social pressure through peer-shape is determinant in explaining anorexia nervosa and distorted self-perception of one’s own body. Bodenhorn and Ruebeck (2003) created models for the influence of identity on ethnic preferences. it has been suggested that the consequent fear of rejection based on physical appearance is behind the increase in the number of persons suffering from eating disorders. seeing herself as fine or too fat. Altogether. at the same time. aesthetic ideals and success (Hill. Lakdawalla and Philipson (2002) referred to an ‘ideal weight’. Blanchflower et al 2008 used Eurobarometer data for 29 countries to show that overweight perceptions and dieting were influenced by the individual relative body mass index (BMI). there is no previous study examining anorexia that uses an economic decision-model perspective combining self-image – or self-identity – formation and individual health production. and the influence of self-image on individual weight. Extreme thinness is different from thinness in that it is influenced by some deviance form a “fine rule of healthiness”. finding that the unobserved factors explaining both processes are correlated. different cultural backgrounds are likely to exert idiosyncratic influences on the prevalence of food disorders. To the authors’ knowledge. Gardner (1996) discussed the role of bodyimage in behavioural reactions in cases where individuals perceived a large gap between their desired image and the one they actually had. there is not much in the literature on the role of social identity as a determinant of health. and these need to be controlled for. Further. one can hypothesize that eating disorders are ‘socially formed’ rather than a biological pathology (Bordo. In a joint-modelling exercise. the power exerted by media stereotypes of beauty and the social norms that individuals are immersed in – especially the association between thinness. Akerlof and Kranton (2000) wrote the seminal paper in this area and included an application to gender attitudes.0). Hutchinson (1982) points out that ‘body image’ refers not only to the description of the body but to the place ‘where body. Accordingly. the paper then estimates the determinants of the probability of a woman being extremely thin and. The Oswald and Powdthavee (2007) study on wellbeing an obesity finds that anorexics typically exhibit a convex utility function with respect to weight. However. A contribution of our study is to introduce identity or self-image to capture the tradeoffs women face between following social pressures to attain a certain body shape and their health. suggesting that this gap gave rise to permanently distorted self-perceptions of the body. The aim of this paper is to build an economic model of eating disorders which relates social and environmental factors to ‘self-image’ and objective weight. and the development of empirical strategies to measure these effects. Some of the implications of this model are taken to the data and the effect of underlying determinants is estimated. and Etile (2007) examined the role of social norms on obesity and concluded that social norms have an effect on ideal body-weight (for women). We restrict attention to women since according to the APA (2000) women account for 90% of all anorexia nervosa. Hence. This paper focuses particularly on the effect of ‘peer weight’ (which is likely to influence self-image or social identity) on the likelihood of anorexia. It then takes the two processes apart and estimates a recursive probit model of being extremely thin and perceiving oneself as being fine or too fat. The structure of the paper is as follows: Section 2 provides some background on the issue of 2 . 1993) – is widely accepted. mind and culture meet’.explanations of health prevention require a better understanding of the effects of social identity and self-image on health. Recent contributions to the economics literature enable baseline modelling.

There is some evidence indicating that eating disorders may run in families. misunderstood and alone. anxieties. dieting. Section 4 sets out the empirical strategy used. which regulate pleasure. Parents who overvalue physical appearance can unwittingly contribute to an eating disorder. there are high expectations of achievement and success. 2001). in an attempt to achieve the appearance of success. describes the data-set and estimates a reduced-form equation derived from the model. Furthermore. purging. these determinants seem to make individuals susceptible to having their food and exercise intake shaped by the strong socio-environmental pressures. which in this paper we define as influencing what an ideal body looks like. This may explain why they feel driven to lose weight but receive no pleasure from shedding pounds (Frank. At the same time. For this group of young women. and imperfections. anorexics reveal great fear of the criticism and rejection that would occur if their perceived flaws and shortcomings should become known (Bachar et al. families that include a person with an eating disorder tend to be rigid and ineffective at resolving conflicts.self-image and healthy eating among women. are at increased risk of becoming eating-disordered (Bordo. and other physical signs of womanhood. with the associated development of breasts. Section 3 proposes an economic model for eating disorders. even if they do not feel successful (Bordo. about their children's bodies. Puberty and anorexia. Girls who achieve sexual maturity ahead of their peers. parental values and socio-cultural influences. genes and predisposition. Some people with eating disorders report having felt smothered in overprotective families. Family and anorexia. In some such cases mothers are emotionally cool while fathers are physically and/or emotionally absent. Some studies suggest that there may be a genetic component in anorexia. These girls often wrongly interpret their new curves as signs of fatness and feel uncomfortable because they no longer look like their peers. as can parents who make critical comments. In the main. and generally evidence does not suggest a clear consistency with the inclusion of social effects. Different factors have been suggested as possible determinants of anorexia. 2. Other factors are more closely related to ‘nurture’ i. Instead they try to solve their problems by manipulating weight and food. gender. who still have childish bodies. Section 5 presents the estimation results and Section 6 contains a discussion and conclusions. Some of these are related to ‘nature’. Generally. Children in this type of family learn not to disclose doubts. There are suggestions that women who develop anorexia nervosa have excess activity in the brain's dopamine receptors. Others have felt abandoned. et al 2005). partly because she wants3 to take control and ‘fix’ her insecurity and importantly because they are under the influence of a culture that equates success and happiness with thinness. According to recent research (Fairburn et al 2005) genetic factors account for more than half (56%) of the risk of developing anorexia nervosa and work on the genetics of bulimia and binge-eating is under way. even in jest. exercising.e. A young woman in this group may ‘tackle’ her body. bingeing. 1993). but the result of a conscious decision process.e. fears. i. 2003). 3 . Parents influence their off-springs’ values and priorities. including those towards food. hips. and other strange forms of behaviour are not random. Genetic Factors. Background Standard health production models fail to explain irregular behaviours such as those of anorexics.

females of similar age.and by socio-cultural environmental factors. SI j . 2006). a  j . a-j is the appearance of the j’s group of reference. some studies have found magazine-reading to be a more consistent predictor than television-viewing (Harrison and Cantor. the utility function can be modelled as: U j  U j (a j . education and background are likely to have been exposed to similar media and social environments and. but also on their self-image (or self-identity) and health. to have similar ideal selfidentities. In westernised countries. successful men and women are interesting. we assume that all others’ actions.Network effects: the media. Studies of undergraduate women have associated reading fashion magazines with having higher preference for lower weight.. 2002). The ‘ideal’ body image portrayed by the media influences social interaction and this may in turn make it more dominant. This circularity only makes the power of social interactions in shaping people’s self-identity more extreme. SIj is j’s self-image.namely their net caloric intake in order to maximize an implicit utility function that depends not only on their net caloric intake. Many people believe media stereotyping helps explain why about 90% of people with eating disorders are women and only 10% are men (Thompson and Heinberg. In order to model anorexia. accordingly. strength and firmness are. feeling frustrated for this reason etc (Turner et al. zj are j’s characteristics. The differences between media images of happy. Besides these individual factors. beautiful and thin. the self-identity model of Akerlof and Kranton (2000) was found to be particularly useful and was adapted to the subject of interest. only 3% of female TV-network characters are obese. 1997). To simplify our model. Despite TV being a dominant media type. in business. while only 5% of the female audience is underweight. cj are embedded in the environmental factors. Similarly. men are young or old. while 25% of US women fall into that category. we assume that utility depends on the rather abridged concept of ‘net 4 . women often experience unrealistic cultural demands for thinness. 3. food seems to need to be modelled as an economic ‘good’ up to a certain caloric intake – which is idiosyncratic due to socially influenced self-perception – and as an economic ‘bad’ thereafter. having lower confidence on their own body image. z j . the utility function of individuals is conditioned by their peers’ net caloric intake and also their appearance and their characteristics . Zj. According to extant literature. Further. c j . in the United States (US) 32% of female TV-network characters are underweight. An Economic Decision Model for Eating Disorders Current empirical evidence makes modelling eating disorders complex as one of the assumptions of consumer-choice theory is the principle of non-satiation. To sum up. Thin is not desirable in men. but strong and powerful in all the areas that matter – physically. characterized by competitive striving for success. Thus. Z j ) (1) where aj is j’s net caloric intake. H j . cj reflects j’s other actions – not related to caloric intake. While women appear young. According to Health magazine (April 2002). power. We assume that individuals choose both food and exercise-related ‘actions’ . and Zj the environmental factors in which j is immersed. Hj is j’s health-production function. and socially.

aj. and any other individual and environmental factors. since nutrition is necessary for survival. Given the empirical evidence.as a person with higher status may have a better self-image than an identical one with lower status. In this special case the utility of net-caloric-intake would achieve a maximum at a much lower level than for a non-anorexic person (see Figure 1). one can expect a negative marginal impact of net caloric intake on self-image after a certain level of net-caloric-intake. z j . and by j’s status’. because the negative effect of eating on self-image is greater for them. it is assumed that a normal net-caloric-intake has a positive marginal impact on health. the ‘bliss point’ of food consumption for anorexics is lower. a health-production function Hj is added.caloric intake’ because food and exercise are a source of satisfaction beyond the body weight they achieve. and is conditioned by j’s individual characteristics and environmental factors. the equation for self-image is written as: SI j  I j (a j . which would make the sign of the second term in equation (4) negative. z j . sj . In contrast. zj and Zj. The healthproduction equation is written as follows: H  H (a j . Note that the sign of the self-image term is idiosyncratic insofar as it depends on the impact on each individual of the societal ideal5 . Thus. Thus. a person with anorexia will have an extraordinarily large negative term associated with the effect of net-caloric-intake on self-image. This depends on j’s net caloric intake. a j . Also. bearing in mind that both anorexic and non-anorexic women will eventually confront the economic principle of non-satiation. from health and also from an improved self-image. aj. a-j. Similarly to Akerlof and Kranton (2000). sj. The opposite would apply for extremely overweight individuals. s j . The net-caloric-intake chosen to optimise overall utility will vary depending on the relative magnitude of the positive and negative signs in equation (4) above. zj and Zj. s j . A person without any eating disorder and in a range of net-caloric-intake would be expected to receive a positive marginal utility from net-caloric-intake. j’s status’. Z j ) (2) Finally. the first two summands in equation (4) are expected to be positive. The difference lies in satiation among anorexic women taking place at lower levels of consumption. self-image SIj depends not only on j’s net caloric intake. In other words. Z j ) (3) Standard utility maximization subject to a budget constraint under the usual regularity assumptions would lead to an associated first-order condition as follows: U j a j U  U j SI j SI j a j SI  U j H j H j a j H   Pa  0 (4) where  is the usual income-multiplier and Pa the monetary price of net caloric intake or the combination of food price and exercise monetary cost including the opportunity cost of the time invested in it. but also on others’ body-weight-related actions or appearance.

The usual sample in standard Eurobarometer Surveys is 1. a j . namely individual-level variables and socioenvironmental variables. From equation (4).000 in Great Britain and 300 in Northern Ireland). Optimal equilibrium with and without anorexia Utility Net caloric intake aj âj Thus. In particular.000 people per country. Eurobarometer 59.g. since Eurobarometer 34. Z j . The former are taken directly from the answers to the Eurobarometer 59.0 questionnaire. 2. c j ) (5) The next section describes how equation (5) is taken to the data.body-shape that is in fashion. we can infer an implicit reduced form of net-caloric-intake that depends on individual status.0 is one of the Eurobarometer Surveys that have been conducted each spring and autumn since autumn 1973. under standard normality and linearity assumptions. adding countries as the European Union has expanded.1 Data We combine two types of variables. individual characteristics and the social environment (peer or network effects). the likelihood of being anorexic. Data and Methods 4.000 in East Germany and 6 . Hence. Figure 1. with the exception of Luxembourg (600) and the United Kingdom (1. Also. This minimum-necessary net-caloric-intake threshold can be thought of as the one that would keep individual j on a body mass index (BMI) considered ‘healthy’. where observed variables are included and others can be controlled for as country specific effects.000 people have been sampled in Germany (1. 4. study number 3903. whilst obesity refers to the negative slope. z j . which includes the appearance/net-caloric-intake of others. the probability that the net-caloric-intake of an individual j is below his/her minimal healthy level âj can be expressed as: ˆ P(a j  a j )   (s j . e. the problem in this paper refers to the positive slope instead. an anorexic individual chooses a net-caloric-intake aj that is under the optimal netcaloric-intake âj associated with his/her characteristics had that individual not been anorexic.

Furthermore. For that purpose. The questions from this Special Eurobarometer centred around attitudes towards life-long learning. we use a sample of younger women between 15 and 34. The prevalence of Anorexia reaches a peak of 2% in some countries when all women’s sample is examined. A second anorexia indicator variable labelled as ‘severe anorexia’ restricted the previous definition by accounting for the fact that the respondent declared herself to be eating adequately (self reported eating).5 and at the same time saw herself as being ‘fine’ or ‘too fat’. gender. in more depth. on women's health and medical tests relating specifically to women's health. to reflect the freedom and quality of the answers. but when we restrict the analysis to a sample of younger women the maximum prevalence rises to almost 5%. childcare and family planning. on long-term treatment. which took a value of 1 if the individual declared she saw herself as fine or too fat. a variable called ‘fine or too fat’ was created. an indicator variable called ‘anorexia’ was created. on dental health and. was carried out in all European Union countries between 15th January and 19th February 2003 on behalf of the European Opinion Research Group. the survey is carried out by national institutes associated with the European Opinion Research Group.000 in West Germany) in order to monitor the integration of the five new Länder into unified Germany and the European Union. health issues.0. given that anorexia is especially prevalent among younger women. which took a value of 1 if a woman had a BMI of less than 17.740 valid observations on women above 15 years of age. safety issues. which can be hypothesized to explain some of the unobserved heterogeneity in the model. In each of the 15 member states. and on general and children’s safety. which altogether make a sample of 2871 women. and of 0 otherwise. in some of the specifications the number of people present during the interview and the level of cooperation is included. This variable was used as a control in the regression. healthiness of eating habits. dietary habits and alcohol consumption. A variable was also created that measured health consciousness through the number of declared gynaecological check-ups received during the previous six months of the interview. Women are categorised as anorexic if they are extremely thin (BMI<17. height. professional category. 1993) this paper focuses on women’s behaviour and thus only evidence on women was selected. The percentage of extremely thin women of all ages ranges from 11% in France to 2% in Germany. Given that the mechanisms that give rise to anorexia and bulimia particularly affect women (Hill. being married. and residence in an urban or rural area.1. This set of variables includes: (self-reported) weight.5) but at the same time perceive themselves as being ‘just fine’ or ‘too fat’. 7 . own-body perception. partnership. Furthermore. on health maintenance (by discussing doctor's visits and various screening tests). A special issue. A larger number of check ups is interpreted to proxy women’s concerns with their own health. Eurobarometer 59. political attitudes. educational level. It focussed particularly on the incidence of chronic illness. household tasks. self image was classified to identify seeing oneself in the ‘right weight range or above it’. Finally. age. This gave a sample of 8. We scrutinise a set of individual variables ranging from socio-demographic characteristics to biometric measures and behavioural attitudes.

41% had finished compulsory secondary education. as defined here. Finally. Other variables such and individual’s income were initially included but were never significant. For this younger group (N=2. Both conditions followed a decreasing pattern till the age of 35. indicating the average number of gynaecological screenings received over the previous 6 months. had a prevalence of 3% for women aged between 15 and 24. Anorexia. just slightly higher than severe anorexia.5% were heads of household and 27% lived in a small town or rural area. 8 . The average value of the variable ‘being health conscious’ for the full sample was 1. Of these.06 0. Only 47% were married (in part due to increasing cohabitation). Roughly 26% had completed primary education.25.00 15-24 25-34 35-44 45-54 55-64 65+ Extreme Thinness Anorexia Severe Anorexia Figure 2 reports the prevalence of the three variables examined in different age groups. 28% were heads of household and 30% lived in a small town or rural area. 24% had studied up to 18 years of age. which account for the inclusion of different controls. The second panel in Table 1 shows the descriptive statistics for young women.871). Eight per cent had completed primary education.10 0. and 9. Extreme thinness in terms of very low BMI was highest during early youth (age group 15-24) and progressively decreased until 55-64 years of age.4% held a university degree. 37. who had a higher incidence of food disorders. after which they remained relatively constant at about 1%. This could be due to attrition alongside later life metabolic conditions. This group included the women between 15 and 34 years of age.12 0. We found that the prevalence of anorexia is just below 4% for younger age groups and just below 2% among women aged 2534. increasing slowly again in the late years of life.08 0.14 0. we report a set of robustness checks.04 0.02 0.740). The first panel in Table 1 provides some overall statistics for the data (N=8. 23% had studied until the age of 18. The average age of the women in the sample was 45. with an average age of 25 years. Therefore. 57% were married. Our analysis includes country fixed effects to explore the variability within countries. and were finally not included in the reported specifications below. the average number of gynaecological check-ups during the last half-year was 1. and 27% held a university degree (indicating some cohort differences with previous generations). 41% compulsory secondary education. women 34 or below are expected to present a different pattern from women over 34.Figure 2: Prevalence of extreme thinness and anorexia among different age groups 0.

9) 2. The last column reports the circulation of women's magazines per 1.268 0. In an attempt to reflect social norms and image patterns. At the bottom of the table.740 WOMEN 15-34 years Mean (Std.0 study number 3903 In order to reflect the woman’s peer effect (pressure felt in terms of acceptable body-shape in her social environment/setting) different variables were created to represent the appearance of others around her. same reference age group (in ten-year groups).3) 24.07 (17. who saw themselves as fine or too fat and believed they were eating adequately (defined as severe anorexic). five to ten years older. the percentage of women with a BMI below 17. a variable called ‘women’s magazines per capita’ was included (measured by subscriptions).871 Age Married Being head of household Living in rural area Having completed primary education Having completed secondary education Having received education up to 18 years Holding a university degree Being health conscious (using number of gynaecological check-ups) Peer effects: group BMI Observations Source: own elaboration using data from Eurobarometer 59.0 (1. Descriptive statistics Variable ALL WOMEN Mean (Std. Dev) 45.083 0. the percentage of women ‘seeing themselves as fine or too fat’. the percentage of women who believed they ate adequately. living in a similar environment (rural or urban) and in the same immediate region of residence – all estimated using a recursive system.469 0.40 (5.59) 8.5) 25.375 0.7 (1.5. the percentage of women who saw themselves as fine or too fat while having a BMI below 17. Table 2 provides a breakdown of average BMI by country and age-group.40 (5. This strategy can be found in other studies such as Grodner and Kniesner (2006) and Etilé (2007) where ideal BMI is used. arguably could pick up to some reverse causality form younger generations to older ones. All these effects would be expected to account for contemporaneous geographical influences and. we report the aggregate average BMI and the standard deviation for these groups of women as categorized above.85) and 9 .260 0.’ that were made available. referring to the number of magazines categorized as ‘for women.411 0.273 1.094 1. the percentage of women with a BMI below 17.234 0. The country with highest average BMI for all women was Greece (25. These measures were: the average BMI of women with the same education level.407 0.9) 0.276 0. Dev) 25.301 0.e.59) 0.239 0.Table 1.3 (1.5. an instrument for peer BMI was developed using the average BMI of women meeting the above criteria but from a different generation – i.568 0.000 persons. and finally.5 (defined here as anorexic). The average BMI by country is about 25 for all ages and 23 for those between 15 and 34 years of age. Since individual BMI and BMI of the group of reference may be affected by common unobserved factors.

Luxembourg and Netherland had the highest percentage with 49. though it was also high in West Germany and Luxembourg. Zj refers to  the socio-environmental factors that individual j faces .29 in Northern Ireland to 35. political affiliation and education. labelled as in equation (5). Significant differences between younger women and women as a whole were found in Austria. where a significant high prevalence of anorexia was found among younger women. That is. suggesting that anorexia is very much a recent phenomenon.59%) and the lowest were found in Greece (64. In the light of both the empirical evidence and the model specification in section 2. Empirical Strategy Given the dual health production and social dimension of the question examined. it allowed us to disentangle the effect of different variables on the two separate processes.41) and the lowest average BMI was that of Italy (21. variables such as magazine circulation are expected to correlate with self-image but not with health production. for young women the highest average BMI was that of Belgium (23.07%) and Austria (78. in a second step we used a bivariate recursive (seemingly unrelated) probit specification to separate the two processes involved the definition of anorexia. Then. Interestingly. equation (5) above.86%). we model the individual’s propensity to be anorexic using a latent variable. aggregate circulation of women’s magazines was particularly high in Northern Ireland and Austria. Also. Finally. we followed a two-step approach. Substantial differences were found between the sample of younger women and the sample of women of all ages. The bivariate recursive probit allows a better empirical representation of the theoretical model as it accounts for common unobserved heterogeneity in these two processes. while Spain had only 22%. unless channelled through self-image. professional status. The country with the highest prevalence of female anorexia (column 4) was Austria (1. which depends on individual and socio-environmental characteristics: A*  w j  Z j  a j   j (6) j where wj are individual-specific controls and determinants of j’s status such as gender. And.e. i.the lowest average BMI was that of Austria (23. Consistently. Italy and Ireland among both young women and the full sample.40). with hardly any exceptions it was found that self-reported perception of eating adequately was higher among older women that among younger ones. In the younger group of women. age.55%). the highest percentages of women who declared they ate adequately were found in Denmark (95. we assume they do so through self-image rather than other routes.67). For instance.60%) and the lowest in West Germany. ranging from 56.42%. the highest prevalence of anorexia was found in Austria. namely being extremely thinness (an outcome of the health production function in equation 3 above) and ‘seeing oneself as fine or too fat’ (self-image formation in equation 2 above). consistently with all the above. Aj*. The lowest prevalence was in Northern Ireland. the highest prevalence of anorexia was in Austria (4. Italy and Luxembourg. For the younger group of women.97% in Ireland. First we estimated the impact of network effects and several covariates on the likelihood of just being anorexic. followed by France and Ireland. Almost all countries contained a population that was generally worried about its body weight. Ireland. although printed health stories influence extreme thinness.including prevalence of women’s 10 .53%) and Finland (90.

21% 0.44% 41.39% 0.88% 90.19% 37.80 21.63% 16.94% 79.54% 1.45% 0.91 21.54% 1.037 0.042 0.98% 2.03 24.00% 0.48% 33.32% 0.41 22.03 25.04% 78.19 22.00% 0.60% 1.07% 26.15% 22.14% 0.42% 0.22% 49.5.77% 24.06 (4.027 1.60% 16.00% 1.04) 0.14% 0.41 25.31 24.40% 16.16 22.54% 0.95% 80.73% 35.42% 49.42% 82.23 22.54% 76.97% 41.94% 56.45% 0.74% 1.35% 1.74% 1.28 24.56% 1.00% 86.91% 27.34% 1.62% 75.052 Magazines All Belgium Denmark Germany East Germany West Greece Spain France United Kingdom Ireland Italy Luxembourg The Netherlands Northern Ireland Austria Portugal Finland Sweden Average BMI (standard dev) 519 494 508 487 467 500 484 492 454 167 309 480 486 592 557 499 517 24.037 0.67 25.29% 34.77% 82.40 22.54% 0.22% 54.66% 2.44% 3.17% 30.54% 0.39% 0.86 23.90 22.62% 3.68 (4.65% 0.33% 0.42 21.00% 0.82) All 48.30% 37.43% 0.32% 0.91% 90.23% 1.45 25.54% 0.71% 82.78% 51.02% 1.63% 76.56% 0.72) Young 78.37 21.10% 51.00% 0.85 24.63 25.66% 1.80% 1.90% 4.00% 0.61% 0.45% 87.63% 28.25% 16.18% 1.044 0.80% 1.043 0.48% 1.55% 0.08) Young 0.159 0.39% 0.053 0.56) All 0.29% 0.022 0.00% 0.66 24.20% 0.214 0.19% 3.72% 2.62% 0. Fine or Too Fat at and Eating Fine All 0.Table 2: Country-specific BMI* average and other measures of thinness and self-image Country obs BMI BMI < 17.79% 64.88 (4.62% 0.00% 1.76% 1.66 (0.00 25.57% 1.56% 45.94% 39.60% 16.21% 0.5 Fine or Too Fat BMI<17.06 23.84) Young 46.81% 0.90% 1.93% 38.14% 0.00% 1.98 24.37% 54.00% 0.60% 0.84% 1.39% 0.98% 0.12% 49.88) Young 23.72% 85.21% 1.34 25.96% 84.29% 0.53% 82.17% 2.14% 2.95% 22.08% 1.66 (1.28) All 1.69 (4.02% 0.01% 82.64 (0.67 24.21% 48.30% 42.19% 76.037 0.62 (1.12) *Body mass index Note: Eurobaometer and World Magazine Trends FIPP/ ZenithOptimedia World Magazine Trends§ 11 .069 0.10% 75.71% 85.59% 84.00% 0.60% 2.31% 44.32 1.84% 87.00% 0.77% 49.57% 37.35% 0.96 23.29% 38.75% 49.40% 40.19% 82.60% 2.48 (4.60 (0.66% 2.15% 75.79% 48.86% 83.57 22.00% 5.26 23.01) Young 0.51% 1.67 25.80% 3.21% 29.00% 0.03 (4.00% 4.06% 38.58% 76.12% 82.21% 0.95) All 85.00% 0.49% 90.50% 16.19 24.62% 0.025 0.01 23.5 and Fine or Too Fat Eating fine BMI<17.62 (1.21% 1.85) Young 1.00% 4.81% 40.63% 0.94 23.54% 5.57% 0.097 0.49% 2.08% 92.62% 2.07% 88.05 24.019 0.22% 2.33% 0.

labelled UWj*. having a selfimage of being in the right weight range or simply too fat. we define another measure of the health production function being extremely thin (BMI<17. takes value 1 when the individual is extremely thin and 0 otherwise. country’s access to the internet. In order to simplify the analysis.Z j .): SI *   w j   a j  e j j (8) Once again.  j represents j‘s unobserved idiosyncratic characteristics. environmental variables. being extremely thin. peer appearance. it is possible to estimate the likelihood of being anorexic in the form of the probit model: P(A j 1| w j . trust in the press. Zj and a-j affect the two different processes involved in anorexia according to the paper’s own definition: 1. which stands for peers’ appearance. a j )  (e j  w j  a j ) (8b) We test different specifications of our health production function outcome.a j )  ( j  w j  Z j  a j ) (7) where is the normal-distribution cumulative-probability function.5) may be thought of as a partial representation of the individual’s latent health-production function Hj. we created a dichotomous variable that takes value 1 if the person can be considered anorexic and 0 otherwise:  1 if A*  0 j A j  1( A* 0)   0 otherwise Assuming normality of the error term in equation (6). The variable Hj. etc. We assumed that seeing oneself on the right weight range or too fat is an outcome reflecting one’s latent (body) ‘self-identity’ as in equation (2). as usual. Under the usual linearity and normality assumptions. As before. so we create a dichotomous variable. 12 .. and also – and very importantly – on the individual’s own-body perception or self-image. We assume that the propensity to being extremely thin. that takes a value of 1 if the individual declares she sees herself to do so and 0 otherwise. and.) and on peers’ appearance (a-j. Based on our own definition of anorexia (see above to address the DSM-IV1 criteria). which is a coding system developed by the American Psychiatric Association to identify mental health disorders.  b) Joint estimation of own-body self image and health-production function This second empirical exercise investigates how wj . associated with this process. for simplicity we assume that Z-j contains c-j all others’ actions not related to a-j.magazines. and. we assume that SI*j depends linearly on individual characteristics and status (summarized in variable wj. depends linearly on individual characteristics and status. this dichotomous variable allows us to obtain the probit model below: P(SI j  1 | w j . and. SI*. 2. Being extremely thin may also depend on additional factors and it is not necessary caused by a 1 The DSM-IV stands for the Diagnostic and Statistical Manual of Mental Disorders. what is being observed is the dichotomous variable. SIj. in the survey we do not observe the ‘propensity’ to see oneself as fine or too fat.

education. It also displays the results of estimating a recursive probit model formed by equations (8) and (9). that the BMI of the group of reference in terms of age. in column 2. For the full sample of women. being the head of the household age (being the excluded group girls aged 18 to 24). These are displayed in column 1 or. if the woman also thought she was eating adequately (Extreme Anorexia). the estimated marginal effects (see Table 3) showed. while allowing the unobserved factors affecting self-image and extreme thinness to be correlated. UWj. Results and Discussion This section contains the results of estimating the probit model for being anorexic according to the definition outlined above in equation (7). a j )  ( j  SI * j  w j  Z j  x j ) (9b) The system formed by equations (8b) and (9b) above is estimated on a recursive probit model by assuming that the idiosyncratic terms  j and ej are jointly normally distributed. The identification of parameters in the recursive probit model defined by equations (8b) and (9b) is satisfied by the inclusion of variables xj in equation (9b) that do not appear in equation (8b) and the triangularity created by the fact that a propensity to being underweight is influenced by a distorted own-body self image but not the other way around. it was possible to investigate how individual and environmental factors influence these two processes. so we create a dichotomous variable. Some potential socio-environmental factors were also included: the circulation of women’s magazines per capita in the country of residence. and a proxy of health-consciousness based on the declared number of gynaecological screenings taken in the last 6 months. in the survey we do not observe the ‘propensity’ to be extremely thin.personal propensity to having a distorted self-image. There were several explanatory variables or potential determinants of such behaviour: marital status.5) . as expected. which made it possible to correlate not only error terms but also self-image and extreme body-weight (thinness) in line with the theoretical prediction displayed in Figure 1. gender and location was 13 . P(UW j  1 | SI j . w j . UW j  SI * j  w j  Z j  x j   j (9) * As before. Z j . which can pick up some of the unobserved cultural influenced on self-image and desired weight. 5. By estimating the recursive probit model above. which consisted of estimating a probit model to determine which observable factors could cause a woman to see herself as fine or too fat whilst being extremely thin (BMI<17. and peer BMI – based on the BMI of women in the same age-bracket living in the same region. Furthermore. A country fixed effect allows exploring the individual’s variability within each country. crosscountry genetic variations in BMI can be controlled for by the inclusion of country dummies and clustering standard errors were always clustered. xj. Table 3 displays the results of the preliminary strategy. that takes a value of 1 if the individual declares is extremely thin and 0 otherwise.which we call Anorexia. living in a rural setting. One probit model was estimated for the full sample and one for women between 15 and 34 years of age. x j .

the more likely it is to suffer anorexia. and this calls for a separate estimation strategy. the lower the probability of suffering from anorexia. i. given that what is being studied is a combination of self-image and thinness. Nevertheless. The non-significance of being the head of the household was sensitive to the inclusion of education. although urban women were expected to be subject to more social pressure with regards to their appearance than those living in rural settings. However. 1999).0026 for the sample of younger women. Surprisingly. The effects of these variables on the probability of being severely anorexic were qualitatively very similar but slightly less marked than the ones commented on above.very significant and negative -0. Results show that having secondary education or having been to university all decreased the likelihood of being anorexic or severely anorexic as defined above. 14 . Ceteris paribus. which remained unobservable due to lack of data. Taking into account that the peers’ BMI is 24.0026 times 2 is 0. Thus. the effect of peer BMI was even more marked.5%. These values reflect cohort effects as explained latter in the results for the recursive probit specification.0015 for the full sample. living in a rural setting. which may be picking up part of the ‘being head of the household’ variation effect. 0. For the younger women.0052. in the case of severe anorexia. 2000). Yet. we will come back to this point when dealing with robustness checks. The result of non-significance for the women’s magazine circulation per capita was quite puzzling as it was not consistent with some specific studies on the subject (Turner et al. in terms of decreasing the probability of being anorexic. Sacerdote. this might have had to do with other household-related variables such as quality and type of parenthood (Fairburn et al. However. Being in a rural environment was found not to be significant. being the head of the family. and in turn followed by women aged 25 to 34 and 35-44. this result should be interpreted with caution because only a crude measure of ‘peer effect’ was used. we can interpret the magnitude of this coefficient in the right context: A change in one standard deviation in the peers’ BMI reduces the probability of anorexia by 0. as opposed to not. However. This may be due to the crudeness of the country measure and the possibility that the categories are not comparable across countries. perhaps better quality data was required to measure the effect of environmental or media-related variables. The signs of the results for being married are influenced by cohort effects which alongside respondent’s age and education are the expected ones. the probability of anorexia is higher for the excluded group (women aged 15 to 24 years of age) followed by women aged 45-55. and the measure of women’s magazine circulation were not significant covariates explaining the probability of anorexia for either group.11 with a standard deviation of about 2. These results suggest the presence of cohort effects. suggesting that the younger the women. 1996. the higher the BMI of the peer group. Network of peer effects are in line with expectation and the literature on social-multiplier effects (Glaeser et al.. -0. Being married. was only significant for the full sample. it might well be that the effects cancel each other out. cohort effects are less marked. 1997).e.

00265 (0.00162 (0.006) Yes 2654 0.0023** (0.001) -0.00076** (0.000493 (0.001) Yes 8012 0.0217 -145 17 Younger women 15-34 -0.002) 0. living in a rural environment.0278 (0.0859 -275.0752* (0.001) -0.0539 (0.0182** (0.01) -0.004) -0.5 17 -0.0788 -242.0256** (0.0404 (0.01.001) -0.006) 0.00459* -(0.004) -0.0370 (0.00214) Younger women 15-34 -0.001) -0.Table 3. * p<0.006) 0. ** p<0.001) -0.0682*** (0.0042) -0.002) -0.005) Yes 2654 0.00457) -0.0327*** (0.0210* (0.001) -0.001) -0.0368** (0.002) 0.001) Yes 8012 0.00681 (0.0535 (0.005) Robust standard errors in brackets *** p<0. being the head of the household. Probit model of the likelihood of suffering from anorexia VARIABLES* Peer effects: group BMI Being married Being health conscious† Between 25 and 34 years old Between 35 and 44 years old More than 45 years of age Having completed secondary education Having a university degree Controlled by country of origin Number of observations Psudo R squared LogLikelihood Number of clusters (countries) Anorexia § All women -0.0134*** (0.5 BMI.01) -0. † Being health conscious is defined as the sum of all gynaecological screenings received in the last 6 months.1 17 -0. considers herself fine or too fat and also thinks she is eating appropriately.0221** (0.8 17 Severe Anorexia §§ All women -0. §§ Severe Anorexia is defined as when the subject is below 17.003) -0.0247** (0.0484 -180. § Anorexia is defined as when the subject is below 17.05. and having education up to 18 years of age.0487*** (0.0026* (0.1 Women’s Magazine circulation per capita.0381* (0.004) -0.002) -0.0015*** (0.0101*** (0.0014*** (0.002) -0.5 BMI and sees herself as a fine or too fat person. were included but were not significant. 15 .

The identification restrictions were that peer BMI and women’s magazine circulation were presumably related to body selfperception but not to own-weight. Having been to university had a significant negative effect on seeing oneself as fine or too fat for both samples (-0. Among younger women. and from -0. Joint estimation of fat self-image and low weight Table 5 shows the marginal effects of a bivariate recursive probit model (Greene. 1998) of being extremely thin and seeing oneself as fine or too fat separately. although the introduction of additional covariates progressively decreased the coefficient from -0.014 for the strictest definition. Overall. 16 . probably because of cohort effects and the limited age-variation in that group.016 to -0.Robustness checks for the probability of being anorexic To test the robustness of the specifications in Table 3. respectively). Living in a rural area had a positive effect on the likelihood of seeing oneself as fine or too fat (0. This result is particularly important as it provides evidence consistent with the idea that self-image and identity do exert an influence on health-production. Finally.345 and -0. and change less than twice a standard deviation.056) but this was not significant for the younger sample.424 and -0. neither being head of the household nor being health-conscious was statistically significant. Age had a non-linear effect on the probability of being extremely thin for the full sample but not on the younger sample. This coefficient was barely more stable for the younger sample. but allows the unexplained variation in both processes to be related. the coefficient representing the correlation of the error terms of both processes. while seeing oneself as fine or too fat (self-identity) was likely to influence the probability of being extremely thin.014 for the less restrictive definition of anorexia.02 to -0. The estimates in the recursive bivariate probit model. and then for those in the 15-34 year age range which allows to pick up the presence of cohort effects. Being married did not explain extreme thinness Seeing oneself as fine or too fat (self-image) had a very negative and significant effect on the probability of being extremely thin (-0. Extreme thinness decreases with age at a decreasing rate. the coefficient decreased from -0.12) and on seeing oneself as fine or too fat (0. the bivariate model was estimated for the full sample of women first.0377 for the younger sample).079 for the full sample) but as expected it revealed a negative effect -0. the peer (network) effects coefficients are fairly stable.010 for both definitions of anorexia. Once again.169). gave rise to some interesting findings. Atrho. where the two processes involved in the paper’s simplistic definition of anorexia were disaggregated. Surprisingly.451 for the full and younger sample. their impact on the peer marginaleffect coefficient was estimated by adding the controls incrementally (see Table 4 below). Results turn out to suggest that unobservable factors influencing the two processes are only significant for the total sample of all women. but not for the subsample of younger women only. corroborating the fact that there are some unobserved factors influencing both women’s body identity and extreme thinness that are positively correlated. as one would expect following the model outlined in Figure 1. Peer BMI had a positive effect on the probability of seeing oneself as fine or too fat (0. The results obtained suggested that the peer marginal-effect was robust to these changes. and increases again after the age of 63.20). is positive and highly significant for both samples. at least in the case of anorexia.014 to 0.214) but only in the full sample did it negatively affect the probability of being extremely thin (-0. Even when additional controls were introduced. being married had a positive effect on the probability of being extremely thin (0.

001) (0.0011 (0.0010 (0.001) Yes 0.0010 (0.001) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 17 .0015 (0.0011 (0.003) Yes Yes Yes 0.001) Yes Yes Yes 0.0015 (0.001) Yes Yes Yes Yes Yes 0.001) Yes Yes 0.001) Yes Yes 0.003) Yes Yes Yes 0.0014 0.0011 (0.0014 (0.0010 0.0014 (0.001) Yes 0.0011 (0.Table 4.0011 (0.0017 (0.001) Yes Yes Yes Yes Yes Yes 0.001) (0.002 (0.001) 0.001) Yes Yes Yes 0.001) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 0.001) Yes Yes 0.0018 (0.0017 (0.0016 (0. Robustness checks using alternative probit model specifications Sample Anorexia (All women) Peer Effects: group BMI Personal characteristics Age variables Education Health consciousness Controlled by country of origin Anorexia (Young Women) Peer effects: group BMI Personal Characteristics Health consciousness Age variables Education Controlled by country of origin Anorexia § Severe Anorexia § § -0.001) Yes 0.0014 (0.0015 (0.001) 0.001) Yes 0.001) Yes Yes 0.

0320 (0.289* (0. were included but were not significant.098) Controlled by country of origin Number of observations Chi-Square for rho=0 Reject null rho=0 Degrees of freedom Loglikelihood Number of clusters (countries) Yes 8740 15.34 17 Robust standard errors in brackets *** p<0.07) Seeing oneself as fine or too fat -0.50) 1. ** p<0.05.036) 0.180*** (0.01*** -(0.03) -0.338*** (0.20*** (0.280) Younger women 15-34 Extreme thinness ‡ -0.07) 0.45 17 Yes 2871 1.01.176*** (0.564 (19.12** (0.110) -0.17) -0.089) 0. Recursive-probit models for being extremely thin and seeing oneself as 'fine’ or ‘too fat' (marginal effects) All women VARIABLES* Peer effects: group BMI Being married Between 25 and 34 years old Between 35 and 44 years old More than 45 years of age Age Age squared Living in a rural area Having been to university Seeing oneself as fine or too fat Constant Atrho -0.1 * Women’s Magazine circulation per capita.451*** -(0.07) 0.162 -(0.01) 0.001) 0.13*** (0.583*** (0.049) 0.309*** (0.569) 7.146) -3.02* -(0.424) -1.424*** -(0.023) -0. being health conscious.036) -0.046 (0.134) 0.5 18 .045) 0.0089 (0.960*** -(0. ‡ Thinness is defined as having a BMI of below 17. † Being health conscious is defined as the sum of all gynaecological screenings had in the last 6 months.18 No 14 -1710.169* -(0.079 (0.Table 5.056** (0.0002 -(0.042) 0. being the head of the household. living in a rural environment.0377** (0.0381 -(0.051) 0.431) Extreme thinness ‡ Seeing oneself as fine or too fat 0.0001* (0.000) 0.214** -(0.345*** -(0.1 Yes 14 -3765.005 (0.012) 0. and having compulsory education and up to 18 years of age. * p<0.039) 0.031) -0.095) -0.0034 -(0.052) -0.

44 (0. the coefficient of fat self image was negative and robust.129) Yes Yes Yes No Yes -0. in line with the identity model of Akerlof and Kranton (2000). we found that weight-related food disorders happen mostly at younger ages and require attention before they extend to older age groups.210) Yes Yes No No Yes -0. Conclusions This paper presents a simple theoretical framework to explain the influence of self image and other people’s body shapes on female caloric intake.44 (0. Note that the findings showed that anorexia primarily affected women aged between 15 and 34. unobservable variables affecting women of different ages may bias upward the relationship between BMI and peer-average BMI (and thus bias downward the peer effect coefficient).130) Yes Yes No No Yes -0.200) Yes No No No Yes -0.130) Yes Yes Yes Yes Yes -0. It then uses an empirical strategy to identify the determinants of food disorders (anorexia) and self-image (seeing oneself as fine or too fat) following first a joint process and secondly a recursive one.44 (0. Table 6.210) Yes Yes Yes No Yes -0. These results have serious 19 . the lower the likelihood of being anorexic.42 (0. Additionally.44 (0. revealing that women that see themselves as ‘fine’ or ‘too fat’ are 44%-46% less likely to be extremely thin. Results show that show that as expected. We find that the larger the peers’ body-mass.210) Yes Yes Yes Yes Yes 6. For instance. additional factors might still be present.44 (0. Our results were consistent with the assumption that individuals trade off health against selfimage.Robustness checks for the joint estimation of fat self-image and low weight Table 6 shows that the impact of self-image on being extremely thin remained almost constant when additional controls were introduced for both the ‘all women’ and ‘younger women’ samples.014) Yes No No No Yes -0.44 (0. Also. The instruments employed performed well following traditional Hausman test methods and were theoretically relevant. in agreement with the epidemiological literature. and that it was primarily socially induced. self-image correlates with body weight even when unobservable factors explaining both processes are controlled for. Robustness checks using alternative specifications All women Seeing oneself as fine or too fat (standard error) Personal characteristics Age variables Education Health consciousness Controlled by country of origin Young Women 15-34 Seeing oneself as fine or too fat (standard error) Personal characteristics Age variables Education Health consciousness Controlled by country of origin Dependent variable: Extreme thinness -0. However.45 (0.

Anorexics. The paper’s findings were the best that could be done with the existing crosssectional data on Europe. Hence. Underlying this debate is the question of time-discount rates and the formation of preferences.are incorporated in people’s utility maximization. it is important to understand how individuals come to value what they do. The distorted self-perception of women with food disorders and the importance or the peer effects may prompt governments to take action to influence role models and compensate for social pressure on women driving the trade-off between ideal weight and health. indicating that socio-environmental factors play an important role. Our findings are in line with the Clark and Oswald (1998) model of comparison utility in that deviant behaviour . by definition. which has important consequences for health-policy given that preferences for health-related activities are likely both to be influenced by and to influence health outcomes. Consistently. eating unhealthy food or avoiding physical activity .with their costs in terms of health and wellbeing . 20 . but these are not available in Europe at the moment. In addition the paper contributes to behavioural economics by using a model for eating disorders that allows for net-caloric-intake being a ‘bad’ instead of a ‘good’ in the consumer utility function above a certain intake. The influence of a crude measure of peer effects is significant and robust throughout the samples. probably while it is still being formed. However. They provide some important results that can act as a basis for future literature. In the light of this study. so as to prevent severe damage to women’s health and in order to improve their well-being and that of their families and friends. given the nature of the data and the absence of natural experiments we can’t prove our results as being causal and should be taken with caution.policy implications.such as anorexia . this implies understanding how preferences for smoking. They call for urgent action on individual identity. want to lose weight. These results go a step beyond Christakis and Fowler (2007) by exploiting self image as an intermediary source of social pressure. In the health policy arena.may occur when an individual attempts to deviate from some social norm using her own-BMI as an instrument. Etile (2007) funds that social norms regarding body shape have significant effects on perceptions of ideal BMI only for those women who want to loose weight. government intervention to adjust individual biases in self-image would be justified to curb or at least prevent the spread of a potential epidemic of food disorders. This result should be corroborated using longitudinal data.

Etile. L. Connor Gorber S. (1998). H. ‘The Mortality Rate from Anorexia Nervosa’. and McDowell. International Journal of Eating Disorders. and Fletcher. Quarterly Journal of Economics 115:3 pp. M. Cohen-Cole. pp.. Elsevier. (2003). Y. N Engl J Med 2007. Doll. Gur. S. Applied Economics. (2008). O. A. 27(5). G.. ‘Social Interactions and the Contemporaneous Determinants of Individuals’ Weight’. J.. (2005). Bodenhorn.. A.. ‘Risk Factors for Anorexia Nervosa’. Sep. Christakis. N. Berry. Western Culture. S. ‘Rejection of Life in Anorexic and Bulimic Patients’. Blanchflower. C. A. Ideal Body Weight and Food Attitudes’... Tremblay. APA (2000). Health Information and Research Division... J. Journal of Public Economics 70 (1998). Cooper. and Gao. and Bonne. Latzer.. 2253-2263. Environmental Factors in the Obesity Epidemic’. Health Economics (in press). C. E. (2000). Bordo. E. 9962. ‘Comparison-Concave Utility and Following Behaviour in Social and Economic Settings’. Journal of Health Economics. D. Statistics Canada. (2001). Bachar. and Ruebeck. Vol. 38: 143-146. (2008). E. Fairburn. J. J. and Oswald. S. M. ‘The Feasibility of Establishing Correction Factors to Adjust Self-Reported Estimates of Obesity’. and Van Landeghem. A.References Akerlof. Unbearable Weight: Feminism. J. H. M. and Welch. L. E. S. (2007). (2003). DSM-IV-TR. M. M. ‘Economics and Identity’. (2007). 21 . 2nd revised edition. (1999). ‘Social Norms. NBER Working Paper No. L.. Clark. Ottawa. 56:468-476. and Gil Transfi. E. B. 36 (20). 357:1866-1868.19(3):71-82. Goldner. J. J. Arch Gen Psychiatry. (2004). 715-753.. H. ‘Imitative Obesity and Relative Utility’. F. E. I. ‘Is Obesity Contagious? Social Networks Vs.. (2008). R. Hlynsky. ‘The Economics of Identity and the Endogeneity of Race’. 31(1): 43–48. Diagnostic on Statistical Manual of Mental Disorders. A. E. Su. University of California Press. American Psychiatric Association: Arlington. 14337. 133–155. VA. Costa-Font. G. NBER Working Paper No. and the Body. Canetti. Health Rep. Shields. International Journal of Eating Disorders. Z. C. A. and Kranton. Oswald. 1382-1387. A. M. and Fowler. American Psychiatric Association. Birmingham. ‘The Spread of Obesity in a Large Social Network over 32 Years’.

. Biol. ‘Why do Most Italian Youths Live with Their Parents? Intergenerational Transfers and Household Structure’. H. (1982). Harrison K. C. C. Pages 40-67. Hill. Hoge. C. (1995). and Scheinkman. (1996). W. C. (1993). S. D. (1993). Psychiatry. Volume 47 Issue 1. Biol Psychiatry. and Suranovic. Fall 1998. Lakdawalla. 87. Barbarich-Marsteller. 58 11. ‘Causes and Consequences of Dieting and Anorexia’. Hunger Strike: The Anorectic's Struggle As A Metaphor For Our Age. Harmondsworth: Penguin. L. ‘The growth of obesity and technological change: a theoretical and empirical examination’. 4(4). Review of Economic Studies. C.. University of Zurich. Appetite. ‘Caught in the Bulimic Trap? On the Incidence. 59-67. Orbach... W. N. S. Frank. and Iorio. ‘Increased Dopamine D2/D3 Receptor Binding After Recovery from Anorexia Nervosa Measured by Positron Emission Tomography and [11C]raclopride’. W. 91: 507-548. B. and Kessler. A. Goldfarb. H.. 60. Journal of Communication. T. M. 291-301. J. Hudson. E. G. J. (2009).. (2006).. (1993). Proceedings of the Nutrition Society 52. Mathis. (2002). J. 908–912 Gardner. Glaeser. T. (2009). G. R. and Moretti. Goeree. Mimeo. Women and Therapy. ‘Transforming Body-Image: Your Body.. British Journal of Psychology. and Cantor. Markowitz.. ‘Can A Rational Choice Framework Make Sense of Anorexia Nervosa?’. Hiripi. NBER Working Paper No. 24(3). (2005). Friend or Foe?’. D. Henry. E. J. and Philipson. Manski. 1. M. U. p. 8946. Bailer.. J. Hutchinson. 327-337. S. ‘The Prevalence and Correlates of Eating Disorders in the National Co morbidity Survey Replication’. I. 61(3): 348-358. Wagner. 531-542. 211-218. (1998). Sacerdote. Ham. G. M. C. Weissfeld. (2006). 231-42. ‘Gender Economics Courses in Liberal Arts Colleges: Further Results’. (1996). S. B. Greene. 22 . Journal of the European Economic Association. Price. R. Ziolko. A. A. F.. Journal of Economic Education..14838. (2007). J. Leonard. and Kaye. Manacorda. J. M.. C. NBER Working Paper No... Drevets. S. ‘Identification of Endogenous Social Effects: The Reflection Problem’. 800-829.. B. Quarterly Journal of Economics. S. C. ‘The Relationship Between Media Consumption and Eating Disorders’. ‘Methodological Issues in Assessment of the Perceptual Component of Body Image Disturbance’. R.. E. E. ‘Crime and Social Interactions’. Meltzer. Persistence and State Dependence of Bulimia Among Blacks and Whites’. Pope.Fine. ‘Towards a Political Economy of Anorexia’.

117(521).Oswald.. (2007). Slade. P. NBER Working Paper No. ‘The Influence of Fashion Magazines on the Body Image Satisfaction of College Women: An Exploratory Analysis’. (1988). Trogdon. ‘Peer Effects in Adolescent Overweight’. British Journal of Psychiatry. 5. L. and Pais.. Volume 55 Issue 2. World Magazine Trends FIPP / ZenithOptimedia World Magazine Trends. M. 23 . J. The Image and Appearance of the Human Body. N. D. G. Journal of Health Economics 27. Angood. (2002). A. Schilder. 32. J. J. L. Adolescence. and Powdthavee. Vol. ‘Obesity. n127: p 603-615. and The Challenge of Affluence: Theory and Evidence’. ‘The Media's Influence on Body Image Disturbance and Eating Disorders: We've Reviled Them. 20-22. (2008). Vol. ‘Body Image in Anorexia Nervosa’.. and Hovde Dwyer. M. H. Pages 339 – 353. Sacerdote.. Economic Journal.7469. P. B. (1997). 441-454. Jacobs. P. Published Online: 17 Dec 2002. (2000). and Heinberg. ‘Peer Effects with Random Assignment: Results for Dartmouth Roommates’. Thompson J. J. Journal of Social Issues. Hamilton. 153 (2). Now Can We Rehabilitate Them?’. 1388-1399. L. K. New York: International Universities Press. Unhappiness. Turner S. Nonnemaker. (1958).

CENTRE FOR ECONOMIC PERFORMANCE Recent Discussion Papers 1097 1096 Michal White Alex Bryson Dominique Goux Eric Maurin Barbara Petrongolo Petri Böckerman Alex Bryson Pekka Ilmakunnas Olivier Marie Judit Vall Castello Claudia Olivetti Barbara Petrongolo Guy Mayraz Francesco Caselli Andrea Tesei Keyu Jin Nan Li Yu-Hsiang Lei Guy Michaels Brian Bell John Van Reenen Amparo Castelló-Climent Ana Hidalgo-Cabrillana Amparo Castelló-Climent Abhiroop Mukhopadhyay Holger Breinlich HRM and Workplace Motivation: Incremental and Threshold Effects Worktime Regulations and Spousal Labor Supply Does High Involvement Management Improve Worker Wellbeing? Measuring the (Income) Effect of Disability Insurance Generosity on Labour Market Participation Gender Gaps Across Countries and Skills: Supply. Schott Elisa Faraglia Albert Marcet Andrew Scott 1083 In Search of a Theory of Debt Management . Redding Peter K. Bernard J. and Political Stability Factor Proportions and International Business Cycles Do Giant Oilfield Discoveries Fuel Internal Armed Conflicts? Firm Performance and Wages: Evidence from Across the Corporate Hierarchy The Role of Educational Quality and Quantity in the Process of Economic Development Mass Education or a Minority Well Educated Elite in the Process of Development: the Case of India Heterogeneous Firm-Level Responses to Trade Liberalisation: A Test Using Stock Price Reactions The Empirics of Firm Heterogeneity and International Trade 1095 1094 1093 1092 1091 1090 1089 1088 1087 1086 1085 1084 Andrew B. Demand and the Industry Structure Wishful Thinking Resource Windfalls. Political Regimes. Bradford Jensen Stephen J.

Becker Yona Rubinstein Camille Landais Pascal Michaillat Emmanuel Saez Klaus Adam Albert Marcet Juan Pablo Nicolini Zsófia L.ac.lse.ac. Serrano Daniel Hale John Coleman Richard Layard David Marsden A Many-Country Model of Industrialization Productivity.1082 1081 1080 1079 1078 Holger Breinlich Alejandro Cuñat Francesca Cornaglia Naomi E.uk .lse. Wages and Marriage: The Case of Major League Baseball The Wealth and Poverty of Nations: True PPPs for 141 Countries Fear and the Response to Terrorism: An Economic Analysis Optimal Unemployment Insurance over the Business Cycle Stock Market Volatility and Learning 1077 1076 1075 The Minimum Wage and Inequality .The Effects of Education and Technology Rising Wage Inequality and Postgraduate Education First Impressions Matter: Signalling as a Source of Policy Dynamics Advertising Expenditure and Consumer Prices Trading and Enforcing Patent Rights 1074 1073 1072 1071 A Model for the Delivery of Evidence-Based PSHE (Personal Wellbeing) in Secondary Schools Individual Voice in Employment Relationships: A Comparison Under Different Forms of Workplace Representation Racial Discrimination and Competition 1070 1069 Ross Levine Alexey Levkov Yona Rubinstein Klaus Adam Albert Marcet 1068 Internal Rationality.uk Web site http://cep. Bárány Joanne Lindley Stephen Machin Stephen Hansen Michael McMahon Ferdinand Rauch Alberto Galasso Mark Schankerman Carlos J. Imperfect Market Knowledge and Asset Prices The Centre for Economic Performance Publications Unit Tel 020 7955 7673 Fax 020 7955 7595 Email info@cep. Feldman Nicholas Oulton Gary S.

and Political Stability Factor Proportions and International Business Cycles Do Giant Oilfield Discoveries Fuel Internal Armed Conflicts? Firm Performance and Wages: Evidence from Across the Corporate Hierarchy The Role of Educational Quality and Quantity in the Process of Economic Development Mass Education or a Minority Well Educated Elite in the Process of Development: the Case of India Heterogeneous Firm-Level Responses to Trade Liberalisation: A Test Using Stock Price Reactions The Empirics of Firm Heterogeneity and International Trade 1095 1094 1093 1092 1091 1090 1089 1088 1087 1086 1085 1084 Andrew B. Demand and the Industry Structure Wishful Thinking Resource Windfalls. Bradford Jensen Stephen J. Body Image and Peer Effects: Evidence from a Sample of European Women HRM and Workplace Motivation: Incremental and Threshold Effects Worktime Regulations and Spousal Labor Supply Does High Involvement Management Improve Worker Wellbeing? Measuring the (Income) Effect of Disability Insurance Generosity on Labour Market Participation Gender Gaps Across Countries and Skills: Supply. Bernard J. Schott .CENTRE FOR ECONOMIC PERFORMANCE Recent Discussion Papers 1098 1097 1096 Joan Costa-Font Mireia Jofre-Bonet Michal White Alex Bryson Dominique Goux Eric Maurin Barbara Petrongolo Petri Böckerman Alex Bryson Pekka Ilmakunnas Olivier Marie Judit Vall Castello Claudia Olivetti Barbara Petrongolo Guy Mayraz Francesco Caselli Andrea Tesei Keyu Jin Nan Li Yu-Hsiang Lei Guy Michaels Brian Bell John Van Reenen Amparo Castelló-Climent Ana Hidalgo-Cabrillana Amparo Castelló-Climent Abhiroop Mukhopadhyay Holger Breinlich Anorexia. Redding Peter K. Political Regimes.

Wages and Marriage: The Case of Major League Baseball The Wealth and Poverty of Nations: True PPPs for 141 Countries Fear and the Response to Terrorism: An Economic Analysis Optimal Unemployment Insurance over the Business Cycle Stock Market Volatility and Learning 1077 1076 1075 The Minimum Wage and Inequality . Serrano Daniel Hale John Coleman Richard Layard David Marsden In Search of a Theory of Debt Management 1082 1081 1080 1079 1078 A Many-Country Model of Industrialization Productivity.ac. Feldman Nicholas Oulton Gary S.lse.uk .1083 Elisa Faraglia Albert Marcet Andrew Scott Holger Breinlich Alejandro Cuñat Francesca Cornaglia Naomi E.The Effects of Education and Technology Rising Wage Inequality and Postgraduate Education First Impressions Matter: Signalling as a Source of Policy Dynamics Advertising Expenditure and Consumer Prices Trading and Enforcing Patent Rights 1074 1073 1072 1071 A Model for the Delivery of Evidence-Based PSHE (Personal Wellbeing) in Secondary Schools Individual Voice in Employment Relationships: A Comparison Under Different Forms of Workplace Representation Racial Discrimination and Competition 1070 1069 Ross Levine Alexey Levkov Yona Rubinstein The Centre for Economic Performance Publications Unit Tel 020 7955 7673 Fax 020 7955 7595 Email info@cep. Becker Yona Rubinstein Camille Landais Pascal Michaillat Emmanuel Saez Klaus Adam Albert Marcet Juan Pablo Nicolini Zsófia L.ac.uk Web site http://cep. Bárány Joanne Lindley Stephen Machin Stephen Hansen Michael McMahon Ferdinand Rauch Alberto Galasso Mark Schankerman Carlos J.lse.

Sign up to vote on this title
UsefulNot useful