This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Social rank and adult male nutritional status: Evidence of the social
gradient in health from a foraging-farming societyq
Victoria Reyes-Garcı ´aa,b,*, Thomas W. McDadec, Jose Luis Molinad, William R. Leonardc,
Susan N. Tannere, Tomas Huancaf, Ricardo Godoyg
aInstitucio ´ Catalana de Recerca i Estudis Avançats (ICREA)
bUniversitat Autonoma de Barcelona, Institut de Ciencia i Tecnologia Ambientals, Facultat de Ciencies, Bellaterra 08193, Spain
cDepartment of Anthropology, Northwestern University, Evanston, Ill 60208, USA
dDepartment of Social and Cultural Anthropology, Universitat Auto `noma de Barcelona, 08193 Bellaterra, Barcelona, Spain
eDepartment of Anthropology, University of Georgia, Athens, GA 30602, USA
fCBIDSI-Centro Boliviano de Investigacio ´n y de Desarrollo Socio Integral, Correo Central, San Borja, Beni, Bolivia
gHeller School for Social Policy and Management, Brandeis University, Waltham, MA 02454-9110, USA
a r t i c l e i n f o
Available online 20 October 2008
a b s t r a c t
Research with humans and non-human primate species has found an association between
social rank and individual health. Among humans, a robust literature in industrial societies
has shown that each step down the rank hierarchy is associated with increased morbidity
and mortality. Here, we present supportive evidence for the social gradient in health
drawing on data from 289 men (18þ years of age) from a society of foragers-farmers in the
Bolivian Amazon (Tsimane’). We use a measure of social rank that captures the locally
perceived position of a man in the hierarchy of important people in a village. In multi-
variate regression analysis we found a positive and statistically significant association
between social rank and three standard indicators of nutritional status: body mass index
(BMI), mid-arm circumference, and the sum of four skinfolds. Results persisted after
controlling for material and psychosocial pathways that have been shown to mediate the
association between individual socioeconomic status and health in industrial societies.
Future research should explore locally-relevant psychosocial factors that may mediate the
association between social status and health in non-industrial societies.
? 2008 Elsevier Ltd. All rights reserved.
Research with humans and non-human primate species
has found an association between social rank, or position in
dominance hierarchies, and individual health (Marmot,
2006; Sapolsky, 2004). When examining the association
between dominance hierarchy and individual health
among humans, researchers have argued that socioeco-
nomic status proxies for social rank (Sapolsky, 2004). A
large and growing literature in industrial societies suggests
that each step down the socioeconomic ladder is associated
with increased morbidity and mortality (Adler, Boyce,
Chesney, Folkman, & Syme, 1993; Marmot, 2004; Marmot,
Ryff, Bumpass, Shipley, & Marks, 1997; Wilkinson, 2000).
qResearch was funded by grants from the Cultural Anthropology and
(BCS-0134225, BCS-0200767, BCS-0322380). We thank M. Aguilar, J. Cari,
S. Cari, E. Conde, D. Pache, J. Pache, P. Pache, M. Roca, and E. Tayo for help
in collecting data and logistical support. Thanks also go to the Tsimane’
and the Gran Consejo Tsimane’ for their continuous support, to ICRI-
SAT-Patancheru for providing office facilities to Reyes-Garcı ´a, and to
Dominica Lizarazu and Jeff Schmit for fruitful discussions on the topic.
* Corresponding author. Universitat Autonoma de Barcelona, Institut de
Ciencia i Tecnologia Ambientals, Facultat de Ciencies, Bellaterra 08193,
Q1 Spain. Tel.: þ34 93 581 4218; fax: þ34 93 581 3331.
E-mail address: email@example.com (V. Reyes-Garcı ´a).
Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
0277-9536/$ – see front matter ? 2008 Elsevier Ltd. All rights reserved.
Social Science & Medicine 67 (2008) 2107–2115
Author's personal copy
Researchers have found evidence for social gradients in
health in many industrialized societies that are indepen-
dent of a wide range of confounders and risk factors (Wil-
Research on the social gradient in health suggests that
socioeconomic status is a strong predictor of health. Both
material and psychosocial causes contribute to health
differences between people. Differences in individual
resources (e.g., income, access to health care) explain about
one third of the variation in health outcomes in industrial
nations (Wilkinson, 2000). Social epidemiologists have
proposed that psychosocial factors related to stress also
mediate the relation between socioeconomic status and
health. For example, research in industrial nations suggests
that job attributes might contribute to stress and poor
health because different jobs have different physical and
psychological demands, risks, and rewards (Bosma, 1998;
Sorenson et al., 1985). In addition, the strongest socioeco-
nomic status gradients occur for diseases with the greatest
sensitivity to stress (Wilkinson, 2000).
Researchers have also shown that the decline of social
capital associated with the rise of income inequality might
contribute to the relation between socioeconomic status
and health (Kawachi, 2002; Kawachi & Kennedy, 1999).
Social capital refers to the trust, safety nets, membership in
local organizations, and other expressions of pro-social
behavior that enable people to act collectively (Coleman,
1990; Kawachi, 2002; Ostrom, 2000). Men with higher
socioeconomic status have a wider and thicker social
network than men with lower socioeconomic status; the
network allows them to gain access to resources and
information that might protect health (Mare, 1990; Moore
& Hayward, 1990). For example, Marmot and Smith (1991)
found that men of lower socioeconomic status were less
likely to have a confidant in whom they could rely in times
of need. Research suggests that social capital may also play
an important role in small-scale societies, wheregift giving,
communal work, and labor contributed to other house-
holds represent pro-social expressions of generosity
(Godoyet al., 2005a). Anthropological evidence fromsmall-
scale societies suggests that people with higher socioeco-
nomic status might be more likely to receive help and gifts
than people with lower socioeconomic status (Henrich &
Prior research on the association between social rank
and individual health comes from industrial societies and
has used economic (i.e., income, occupation) rather than
social measures to assess social status (Deaton, 2003;
Kawachi, 2002; Marmot, 2004, 2005; Subramanian &
Kawachi, 2004; Wilkinson, 1996). Due to the paucity of
data, little is known about howsocioeconomic status might
affect individual health in small-scale, rural societies.
The use of data from a small-scale society to study the
social gradient in health allows one to circumvent relations
that become harder to identify once societies grow in
complexity. For instance, in small-scale societies people
enjoy high levels of job autonomy and social capital
(Johnson, 2003; Kaplan & Hill, 1985). Such societies have
relatively little division of labor beyond the sexual division
of labor. As a result, one can largely ignore the confounding
role of occupations and stressful events associated with
occupations and focus on how social status itself might
Here, we contribute to the literature on the social
gradient in health by exploring whether results from
industrial societies hold up in a very different socioeco-
a foraging-farming society of indigenous Amazonians in
Bolivia, we estimate the association between socioeco-
nomic status and individual-level anthropometric indices
of short-run nutritional status among adult men. We use a
locally-relevant measure of social rank that stresses indi-
vidual position in the social hierarchy. We examine
the association between social rank and anthropometric
indices of nutritional status controlling for three paths –
individual resources, job autonomy,and social capital– that
prior research suggests may mediate the relation between
socioeconomic status and individual health.
We hypothesizea positive
anthropometric indices of short-run nutritional status and
individual social rank. Research in industrial populations
has found that higher levels of overweight and obesity are
generally associated with
whereas in developing nations obesity and socioeconomic
status are positively associated (McLaren, 2007; Sobal &
Stunkard, 1989). Previous research among the Tsimane’
suggests that the nutritional status of adults is low relative
to reference values from industrial nations (Godoy, Reyes-
Garcı ´a, Byron, Leonard, & Vadez, 2005b), and in this context
of marginal nutrition higher levels of body mass index
(BMI) indicate better health. We therefore expect to find
a positive association between social rank and indices of
short-run nutritional status: If higher social rank is asso-
ciated with higher BMI, but without reaching levels asso-
ciated with overweight and obesity, one could argue that
social rank is associated with higher levels of caloric and
nutrient reserves, an important safety net that these indi-
viduals can draw on during times of nutritional adversity.
low socioeconomic status,
Social hierarchies, job autonomy, and social capital
One of the largest native Amazonian groups in Bolivia,
the Tsimane’number w8000people(CensoIndı ´gena,2001),
live in w100 villages mostly in the department of Beni, and
have been in continuous contact with Westerners since the
1950s (Daillant, 2003; Huanca, 2008). Tsimane’ in our
sample reside in villages of w20 households settled along
riverbanks and logging roads. Villages consist of a loose
number of households related by blood and marriage. Tsi-
mane’ rely on slash-and-burn farming supplemented by
hunting, gathering, and wage labor in logging camps, cattle
ranches, and in the homestead of colonist farmers. Their
chief sources of cash come from the saleof thatch palm from
the forest and cultivated rice from their farms.
Tsimane’ nutritional status is comparable to other
lowland South American indigenous peoples (Dufour,1994;
Orr, Dufour, & Patton, 2001). Prior research suggests that
among the Tsimane’ neither village income inequality nor
individual income correlate systematically with nutritional
status, although individual variables (e.g., schooling)
correlate with improved nutritional status (Godoy et al.,
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152108
Author's personal copy
2005a, 2005c; Reyes-Garcı ´a et al., 2008a, 2008b). Village
social capital and village income complement each other
and are associated with higher BMI: the rich who are stingy
have lower BMI than the rich who display generosity
(Brabec, Godoy, Reyes-Garcı ´a, & Leonard, 2007).
Living forcenturies at the doorsteps of a highly stratified
empire, the Inkas, the Tsimane’ nonetheless managed to
remain highly autarkic and egalitarian until recently (Ellis,
1996). The only evidence of past formal, structured social
hierarchy is the presence of shamans – healer, priest, and
political leaders – called cocojsi’ in the Tsimane’ language
(Huanca, 2008). Tsimane’ respected the cocojsi’ for their
power to cure and communicate with humans, plants, and
animal spirits (Daillant, 2003; Huanca, 2008). However, the
powerof the cocojsi’ declined with the arrival of traders and
missionaries in the 1950s. The last cocojsi’ died during the
At present, there are two sources of social status among
the Tsimane’. First, informal social evaluations create social
hierarchies. For example, Tsimane’ show deference to
people with specific skills and qualities. Tsimane’ respect
old, knowledgeable people (baba) whom they ask for
advice on multiple issues. Tsimane’ also show deference to
some young people (e.g., Protestant preachers) or to good
hunters. Second, new leaders have emerged as Tsimane’
gain greater exposure to the outside world. During the
1980s, the Tsimane’ started to organize politically. To elect
members for the umbrella Tsimane’ government, the Great
Tsimane’ Council, each village elects a representative (cor-
regidor) who mediates the relations between the village
and the outside world. Tsimane’ usually elect as village
representatives schooled men who can speak Spanish,
Bolivia’s national language.
Like other native Amazonian societies (Johnson, 2003;),
Tsimane’ individuals and households have traditionally
enjoyed high levels of autonomy in production and
household activities. Until the late1940s, the Tsimane’ lived
like any other pre-contact Amazonian society. They hunted,
fished, gathered wild plants, and practiced slash-and-burn
agriculture with stone and wood tools. To this day, they
remain highly autonomous, freely allocating their time to
various activities throughout the day.
With the arrival of highland colonists to the area in the
early 1970s, some Tsimane’ gave up some of their job
autonomy in exchange for jobs that provided a monetary
wage. At present, some Tsimane’ seek employment in cattle
as professionals for various organizations active in the
Tsimane’ territory. Work for loggers, cattle ranchers,
and colonists partially restricts the autonomy of Tsimane’,
but the jobs are temporary, both because the jobs them-
selves come and go in response to market prices for logs,
and because Tsimane’ often take up the jobs temporarily,
and do so chiefly during lulls in the agricultural season.
As is true of other native Amazonian societies (Kaplan &
Hill, 1985, Winterhalder, 1997), social capital is strong
among Tsimane’ (Brabec et al., 2007; Reyes-Garcı ´a, Godoy,
Vadez, Huanca, & Leonard, 2006). For example, Tsimane’
routinely share drinks and food with other Tsimane’ and
outsiders. In a previous study it was found that only 7% of
households reported not making any gifts to other house-
holds, and only 39% of households did not do any
communal work or offer any labor help during the week
before the day of the interview (Godoy et al., 2005a).
However, social capital does not seem to get activated to
help people cope with large-scale, covariant shocks (Godoy
et al., 2007).
In sum, the Tsimane’ represent an ideal case to study the
relation between social status and health for three reasons.
First, social hierarchies are not new in Tsimane’ society, but
they are growing in importance as Tsimane’ increase their
participation in the national society. Second, Tsimane’
enjoy high levels of autonomy inproduction and household
activities and widespread sharing and reciprocity, so one
can test whether the social gradient holds up after
controlling for those explanatory paths. Third, Tsimane’
share characteristics in common with other small-scale
indigenous groups, which suggests that results from this
research might apply to other non-industrial, indigenous
populations around the world.
Our main aim is to estimate the association between
social rank and anthropometric indices of short-run
nutritional status among adult men while controlling for
three paths that may explain the social gradient in health:
individual resources, job autonomy, and social capital. For
the empirical analysis, we assess the association between
(a) three outcome variables: adult body mass index (weight
in kg/height in m2), mid-arm circumference, and sum of
four skinfolds (biceps, triceps, subscapular, and supra-iliac)
and (b) individual social rank while controlling for the
three paths just noted. In subsequent analysis, we test the
robustness of our findings by including (i) self-reported
poor health, a standard covariate of BMI, (ii) smiles, an
indicator of level of stress, and (iii) alcohol consumption,
a risk factor for health. We use the following expression to
model the association between anthropometric indices of
nutritional status (Y) and covariates:
Assume, first, that Y captures the BMI of an adult –
transformed to logarithms – where i is the subject, h the
household, and v the village. We use BMI for ease of
exposition, but the expression also applies to the other
indicators of nutritional status. log Sihvrefers to the loga-
rithm of the social rank of the subject, defined as the
position of an individual within the social hierarchy in the
village. Pihvis a vector of variables for the three paths that
researchers have proposed to explain the social gradient in
health in industrial nations (i.e., income, occupation, and
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152109
Author's personal copy
social capital). Iihvis a vector of variables for the subject
(e.g., age and schooling) that directly affects nutritional
status. Hhvstands for household size. Cvstands for a set of
village indicator variables to control for factors that could
directly affect nutritional status and social rank (e.g.,
proximity to market towns). 3ihvis a random error term
with standard properties. If the coefficient of the variable
for social rank (g) bears a positive association with indices
of nutritional status it will support the hypothesis that
socioeconomic status might influence health through other
pathways beyond the ones examined here.
Potential biases and limitations of the study
Potential biases in our estimations relate to (1)
measurement errors of dependent and explanatory vari-
ables, (2) omitted variables, and (3) possible reverse
causality. First, we find random measurement error in
height (Godoyet al., 2006a). Random measurementerrorin
the outcome variable would inflate the standard error.
Since we find measurement error in height, we cannot
discard the possibility that other outcome and explanatory
variables might also be affected by random or systematic
measurement error. For example, assume that in the study
context power and prestige are different. Informants might
have systematically named powerful (but not prestigious)
people in the village, thus systematically increasing the
number of nominations of powerful people. Unfortunately,
we do not have information to detect measurement bias in
the variables used in the analysis.
Second, our estimations might be biased by the role of
omitted variables. For example, people with higher social
rank might have more access to outside resources (e.g.
hospitals), which might improve their health or nutri-
tional status. Failure to control for access to outside
resources might bias our estimation. It is also possible that
the covariates we use are inadequate to measure the
association between social status and health. For example,
it is possible that social capital has a non-linear relation
with health, and by not including a quadratic term for
social capital we might accidentally attribute some of the
effects of social capital to the social status measure,
assuming they are positively associated. We include some
of those confounders (i.e., self-reported health, mother’s
height, age squared), but we cannot rule out the possi-
bility the existence of other omitted variables (i.e.,
Third, we do not have convincing instrumental variables
to control for the potential endogeneity of social rank. It is
possible that social rank contributes to improve nutritional
status, but the causality could also run in the other direc-
tion. Therefore, we cannot speak about causality and limit
our discussion to the association between the variables
Two main limitations in our measurements make it
difficult to compare our results with previous research on
the social gradient of health. First, we focus on indices of
short-run nutritional status rather than on mortality or
morbidity. Second, we use measures of psychosocial factors
adapted from research in industrial nations, but those
measures might be less relevant in small-scale societies.
The first limitation of our study is that we use anthro-
pometric indices of nutritional status as a proxy for health
whereas most of the previous literature on the social
gradient of health has used mortality or morbidity as
outcomes (Adler et al., 1993; Marmot, 2004; Marmot et al.,
1997; Wilkinson, 2000). Differences in the outcomes
measured make it difficult to compare the results of our
study with results from previous research. Furthermore, at
least one of our indices of short-run nutritional status, BMI,
is not a good universal measure of under-nutrition and
obesity (Garn, Leonard, & Hawthorne,1986; Norgan,1994).
Previous research among indigenous peoples throughout
the world suggests that the relationship between BMI and
percent body fat varies according to differences in body
proportions (Charbonneau-Roberts, Saudny-Unterberger,
Kuhnlein, & Egeland, 2005, Godoy et al., 2005b, Norgan,
1994; Rode & Shephard,1994; Shephard & Rode,1996). For
example, in a study among the Inuit population of Canada
(Charbonneau-Roberts et al., 2005) Inuits were found to
have shorter legs than non-Caucasianpeople, which results
in an overestimation of percent of body fat when calcu-
lating BMI. Similarly, Shephard and Rode (1996) have
shown that among the traditionally-living Ingloolik Inuit of
the Northwest Territories, percent body fat levels are quite
low, despite having BMIs of 24 kg/m2and above. In
contrast, indigenous populations with more linear builds
(e.g., Australian aboriginal and South Asian populations)
appear to have relatively higher body fat levels than sug-
gested by their BMIs (Deurenberg-Yap & Deurenberg, 2003;
Norgan, 1994). Our own data suggest that BMI probably
overestimates Tsimane’ percent of fatness and risk of
obesity (that is, they are actually leaner than their BMI
would suggest). For example, the mean BMI of Tsimane’
men falls at about the 25th percentile relative to the US
reference values (Frisancho, 1990), whereas their sum of
four skinfolds (a direct measure of fatness) approximates
the 15th. At the same BMI, Tsimane’, on average, have less
body fat than their peers in the US. Because of the limita-
tions of BMI, the results of this measure should be read
The second important limitation of our study relates to
the measures chosen to proxy the psychosocial factors that
might mediate the relation between social status and
health. We use measures of psychosocial factors adapted
from research in industrial nations, i.e., job autonomy and
social capital. Both measures might not be relevant in
small-scale societies. The measure of job autonomy is
problematic because it might clump together people with
real autonomy to decide how to allocate their time and
people who want – but are unable – to obtain paid jobs.
People unable to enter the wage system might feel
marginalized, and feelings of exclusion might offset the
predictable positive effectof job autonomyon health. Social
capital might also be problematic. Social capital is strong
among the Tsimane’ and might not show enough variation
across individuals. Low variation in social capital limits the
ability of the measure to capture the lack of social inte-
gration that is believed to be a risk factor for health in
industrialized nations. Lack of adequacy of our measures of
psychosocial factors does not allow us to fully control for
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152110
Author's personal copy
Materials and methods
Data for this analysis come from a panel study in
progress among the Tsimane’ (2002–present). Experienced
interviewers and translators conducted the 2005 survey.
Previous publicationsdescribe methods usedtocollect data
on anthropometric measures (Godoy et al., 2005a), income
(Godoy et al., 2005c), and social capital (Reyes-Garcı ´a et al.,
2006). Those methods are covered briefly, and we explain
in more detail methods used to collect information on
We collected data through a survey that took place
during June–September 2005 among nearly all households
(n¼252) in 13 Tsimane’ villages straddling the Maniqui
river. The villages surveyed differed in their proximity to
(mean¼25.96 km; SD¼16.70). We selected villages at
varying distances from market towns because previous
research suggests that village-to-town distance might
affect the nutritional status (Folle ´r, 1995; Leonard &
Thomas, 1988) and notions of social status (Reyes-Garcı ´a
et al., 2008a) of indigenous populations.
We asked every person over 16 years of age (or younger
if theyheaded a household) (n¼611) tolist ‘‘the name of all
the important people in the village’’. In results reported
shortly we found that participants were more likely to
name men. We are not sure why women received few
nominations but this might reflect our poor understanding
of how to ask about female social rank. Because women
received so few nominations, we limit the multivariate
analysis to men. The total sample with complete informa-
tion included 289 adult men.
Anthropometric indices of short-run nutritional status
We trained surveyors on how to measure physical
stature, weight, mid-arm circumference, and skinfold
thickness (see Godoy et al., 2005a for a description of the
protocols). We use three different anthropometric indices
of short-run nutritional status as outcome variables: (a)
body mass index, (b) mid-arm circumference (cm), and (c)
sum of four skinfolds (biceps, triceps, subscapular, supra-
iliac; mm). BMI is a measure of body composition and the
most widely used measure of nutritional status among
adults (National Institutes of Health, 1998; Shetty & James,
1994). Mid-arm circumference provides an index of both
protein and energy status (Frisancho, 1990). The skinfold
measures are sensitive to short-term change in subcuta-
neous fat stores and are thus good measures of energy
reserves (Frisancho, 1990).
For (b) and (c), we use raw scores rather than age-stan-
dardized z scores becausemany Tsimane’ do notknowtheir
error. Using age-standardized anthropometric indices
power. Instead, we include age as a covariate. Partial
correlation coefficients between the three indices of short-
term nutritional status were as follows: BMI and mid-arm
circumference¼0.74 (p<0.0001), BMI and sum of four
skinfold¼0.69 (p<0.0001), and mid-arm circumference
and sum of four skinfold¼0.56 (p<0.0001).
To measure a person’s position in the social hierarchy
we collected relational data. We asked participants to
provide an exhaustive list with ‘‘the name of all the
important people in the village’’. To analyze the list of
nominations of ‘‘important people in the village’’ we used
social network analysis and calculated the centrality of
each person in the village network of influence, defined as
the number of nominations received by a person. The
measure, known as in-degree centrality, has been previ-
ously used to identify influential people in communities
(Costenbader & Valente, 2003). We then used the measure
of in-degree centrality to construct a measure of social
rank. The variable social rank refers to the relative rather
than to the absolute position in the hierarchy of nomina-
tions in the village, where more nominations are equated
with a higher value in our measure of social rank, and equal
observations are assigned the average rank. All the people
in a village who were not nominated received the lowest
ranking in the village. For example, if four persons received
15, 4, 0, and 0 nominations, the person with the most
nominations will be assigned rank 4, the person with the
second highest number of nominations will be assigned
rank 3, and the two people with the fewest nominations
will be assigned rank 1.5.
Because Tsimane’ market purchases account for less
than 3% of the value of household consumption (Godoy
et al., 2002), production and consumption almost fully
overlap. Therefore, we defined personal daily income as the
sum of (a) the average monetary value of a basket of farm
and forest items consumed during a day (i.e., excluding
items that were acquired through the market), which on
average represented 73% of the daily income value, (b) the
average monetary value of goods sold or bartered during
a day (17%), and (c) the daily average monetary earnings
from wage labor (11%).
To control for job autonomy we asked people whether
they had engaged in any type of wage labor during the 2
months before the dayof the interview. Becausewage labor
is an uncommon activity, it is a well-remembered event
among the Tsimane’. For the regression analysis we created
a dummy variable that took the value of one if the partic-
ipant had earned any cash from wage labor during the 2
months before the interview, and zero otherwise.
We asked participants to report all the times they had
received gifts from kin or other Tsimane’ inside or outside
the village (but not gifts received from people in their
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152111
Author's personal copy
household) during the week before the day of the inter-
view. We collected information on the receiver of the gift,
independently of whether the gift was later shared in the
We controlled for age, schooling, household size, village
of residency, and hereditary factors. Previous studies have
inequality, institutions, or residential segregation con-
tribute to health (Marmot, 2005). We could not control for
this wide range of variables, but village dummy variables
have the advantage of accounting for all of these commu-
nity-level fixed confounders. We also attempted to control
for hereditary factors. Since BMI includes weight and
height, and height has a strong hereditary component
(Henneberg & van den Berg, 1990) we included mother’s
height as a control.
Table 1 contains definitions and summary statistics for
the variables used in the regression analyses.
Table 2 contains the main regression results. We find
a positive and statistically significant association between
social rank and the three anthropometric indicators of
nutritional status. Since we express social rank and the
indicators of nutritional status in natural logarithms, we
can read the coefficients as elasticities (%D outcome/1% D
social rank). Given that our measure of social rank only
gives us ordinal rankings, the magnitude of the associations
found should be read with caution. For example, one
cannot compare the relative magnitude of the coefficients
across the three indicators of nutritional status, as
a percentage variation of the average might represent
different percent of the actual variation in the sample for
each of the three outcome variables.
InTable 2 column , we find that a 1% increase in social
rank is associated with an increase of BMI of 0.032%
(p¼0.10). The association implies that moving the lowest
ranked men from position one to position two in the rank
hierarchy would increase his BMI by 3.2%.
In column  we include the height of the man’s
mother. We could not obtain data from the mothers of all
men in the sample, so the sample size of the regression
using mother’s height is smaller (n¼131) than the sample
size of regression  (n¼289). Once we condition for
mother’s height, the point elasticity has a three-fold
increase from 0.032 in model  to 0.090 in model .
We tested the association between social rank and mid-
arm circumference (columns –) and sum of skinfold
thickness (columns –). We found statistically signifi-
cant, positive, and meaningful associations. For instance,
doubling a person’s social rank was associated with a 4.1%
increase in mid-arm circumference in model  that does
not control for mother’s height, and with a 8.5% increase in
mid-arm circumference in the model that controls for
mother’s height (column ). Doubling a person’s social
rankwasassociated with a 14.0% increase in the sum of four
skinfolds in model  that does not control for mother’s
height, and with a 24.1% increase in the sum of four sk-
infolds in model  that controls for mother’s height
In sum, social rank bears a positive association with the
three anthropometric indices of nutritional status that we
measured. For the three outcomes, the magnitude of the
association was at least double in the model controlling for
mother’s height. All the results, except the association
between social rank and BMI without controlling for
mother’s height, were statistically significant at the 95%
We tested the robustness of our results in four different
ways (results not shown). First, we ran the regressions of
Table 2 using a different proxy for social status. We
generated a dummy variable that took the value of one if
the person was nominated as important in the village at
Definition and summary statistics of variables used in regressions, men over 18 years of age
Variable DefinitionN Mean Standard deviation
I. Outcome variables (in regressions entered in natural logarithm)
BMI Body mass index (weight in kg/height in m2)
Mid-arm circumference Mid-arm circumference of participant (cm)
Sum of four skinfoldsSum of biceps, triceps, subscapular, supra-iliac skinfold thickness (mm)
II. Explanatory variable (in regressions entered in natural logarithm)
Social rank Person’s rank in a village, from the lowest number of nominations
(lower social rank) to the person with the highest number of nominations
III. Control variables
Daily income Average personal monetary income from (1) value of a basket of foods
consumed during a day and (2) monetary income from sale, barter, and
wage labor; in bolivianos (Bs) (1US $¼7.98 Bs)
Total value in bolivianos of gifts received by subject in last 7 days
Dummy variable. 1¼the person has not worked for wage labor during the
last 2 months, 0¼the person has done some wage labor
Age of participant (years)
Maximum school grade achieved by participant
Number of people in the household
Measured standing physical stature of subject’s mother (cm)
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152112
Author's personal copy
least once and zero if the person was never nominated.
Men who were nominated at least once had between 2.6%
and 8.3% higher BMI, between 3.2% and 7.3% higher mid-
arm circumference, and between 11.2% and 23.7% higher
sum of four skinfold than men who were never nominated.
All the associations were statistically significant at the 95%
confidence level or higher except the regression using BMI
as outcome but without controlling for mother’s height
Second, we tested whether the association persisted
after controlling for (a) self-reported poor health, a stan-
dard covariate of BMI, (b) smiles, an objective indicator of
level of stress, (c) alcohol consumption, a standard risk
factor for health, (d) daily individual consumption (rather
than income), and non-linear effects of (e) age and (f)
household size. To proxy health, we collected information
on the total number of days subjects reported being bed-
ridden during the 2 weeks before the day of the interview.
To proxy the level of stress of respondents, we used a direct
measure of happiness, intensity of smiling during inter-
views, as coded by surveyors. Experimental and observa-
tional studies suggest that smiling bears a positive
association with self-reported happiness across cultures
(Pavot, Diener, Colvin, & Sandvik, 1991). We collected
information on alcohol consumption by asking about the
amount of alcohol and beer consumed over the 7 days
before the day of the interview. We defined daily individual
consumption as the sum of the average monetary vale of all
money spent during a day, the average monetary value of
goods bartered during a day, and the average monetary
value of a basket of farm and forest items consumed during
a day (i.e., excluding items that were acquired through the
market). To control for non-linear effects of age and
household size we included squared terms. We ran a set of
regressions similar to those in Table 2 adding the different
variables just discussed one at a time. We found essentially
the same results, with weaker significance for BMI than for
the other two outcomes. We also ran a regression with all
the variables in the benchmark model and all the covariates
described in the previous paragraph (notshown). We found
weaker results to the ones presented in Table 2 for the
regression that excluded maternal height, and similar
results for the model controlling for maternal height.
Third, we used paternal, rather than maternal stature as
control for hereditary factors. The coefficients for social
rank in regressions including paternal stature are slightly
lower than the coefficients in regressions using maternal
stature, but statistically significant at the 95% confidence
In our final test of robustness, we ran the analysis of
Table 2 columns [a], [c], and [e] but limiting the sample to
men for whomwe lack information on mother’s height. We
did so to test whether the effects we find when including
mother’s height are driven by changes in sample compo-
sition. We found that for this part of the sample none of our
outcome variables were associated in a statistically signif-
icant way to social rank.
Discussion and conclusion
We organize the discussion around three findings. First,
we found that a locally-relevant measure of social rank has
a positive association with anthropometric indicators of
short-run nutritional status. Second, the association was
limited to the part of the sample for which we had data on
mother’s height. Third, for this part of the sample, the
confounders and standard pathways examined in research
in industrial nations.
First, we found that the association between social rank
and anthropometric indicators of short-run nutritional
status is robust to the three indicators of nutritional status
used. The magnitude of the association and the level of
statistical significance were lower when using BMI than
when using mid-arm circumference or the sum of four
skinfolds as outcomes. As discussed in a previous section,
a possible explanation for the lower significance of BMI
relates to limitations of BMI as a universal measure of
under-nutrition and obesity (Garn et al., 1986; Norgan,
Regression results of indicators of nutritional status (outcome variable) against social rank among Tsimane’ men (>18 years)
Explanatory variables: Outcome variable. Natural logarithm of:
BMI Mid-arm circumferenceSum four skinfold
   
Social rank (log)
0.032 (0.018)* 0.090 (0.018)***0.041 (0.013)***0.085 (0.021)*** 0.140 (0.043)*** 0.241 (0.093)**
Note: Ordinary least squares (OLS) regressions with robust standard error and clustering by village. Regressions contain a full set of village dummies
(13 ?1¼12) notshown. Fordefinition of variables see Table 1. Robust standard errorsused whenprobabilityof exceeding critical value in Cook-Weisberg test
of heteroskedasticity <0.10. In cells we show coefficients and, in parenthesis, standard errors. *, **, And *** significant at the 10%, 5%, and 1% level. ^ Variable
intentionally excluded from the analysis.
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152113
Author's personal copy
Our first finding, the association between social rank
and nutritional status, is consistent with findings from
industrial and non-human primate societies (Marmot,
2006; Sapolsky, 2004; Wilkinson, 2000). The magnitude of
the association was important in real terms, but probably
not equally significant for the entire sample. Since less than
a third of the sample was nominated as important, we are
unsure of how hard would it be for a person to go from the
group of non-nominated to the group of people who at
least received one nomination.
Second, when splitting the sample between informants
with and without data on mother’s height, we found that
the association between social status and indicators of
short-run nutritional status persisted only for the sample of
people for whom we have data on mother’s height. The
subsample of people for whom we do not have data on
mother’s height. Why would the results of the two
subsamples be different? Using a series of t-test compari-
sons of means, we analyzed the socio-demographic and
physical characteristics of the two subsamples (not shown).
Men for whom we do not have data on mother’s height
were older than men for whom we do have data (45 versus
28 years of age, p<0.00001), probably because it is more
likely that the mothers of older people were dead. We also
found a positive and statistically significant association
between age and social rank for the two subsamples. Our
results, suggest that the association between social rank
and indices of short-run nutritional status might decrease
across the life span.
The last finding that deserves discussion is the robust-
ness of the association after controlling for standard path-
ways. The finding supports the idea that dominance
hierarchies might influence health through other channels
beyond the channels examined in industrial societies, i.e.,
perceived stress, social capital, and own resources. Because
of the limitations of our proxies for psychosocial factors
(discussed before), our results do not allow us to conclude
that the psychosocial path does not explain the association
between social rank and nutritional status in a foraging-
farming society, but rather it is possible that the proxies for
the psychosocial factors used might not be adequate in this
context. Which other channels might mediate the social
We suggest two other potential paths: exemption from
social obligations, and psychological well-being associated
experiments suggest that individuals with higher social
rank receive privileges and are excused from some social
obligations (Henrich & Gil-White, 2001). Using behavioral
high social rank are more likely to be exempted from social
sanctions (Bickman, 1971) and are more likely to be
conferred special privileges (Ungar, 1981) than individuals
of low social rank. Thus, it is possible that social rank
changes the allocation of tasks and occupations and, in so
doing, changes energy expenditures, which influences
anthropometric indices of short-run nutritional status.
Second, in a Foraging-farming society, the sense of being
respected by the community might be a better measure of
the psychosocial factors that mediate the relation between
socioeconomic status and health than job autonomy and
social capital. Researchers have argued that poor health
reflects both the feeling of being poor and the objective
status of poverty. Subjective indicators of socioeconomic
status are as good or better predictors than objective
indicators of socioeconomic status for stress-related
outcomes (Goodman et al., 2003). Along the same lines,
a subjective, culturally relevant indicator of respect might
be a better measure of the pathway for the association
between social rank and health.
In conclusion, this study opens three lines for future
research on the social gradient in health among small-scale
societies. First, further research should address the role of
culturally defined notions of respect as potential pathways
to explain the relationship between social rank and health.
Second, further research should also address causality
between the variables explored.Longitudinal surveysof the
same informants have the potential to control for unob-
served person-level fixed effects correlated with the
regressors and outcomes, and could therefore move the
analysis towards a causal interpretation. Last, results from
this research point tothe need fora better understandingof
the association between social rank and indices of short-
run nutritional status at different points in a lifetime.
Further empirical research among small-scale societies
should address the longitudinal dimension of the social
gradient in health.
Adler, N., Boyce, T., Chesney, M., Folkman, S., & Syme, S. (1993). Socio-
economic inequalities in health: no easy solution. Journal of the
American Medical Association, 269, 3140–3145.
Bickman, L. (1971). The effect of social status on the honesty of other.
Journal of Social Psychology, 85, 87–92.
Bosma, H. (1998). Two alternative job stress models and risk of coronary
heart disease. American Journal of Public Health, 88, 68–74.
Brabec, M., Godoy, R., Reyes-Garcı ´a, V., & Leonard, W. (2007). BMI,
income, and social capital in a native Amazonian society: interaction
between relative and community effects. American Journal of Human
Biology, 19, 459–474.
Censo Indı ´gena. (2001). Censo Indı ´gena del Oriente, Chaco, y Amazonı ´a.
Secretaria de Asuntos E´tnicos, de Ge ´nero y Generacionales. La Paz:
Ministerio de Desarrollo Humano.
Charbonneau-Roberts, G., Saudny-Unterberger, H., Kuhnlein, H. V., &
Egeland, G. M. (2005). Body mass index may overestimate the prev-
alence of overweight and obesity among the Inuit. Journal of
Circumpolar Health, 64(2), 163–169.
Coleman, J. S. (1990). Foundations of social theory. Cambridge, Massa-
chusetts: Harvard University Press.
Costenbader, E., & Valente, T. W. (2003). The stability of centrality
measures when networks are sampled. Social Networks, 25, 283–307.
Daillant, I. (2003). Sens Dessus Dessous. Organization sociale et spatiale des
Chimane d’Amazonie boliviane. Nanterre: Societe d’ethnologie.
Deaton, A. (2003). Health, inequality, and economic development. Journal
of Economic Literature, 41, 113–158.
Deurenberg-Yap, M., & Deurenberg, P. (2003). Is a re-evaluation of WHO
body mass index cut-off values needed? The case of Asians in Sin-
gapore. Nutrition Review, 61, S80–S87.
Dufour, D. (1994). Diet and nutritional status of Amazonian peoples. In A.
Roosevelt (Ed.), Amazonian Indians from prehistory to the present.
Tuczon: University of Arizona Press.
Ellis, R. (1996). A taste for movement: an exploration of the social ethics
of the Tsimane’ of Lowland Bolivia. Ph.D. Thesis. St Andrews Univer-
Frisancho, A. (1990). Anthropometric standards for the assessment of growth
and nutritional status. Ann Arbor, MI: University of Michigan Press.
Folle ´r, M. L. (1995). Future health of indigenous peoples: a human ecology
viewand the case of the Amazonian Shipibo-Conibo. Futures, 27(9/10),
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–21152114
Author's personal copy Download full-text
Garn, S., Leonard, W., & Hawthorne, V. (1986). Three limitations of the
Body Mass Index. American Journal of Clinical Nutrition, 44, 996–997.
Godoy, R., Byron, E., Reyes-Garcı ´a, V., Vadez, V., Leonard, W., Apaza, L.,
et al. (2005a). Income inequality and adult nutritional status:
anthropometric evidence from a pre-industrial society in the Bolivian
Amazon. Social Science & Medicine, 61, 907–919.
Godoy, R., Leonard, W., Reyes-Garcı ´a, V., Goodman, E., McDade, T.,
Huanca, T., et al. (2006a). Physical stature of adult Tsimane’ Amer-
indian of the Bolivian Amazon in the 20th century. Economics and
Human Biology, 4, 184–205.
Godoy, R., Overman, H., Demmer, J., Apaza, L., Byron, E., Huanca, T., et al.
(2002). Local financial benefits of rain forests: comparative evidence
from Amerindian societies in Bolivia and Honduras. Ecological
Economics, 40(3), 397–409.
Godoy, R., Reyes-Garcı ´a, V., Byron, E., Leonard, W., & Vadez, V. (2005b).
The effect of market economies on the well-being of indigenous
peoples and on their use of renewable natural resources. Annual
Review of Anthropology, 34, 121–138.
Godoy, R., Reyes-Garcı ´a, V., Vadez, V., Leonard, W., Huanca, T., &
Bauchet, J. (2005c). Human capital, wealth, and adult nutritional
status among a foraging-horticultural society of the Bolivian Amazon.
Economics and Human Biology, 3, 139–162.
Godoy, R., Reyes-Garcı ´a, V., Leonard, W., Huanca, T., McDade, T., Tanner, S.,
et al. (2007). On the measure of income in autarky and the economic
unimportance of social capital. Journal of Anthropological Research,
Goodman, E., Adler, N., Daniels, S., Morrison, J., Slap, G., & Dolan, L. (2003).
Impact of objective and subjective social status on obesity in a biracial
cohort of adolescents. Obesity Research, 11, 1018–1026.
Henneberg, M., & van den Berg, E. R. (1990). Test of socioeconomic
causation of secular trend: stature changes among favored and
oppressed South Africans are parallel. American Journal of Physical
Anthropology, 83, 459–465.
Henrich, J., & Gil-White, F. (2001). The evolution of prestige. Freely
conferred deference as a mechanism for enhancing the benefits of
cultural transmission. Evolution and Human Behavior, 22, 165–196.
Huanca, T. (2008). Tsimane’ oral tradition, landscape, and identity in tropical
forest. La Paz: Imprenta Wagui.
Johnson, A. (2003). Families of the forest: Matsigenka Indians of the Peru-
vian Amazon. University of California Press.
Kaplan, H., & Hill, K. (1985). Food sharing among ache foragers: tests of
explanatory hypotheses. Current Anthropology, 26, 223–245.
Kawachi, I. (2002). The health of nations. Why inequality is harmful to your
health. New York: The Free Press.
Kawachi, I., & Kennedy, B. (1999). Income inequality and health: pathways
and mechanisms. Health Services Research, 34, 215–217.
Leonard, W., & Thomas, R. B. (1988). Changing dietary patterns in the
Peruvian Andes. Ecology of Food and Nutrition, 21, 245–263.
Mare, R. (1990). Socio-economic careers and differential mortality among
older men in the United States. In J. Vallin, S. D’Souza, & A. Palloni
(Eds.), Measurement and analysis of mortality (pp. 362–387). Oxford:
Marmot, M. (2004). The Status Syndrome: How social standing affects our
health and longevity. Times Books.
Marmot, M. (2005). Social determinants of health inequalities. Lancet,
Marmot, M., Ryff, C., Bumpass, L., Shipley, M., & Marks, N. (1997). Social
inequalities in health: next questions and converging evidence. Social
Science & Medicine, 44, 901–910.
Marmot, M., & Smith, G. (1991). Health inequalities among British civil
servants: the Whitehall II study. Lancet, 337, 1387–1394.
McLaren, L. (2007). Socioeconomic status and obesity. Epidemiologic
Review, 29, 29–48.
Moore, D., & Hayward, M. (1990). Occupational careers and the mortality
of elderly men. Demography, 27, 31–53.
National Institutes of Health. (1998). Clinical guidelines on the identifica-
tion, evaluation, and treatment of overweight and obesity in adults: The
Evidence Report. 98-4083. Bethesda, MD: NIH.
Norgan, N. (1994). Population differences in body composition in relation
to the body mass index. Journal of Clinical Nutrition, 48, S10–S25.
Orr, C. M., Dufour, D., & Patton, J. Q. (2001). A comparison of anthropo-
metric indices of nutritional status in Tukanoan and Achuar Amer-
indians. American Journal of Human Biology, 13, 301–309.
Ostrom, E. (2000). Social capital: a fad or a fundamental concept? In P.
Dasgupta, & I. Sarageldin (Eds.), Social capital: A multifaceted
perspective (pp. 172–214) Washington, DC: The World Bank.
Pavot, W., Diener, W., Colvin, R., & Sandvik, E. (1991). Further validation of
the satisfaction with life scale: evidence for the cross-method
convergence of well-being measures. Journal of Personality Assess-
ment, 57, 149–161.
Reyes-Garcı ´a, V., Godoy, R., Vadez, V., Huanca, T., & Leonard, W. (2006).
Individual and group incentives to invest in prosociality: a empirical
study in the Bolivian Amazon. Journal of Anthropological Research, 62,
Reyes-Garcı ´a, V., Molina, J. L., Broesch, J., Calvet, L., Huanca, T., Saus, J.,
et al., TAPS study team. (2008a). Male ethnomedicinal plant knowl-
edge, age, and prestige: a test of predictions from the prestige-bias
model of cultural transmission. Evolution and Human Behavior, 29(4),
Reyes-Garcı ´a, V., Vadez, V., Godoy, R., Huanca, T., Leonard, W., McDade, T.,
et al. (2008b). Non-market returns to traditional and modern human
capital: nutritional status in a native Amazonian society. Journal of
Development Studies, 44(2), 206–221.
Rode, A., & Shephard, R. J. (1994). Prediction of percent body fat content in
an Inuit community. American Journal of Human Biology, 6, 249–254.
Sapolsky, R. (2004). Social status and health in humans and other animals.
Annual Review of Anthropology, 33, 393–418.
Shephard, R. J., & Rode, A. (1996). Health consequences of ‘modernization’:
Evidence from circumpolar peoples. Cambridge University Press.
Shetty, P. S., & James, W. P. T. (1994). Body mass index: A measure of chronic
energy deficiency in adults. Food and nutrition paper 56. Rome: Food
and Agriculture Organization.
Sobal, J., & Stunkard, A. J. (1989). Socioeconomic status and obesity:
a review of the literature. Psychological Studies, 105(2), 260–275.
Sorenson, G., Pirie, P., Folsom, A., Luepker, R., Jacobs, D., & Gillum, R.
(1985). Sex differences in the relationship between work and health:
the Minnesota Heart Survey. Journal of Health and Social Behavior, 26,
Subramanian, S. V., & Kawachi, I. (2004). Income inequality and health:
what have we learned so far? Epidemiologic Review, 26, 78–91.
Ungar, S. (1981). The effects of status and excuse on interpersonal rela-
tions to deviant behavior. Social Psychology Quarterly, 44, 260.
Wilkinson, R. G. (1996). Unhealthy societies: The affliction of inequality.
Wilkinson, R. G. (2000). Mind the gap: Hierarchies, health, and human
evolution. London: Weidenfeld & Nicolson.
Winterhalder, B. (1997). Gifts given, gifts taken: the behavioral ecology of
nonmarket, intragroup exchange. Journal of Archeological Research,
V. Reyes-Garcı ´a et al. / Social Science & Medicine 67 (2008) 2107–2115 2115