Why do females live longer than males?

Why Do Females Live Longer Than Males?
Jean Lemaire1
Abstract. In most countries, females live several years longer
than men, due to genetic and hormonal differences; they than males. Many biological and behavioral reasons have been benefit more from advances in medical science and economic presented in the scientific literature to explain this "female ad- vantage." A cross-sectional regression study, using 50 explana-tory variables and data collected from 169 countries, provides Behavioral differences: the lifestyle of men is damag- support to the behavioral hypothesis. Four variables, unrelated ing to their health; the FA increases as discrimination against to biological sex differences, explain over 61% of the vari- females subsides, following changes in cultural and religious ability of the life expectancy differential. One variable (the number of persons per physician) summarizes the degree ofeconomic development of a country. The three other selected Identifying the causes of the FA is critical for an accurate variables (the fertility rate, the percentage of Hindus and Bud- forecast of mortality in the 21st century. If the larger part dhists, and Europeans living in the former Soviet Union) are of the FA is the result of behavior (such as smoking, stress, social/cultural/religious variables.
exposure to AIDS, driving patterns), the FA should reduce, asbehaviors of the two genders tend to become similar in manysocieties. If the larger part of the FA is due to biological causes, Keywords:
Life expectancy, cross-country regression, fe- a significant difference will persist, barring any spectacular A vast body of literature, from many different disciplines Introduction — Literature Overview
(medicine, biology, sociology, demography, and epidemiol-ogy) addresses the issue. Excellent summaries are Waldron Mortality rates have decreased markedly in the twentieth cen- (1985) and Nathanson (1984). A comprehensive survey by tury. The gap in life expectancies between rich and poor, whites Kalben (2000), concludes that the causes of the FA supported and non-whites, educated and less educated, has narrowed sig- by evidence are (i) biological, (ii) the greater prevalence of nificantly. However, the gender gap has become wider. In most smoking among males, and (iii) the better ability of females to countries, male fetuses, infants, children, and adults, exhibit take advantage of socioeconomic and medical advances of the greater mortality. This directly affects the sex ratio of the popu- last 150 years. The theory that the FA is the result of greater lation, and social and demographic factors such as the chances male labor force participation and occupational risk is not sup- of marriage, the duration of widowhood, the stability of social ported by evidence. There is no confirmation of the widely held security systems, the construction of unisex actuarial tables, assumption that the FA will progressively disappear as women the pricing of annuities and second-to-die policies, and the achieve equality with men, particularly in employment.
valuation of pension plans. The male/female sex ratio at con-ception is estimated to range between 1.2 and 1.5 -the first and Studies supporting the biological hypothesis
in some respects the only biological advantage of the males.
One hundred years (and nine months) later, females outnumber Wingard (1982) performs a multiple logistic analysis of the males by a ratio of four to one. This results in life expectancy mortality of 6,928 adults from California, followed-up dur- differentials at birth averaging 4.51 years worldwide, with a ing nine years. The study controls death rates for sixteen maximum of 12.3 years in Belarus. Males also have higher health, social, demographic, and psychological factors: age, mortality rates in the vast majority of animal species (Rether- race, socioeconomic status, occupation, health, use of health services, smoking and alcohol use, physical activity, weight, A wide variety of variables, and interactions among them, sleeping patterns, marital status, social contacts, church and influence sex differences in mortality. Factors explaining group membership, and life satisfaction. The unadjusted ratio the "female advantage" (FA) can be broadly subdivided into of men to women mortality is 1.5. Controlling for factors such as smoking and alcohol use decreases this ratio, as more mensmoke and drink. Controlling for other factors, such as physi- Biological differences: women are biologically more fit cal activity, increases the ratio, as more women than men arephysically inactive. The adjustment for all 16 factors increases 1 Wharton School, University of Pennsylvania, Insurance and Risk Man- the mortality ratio to 1.7. So this large set of demographic and agement Department, 3641 Locust Walk, CPC 310, Philadelphia, PA19104-6218, USA, Tel: 1 (215) 898-7765, Fax: 1 (215) 898-0310, behavioral factors does not explain the FA. Other behaviors that differ among men and women, such as suicides, homi- c BELGIAN ACTUARIAL BULLETIN, Vol. 1, No. 1, 2001 cides, and fatal accidents, only account for a small proportion of both parents. Males do not have a second X chromosome of total deaths and cannot explain the FA either. Men in the to provide extra protection. If an abnormal gene turns up in a 15-24 age group exhibit excess mortality from motor vehicle male’s single X, he is at its mercy, while a female will have a accidents, but this only explains a small fraction of the overall normal gene in her extra X to counteract the defective gene.
sex mortality differential. It is concluded that an explanation Waldron (1976) provides an hormonal explanation of the of the FA needs to incorporate biologic factors. Also, inter- FA. Males produce more androgens than estrogens, while actions between biologic and behavioral factors need to be in females proportions are reversed. Androgens, particularly considered, as the impact of a given genetic factor on the FA testosterone, raise blood pressure and increase liver produc- can vary considerably according to environmental conditions.
tion of LDL, the bad cholesterol. Estrogens act on the liver Madigan and Vance (1957) study a group showing little be- to produce more immune globulin and more HDL, the good havioral differences between males and females: the teachers cholesterol. This makes the female biochemical environment and staff of Roman Catholic Brotherhoods and Sisterhoods, better able to fight bodily stresses. After menopause, the de- who lead very similar lives as regards diet, housing, work, crease in estrogen levels seems to have an immediate impact recreation, and medical care. Many sources of mortality differ- on the cardiovascular risk. The male to female ratio for my- entials, such as pregnancies, employment differences, service ocardial infraction drops from 3:1 before age 50 to less that in armed forces, hazardous leisure and work activities, do not 2:1 after. The hormonal explanation is supported by Hamilton exist in this group. Variables that could not be controlled in- and Mestler (1969), who compare the life expectancy at age clude smoking, alcohol consumption, and obesity. Over 41,000 8 of castrated and intact men. Castrated men live 10.2 years subjects were observed during a 54-year period. Because of their lifestyle, both Brothers and Sisters experience lower mor- The greater adaptability of the female body may arise from tality rates than the general population. However, sex mortality the need to adjust to the huge changes that take place during differentials are similar, and even greater after the age of 45.
menstruation, childbearing, and menopause. Graney (1979) An analysis of the causes of death shows that women may suggests that biological differences between the sexes are re- be no more resistant than men to infectious and contagious lated to their differentiated social roles. To support the intense diseases, so that the gains achieved by women this century demands of pregnancy, childbirth, and nursing, females have may be explained by a better constitutional resistance to de- developed biological resources that are available at other times generative diseases. The increasing FA may result from the as emergency reserves. Menopause promotes longer life by transition from times where infectious diseases were the main eliminating the mortality risk from childbirth. Male biologic cause of death, to modern times where death mostly results characteristics have evolved to meet long-term demands of from degenerative diseases. The disappearance of infectious hunting, shelter-building, and even combat with other males; diseases unmasks an innate male survival disadvantage from so size and muscle mass are maximized in males, leaving less reserves to combat emergencies such as acute infections.
There is substantial evidence that males have not benefited from medical advances as much as females. Graney (1979) Studies supporting the behavioral hypothesis
compares pre- and post-1946 mortality rates showing that in-fant mortality dropped drastically for both sexes after the in- There is considerable evidence that changes in smoking habits troduction of antibiotics; by far the greatest decline occurred this century have contributed to the evolution of the FA. Rether- for females. The period from 1950 to 1969 saw a decline of ford (1975) finds that the sex difference in life expectancy be- 17% of death rates due to chronic heart disease in the United tween the ages of 37 and 87 was 2.71 years for nonsmokers States, due to a 22% reduction among women, and a 7% de- and 5.13 years for smokers in 1962. He concludes that nearly cline among men. The improvement in maternal mortality, half of this difference is due to tobacco smoking, and that from 66 per 10,000 in the 1920’s to 1.5 per 10,000 in 1969, about 75% of the increase of the life expectancy difference has benefited females exclusively. Mortality from cancer of between 1910 and 1962 is due to changes in smoking habits.
the reproductive organs is lower for males, so that females However, cigarette consumption is correlated with alcoholism, are enjoying more benefits from the improvement of cancer socioeconomic status, psychological type, marital status, and detection and treatment (Retherford, 1975).
no attempt is made to control for these variables. This results Graney (1979) provides a genetic explanation of the FA.
in an overestimation of the effect of smoking.
While any X chromosome contains a large amount of genetic There is some evidence that the greater participation of men information, the Y chromosome carried by males is smaller, in the labor force, and the subsequent exposure to occupational has fewer genes, and carries less information. It has even been hazards, may contribute to the FA. Men are more likely to be suggested that a Y chromosome may act as no more than a employed in jobs exposed to carcinogens, and have a higher blank (Scheinfeld, 1958). A Y chromosome-bearing sperm rate of fatal work accidents. This can only account for at most is lighter and swims faster than its X-bearing counterpart, 5% of the male excess mortality, of which about half can be resulting in more male conceptions. However, males lack the explained by exposure to asbestos (Waldron, 1991). The effect genetic advantage of a second X chromosome. With their two of occupational hazards has decreased substantially today, as X chromosomes, females use the entire immunology system safety measures, better hygiene and reduced working hours, have improved work conditions, while most jobs with exposure to carcinogens have been eliminated. The decrease in cigaretteconsumption will further reduce any effect of occupational Expectation of Life at Birth in Sri Lanka, 1920-2000.
hazards, given the interaction between smoking and carcino- gens. Consequently, men’s employment in riskier occupations contributes very modestly to the FA.
The FA is smaller in most developing countries. A first ex- planation of this phenomenon is that some causes of death favor males (who are less vulnerable to intestinal infections and tuberculosis, for instance) while some favor females (who are less prone to die from violence and accidents). An ex- cess mortality for women will automatically result in coun- tries where the former causes are more prevalent (death fromintestinal infection is more frequent in developing societies,for instance).
Nadarajah’s comparison of causes of death in the age group A second explanation is son preference (Das Gupta and 15-44 in Sri Lanka in 1952-54 and 1970-72, summarized in Bhat, 1966). In many societies, particularly those with strong Table 2, supports the theory that women have taken more Hindu or Confucian traditions, the patriarchal family struc- advantage of medical improvements. Death rates for diseases ture and the low status of women induce a preference for and causes that affect females more (tuberculosis, pneumonia, sons over daughters. Son preference is strong in Jordan, Syria, infectious and parasitic diseases, and maternal deaths) have Bangladesh, Nepal, India, Pakistan, and only slowly fading in drastically declined during the period under study. Death rates China, South Korea, Taiwan (Arnold and Zhaoxiang, 1986).
for causes that affect males disproportionately (diseases of Females get discriminated against throughout their lives, are the circulatory system, accidents, suicide and violence) have weaned earlier than boys, have less access to education, health care, food supplies, and other goods and benefits scarce in apoor society. Reasons for son preferences are numerous, and Death Rates in the Age Group 15-44 by Sex and Cause, Sri Lanka.
summarized by the south Indian proverb "Raising a daugh- ter is like watering your neighbor’s plant." Males are valued in agricultural areas because of their larger contribution to house- hold production and the support of aging parents. Educationof female children is perceived as an investment that will shift outside the family after marriage, after payment of dowry and wedding costs. Hindu sons have to perform religious functions,such as the cremation of deceased parents.
Many other reasons have been put forward to explain the FA.
They include the loss of iron during menstruation, the tendencyof women to visit doctors more often, pressure on men not to miss work, the higher use of preventive care by women, type A behavior, the fear of men to survive their spouse, even the International Comparisons
Evolution is a fairly rapid and effective process of adaptation to changes in the environment. However, the recent increaseof the FA has been way too fast to be explained by evolutiononly; it proves the importance of social, economic, and envi- Few papers provide a comprehensive analysis of the secular ronmental influences on mortality. Historically, males tended trend in the FA. They usually report the results of a longitudinal to survive longer than females, a pattern that seems to have study, analyzing the evolution over time of the causes of death persisted from the origins of our species until well into the in a given country. International comparisons are usually de- modern era. Survival rates only began to change 150 years scriptive, analyzing sex differentials by age groups and causes ago. Around the turn of the century, the FA was small in a of death (Stolnitz, 1955, 1956, UN Secretariat, 1988.) Stud- number of countries. It has grown significantly since (Berin, ies focus on immediate medical causes of death, and do not Stolnitz, and Tenenbein, 1990). Only recently has some sta- explore the reasons for heart diseases, cancers, and violence.
bilization of the FA occurred in developed countries. Table 1, A notable exception is a regression analysis by Preston from Nadarajah (1983) and recent data, shows the evolution (1976), based on mortality data from 43 countries, most of them developed, during the period 1960-64. Preston’s conclu- tions would have given too much importance to the ten largest sions are as follows: the variable most strongly related to FA countries, with a combined population exceeding 60% of the observed by Preston is the percentage of the labor force in world’s total. Also, it was felt that small countries, like Lux- agriculture (males and females), with a correlation of -0.574.
embourg or Norway, have their own specific cultural values Variables evaluating sex differentiation in education or in the and health care systems. Using weighted correlations would labor force are poorly correlated with the FA. Stepwise regres- have amounted to disregard that information.
sion results in the selection of three variables to explain the In an independent sample, correlations exceeding 0.18 sex mortality differential: the percentage of the labor force in would be significant at the 1% level. However, any cross- agriculture, forestry, hunting, and fishing; the percentage of sectional data may be subject to some degree of spatial corre- population residing in cities of more than 1 million inhabi- lation, which would make correlation coefficients less signifi- tants; and an interaction term, the reciprocal of daily grams cant than they appear to be. The problem of spatial correlation of animal protein per capita times the percentage of males in level 1 school enrollment. All three regression coefficients are Variables have been subdivided into four categories, mea- significant at the 5% level. The square of the multiple corre- suring economic modernization, social/cultural/religious be- lation coefficient is 0.541. To date, the Preston study is the havior, health care quality, as well as geographic dummy vari- most persuasive published demonstration of the influence of ables. There is some overlap between categories. For instance, socio-environmental factors on mortality.
a decrease of infant mortality results not only from an im- In this article, we update and extend Preston’s work. We provement in health care facilities, but also from better female perform a cross-sectional study, analyzing the FA today across the world, using regression techniques. We incorporate data Given the extreme skewness of the distribution of some from 169 countries in the world, in various stages of develop- variables such as persons by car, doctor, and hospital bed, and ment. We use a much larger set of explanatory variables. We maternal mortality, variance-stabilizing logarithmic transfor- also investigate spatial autocorrelation.
mations were applied, resulting in all cases in a significantincrease of the correlation coefficient.
Variables and Correlations
Variables measuring the degree of economic modernization Data on the possible causes of the FA were collected from 169 1. Percentage of population living in urban areas (correlation countries, with a total population of 5.964 billion. The latest available data were used, 1999 or 2000 for most variables. Ad-mittedly, there can be wide variations of demographic variables Urbanization is strongly correlated with FA and with most within large countries. For instance, India exhibits striking di- measures of economic development. It is an indicator of the versity. The state of Kerala has features that are typical of a degree of economic modernization, but also a proxy for gender middle-income country: a life expectancy of 72 years, an in- bias, as discrimination occurs mainly in rural areas, due to fant mortality rate of 17 per thousand, a fertility rate (1.8 births the perceived larger value of men in an agricultural setting per woman) under replacement level, and a female/male ratio (Williamson, 1973, concludes that urbanization is the strongest above unity (1.04). In Uttar Pradesh, the infant mortality rate is six times as high as in Kerala, the fertility rate is 5.1, andthe female/male ratio stands at 0.88, lower than any country in the world (Murthi, Guio and Dr`eze, 1995.) 3. GNP per capita, converted to international dollars using The values taken by 50 potential explanatory variables have purchasing power parity rates. An international dollar has been recorded. Sources of data are the World Fact Book of the same purchasing power as a $US in the United States the Central Intelligence Agency, the Encyclopaedia Britannica Book of the Year 2000, the Food and Agriculture Organization, the United Nations, the World Bank’s Development Indicators, 5. Percentage of GDP from the services sector (0.3800) The FA, defined as the difference between the life ex- pectancy at birth of women and men, in years, is the dependent 8. Difference between daily calorie intake and requirements variable of this research. This measure of the overall sex differ- (%). The FAO determines calorie requirements per coun- ential in mortality is the most commonly used, as it summarizes try, as a function of age and sex distribution, average body mortality at all ages. It is suitable to make comparisons among weight, and temperature. The variable expresses as a per- populations with different age structures, as it is not affected centage, the difference (positive or negative) between ac- All variables are listed in this section, along with the cor- 9. Prevalence of malnutrition among children under the age relation coefficient with FA, and comments. It was decided to use unweighted correlations rather than to assign each coun- 10. Percentage of economically active females working in try a weight proportional to its population. Weighted correla- 11. Percentage of individuals who have access to safe water that education-related variables such as female literacy have a more profound influence. Always accompanying economicmodernization is a transformation of value and belief systems.
Social, cultural, religious variables Education may be the most effective agent of change in thebelief system.
12. Percentage of smokers in the male population (0.1417) Female education is considered to be crucial even if family 13. Percentage of smokers in the female population (0.4852) income is controlled. It reduces the desired family size, while 14. Difference between male and female smoking rates improving the ability to achieve the planned number of births.
The desire for a large family reduces, as educated women are 46.84% of the world’s male population smokes, versus 11.29% more likely to resist the burden of repeated pregnancies. They of the female population. Male and female smoking are uncor- have other sources of fulfillment. They are less dependent on related. There is little variability among male smoking, which their sons for old-age security. Time has a higher opportunity leads to a low correlation with FA. The positive relationship cost that reduces the value of the time-intensive activity of between female smoking and FA does not mean that smoking child bearing and education. Educated women are more likely increases life expectancy, but rather that the FA is larger in to work, which reduces fertility, due to the burden of household countries where more women smoke. Social pressures against work and employment. Child mortality reduces as educated female smoking exist in countries with a low FA. Higher cor- women are more knowledgeable about nutrition and health relations may have been obtained, had we been able to factor care. A lower fertility reduces child mortality through longer in the lag time between smoking and death. Smoking patterns have changed drastically in some countries, but this has yet It is likely that policies to improve female literacy will prove to affect mortality rates. The 32% smoking rate among males to be more efficient in reducing mortality rates than measures from Singapore represents a decrease from 74% due to legisla- aiming to change the nature of marriage systems or the in- tion introduced in the 1970s. Cigarette consumption in China grained discrimination against women. Laws outlawing dowry has increased more than three-fold since the 1950s; this is ex- or arranged marriages for minors, specifying inheritance rights pected to increase the proportion of deaths due to smoking for women, or forbidding the use of ultrasound to determine from 13% in 1987 to about 33% (Pokorski, 2000).
gender of fetuses, have generally proved to be useless in chang-ing century-old habits.
15. Illiteracy rate (%) for women above the age of 15 (-0.6288) A comparison of correlations for all education variables 16. Illiteracy rate (%) for men above 15 (-0.5895) demonstrates the importance of basic education. Enrollment 17. Difference between female and male illiteracy rates (%) figures at the second and the third level are not as related to FA and illiteracy as variables measuring education at the first 18. Enrollment ratio for women. Total school enrollment at first and second levels divided by the population of thecorresponding age groups (0.5053) 26. Females as a percentage of the labor force (0.2515) 27. Female contribution to the service industry, measured by 20. Difference between male and female enrollment ratio percentage of females in the service industry, out of work- ing females, divided by percent of GDP from the service 21. Expected number of years of education for males (0.4821) 22. Expected number of years of education for females 28. Percent of economic activity due to female labor (-0.0884) Higher levels of female labor participation reduce sex dis- 23. Females per 100 males enrolled, second level (0.4656) crimination, as they raise the status of women in society, lower 24. Females per 100 males enrolled, third level (0.2789) dowry levels, (and consequently the costs of rearing daugh- 25. Difference in school life expectancies (-0.4534) ters), make women less dependent on sons for old-age secu- In four cases (literacy, enrollment, school life expectancy, and rity, and make women more able to resist male pressure to smoking) data enabled us to compute separately correlations for men and women. In all cases links were found to be weaker 29. Fertility: number of children per childbearing woman for men. For instance, an improvement in female literacy con- tributes more to a decrease of child mortality than a male increase, as females are the main providers of care for chil- Variables related to education seem essential to understand Homicides account for a small percentage of deaths, which the FA phenomenon. While all authors agree that the "de- explains the low correlation. The divorce rate shows no cor- mographic transition" to lower levels of mortality and fer- relation with FA, despite the vast body of literature proving tility is linked to economic development, recent research that marital status strongly influences survival, with divorced (Murthi, Guio and Dr`eze, 1995) and our correlation coeffi- men appearing to be more vulnerable to the disruption of so- cients show that the income effect can be slow and weak, and cial relationships (Retherford, 1975, Trowbridge, 1994) Social ties seem to protect people against mortality, and women have Regression results
more alternative ties outside marriage that they can fall back The literature review suggests that a combination of biologic, upon. The weak relationship may be explained by the fact that social, economic, medical, and behavioral factors, and the in- many emerging countries do not report a divorce rate, and by terplay between them, can explain the FA. Selection techniques the heterogeneous cultural approaches toward divorce.
of regression analysis were applied to identify the most signif- icant variables, among the 50 variables introduced in section The selected regression model contains four variables: the logarithm of the number of persons per physician, the fertility 36. Percentage of people with indigenous beliefs, Africa rate, the percentage of people with Hindu or Buddhist beliefs, and the dummy variable representing European countries that 37. Percentage of non-religious people (0.4004) belonged to the Soviet Union. The regression equation is Religious beliefs are subject to much uncertainty. The CIA re- ports 86% of Christians in Bulgaria, Britannica 39%! Figuresare probably underestimated in communist countries, over- estimated in Western Europe, and approximated in Eastern Europe. There is no measure of intensity of beliefs, and on the degree of religious practice. Figures from Britannica ap- pear to be more reliable, and have been used here. Despiteuncertainties, significant correlations appear. Christianity and The p-values for the four selected variables are, respectively, atheism are more prevalent in countries with a high FA. Hin- 0.1591%, 0.0647%, 0.1723%, and 1.72E-11. With these four duism, Buddhism, and Islam are associated with a lower FA, variables selected, no other variable is significant, even at the probably a consequence of gender discrimination.
Variables measuring the quality of health care Three countries (Afghanistan, Bangladesh and Namibia) ex- hibit a standardized residual under -2; they are among the six 38. Health care expenses, as a proportion of GNP (0.2010) countries for which the FA is negative. Brazil, Kazakhstan, 39. Cost of health care per capita, in international dollars and Mauritania are the only three countries with a standard- ized residual exceeding 2. Brazil and Kazakhstan are the only countries out of the former Soviet Union European zone with 41. Log(persons per hospital bed) (-0.6520) 42. Percentage of children under age one immunized by vac- Several interaction terms were considered. It is for instance cination against diphtheria, pertussis and tetanus (0.3017) likely that gender discrimination has more effect when food is scarce and doctors rare. Also, female education may help 44. Log(maternal mortality rate) (-0.6170) reduce child mortality by enabling women to take better ad- 45. Index measuring the overall performance of the health care vantage of available medical facilities. Female literacy and availability of doctors and hospital beds could have a syner- The 2000 World Health Report, published by the WHO, ranks gistic effect. While several interaction terms prove to be mildly all countries according to the performance of their health care correlated with FA, none remains significant after the inclusion system. This ranking is a source of controversy, mainly because the United States spends far more per person than any other The most significant variable selected is by far the dummy country, yet only ranks 37th in health care quality. Yet, the variable characterizing the eastern European countries than index has a correlation exceeding 0.85 with life expectancy! formerly belonged to the Soviet Union. Living in one of these Our correlation coefficients show that health care quality is a six countries increases the FA by close to five years. None better predictor of the FA than health care cost.
of the 50 variables recorded in this study explains the col-lapse of the male life expectancy in these countries following the demise of communism, which has been dramatic: the FA 46. Dummy variable for the Asia and Pacific regions (-0.1517) exceeds 11 years in these six countries, and only in these countries. Alcoholism has been proposed as a major reason 48. Dummy variable for Latin America and the Caribbean for this phenomenon. Other factors commonly mentioned are rapidly declining social and economic conditions, increasing 49. Dummy variable for Europe, North America, Israel, Aus- crime and corruption, a deteriorating health care system, ra- diation produced by decades of nuclear irresponsibility, and 50. Dummy variable for the six former Soviet Union European birth defects provoked by environmental disasters.
countries: Belarus, Estonia, Latvia, Lithuania, Russia, and The next most significant variable selected is fertility, the number of births per woman. The FA increases by 0.44 years for each unit decrease of the number of children. Fertility is Conclusions
highly correlated with many variables such as female illit- The increase in the sex mortality differential during the 20th eracy (correlation = 0.866), female school enrollment ratio century has paralleled important events: huge declines in (1) (-0.8084), maternal mortality (0.8277), employment in agri- deaths from infectious and parasitic diseases, (2) the size of culture (0.7746), even female smoking (-0.6528.) It is the best the family, (3) illiteracy rates, (4) improvements in gender dis- summary measure of the transformations in beliefs and values crimination, and (5) increased urbanization. As these changes that result from increased female education and the parallel occurred simultaneously, a high degree of multicollinearity decrease of employment of women in agriculture. Fertility is between explanatory variables results, which makes it unreal- also linked to son preference. In areas where the education of istic to expect a definitive answer to the question "Is the female daughters has little value, they are likely to be married earlier advantage a consequence of biological or behavioral causes?" and start childbearing sooner. Where women have a substan- While biological differences are undeniable, the fact that tial productive role, fertility decreases, as females marry later, the sex mortality differential has changed rapidly over time is have a base of power other than reproduction, and are more an indication that they are not the sole reason for the FA. Our regression study, that incorporates data from 169 countries, at The number of persons per physician is also highly sig- various stages of development, emphasizes the importance of nificant; it is strongly correlated with many of the variables behavior, as three of the four selected variables are based on measuring economic modernization: urbanization (correlation social/cultural/religious values. Our fourth selected variable = -0.7269), number of cars (0.7486), percentage working in summarizes the degree of economic modernization of a coun- agriculture (0.8290), malnutrition (0.7842). It is also corre- try. Together, the four variables, that are totally unrelated to lated with most health care variables such as infant mortality biological differences, explain over 61% of the variability of (0.7903), maternal mortality (0.8415), and health care quality the FA, a strong support of the behavioral hypothesis.
index (-0.6974). Therefore we consider the number of persons The impact of the interactions between biological and be- per physician to be the best variable summarizing, not only of havioral factors should also not be neglected. The change of the quality of a country’s health care system, but also of its the sex mortality differential indicates that, if indeed there is an innate male survival disadvantage (for degenerative and Also highly significant is the percentage of people with heart diseases, for instance), it has a significant effect only in Hindu or Buddhist beliefs. There is a strong body of evidence combination with factors emerging in the course of socioeco- of discrimination against women in countries where these two nomic modernization. Highly differentiated death rates from religions are prevalent. The Indian Medical Association esti- cardiovascular diseases can only appear in conditions of low mates that three million female fetuses are aborted each year infant mortality and high life expectancy.
after sex-selection sonograms. Estimates of the number of Finally, it should be acknowledged that different explana- annual female infanticides in India reach several million. Mid- tions may account for the FA at different ages: differentials wives in India earn the equivalent of $0.50 and a sack of grain in infants may be primarily biological, while social/behavioral for each live delivery of a girl, twice as much plus a sari if reasons may explain more of the FA in adults.
it is a boy - and $5 to get rid of a newborn female (WallStreet Journal, May 9, 2000.) A comparative study of maleand female mortality during childhood by D’Souza and Chen ACKNOWLEDGEMENTS
(1980) shows abnormally high death rates among girls living This work was supported by an unrestricted grant to the in Bangladesh rural areas, between the ages of one month and Leonard David Institute of Health Economics provided by the 15 years. Some countries officially recognize son preference; Merck Company. Many thanks to Narumon Saardchom for in several Chinese provinces, couples accepting a one-child numerous suggestions, and to Jianhua Huang for outstanding certificate are compensated by a larger monetary bonus if their child is a daughter; in some rural areas, couples can have twochildren if the first is a girl. The maximum penalty for female REFERENCES
infanticide is China is only 13 years of prison (Arnold andZhaoxiang, 1986.) [1] L. Anselin and A. K. Bera (1998). "Spatial dependence in Linear Regres- sion Models with an Introduction to Spatial Econometrics." In Hand- The multiple correlation coefficient is 0.7815: over 61% of book of Applied Economic Statistics, A. Ullah and D.E.A. Giles, Editors, the variability of the FA can be explained by the four selected variables (this is significantly higher than Preston’s 54.1%).
[2] F. Arnold and L. Zhaoxiang (1986). "Sex Preference, Fertility, and Family Planning in China." Population and Development Review, 12, Therefore this study provides support to the hypothesis that be- havior strongly influences differential mortality between males [3] B.N. Berin, G.J. Stolnitz and A. Tenenbein (1990). "Mortality Trends and females. Three of the selected variables are mostly deter- of Males and Females over the Ages". Transactions of the Society ofActuaries, 41, 9-32.
mined by beliefs and family values. The fourth variable is [4] M. Das Gupta and P.N Mari Bhat (1966). "Intensified Gender Bias in India: A Consequence of Fertility Decline." Annual Meeting of thePopulation Association of America.
[5] S. D’Souza and L.C. Chen (1980). "Sex Differentials in Mortality in Rural Bangladesh." Population and Development Review, 6, 257-270.
[6] M.J. Graney (1979). "An Exploration of Social Factors Influencing the relation, until recently. A recent review by Anselin and Bera Sex Differential in Mortality." Social Symposium, 28, 1-26.
(1998) points to Paelinck and Klaasen (1979) as the pioneer- [7] J.B. Hamilton and G.E. Mestler (1969) "Mortality and Survival: Com- parison of Eunuchs with Intact Men and Women in a Mentally Retarded ing approach to spatial econometrics. Still, the most widely Population". Journal of Gerontology, 24, 395-411.
available statistical software do not emphasize spatial prob- [8] B.B. Kalben (2000). "Why Men Die Younger: Causes of Mortality lems; they contain no specific routines to perform maximum Differences by Sex." North American Actuarial Journal [9] F.C. Madigan and R.B. Vance (1957). "Differential Sex Mortality: A likelihood estimation of spatial processes or specific tests for Research Design." Social Forces, 35, 193-199.
the presence of spatial autocorrelation. S+ is among the few [10] M. Murthi, A-C. Guio and J. Dr`eze (1995). "Mortality, Fertility, and packages that have a spatial module, SpatialStats.
Gender Bias in India: A District-Level Analysis." Population and De-velopment Review, 21, 745-780.
Spatial autocorrelation between errors in adjoining loca- [11] T. Nadarajah (1983). "The Transition from Higher Female to Higher tions is the result of a mismatch between the spatial unit of Male Mortality in Sri Lanka." Population and Development Review, 9, observation (the country, in this case) and the spatial extent of [12] C.A. Nathanson (1984). "Sex Differences in Mortality." Annual Reviews the variable under study (son preference, for instance.) High values for a random variable tend to cluster in space: loca- [13] J. Paelinck and L. Klaasen (1979). "Spatial Econometrics." Saxon tions tend to be surrounded by neighbors with similar values.
[14] R. J. Pokorski (2000). "Excess Mortality in Asia Associated with As a result, the sample contains less information than an un- Cigarette Smoking." North American Actuarial Journal, 4, 101-113.
correlated counterpart. This loss of information needs to be [15] S.H. Preston (1974). "Mortality Patterns in National Populations". Aca- explicitly acknowledged in estimation and tests.
[16] R.D. Retherford (1975). "The Changing Sex Differential in Mortality".
A crucial issue in modelling spatial correlation lies in the Greenwood Press, Westport, Connecticut.
specification of "locational similarity," the determination of [17] A. Scheinfeld (1958). "The Mortality of Men and Women". Scientific those locations for which values of random variables are cor- [18] G.J. Stolnitz (1955, 1956). "A Century of International Mortality related. Such locations are referred to as "neighbors," even Trends" Population Studies, IX, 24-55, and X, 17-42.
though this does not necessarily mean than they are physically [19] W. Tobler (1979). "Cellular Geography" In Philosophy in Geography, adjacent. Since, in a set of n observations, it is impossible S. Gale and G. Olsson, Editors, Reidel, Dordrecht, 379-386.
[20] C. Trowbridge (1994) "Mortality Rates by Marital Status." Transactions to estimate n × n covariance terms, the structure of spatial of the Society of Actuaries, XLVI, 321-344.
dependence has to be specified through an exogenous model.
[21] UN Secretariat (1988). "Sex Differentials in Life Expectancy and Mor- This is usually achieved through the construction of an n by tality in Developed Countries: An Analysis by Age Groups and Causesfrom Recent and Historical Data." Population Bulletin of the United n positive and symmetric matrix W which expresses for each observation (row) the locations (columns) that are neighbors.
[22] I. Waldron (1976). "Why Do Women Live Longer Than Men?" Social Science and Medicine, 10, 349-362.
ij = 1 when i and j are neighbors, wij = 0 [23] I. Waldron (1985). "What Do We Know about Causes of Sex Differences otherwise. To facilitate interpretation, this weights matrix is in Mortality?" Population Bulletin of the United Nations, 18,59-76.
often standardized so that the elements of a row add up to one: [24] I. Waldron (1991). "Effects of Labor Force Participation on Sex Differ- ences in Mortality and Morbidity." In Women, Work, and Health, edited by M. Frankenhaeuser, U. Lundberg, and M. Chesney, Plenum Press, [25] N.E. Williamson (1973). "Preferences for Sons Around the World." The selection of the elements that are nonzero in W is a Ph.D. dissertation, Department of Sociology, Harvard University.
[26] D.L. Wingard (1982). "The Sex Differential in Mortality Rates. Demo- matter of considerable arbitrariness. The traditional approach, graphic and Behavioral Factors." American Journal of Epidemiology, based on geography only, designates observations as neighbors if they have a physical border in common. Another approachconsists in using distances: wij = 1 if the distance between APPENDIX
the observations dij < δ, where δ is a cutoff value. Othersuggestions are wij = 1/dα Spatial Dependence in Regression Models
ij , where bij is the share of the common border Regression analysis is based on the assumption that errors between i and j in the perimeter of i. In economic applica- are independently distributed. In a cross-sectional study, er- tions, weights are sometimes based on selected socioeconomic rors between neighboring countries are often spatially corre- characteristics such as per capita income. In social sciences, lated. This may result from the influence of unobserved, or weights may reflect whether or not two individuals belong to unobservable, variables that exhibit spatial dependence. For the same social network. The weights need to be specified instance, unobserved cultural factors may lead Pakistan, In- exogenously, to express any measure of potential interaction dia and Bangladesh to common approaches to gender bias or fertility. Or unobserved dietary similarities between Latin Once matrix W has been built, the spatial dependence is American countries may impact mortality. As summarized by incorporated into a linear regression model with k explanatory Tobler’s (1979) "first law of geography," "everything is related to everything else, but closer things more so." While there is a voluminous literature on serial dependence over time in the analysis of time series, little attention has been where y is the n by 1 vector of observations of the depen- paid to its counterpart in cross-sectional data, spatial autocor- dent variable, X is a n by k matrix of observations of the explanatory variables, ε is a n by 1 vector of error terms, β is and H0 : λ = 0 are presented. Numerical procedures are a k by 1 vector of regression coefficients, and ρ is the spatial usually complicated, as many simplifying results from serial autoregressive parameter; it measures true contagion between correlation in time series do not hold in the case of spatial cor- countries such as diffusion of beliefs. The product W y results relation. Estimation techniques require nonlinear optimization in a weighted average of the y values in their neighborhood of the likelihood function, and the manipulation of matrices of set. This concept of a spatial lag operator is similar to the in- dimension equal to the number of observations.
clusion of an autoregressive term for a dependent variable in a In order to test for the presence of spatial autocorrelation in time-series analysis. The spatial lag for a given observation is our final regression equation, a simple neighbors matrix W was always correlated, not only with its error, but also with error built, in which entries are 1 when two countries are physical terms at all other locations. Each country is correlated with ev- neighbors, and 0 otherwise. It was felt that a more sophisticated ery other country, in a relationship that decays with the order approach was not necessary: the spread of unobserved cultural values across countries was not deemed to depend much on This model for spatial lag dependence can be re-formulated distance between capitals or length of the border. The most common test for the assumption H0 : λ = 0, Moran’s test,the spatial equivalent of Durbin Watson’s test, was applied to the residuals of the selected regression equation. This test ispowerful against a wide range of forms of spatial dependence.
(I − ρW )y is called a spatially filtered dependent variable: the effect of spatial autocorrelation has been taken out. This isroughly similar to the process of first differencing a dependent An alternate way to incorporate spatial autocorrelation in regression is to specify a spatial process for the error term.
where e is the vector of OLS residuals, n the number of obser- vations, and S0 a standardization factor equal to the sum of the spatial correlation among residuals, I is asymptotically nor- mally distributed with known first two moments (see Anselin Moran’s statistic, the estimated value of the spatial correla- where λ is the spatial autoregressive coefficient for the error lag tion coefficient, was 0.163, indicating a modest but significant W ε, and ξ is an uncorrelated error term. λ is often called "the degree of unexplained correlation across neighbors.
nuisance parameter," reflecting the interpretation of spatial Matrix W was then standardized so that the sum of its coef- error dependence as a nuisance, resulting from correlation in ficients along each row adds up to 1. This approach may better measurement errors or in variables that are not crucial to the reflect the contagion of values and beliefs across countries: the model (the unobserved variables spillover across countries).
more neighbors a country has, the smaller the spillover of cul- A consequence of correlation among error terms is that or- ture across the border. China has more influence on Mongolia dinary least-squares estimators are biased and inconsistent; a than Mongolia on China. Moran’s revised spatial correlation maximum likelihood approach is necessary. The traditional for this modified matrix is hardly changed, at 0.166.
R2 is not an appropriate measure of fit in the presence of spa- It is concluded that a modest spatial correlation of errors is tial autocorrelation. Models can be compared using the maxi- present in our data, given the four selected variables. Conse- mized log-likelihood, or an adjusted form such as the Akaike quently, the p-values for the four variables might be slightly Information Criterion, that takes into account the number of understated, and the multiple correlation coefficient might be parameters in the models. Anselin and Bera (1998) provide a misleading. Nevertheless, the four selected variables are so sig- summary of estimation techniques and tests for spatial depen- nificant that our conclusions are very unlikely to be challenged dence. In particular, tests for the null hypotheses H0 : ρ = 0 by a more sophisticated maximum log-likelihood approach.

Source: http://www.belgianactuarialbulletin.be/articles/vol01/02-Lemaire.pdf