DEMOGRAPHIC RESEARCH
VOLUME 38, ARTICLE 3, PAGES 95,108
PUBLISHED 5 JANUARY 2018
http://www.demographic-research.org/Volumes/Vol38/3/
DOI: 10.4054/DemRes.2018.38.3
Descriptive Finding
High life expectancy and reversed socio-
economic gradients of elderly people in
Mexico and Costa Rica
Luis Rosero-Bixby
© 2018 Luis Rosero-Bixby.
This open-access work is published under the terms of the Creative Commons
Attribution 3.0 Germany (CC BY 3.0 DE), which permits use, reproduction,
and distribution in any medium, provided the original author(s) and source
are given credit.
See https://creativecommons.org/licenses/by/3.0/de/legalcode.
Contents
1 Introduction 96
2 Methods 98
3 Results 101
4 Discussion 103
References 105
Demographic Research: Volume 38, Article 3
Descriptive Finding
High life expectancy and reversed socioeconomic gradients of elderly
people in Mexico and Costa Rica
Luis Rosero-Bixby1
Abstract
BACKGROUND
Some existing estimates suggest, controversially, that life expectancy at age 60 (LE60)
of Latin American males is exceptionally high. Knowledge of adult mortality in Latin
America is often based on unreliable statistics or indirect demographic methods.
OBJECTIVES
This study aims to gather direct estimates of mortality at older ages in two Latin
American countries (Mexico and Costa Rica) using recent longitudinal surveys and to
determine the socioeconomic status (SES) gradients for LE60.
METHODS
Data were collected from independent panels of approximately 7,000 older adults
followed over more than a decade ‒ the MHAS and CRELES surveys. The age-specific
death rates were modeled with Gompertz regression, and thousands of life tables were
simulated to estimate LE60 and its confidence interval.
RESULTS
LE60 estimates obtained from MHAS and CRELES are similar to those obtained from
traditional statistics, confirming the exceptionally high LE60 of men in the two
countries. The expected gradients of higher LE60 with higher SES are not present,
especially among males, who even show reverse gradients (some exaggerated by data
issues).
CONCLUSIONS
Vital statistics correctly estimate elderly mortality in Mexico and Costa Rica. The
higher-than-expected LE60 among Latin American males in general, and particularly
among low-SES individuals, seems to be real; their determinants should be thoroughly
investigated.
1 Centro Centroamericano de Población (CCP), Universidad de Costa Rica. Email: lrosero@mac.com.
http://www.demographic-research.org 95
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
CONTRIBUTION
This study shows with hard, reliable data, independent of traditional statistics, that
elderly males in tropical Latin America enjoy an exceptionally high life expectancy and
that SES gradients are absent or even reverse.
1. Introduction
In each cohort of Latin Americans, 84% of life table deaths occur after age 60 (UNPD
2015). Despite this preponderance, estimates of old-age mortality in Latin America are
controversial, often based on unreliable statistics or indirect demographic methods.
Only five countries2 have vital statistics that are considered adequate to allow valid
estimates of mortality (CELADE 2010); only Chile is included in the Human Mortality
Database of populations with complete data (HMD 2017). In most Latin American
countries, adult mortality estimates are based on indirect demographic methods using
questionable assumptions (Hill, Choi, and Timæus 2005).
A peculiarity of Latin America is that elderly males show better-than-expected life
expectancy given the level of development of the countries. Figure 1 shows the life
expectancy of males at age 60 (LE60-M) (UNPD 2015) in relation to the per capita
gross domestic product parity purchasing power (GDP-PPP) (World Bank 2013) in 170
countries across the world. As expected, countries with higher GDP show higher LE60-
M. Strikingly, all tropical Latin American countries are above the regression line that
best fit the data, which means that these countries have a higher-than-expected LE60-
M.3 Costa Rica, Cuba, Ecuador, and Mexico, with a GDP-PPP per capita of about
$10,000, show LE60-M of approximately 22 years, which would normally agree with
economies of about $70,000.
Although the exceptionally low Latin American mortality at older ages has been
documented (UNPD 1982), it has been dismissed as a result of faulty data, particularly
age-misreporting in censuses (Coale and Kisker 1986; Dechter and Preston 1991;
Preston, Elo, and Stewart 1999). Recent evidence, however, suggests that health at old
ages in Latin America may be as good as in the United States (Payne 2015).
2 Argentina, Chile, Costa Rica, Cuba, and Uruguay.
3 An analogous figure (provided as supplemental material to this article) with child mortality instead of GDP
on the x-axis shows that LE60-M is also higher than expected in tropical Latin American countries given their
child mortality levels.
96 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
Figure 1: Life expectancy of males aged 60 years by GDP-PPP. World’s
countries circa 2010
BRA=Brazil, COL=Colombia, CRI=Costa Rica, CUB=Cuba, DOM=Dominican Republic, ECU=Ecuador, SLV=El Salvador,
GTM=Guatemala, HND=Honduras, MEX=Mexico, NIC=Nicaragua, PAN=Panama, PRY=Paraguay, PER=Peru, VEN=Venezuela.
Latin America has the highest income distribution inequality in the world
(Ravallion 2014). If mortality were determined exclusively by income, the region
should also show extreme social inequalities in life expectancy. Indeed, Behm and
collaborators documented extreme socioeconomic inequalities in child mortality in the
1970s (Behm 1980). In Guatemala, for example, the children of uneducated mothers
died at a rate four times higher than that for the children of mothers with ten or more
years of education (Chackiel and Plaut 1996). It has often been assumed that similar
inequalities occur in adult mortality, although the statistical evidence supporting this
assumption is exiguous. For example, in a book on adult mortality in Latin America
that compiles 18 papers, not a single item of hard data shows socioeconomic status
(SES) inequalities (Timaeus, Chackiel, and Ruzicka 1996). Only in Chile and Argentina
has it been documented that low-educated or low-income adults tend to die at higher
rates (Sandoval and Turra 2015; Peláez and Acosta 2011; Rofman 1994). By contrast,
http://www.demographic-research.org 97
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
studies of Hispanics in the United States and of adult Costa Ricans have questioned the
existence of SES gradients in adult mortality, especially at older ages (Rosero-Bixby
and Dow 2016; Turra and Goldman 2007; Lariscy, Hummer, and Hayward 2015).
This report, based on longitudinal surveys in Mexico and Costa Rica, has the
double purpose of (1) directly and independently determining adult mortality in those
countries and to assess the validity of existing estimates and (2) documenting the SES
gradients for mortality as a first step toward understanding its determinants.
Both Mexico and Costa Rica are middle-income economies with large income
distribution inequality (Gini index of 0.5 (Underwood 2014)). While 70% of Mexicans
had health insurance in 2008 (Gómez-Dantés et al. 2011), 85% of Costa Rica had it
according to the 2011 census.
2. Methods
This report uses existing databases from two independently conducted longitudinal
surveys: (1) the Mexican Health and Aging Study (MHAS 2004; Wong et al. 2015) and
(2) the mother sample of the Costa Rican Study of Longevity and Healthy Aging
(CRELES) (Rosero-Bixby and Dow 2009).
MHAS is a representative panel of the older Mexican population initiated in 2001
with follow-up waves in 2003 and 2012. Its micro-databases are publicly available on
the project’s website (MHAS 2012). For comparability purposes, ages younger than 55
and older than 99 as well as interviewed spouses were excluded. The analytical sample
size used here is 6,700 individuals (Table 1). Relatives and neighbors provided
information on the date of death for 87% of the deceased participants; 4% of the
missing dates were randomly imputed for 2001‒2003 and 9% for 2003‒2012. The
sample includes 4% of the deaths recovered from the MHAS data files of no-interviews.
The analytical panel has an attrition rate of 7% that clearly increases with SES.
Dropouts were censored at the date of the last contact. A sensitivity analysis to attrition-
mortality scenarios is provided as supplementary material. The observation time started
on  January  1,  2002,  several  months  after  the  baseline  interviews,  and  stopped  on  the
date of the 2012 interview.
98 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
Table 1: Sample sizes and attrition rate by gender, age, and SES groups.
Persons Person-years Deaths % attrition
Groups
Mexico Costa Rica Mexico Costa Rica Mexico Costa Rica Mexico Costa Rica
Total 6,748 7,629 57,995 59,147 2,049 3,290 6.9 1.5
Gender
Male 2,908 3,620 24,541 27,742 969 1,651 6.7 1.1
Female 3,840 4,009 33,454 31,405 1,080 1,639 7.0 1.9
Baseline age
55–74 5,508 4,076 49,829 39,402 1,275 975 6.9 0.7
75–99 1,240 3,553 8,166 19,745 774 2,315 6.9 2.4
Observed age
55–74 39,853 28,521 768 509
75–99 18,142 30,626 1,281 2,781
Baseline residence
<100,000 inhab. 2,825 3,488 25,103 27,226 864 1,486 4.0 1.4
100,000+ inhab. 3,923 4,141 32,892 31,921 1,185 1,804 8.9 1.6
Education
None 2,017 1,500 16,858 10,767 738 724 5.1 1.9
Primary 3,566 4,979 31,186 38,705 1,059 2,152 6.0 1.5
Secondary 833 712 7,196 5,954 184 266 11.3 0.8
Post-secondary 332 438 2,755 3,722 68 148 15.4 1.4
Wealth tercile
Low 1,948 2,957 16,710 22,479 672 1,332 5.1 1.4
Medium 2,409 2,162 20,510 17,129 771 911 6.4 1.7
High 2,391 2,510 20,775 19,539 606 1,047 8.8 1.5
The CRELES panel is a national sample of residents aged 55 or more drawn from
2000 census files linked to the death registry (details in Rosero-Bixby, Brenes, and
Collado 2004). After excluding foreigners (3%) and centenarians the analytical sample
size is 7,200. In addition to the follow-up in the death registry, the survival of
individuals was verified using the voting lists for the presidential elections of 2002,
2006, 2010, and 2014, resulting in an attrition rate of 1.5% of non-death individuals
who disappeared from the voting lists. These lost individuals were excluded from
observation on the closing date of the voting registry in which they first disappeared.
Observation started on January 1, 2001 (six months after the census) and stopped on
December 31, 2011. A nested subsample of 3,000 individuals was contacted in person
in three waves of visits mostly in 2005, 2007, and 2009. In waves 2 and 3, relatives or
neighbors reported 566 deaths in this subsample. Only five of these deaths (<1%) were
missing in the death registry follow-up, suggesting that the Costa Rican registry is
essentially complete.
SES groups were defined with three indicators at baseline: (1) educational
attainment (no formal education, some elementary, some secondary, and postsecondary
education); (2) tercile of household wealth (measured by the count of eight household
assets: tap water inside the house, toilet, television, refrigerator, washer, telephone, hot
http://www.demographic-research.org 99
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
water, and car); and (3) whether the place of residence is a city of more than 100,000
inhabitants. People in cities usually enjoy better economic opportunities and services
(clean water, electricity, transportation, banking, health care, and so on) than in rural
areas, and thus they are considered of a higher SES. Table 1 shows the analytical
sample sizes of the 40 groups defined with these variables.
Age in each observation segment was established from the date of birth (DoB) in
months. DoB in CRELES was taken from the linked birth registry, which makes it
error-free. Participants in MHAS reported their DoB in the 2001 wave. An assessment
of the accuracy of DoB in MHAS, based on confirmatory reports during the 2012 wave,
is included as supplementary material. Only 13% of MHAS participants changed their
reported birth year; 8% changed their five-year bracket.
A two-parameter Gompertz function (Pollard 1991) was estimated for each SES
group using hazard regression (Hosmer and Lemeshow 1999). The LE60 and its 95%
confidence interval (95% CI) were estimated for each group using the Gompertz
parameters (and their standard errors) to generate 1,000 sets of death rates, and the
corresponding life tables with Monte Carlo simulation. The median value of LE60 in
1,000 simulations is taken as the point indicator of mortality in a group along with the
2.5 and 97.5 percentiles as estimates of the 95% CI.4
The national estimates of LE60 in this report are compared to estimates5 from the
following other sources:
∂ Vital statistics: deaths in the last intercensus period and, as rates denominator,
the population average of the census of 2000 and 2010 in Mexico and of 2000
and 2011 in Costa Rica (Palloni, Pinto, and Beltrán-Sánchez 2014).
∂ Official estimates: the age-specific death rates from 2000‒2010 used by the
Consejo Nacional de Población (CONAPO 2012) in the population
projections of Mexico and the average of the Costa Rican death rates of 2000‒
2005 and 2005‒2010 in the life tables by the Pension Superintendent (SUPEN)
(CCP 2014).
∂ LAMBdA project: death rates in the life tables estimated by the project “Latin
American Mortality Database” (LAMBdA), Mexico 2000‒2010 and Costa
Rica 2000‒2011 (Palloni, Pinto, and Beltrán-Sánchez 2014).
4 Supplementary material includes a STATA data file with the microdata used to estimate death rates and the
hazard regression models. Table S1 shows the two Gompertz parameters for each of the 40 groups and their
standard errors as well as the corresponding LE60 estimate and its 95% CI, along with estimates of life
expectancy for ages 55 and 65.
5  To avoid discrepancies due to differences in the method used by each source to compute LE60, this
indicator was re-estimated with the set of age-specific death rates in ages 55‒99 years for each source
modeled with a Gompertz function to obtain smoothed rates for ages 55–114 years and the corresponding life
table.
100 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
∂ GBD: average of the death rates reported for 2000 and 2010 by the project
“Global Burden of Disease” (GBD) (IHME 2015).
∂ UNPD: Average of death rates of 2000‒2005 and 2005‒2010 in (UNPD 2015).
3. Results
The resulting LE60 for males (21.2 years in Mexico and 21.9 years in Costa Rica) is
essentially the same as those from vital statistics considering the 95% CI (Table 2).
Among women, the MHAS estimate (23.4 years) and its 95% CI is slightly higher than
that from the Mexican vital statistics, whereas the CRELES estimate (24.4 years) is
slightly lower. The UNPD estimates for these two countries, as well as the official life
tables of Costa Rica, also yield estimates similar to those of MHAS and CRELES. The
official mortality estimate for Mexico found in the CONAPO seems slightly up-biased
(down-biased LE60).
Table 2: Comparing estimates of life expectancy at age 60
Country, period, and source Males Females
Mexico 2000‒2010
MHAS 2002‒2011 21.2 (20.5‒21.8) 23.4 (22.8‒24.0)
Vital statistics 20.4 22.4
Official CONAPO 19.9 22.1
LAMBdA 18.0 19.6
GBD 21.8 24.7
UNPD 20.8 22.6
Costa Rica 2000‒2010
CRELES 2002‒2012 21.9 (21.5‒22.2) 24.3 (23.9‒24.8)
Vital statistics 21.8 24.8
Official SUPEN 21.8 24.4
LAMBdA 19.3 21.4
GBD 20.7 24.0
UNPD 21.5 24.2
In parenthesis the 95% confidence interval.
Estimates  from  the  multicountry  projects  LAMBdA  and  GBD  tend  to  be  out  of
range. LAMBdA yields substantially higher mortality (lower LE60) in the two
countries. For example, LE60 of Mexican males is just 18.0 years according to
LAMBdA. The LE60 GBD estimate is too high for Mexico and too low for Costa
Rica.6
6 A more detailed comparison of the age-specific death rates (instead of the summary LE60) is shown in
Figure S2, included as supplementary material.
http://www.demographic-research.org 101
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
Figure 2 shows the resulting SES gradients in LE60. Strikingly, the expected
gradients of higher LE60 with better SES do not show up in Figure 2, except among
Costa Rican women. There are no significant differences in LE60 by education among
Mexican women. Males in the two countries clearly show declining LE60 with
increased education ‒ a reverse gradient. Mexicans with postsecondary education, for
example, have just 16.0 years (13.6–18.5 CI) of LE60 compared to 22.9 years (21.7‒
23.9 CI) for those with no education.
Figure 2: Life expectancy at age 60 (LE60) by gender and SES groups. Mexico
and Costa Rica
102 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
Residents in large cities tend to have lower LE60, with the exception of Costa
Rican women. For example, LE60 among Mexican males residing in cities larger than
100,000 inhabitants is 20.0 years compared to 21.7 years for people living in small
towns or rural areas.
The stratification by three groups (terciles) of household wealth does not yield
gradients, negative or positive, in LE60.
4. Discussion
Panels of approximately 7,000 older adults followed over more than a decade in Mexico
and Costa Rica yielded direct and independent estimates of adult mortality that allow an
assessment of existing estimates. Two results are apparent: (1) LE60 from these two
panels is similar to the raw estimates from vital statistics in the two countries, and (2)
SES gradients are lacking; moreover, some ‘reverse inequality’ shows up among males,
with LE60 declining with increasing education or in large cities.
The lack of SES gradients challenges the assumption that SES-based inequality of
mortality at older ages is similar to the inequalities documented for child mortality in
Latin America (Behm 1980) or those observed among adults in developed countries
(Mackenbach et al. 2008). Earlier Costa Rican studies had noted this lack of SES
gradients among elderly Costa Ricans (Rosero-Bixby and Dow 2009; Rosero-Bixby
and Dow 2016) as well as among the Hispanic population (which has a large Mexican
component) in the United States (Turra and Goldman 2007; Lariscy, Hummer, and
Hayward 2015). Consistent with these results, an analysis of cardiovascular risk factors
(obesity, smoking, hypertension, and diabetes) with data from CRELES and MHAS
found a weak or null association between the prevalence of these factors and education,
particularly in rural areas, where reverse gradients even occur (Hummer et al. 2014).
Another study using partial data from CRELES and MHAS found functional health
levels comparable to the United States (Payne 2015).
The paradoxically high life expectancy of older Latin American males could very
well be an expression of another paradox unveiled in this study: low-SES adults do not
endure lower life expectancy.
Additional research is needed to determine the origin of these paradoxes. Part of
the explanation could be the relatively low prevalence of obesity among older males in
the region, particularly in rural areas. For example, the prevalence of obesity (BMI ≤ 30
kg/m2) among elderly males in Costa Rica is 20%, compared to 38% in the United
States (Rosero-Bixby and Dow 2016).
A different approach from the explanation that focuses on a risk factor would be
explanations that focus on past survival to very high mortality rates in childhood and
http://www.demographic-research.org 103
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
the corresponding selection of the fittest, which would have resulted in cohorts of older
adults with gene mutations that are protective against some diseases. Examples of
promising studies that take this approach are those of polymorphisms in the enzymes
ACP1 (Gloria-Bottini et al. 2010) and G6PD (Manganelli et al. 2013) that protect
against both cardiovascular diseases and some types of cancer. The key to these
polymorphisms is that they are gender specific and more prevalent among malaria
survivors, which would explain the concentration of exceptional longevity only in
males from tropical countries. The survival-selection argument has been used to explain
the black/white crossover in old-age mortality in the United States (Manton, Poss, and
Wing 1979).
A third explanation is faulty data: age-misreporting and attrition-caused biases
from excluding healthy out-migrants or individuals who were lost because of death. As
mentioned before, age-misreporting does not exist in the CRELES data. Age errors in
MHAS, assessed from comparing DoB reports in waves of 2001 and 2012, are lower
than errors documented in census data (Preston, Elo, and Stewart 1999). Corrections of
age errors in MHAS reduce little the LE60 estimates and essentially do not change SES
gradients. Attrition of healthier out-migrants could be exaggerating some of the reverse
SES gradients found especially in Mexico. In an extreme scenario of zero-mortality of
dropouts, the contrast between education groups would cut by half in Mexican males.
Attrition of high-mortality participants may be up-biasing LE60 estimates only in a few
decimal points. Sensitivity analysis of these potential errors is included as
supplementary material.
Strengths of estimates in this report are that rates were computed using information
from the same source for the numerator (deaths) and denominator (population), which
may be crucial in studying SES gradients, and the similarity of the results from two
independent surveys. The estimates in this report also have the advantage that they did
not require assumptions about population dynamics as those required by indirect
methods. Notably, the estimates excluded the approximately six-month period
immediately following the baseline contact, thus avoiding the possibility that ill
individuals close to death were omitted from the sample.
A weakness  of  the  estimates  in  this  report  is  the  noise  from sampling  errors  that
necessitate smoothing out the age-specific rates, which was performed with a Gompertz
function. It must be noted, however, that this function is reputed to be a very good fit in
the age range studied (Bongaarts and Feeney 2002).
104 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
References
Behm, H. (1980). Socio-economic determinants of mortality in Latin America.
Population Bulletin of the United Nations 13: 1‒15.
Bongaarts, J. and Feeney, G. (2002). How long do we live? Population and
Development Review 28(1): 13‒29. doi:10.1111/j.1728-4457.2002.00013.x.
CCP (2014). Tablas de vida completas quinquenales [electronic resource]. San Pedro:
Centro Centroamericano de Población (CCP), Universidad de Costa Rica.
http://ccp.ucr.ac.cr/observa/CRindicadores/TVcompletas.html.
CELADE (2010). América Latina y el Caribe: Mortalidad. Santiago de Chile: Centro
Latinoamericano y Caribeño de Demografía (Observatorio Demográfico N. 9).
Chackiel, J. and Plaut, R. (1996). Demographic trends with emphasis on mortality. In:
Timæus, I.M., Chackiel, J., and Ruzicka, L. (eds.). Adult mortality in Latin
America. Oxford: Oxford University Press: 14‒41.
Coale, A.J. and Kisker, E.E. (1986). Mortality crossovers: Reality or bad data.
Population Studies 40(3): 389‒401. doi:10.1080/0032472031000142316.
CONAPO (2012). Documento metodológico: Proyecciones de la población de México
2010–2050 [electronic resource]. México: Consejo Nacional de Población
(CONAPO). https://www.gob.mx/cms/uploads/attachment/file/63977/Documen
to_Metodologico_Proyecciones_Mexico_2010_2050.pdf.
Dechter, A.R. and Preston, S.H. (1991). Age misreporting and its effects on adult
mortality estimates in Latin America. Population Bulletin of the United Nations
31‒32: 1‒16.
Gloria-Bottini, F., Saccucci, P., Magrini, A., and Bottini, E. (2010). Is there a role of
ACP1-ADA1 genetic complex in immune reaction? Association with T1D and
with past malarial morbidity. American Journal of the Medical Sciences 340(4):
268‒270. doi:10.1097/MAJ.0b013e3181e62ae6.
Gómez-Dantés, O., Sesma, S., Becerril, V.M., Knaul, F.M., Arreola, H., and Frenk, J.
(2011). Sistema de salud de México. Salud Pública de México 53(Suppl. 2):
S220‒S232.
Hill, K., Choi, Y., and Timæus, I.M. (2005). Unconventional approaches to mortality
estimation. Demographic Research 13(12): 281‒299. doi:10.4054/DemRes.
2005.13.12.
http://www.demographic-research.org 105
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
HMD (2017). Human Mortality Database [online database]. Berkeley and Rostock:
University of California, Berkeley, and Max Planck Institute for Demographic
Research. http://www.mortality.org.
Hosmer, D.W. and Lemeshow, S. (1999). Applied survival analysis.  New  York:  John
Wiley.
Hummer, R.A., Dondero, M., Rosero-Bixby, L., Dow, W.H., and Wong, R. (2014).
Educational attainment and older adult cardiovascular health: A comparison
between Mexico and Costa Rica.  Paper presented at the Annual Meeting of the
Population Association of America, Boston, May 1–3, 2014.
IHME (2015). GBD results tool [online database]. Seattle: Institute for Health Metrics
and Evaluation (IHME). http://ghdx.healthdata.org/gbd-results-tool.
Lariscy, J.T., Hummer, R.A., and Hayward, M.D. (2015). Hispanic older adult
mortality in the United States: New estimates and an assessment of factors
shaping the Hispanic paradox. Demography 52(1): 1‒14. doi:10.1007/s13524-
014-0357-y.
Mackenbach, J.P., Stirbu, I., Roskam, A.J.R., Schaap, M.M., Menvielle, G., Leinsalu,
M., and Kunst, A.E. (2008). Socioeconomic inequalities in health in 22
European countries. New England Journal of Medicine 358(23): 2468‒2481.
doi:10.1056/NEJMsa0707519.
Manganelli, G., Masullo, U., Passarelli, S., and Filosa, S. (2013). Glucose-6-phosphate
dehydrogenase deficiency: Disadvantages and possible benefits. Cardiovascular
and Haematological Disorders – Drug Targets 13(1): 73‒82. doi:10.2174/18715
29X11313010008.
Manton, K.G., Poss, S.S., and Wing, S. (1979). The black/white mortality crossover:
Investigation from the perspective of the components of aging. The
Gerontologist 19(3): 291‒300. doi:10.1093/geront/19.3.291.
MHAS (2004). Estudio Nacional de Salud y Envejecimiento en México (ENASEM):
Documento metodológico, reporte de proyecto, versión 2 [electronic document].
Galveston: Mexican Health and Aging Study (MHAS). http://mhasweb.
org/Resources/DOCUMENTS/2001/Methodological_Document_2001.pdf.
MHAS (2012). Data [online database]. Galveston: Mexican Health and Aging Study
(MHAS). http://www.mhasweb.org/Data.aspx.
106 http://www.demographic-research.org
Demographic Research: Volume 38, Article 3
Palloni, A., Pinto, G., and Beltrán-Sánchez, H. (2014). Latin American Mortality
Database (LAMBdA) [online database]. Madison: University of Wisconsin–
Madison. https://www.ssc.wisc.edu/cdha/latinmortality/.
Payne, C.F. (2015). Aging in the Americas: Disability-free life expectancy among
adults aged 65 and older in the United States, Costa Rica, Mexico, and Puerto
Rico. Journals of Gerontology Series B gbv076: 1‒12. doi:10.1093/geronb/
gbv076.
Peláez, E. and Acosta, L. (2011). Educación y mortalidad diferencial de adultos:
Provincia de Córdoba, República Argentina. Papeles de Población 70: 9‒31.
Pollard, J.P. (1991). Fun with Gompertz. Genus 47(1‒2): 1‒20.
Preston, S.H., Elo, I.T., and Stewart, Q. (1999). Effects of age misreporting on
mortality estimates at older ages. Population Studies 53(2): 165‒177. doi:10.108
0/00324720308075.
Ravallion, M. (2014). Income inequality in the developing world. Science 344(6186):
851‒855. doi:10.1126/science.1251875.
Rofman, R. (1994). Diferenciales de mortalidad adulta en Argentina. Notas de
Población 22(59): 73‒91.
Rosero-Bixby, L., Brenes, G., and Collado, A. (2004). Tablas de vida para cálculo
actuarial de rentas vitalicias y retiro programado: Costa Rica circa 2000.
Población y Salud en Mesoamérica 1(2): Art. 4.
Rosero-Bixby, L. and Dow, W.H. (2009). Surprising SES gradients in mortality, health
and biomarkers in a Latin American population of adults. Journal of
Gerontology Social Sciences 64(1): 105‒117. doi:10.1093/geronb/gbn004.
Rosero-Bixby, L. and Dow, W.H. (2016). Exploring why Costa Rica outperforms the
United States in life expectancy: A tale of two inequality gradients. Proceedings
of the National Academy of Sciences 113(5): 1130‒1137. doi:10.1073/pnas.
1521917112.
Sandoval, M.H. and Turra, C.M. (2015). El gradiente educativo en la mortalidad adulta
en Chile. Revista Latinoamericana de Población 17: 7‒35.
Timæus, I., Chackiel, J., and Ruzicka L. (eds.) (1996). Adult mortality in Latin
America. New York: Oxford University Press.
http://www.demographic-research.org 107
Rosero-Bixby: High life expectancy and reversed socioeconomic gradients of elderly Latin Americans
Turra, C.M. and Goldman, N. (2007). Socioeconomic differences in mortality among
US adults: Insights into the Hispanic paradox. The Journals of Gerontology
Series B: Psychological and Social Sciences 62(3): S184‒S192. doi:10.1093/
geronb/62.3.S184.
Underwood, E. (2014). A world of difference: Countries vary widely in inequality.
Science 344(6186): 820‒821. doi:10.1126/science.344.6186.820.
UNPD (1982). Model life tables for developing countries. New York: United Nations
Publications. http://www.un.org/esa/population/publications/Model_Life_Tables
/Model_Life_Tables.htm.
UNPD (2015). World Mortality Report 2015 [CD-ROM edition]. New York: United
Nations Population Division. http://www.un.org/en/development/desa/populat
ion/publications/mortality/world-mortality-cdrom-2015.shtml.
Wong, R., Michaels-Obregón, A., Palloni, A., Gutiérrez-Robledo, L.M., González-
González, C., López-Ortega, M., Téllez-Rojo, M.M., and Mendoza-Alvarado,
L.R. (2015). Progression of aging in Mexico: The Mexican Health and Aging
Study (MHAS) 2012. Salud Publica de México 57(Suppl: 1): S79‒89.
doi:10.21149/spm.v57s1.7593.
World Bank (2013). World development indicators 2013. Washington, D.C.: World
Bank. http://databank.worldbank.org/data/download/WDI-2013-ebook.pdf.
108 http://www.demographic-research.org