Sot. Sci & Med,. Vol. 12C. pp. 89 to 98. Pergamon Press Ltd, 1978. Printed in Great Britain
S O C I O E C O N O M I C D E T E R M I N A N T S O F FETAL A N D C H I L D DEATHS IN LATIN AMERICA: A COMPARATIVE
STUDY OF BOGOTA, CARACAS AND RIO DE JANEIRO
MANUEL CARVAJAL and
P A U L BURGESS
Nova University, Fort Lauderdale, Florida
Abstract--This study uses data from the Urban Fertility Studies sponsored in Bogota, Caracas and Rio de Janeiro by the United Nations Center for Latin American Demography to test the direction and magnitude of the effect of five socioeconomic variables on fetal and child mortality. The five independent variables in the ordinary-least-squares model are: (I) household income; (2) mother's education; (3) mother's participation in the labor force; (4) migration; and (5) incidence of consensual union. Surveyed women in each city are divided into six five-year age groups ranging from 20-24 to 45-49 years of age. The empirical results indicate that the explanatory variables do explain a substantial portion of the cross-sectional variation in both fetal and child mortality.
The last decade has witnessed a growing enthusiasm for emphasizing, theoretically as well as empirically, mechanisms of interaction between parental environment and observed family size. The microeconomic theory of fertility associated with the early work of Becker [1] and Mincer [2], subsequently extended by Ben-Porath [3] and Willis [4], is essentially an elaboration of the concept of human capital. More recently Blandy [5], Leibenstein [6], and Carvajal and Geithman [7] have advanced alternative approaches to strengthen the understanding of the nature and scope of human fertility differentials. While the human capital approach implicitly holds that every couple is free to choose, within the limits set by its income and time constraints, any combination it wishes of number of children, expenditures on child services, and expenditures on nonchild goods and services, the more recent theoretical formulations stress that, insofar as the couple's tastes and preferences are partly determined by the socioeconomic reference group with which the couple identifies itself, consumption expenditures on children as well as nonchild goods and services are actually dictated by peer group standards, thus transforming consumption patterns from the category of purely voluntary activities into almost obligatory behaviour. Furthermore, since children commonly share their parents' socioeconomic reference group, the willingness to share on the part of parents is viewed as a function of the degree of parental psychic interdependence of utilities toward their offspring. Surprisingly, research on the socioeconomic differentials of fetal and child mortality has been limited, at best, to the impact on observed family size of anticipated offspring survival rate. On the premise that realistic analysis of family size decision making must be as comprehensive as possible, and expected surviving progeny having been established as a primary determinant in the process of family formation [8], more knowledge of the correlates and frequency with s.s.~. 1 2 - - 3 4 c c
which fetal and child deaths occur has the potential of shedding further light into the human reproductive process. The ultimate indicator of the value of this type of analysis should lie not in its theoretical elegance but in its judgments about which forces are most central, and which are secondary or tertiary, in explanatory power. While severe limitations imposed by the theorist may not damage a model's ability to explain observed data, great theoretical elegance will not save a model's basic unfitness to adjust to empirical evidence. One factor that may have contributed to limit research on the socioeconomic determinants of fetal mortality is the difficulty of obtaining reliable data for the early months of pregnancy. Not only hospital records are often incomplete, but in many cases women are unaware of pregnancy wastage or reluctant to disclose information, especially if pregnancy occurs because of premarital or extramarital contact. P U R P O S E OF THE S T U D Y
Traditionally, prenatal and neonatal deaths have been treated as induced almost exclusively by intrauterine or endogenous factors such as congenital malformations or conditions of the delivery [9]. In other words, genetic or developmental variables have preponderated in explaining fetal and neonatal mortality differentials. Recent evidence seems to indicate, however, that environmental factors do play a significant role in determining variations in fetal and child deaths [10]. This paper seeks to: (1) identify several socioeconomic variables that are expected to generate fetal and child mortality differentials; (2) hypothesize the direction and magnitude of the effect of each variable; and (3) test empirically the model using data from the Urban Fertility Surveys sponsored in Bogota, Caracas and Rio de Janeiro by the United Nations Center for Latin American Demography (CELADE) during the 1960s [11].
90
MANUEL CARVAJALand PAUL BURGESS
Bogota, Caracas and Rio de Janeiro have been METHODOLOGY selected as case studies for two reasons: (1) they rank among the most rapidly developing cities in Latin To measure and evaluate the degree to which America during the last few decades; and (2) each socioeconomic variables influence fetal and infant city represents a distinct philosophy with respect to mortality, surveyed women in each city are divided governmental attitude toward family planning and into six five-year age groups ranging from 20-24 to maternal-child health. Since the establishment of the 45-49 years of age. The Urban Fertility Surveys confirst family planning and maternal-health clinic in sist of 2259 women in Bogota, 2087 in Caracas, and Caracas in 1962, the Venezuelan birth-control move- 2512 in Rio de Janeiro. Of these, only legally or conment has received widespread acceptance by both sensually ever-married women who have experienced public and private sectors, including the Catholic at least one live birth are included in the analysis, Church hierarchy. As early as 1966 a Population each woman representing a household. Single women Division was created within the Ministry of Public are excluded because the surveys do not contain inHealth, whose function was to integrate all popula- formation about their observed fertility; ever-married tion activities into the national health effort. Thus, women who have not experienced at .least one live the provision of family planning and maternal--child birth also are excluded since their probability of health services has experienced a smooth evolution reporting child deaths is zero. Table I presents the process in this country [12]. classification of observation units by type of woman. In Colombia, however, governmental attempts to city and age group. Ever-married women with at least curb growing population pressures initially met resist- one live birth represent approximately 75, 72 and 69~o ance by powerful interest groups such as the Catholic of all women interviewed in Bogota, Caracas and Rio Church, the extreme right, and the extreme left. As de Janeiro, respectively, while single women account a result, the government acted with extreme caution for 19, 16 and 19~o of the observation units in each in its involvement with family planning activities dur- city. Ever-married women who reported no live births ing the 1960s. It was not until 1969 (i.e. after the constitute the remaining 6, 12 and 12~o of the CELADE survey was taken) that a Department of samples. In interpreting the empirical results of the Maternal Child Health Services was founded under subsequent statistical analysis, the reader must bear the auspices of the Ministry of Health. In the 1970s in mind that such results may be biased insofar as the Colombian family planning program has become they are not necessarily representative of a sizable more firmly established, receiving a substantial por- portion of the population. In other words, the statistition of its operational budget from the public sector. cal inferences are valid only for ever-married women Political controversy has shifted away from popula- who have experienced at least one five birth. tion control into abortion and divorce [13]. The rationale for classifying women into five-year In Brazil the situation is altogether different, as the age groups lies in the sharp increase in fetal and ingovernment has traditionally favored a rapid popula- fant deaths reported by various studies [15] from one tion growth policy in light of vast areas that remain age group of mothers to the next. Contrary to the unpopulated or underpopulated throughout the findings of these studies, fetal mortality rates per 1000 country. At the time the CELADE survey was con- pregnancies calculated from the Urban Fertility Surducted in Rio de Janeiro, the government was ada- veys in all three cities tend to decrease as the age mantly opposed to all family planning activities, in- of the mother increases (see Table 2). This rather uncluding those of the private sector. Subsequent to the orthodox finding probably can be partly attributed Bucharest Conference of 1974, however, the official to the nature of the data, since women are classified policy has been revised insofar as family planning and into age groups according to their age at the time maternal-child health services are provided to low- of the interview and not at the time when they experiincome families who voluntarily request such services enced pregnancy, fetal death, live birth or child death. [143. But even if it is assumed that most pregnancies occur
Table 1. Distribution of observation units by type of woman, city. and age group Type of woman and age group Ever-married with at least one live birth 20-24 25-29 30-34 35-39 4(I-44 45~49 Single Ever-married with no live'births
Bogota
Number of observations Caracas Rio de Janeiro
1688 262 380 371 313 205 157 433
1505 210 345 314 267 190 179 335
1741 204 323 376 340 259 239 48 l
138
247
290
Determinants of fetal and child deaths in Latin America
91
Table 2. Fetal and infant mortality rates by city and age group
Age group 20-24 25-29 30-34 35-39 40-44 45-49 All age groups
Fetal mortality (per 1000 pregnancies) Rio de Bogota Caracas Janeiro
Bogota
Caracas
Rio de Janeiro
195.6 125.9 91.6 82.9 76.3 62.8
165.8 121.7 94.2 82.1 79.3 79.2
178.6 178.2 142.6 149.8 124.7 147.2
71.4 61.9 74.3 63. I 100.7 116.4
59.4 43.0 49.7 51.5 80.4 101.1
81.5 56.7 86.6 91.0 120.9 156.0
109.3
104.7
152.9
76.1
59.8
95.9
when women are between 20 and 29 years old, the data seem to suggest that the probability of a fetal death declines with age. Table 2 also presents child mortality rates per 1000 live births for Bogota, Caracas and Rio de Janeiro. The child mortality rate of younger women is considerably lower than their fetal mortality rate. The values for the child mortality rate are lowest for the 25 29 age group and tend to increase with age. Unfortunately, the child mortality rate computed here includes in its numerator all offspring deaths occurring throughout the mother's reproductive history. Thus, the child mortality rate is upwardly biased for older women, say 35 years of age and older, although it is mitigated by the fact that almost one-half of all deaths generally occur among children less than I0 years old, while less than 10% of total deaths occur in the population 10-34 years of age [16]. Overall, Caracas records the lowest fetal and child mortality rates, followed by Bogota and Rio de Janeiro; it is worth noting that these mortality differentials correlate with the attitude on the part of the government of each country toward supporting and/or implementing family planning programs. Table 3 shows a somewhat different mortality indicator, namely, the average number of fetal and child deaths per woman. This is the indicator which will be used in the subsequent statistical analysis. In Bogota and Caracas the average number of fetal
Child mortality (per 1000 live births)
deaths tends to decrease with age, while the opposite occurs in Rio de Janeiro. The average number of child deaths per woman increases with age in all three cities. VARIABLES AFFECTING FETAL AND CHILD MORTALITY The decline in mortality (including fetal and child mortality) experienced in developing countries since World War II is probably a reflection of the progress made by medical science in combating the infectious causes of death, essentially as a result of the impact of external medical help. It is logical to expect that people in higher-income brackets are more likely to benefit from the availability of increasingly better mother and child health care facilities in these countries than are people in lower-income groups. Income also tends to vary positively with the levels of sanitation and nutrition, which in turn vary inversely with morbidity. To measure its impact on fetal and child mortality differentials the level of household income is approximated in this study by a discrete variable measuring household expenditure. This variable measures along a 0-9 discrete scale and is used for standardization purposes, since the currency unit varies from city to city. In order to obtain values for the income scale, households are distributed in deciles ranking from
Table 3. Average number and standard deviation of fetal and child deaths per woman, by city and age group
Age group 20-24 25-29 30-34 35-39 40-44 45~19 All age groups
Fetal deaths per woman Rio de Bogota Caracas Janeiro
Child deaths per woman Rio de Bogota Caracas Janeiro
0.481 (0.742) 0.471 (0.918) 0.439 (0.925) 0.444 (0.913) 0.424 (0.939) 0.318 (0.831)
0.481 (0.764) 0.429 (0.864) 0.439 (0.9191 0.423 (0.912) 0.395 (0.902) 0.380 (0.890)
0.392 (0.533) 0.533 (0.863) 0.521 (1.127) 0.559 (1.168) . 0.541 (1.078) 0.611 (1.317)
0.141 (0.433) 0.203 (0.594) 0.323 (0.705) 0.310 (0.797) 0.517 (0.973) 0.529 (1.036)
0.138 (0.3961 0.133 (0.428) 0.210 (0.621) 0.243 (0.657) 0.368 (0.739) 0.447 (0.870)
0.147 (0.384) 0.139 (0.427) 0.271 (0.751) 0.288 (0.788) 0.459 (1.027) 0.552 (1.040)
0.441
0.425
0.531
0.308
0.237
0.302
92
MANUEL CARVAJAL and PAUL BURGESS
lowest to highest tenth. Households in the lowest decile are assigned a value of zero, those in the nextto-lowest decile are assigned a value of one, and so forth; households in the highest income decile are assigned a value of 9. In all three cities the average level of household income increases with age. Education also seems to be a factor influencing variations in fetal and child mortality. As Sauvy points out, "...child mortality varies inversely with education. Ignorance is more deadly than poverty" [17]. A higher level of educational achievement presumably expands parental knowledge of more efficient techniques to care for their fetuses and newborns. Education is measured in this study by the number of years of schooling completed by the mother. Women in Rio de Janeiro seem to possess consistently higher levels of educational attainment than women in the other two cities. In addition, except for women in the youngest age group, Bogota women are more educated than women in Caracas. Education generally decreases with age in all three cities. Female labor force participation is the third variable hypothesized to exert a systematic influence on fetal and child mortality differentials in Bogota, Caracas and Rio de Janeiro. Bearing and rearing children generally are labor-intensive activities on the part of the mother. More time devoted to activities competitive with caring for children, such as working outside the home, may increase the incidence of fetal or child death by virtue of an increase in the probability of accident occurrence. The data for the Urban Fertility Surveys indicate that female labor force participation in all three cities tends to increase as age rises, with Bogota recording the highest participation rates, followed by Caracas and Rio de Janeiro. Another socioeconomic variable that may explain differentials in the incidence of fetal and child death pertains to the household's migration history. Migration is measured in this study with a dummy variable recording whether or not the mother was born in the metropolitan area where the survey took place. The percentage of mothers migrating from other areas increases with age, thus reflecting the fact that older people have had a longer period of time to respond to migration opportunities. Less than one-fourth of all households in Bogota and Caracas and less than one-half in Rio de Janeiro are nonmigrant, which implies the existence of a considerable influx of people in all three cities. The fifth and last socioeconomic force hypothesized to exert a systematic influence on fetal and child mortality differentials is the presence of consensual marriages or de facto unions. If the prospect for pregnancies resulting from a de facto union is illegitimacy for the child, and if illegitimacy is negatively sanctioned by society, such pregnancies are more likely to end in induced abortion or stillbirth than they would otherwise. Similarly, to the extent that consensual marriages generally do not last as long as legal marriages, children born from de facto unions may be more prone to die at an early age than other children because of lack of appropriate care. The Urban Fertility Survey data suggest that consensual unions are rather common in Caracas, where almost one out of every five women in the sample lives in
concubinage, and slightly more common in Rio de Janeiro than in Bogota, with levels of 7.4 and 5.7%, respectively. In all three cities the incidence of de facto marriages declines with age. STATISTICALMODEL The statistical model consists of 36 equations (3 cities, 6 age groups and 2 dependent variables). In the first 18 equations fetal deaths appear as the dependent variable, while the other 18 equations attempt to explain variations in child deaths. The empirical models of fetal and child mortality that are tested here within the hmitations of the data provided by the Urban Fertility Surveys interpret fetal and child deaths as linear functions of five independent variables. Each of these mortality determinants has been discussed in the preceding section: household income, mother's education, mother's participation in the labor force, migration, and incidence of consensual marriage. Thus,
Mfij = aoij + alijYij + a2ijEij + a3ijLij + a4ijGij + a5ijCij + Vfij and
Mcij = bolj + blijYij + b2ijEij + baoLij + b4ijGij + bsijCij + [/cij where
Mflj Mci j Y~j
E~j Lit
Gij -
Cij
Vy~jand
V~j
a0~, .... as~
is the number of fetal deaths reported by cach woman in the jth age group and the ith city; is the number of child deaths reported by each woman in the jth age group and the ith city; is a 0-9 discrete income scale for each household in the jth age group and the ith city; is the number of years of formal education completed by each woman in the jth age group and the ith city; is a dummy variable for female participation in the labor force, assigned a value of 1 if the woman in the jth age group and the ith city is a member of the labor force, a value of 0 otherwise; is a dummy variable for migration, assigned a value of 1 if the woman in the jth age group and the ith city was born and raised in the ith city, a value of 0 otherwise; is a dummy variable for consensual marriage, assigned a value of 1 if the woman in the jth age group and the ith city is consensually married, a value of 0 otherwise; are normally, independently distributed stochastic disturbance terms for the jth age groups and the ith city when fetal deaths and child deaths, respectively, are the dependent variable; are the least-squares coefficients to be estimated for the jth age group and the ith city in the set of equations explaining variations in fetal deaths;
Determinants of fetal and child deaths in Latin America
boij..... bsi #
are the least-squares coefficients to be e s t i m a t e d for the j t h age g r o u p and the ith city in the set o f e q u a t i o n s explaining variations in child d e a t h s ;
and where i = 1, 2, 3
for Bogota, Caracas and Rio de
j = 1..... 6
Janeiro; and for the six age groups.
EMPIRICAL
RESULTS
This section presents and discusses least-squares estimates for the 36 equations. First, the estimated coefficients for fetal mortality are analyzed, followed by a discussion of estimated coefficients for child mortality. The standard deviations of the estimates are
shown in parentheses, with the levels of significance denoted as follows: * * = 0.99 level, * = 0.95 level, t t = 0.90 level, a n d t = 0.80 level.
Fetal mortality
City and age group Bogota 20-24
Intercept
Income
0.663
- 0.030tt
(O.Ol7) 25-29
0.735
30-34
0.755
35-39
0.788
40-44
0.799
45-49
0.628
Caracas 20-24
0.779
- 0.028t¢ (0.016) -0.025? (0.016) - 0.027t (0.019) --0.026f (0.017) -0.023? (0.017) - 0.066**
(0.020) 25-29
0.833
-0.046** (0.014) - 0.045 (0.O2O) - O.033tt (0.018) - 0.031 t¢
30-34
0.895
35-39
0.819
40-44
0.806
45-49
0.793
-- 0.031 t (0.021)
0.324
- 0.037* (0.018) - 0.035* (0.016) --0.034?t
(0.o18)
Rio de Janeiro 20-24 25-29
0.546
30-34
0.576
35-39
0.640
40-44
0.645
45-49
0.784
(0.020) - 0.026?t (0.014) --O.021t (0.014) - 0.024¢ (0.014)
93
Discussion Income. All i n c o m e coefficients are statistically significant at least at the 0.80 level a n d s h o w a negative sign, which lends s u p p o r t to the hypothesis that inc o m e varies inversely with fetal mortality by virtue of: (1) p r e v e n t i n g l o w e r - i n c o m e m o t h e r s from seeking a d e q u a t e medical assistance when they are p r e g n a n t ; (2) lack of s a n i t a t i o n facilities a m o n g l o w e r - i n c o m e earners, which translates into higher incidence of m o r b i d i t y ; a n d / o r (3) m a l n u t r i t i o n , which occurs m o r e often in lower- t h a n h i g h e r - i n c o m e brackets. T h e negative sign of the i n c o m e coefficients accords with findings that l o w e r - i n c o m e w o m e n tend to see physicians less often t h a n h i g h e r - i n c o m e w o m e n , even if p r e n a t a l care is available w i t h o u t charge [18]. It seems that p o o r people are c o n s t r i c t e d in their ability to use c o m m u n i t y health a n d welfare services, a n d thus d o n o t benefit as m u c h as they could from scientific a n d technical d e v e l o p m e n t s which m a y increase their incidence o f live births. T h e levels of significance of the i n c o m e coefficients reveal two definite oatterns. First, the level o f statisti-
Fetal Mortality Regression Coefficients Female labor force Education participation Migration
Consensual union
-0.019"* (0.005) - 0.030~t (0.015) - 0.034"* (0.006) -0.038** (0.008) - 0.043"* (0.012) - 0.043 (0.010)
-0.157"* (0.053) - 0.177"* (0.057) - O.188 ** (0.050) -0.195"* (0.064) - 0.205* (0.087) - 0.035 (0.090)
0.199"* (0.054) 0.155** (0.057) 0.090¢ (0.068) 0.110¢ (0.072) 0.119it (0.067) 0.104t (0.063)
0.267** (0.092) 0.345"* (0.070) 0.649 ** (0:144) 0.618"* (0.128) 0.821 ** (0.221 ) 0.883" * (0.281 )
-0.021" (0.009) - 0.032** (0.011) -0.037* (0.010) -0.042** (0.014) - 0.047* (0.020) -0.050* (0.022)
-0.195" (0.093) - 0.227** (0.054) -0.212** (0.081) -0.216** (0.067) - 0.246* (0.099) -0.232* (0.102)
0.113+ (0.070) 0.092t (0.058) O.Ol 5 (0.092) 0.011 (0.090) 0.085? (0.053) 0.076? (0.050)
0.222* (0.090) 0.217* (0.095) 0.217t¢ (0.128) 0.258?? (0.155) 0.243¢? (0.125) 0.204* (0.100)
-0.008t (0.005) --0.012" (0.005) -0.015it (0.008) -0.013t (0.009) -O.O17?t (0.010) - 0.022? (0.014)
- 0.125?? (0.067) -0.118t (0.073) -0.140t+ (0.082) -0.152t (0.095) -O.140t (0.098) - O.148t (0.094)
0.473** (0.138) 0.399** (0.111) 0.338** (0.121 ) 0.253* (0.101) 0.152?? (0.085) O.167* (0.079)
0.258* (0.084) 0.280* (0.100) 0.467** (0.161 ) 0.428* (0.214) 0.613"* (0.195) 0.681 * (0.307)
94
MANUEL CARVAJAL and PAUL BURGESS
cal significance in all three cities is higher for younger than for older women, which implies that the negative effect of income on fetal mortality is better specified for age groups in the process of family formation. Second, the levels of significance tend to be highest for Caracas and lowest for Bogota. The values of the least-squares coefficients depend on the variables' unit of measurement and on the range of the variables. As such, the least-squares coefficients are not strictly comparable among cities or age groups. A unit of standardization, such as the elasticity evaluated at the mean of the variables, is needed to make the empirical results truly comparable. Elasticity is defined as the ratio of a percentage change in the dependent variable to an infinitesimal percentage change in the independent variable. Specifically, income elasticity of fetal mortality refers to the limit of the ratio of the percentage change in observed fetal deaths to a percentage change in income [19]. These elasticities (see Table 4) show, for example, that in Bogota a 10~o increase in the income scale reduces fetal deaths in age group 20-24 by approximately 2.2~o. An analysis of Table 4 reveals that fetal deaths in all age groups for the three cities are relatively income inelastic; i.e. the percentage decline in fetal deaths resulting from a given percentage increase in income is well below unity. Fetal deaths in Bogota tend to become more income elastic as the age of the mother increases, while in Caracas and Rio de Janeiro there seems to exist an inverse association between mother's age and the income elasticity of fetal deaths. In general, these elasticities show their highest values for Caracas, where the provision of family planning and maternal-child health services has not been at all a controversial issue+ and lowest values for Rio de Janeiro, where attempts to provide such services have traditionally met stern governmental opposition. Education. The negative sign of the education coefficients support the contention that more educated mothers experience lower levels of fetal mortality through engaging in healthier prenatal practices and reaching out more often for medical help than do less educated mothers. Although all 18 coefficients are significantly different from zero with a probability of at least 80°r/o, the levels of statistical significance for Bogota and Caracas are generally higher than for Rio de Janeiro, therefore suggesting that the impact of education on fetal mortality is better specified for the two cities in which family planning and maternalchild health services were provided during the 1960s.
The education elasticities of fetal mortality [20] can be observed in Table 5. Fetal deaths in both Bogota and Caracas are also more education elastic than in Rio de Janeiro. Contrary to the pattern observed for income elasticities in Caracas and Rio de Janeiro, fetal mortality in these cities, as well as Bogota, tends to become more education elastic as the age of the mother increases. This phenomenon finds a plausible explanation in the fact that average educational attainment varies inversely with age, while the level of income generally increases with age. Thus, it seems that the values of both education and income elasticities are considerably influenced by the average values of the independent variables. Female labor force participation. The estimated coefficients for this variable imply, for example, that working women in Bogota aged 20-24 experience, on average, 0.16 fewer fetal deaths than their nonworking counterparts. Regardless of age group and in all three cities, working mothers systematically report fewer fetal deaths than nonworking mothers, all coefficients except one being statistically significant at the 0.80 level or higher. The negative effect of female labor force participation on fetal mortality is generally greater for Caracas than for Bogota and for Bogota than for Rio de Janeiro; in other words, female labor force participation seems to reduce fetal mortality more effectively under conditions of availability of family planning and maternal--child health services. In all three cities the absolute value of the coefficients tends to become larger with age. Migration. Household migration history is the fourth variable hypothesized to explain fetal mortality differentials. Other things being equal, it is expected that migration imposes hardships on households that lead to fetal death as a result of difficulties that may arise in adapting to a new environment at the migrants' point of destination. The generally positive and statistically significant least-squares coefficients of the migration variable, however, do not conform to this expectation. On the contrary, nonmigrant mothers seem to experience a greater fetal death incidence than their migrant counterparts. A plausible explanation for this phenomenon may lie in the fact that migration is highly selective of the young and healthy, whose good health accounts for lower levels of fetal death. This explanation finds support in the decreasing values of the migration coefficients for all three cities as age increases. For each age group the migration effect in explaining fetal mortality differentials, as measured by the size of the coefficients, is
Table 4. Income elasticities of fetal mortality by city and age group
Table 5. Education elasticities of fetal mortality by city and age group
Income elasticity of fetal mortality Age group
Bogota
Caracas
Rio de Janeiro
Age group
20-24 25-29 30-34 35-39 40~4 45-49
-0.22 - 0.26 -0.26 -0.30 -0.31 -0.38
-0.55 - 0.57 -0.56 -0.44 -0.44 -0.46
-0.32 - 0.24 -0.26 -0.20 -0.17 -0.16
20-24 25-29 30-34 35-39 4~44 45-49
Education elasticity of fetal mortality Rio de Bogota Caracas Janeiro -0.21 -0.36 -0~42 -0.46 -0.54 -0.72
-0.25 -0.40 -0.44 -0.47 -0.54 -0.55
-0.13 -0.14 -0.17 -0.14 -0.18 -0.20
95
Determinants of fetal and child deaths in Latin America more i m p o r t a n t for Rio de Janeiro than for Bogota and for Bogota than for Caracas. Consensual union. The last variable employed here to ascertain the degree to which parental socioeconomic e n v i r o n m e n t determines the incidence of fetal deaths is the presence of consensual marriages or de facto unions. The positive sign and high levels of significance of these coefficients accord with the expectation that consensually married women experience higher levels of fetal mortafity than legally married women. The relatively large values of the estimated coefficients indicate that fetal death differentials attributed to type of marriage are of considerable importance. The impact of de facto unions on fetal deaths is strongest for Bogota a n d weakest for Caracas. It will be recalled that the percentage of de facto unions in Caracas is higher than in Rio de Janeiro, while the percentage in Rio de Janeiro is slightly higher than in Bogota. Thus, the evidence seems to imply that consensually married women experience higher levels of fetal mortality where stronger negative sanctions
on de facto marriages are not so severely sanctioned, as evidenced by a higher incidence, of concubinage. The fact that in all three cities a lower percentage of women are consensually married, and the influence of de facto unions becomes stronger, as the age of the w o m a n increases further suggests that relatively more abortions and stillbirths occur under conditions in which concubinage is more severely sanctioned.
Discussion Income. The estimated values of the coefficients are negative and statistically significant at least at the 0.90 probability level, thus supporting the hypothesis that higher-income households tend to experience fewer losses of children through death relative to poorer households. The levels of significance are lower for Bogota than for the other two cities, which suggests that in Bogota the level of income is not as i m p o r t a n t in determining child death differentials as in Caracas or Rio de Janeiro. Age group patterns of statistical significance do not appear in any of the three cities.
Child mortality
City and age group
Child Mortality Regression Coefficients Female labor force Education participation Migration
Intercept
Income
Bogota 20-24
O.158
25-29
o.316
30-34
0.526
35-39
0.516
40-44
0.960
- 0.010** (0.006) -O.Ol3tt (0.007) - O.Ol9 t t (0.010) - 0.027tt (0.014) -0.04177
45-49
1.071
- 0.049* (0.024)
Caracas 20-24
0.304
- 0.036** (0.014) - 0.026* (0.012) -- 0.030* (0.013) -- 0.032* (O.Ol5) - 0.039* (0.017) - 0.048* (0.021)
(0.005) -0.017" (0.008) - 0.031 * (0.012) - 0.045* (0.019) - 0.066** (0.021 ) -0.082** (0.026)
- 0.028 * (0.013) -- O.025tt (0.013) - 0.042* (0.020) -0.042* (0.019) -0.050* (0.021) - 0.050* (O.O2O)
- 0.008tt (0.004) -0.009t (0.007) --0.018" (0.007) - 0.027* (0.013) -0.052** (0.018) --0.067** (0.018)
(0.023)
25-29
0.305
30-34
0.511
35-39
0.576
40-44
0.820
45-49
1.002
Rio de Janeiro 20-24
- 0.024
25-29
0.009
30-34
0.294
35-39
0.424 0.821
45-49
0.996
- 0.008tt (0.004) --O.Ol5tt (0.0o8) -0.026** (0.008) - 0.029"* (0.010) -0.052** (0.018) - 0.074** (0.024) -0.014"*
Consensual union
0.055t (0.031) -0.036 (0.o40) -0.001 (0.049) o. 111 t t (0.064) -0.136t (0.093) o. 153 (0.122)
0.148** (0.031) 0.130"* (0.o40) 0.109?¢ (0.064) o. 119tt (0.067) -0.084 (0.074) o. 129tt (0.068)
0.135* (0.063) 0.045 (0.061 ) 0.20it (0.144) 0.263tt (0.137) o. 122 (0.233) 0.541 * (0.213)
0.062t 0.046) -0.023 (0.024) 0.077¢ (0.049) 0.004 (0.060) 0.085~ (0.059) 0.091t (0.057)
o. 145"* (0.034) 0.145"* (0.034) O.124* (0.056) O.111 f (0.069) 0.106+ (0.059) 0.088** (0.033)
- 0.017 (0.064) o.120t+ (0.062) - O.127tt (0.066) O.195* (0.080) 0.183tt (0.097) 0.146 (0.132)
0.611 ** (0.073) 0.559** (0.068) 0.446** (0.080) 0.328"* (0.086) 0.224** (0.074) 0.230** (0.085)
0.008 (0.061) -0.007 (0.055) 0.15lit (0.090) 0.212t+ (0.117) -0.017 (0.176) -0.006 (0.182)
0.019 (0.034) 0.020 (0.036) 0.094tt (0.057) 0.132* (0.065) 0.140t (0.086) 0.136t (0.096)
96
MANUEL CARVAJAL and PAUL BURGESS
Table 6. Income elasticities of child mortality by city and age group
Table 7. Education elasticities of child mortality by city and age group
Income elasticity of child mortality Age group
Bogota
Caracas
Rio de Janiero
Age group
20-24 25-29 30-34 35-39 40-44 45-49
-0.25 -0.28 -0.28 -0.42 - 0.40 -0.48
-0.99 - 1.02 -0.77 -0.74 - 0.60 -0.59
-0.64 -0.66 -0.62 -0.63 - 0.48 -0.37
20-24 25-29 30-34 35-39 40-44 45-49
The income elasticities of child mortality [21], which can be observed in Table 6, show that this dependent variable is more responsive to changes in income than fetal mortality, although the estimated values of all elasticities, except for the 25-29 age group in Caracas, still are below unity. As with fetal deaths, child deaths in Bogota become more income elastic as the age of the mother increases, while in Caracas and Rio de Janeiro the elasticity of child deaths varies inversely with age. Education. The incidence of child deaths also varies inversely with the level of mother's education, as evidenced by the negative sign of the statistically significant coefficients. In all three cities the level of significance tends to increase with age. Since educational attainment of younger women exceeds that of older women, the positive relationship between mothers' age and the coefficients' level of significance seems to be indicative of the impact on child mortality of an increasingly widespread exposure to education. In other words, insofar as higher levels of educational achievement by younger mothers expand their knowledge of more efficient techniques to care for their offspring, the effect of narrower differentials may translate into lower levels of statistical significance relative to their older counterpart. This hypothesis receives further support from the finding that education elasticities of child mortality [22] in all three cities generally increase with age (see Table 7), thus implying that observed child deaths among older women are more responsive to changes in educational attainment than among younger women. In general, education elasticities of child mortality show higher absolute values than education elasticities of fetal mortality, which indicates that as the level of education continues to increase as a result of further economic development, expected reduction in child deaths will occur at a faster rate than expected reduction in fetal deaths. Female labor force participation. Only 11 of the 18 coefficients for this variable possess statistical significance. The positive sign of the coefficients accord with the hypothesis that female labor force participation tends to increase the incidence of child deaths by virtue of an increase in the probability of accident occurrence. To the extent that rearing children is a highly labor-intensive activity on the part of the mother, the additional time devoted by nonworking women to their offspring seems to be reflected in fewer losses of children through death. The fact that the size of the statistically significant
Education elasticity of child mortality Rio de Bogota Caracas Janeiro - 0.31 - 0.42 - 0.44 - 0.50 -0.53 - 0.73
- 0.57 - 0.68 - 0.78 - 0.88 -0.81 - 0.76
- 0.35 - 0.43 - 0.39 - 0.54 -0.63 -0.66
labor force participation coefficients increases with age implies that the danger of child death for working mothers has declined in recent years. To be sure, part of the age differential must be attributed to the fact that older women have been exposed for a longer period of time to the probability of offspring death. The growing institutionalization of child-care centers, as well as increasing school attendance rates, however, also seem to have curtailed the apparent incompatibility between female participation in the labor force and child supervision. Migration. With the exception of the 40-44 age group in Bogota, all migration coefficients are positive and statistically significant at least at the 80~o level. This implies that child mortality occurs less frequently among migrant mothers than among those who were born and raised in the cities in which the surveys were conducted. The differences tend to decrease with age, thus following the same pattern found for fetal mortality. Apparently the effect of the selectivity of the migration process, in terms of the young and the healthy, continues to be felt throughout the lives of their offspring. Consensual union. Although eight out of the 18 estimated coefficients for this variable are not significantly different from zero, the positive sign of the remaining 10 coefficients indicates that consensually married women experience higher levels of child mortality than legally married women. The nature of the differential, however, is relatively unimportant as evidenced by the rather small coefficient values. The positive influence of de facto unions on child death incidence becomes more intensive with older age groups, partly because older women have been exposed longer to the danger of losing their children through death, and partly because of more severe negative sanctioning of concubinage in the past when these women were of childrearing age.
Comparison of fetal and child mortality equations The coefficient of multiple determination, or R 2, provides a good indicator of whether or not the model is correctly specified. It measures the percentage of variation in the dependent variable accounted for by variation in the independent variables. The R 2 values (see Table 8) show that the model as specified is more appropriate for explaining fetal than child deaths. They also show that the explanatory power of the equations for younger age groups is superior to that of older age groups.
97
Determinants of fetal and child deaths in Latin America Table 8. R z by city and age group Dependent variable: child deaths
Dependent variable: fetal deaths Age group
Bogota
Caracas
Rio de Janeiro
Bogota
Caracas
Rio de Janeiro
20-24 25-29 30-34 35-39 40-44 45-49
0.433 0.321 0.310 0.248 0.227 0.170
0.320 0.344 0.223 0.228 0.176 0.200
0.251 0.263 0.213 0.210 0.161 0.156
0.333 0.235 0.178 0.218 0.154 0.226
0.277 0.252 0.203 0.217 0.152 0.163
0.231 0.151 0.267 0.265 0.160 0.138
CONCLUSION
This paper has sought to test the basic hypothesis that occurrence of fetal and child deaths in Bogota, Caracas and Rio de Janeiro is systematically influenced by parental socioeconomic environment. It has dealt with attempting to connect such occurrence, beyond the traditional congenital or developmental approach, to socioeconomic p h e n o m e n a on the proposition that even though biological forces may be important in determining fetal and child mortafity differentials, these differentials also are the product of socioeconomic forces, either directly or t h r o u g h their effect on biological variables, which in turn translates into death differentials of fetuses a n d children. Although it c a n n o t be claimed that the model used in this paper even remotely approaches a completely specified fetal or child mortality function, the empirical results indicate that the explanatory variables do explain a substantial portion of the cross-sectional variation in both dependent variables. Despite its many limitations, the a p p r o a c h and the findings broadly c o r r o b o r a t e the view that parental socioeconomic e n v i r o n m e n t does exert a measurable impact on fetal and child mortality differentials. As the levels of income and education increase in Latin American cities as a result of the development process, poorer and less educated people will increasingly benefit from m o d e r n scientific and technical developments available through c o m m u n i t y health a n d welfare services. Such accessibility to medical help, coupled with improvements in nutritional diets, ultimately will lead to further reductions in the incidence of stillbirths a n d child deaths. Insofar as parents tend to compensate for the loss of their progeny t h r o u g h death with additional pregnancies in order to achieve a desired n u m b e r of survivors, expectations of lower fetal and child mortality levels will lead to lower fertility as fewer pregnancies are required to attain a given desired completed family size with a higher offspring life expectancy. REFERENCES
l. Becket G. S. An economic analysis of fertility. In Demographic and Economic Change in Developed Countries, pp. 209-231. Princeton University Press, Princeton, 1960; Human Capital: A Theoretical and Empirical Analysis. National Bureau of Economic Research, New York, 1964; and A theory of the allocation of time. Econ. J. 75, 493, 1965. ~.SM. 12
3/4C--D
2. Mincer J. Market prices, opportunity costs, and income effects. In Measurement in Economics: Studies in Mathematical Economics and Econometrics in Memory of Yehuda Grunfeld, pp. 67 82. Stanford University Press, Stanford, 1963. 3. Ben-Porath Y. Economic analysis of fertility in Israel: point and counterpoint. J. Polit. Econ. 81, $202, 1973. 4. Willis R. J. A new approach to the economic theory of fertility behavior. J. Polit. Econ. 81, S14, 1973. 5. Blandy R. The welfare analysis of fertility reduction. Econ. J. 81, 109, 1974. 6. Leibenstein H. L. An interpretation of the economic theory of fertility: promising path or blind alley'? J. Econ. Lit. 12, 457, 1974~ and On the economic theory of fertility: a reply to Keeley, J. econ. Lit. 13, 469. 1975. 7. Carvajal M. J. and Geithman D. T. Socioeconomic fertility differentials in Costa Rica 1963-1973. In New Perspectives on the Demographic Transition, pp. 95-162. Interdisciplinary Communications Program, Smithsonian Institution, Washington, D.C., 1976. 8. Carvajal M. J. and Geithman D. T. Family Planning and Family Size Determination: The Evidence from Seven Latin American Cities, pp. 49-51. The University of Florida Press, Gainesville, 1976; and Schultz. T. P. An economic analysis of family planning and fertility. J. polit. Econ. 77, 153, 1969. 9. Pressat R. Demographic Analysis p. 84. Aldine-Atherton, Chicago, 1972; and Thomlinson R. Population Dynamics p. 138. Random House, New York, 1965. 10. Resseguie L. J. Changes in stillbirth ratios resulting from changing fashions in age of childbearing. Soc. Biol. 20, 173, 1973; Shah F. K. and Abbey H. Effects of some factors on neonatal and postneonatal mortality. Milbank Metal Fund q. Bull. 49, 33, 1971; Shapiro S., Jones E. W. and Desen P. M. A life table of pregnancy terminations and correlates of fetal loss. Milbank Metal Fund q. Bull. 40, 7, 1962; and Willie C. V. and Rothney W. B. Racial, ethnic, and income factors in the epidemiology of neonatal mortality, in Social Demography (Edited by Ford T. R. and DeJong G. F.) pp. 501-507. Prentice Hall, Englewood Cliffs, N.J., 1970. I I. CELADE also sponsored during the 1960s similar surveys in Buenos Aires, Mexico City, Panama City and San Jose. 12. Gonzalez R. and Kar S. B. Venezuela, Perfiles de Pqises Bogota: Asociacion Colombiana para el Estudig de la Poblacion, pp. 14-15, June 1975. 13. Perez E. Perfil de Colombia. Estudios de Poblaeion Bogota: Asociacion Colombiana para el Estudio de la Poblacion, pp. 12-15, December 1976. 14. Rodrigues W. Brasil. In J. Lapham Programas de Planificacion Familiar: Revision Mundial 1974 (Edited by Watson W. B. and Lapham R. J.) Bogota: Asocia-
98
MANUEL CARVAJAL and PAUL BURGESS
cion Colombiana para el Estudio de la Poblacion, pp. 122-26, August 1975. 15. Chase H. International Comparison of Perinatal and Infant Mortality: The United States and Six West European Countries Washington, D.C.: U.S. National Center for Health Statistics, Series 3, No. 6, 1967; Shapiro S., Jones E. W. and Densen P. M. op. eit.; and Tabah L. and Sutter J. Influence Respective de l'Age Maternel et du Rang de Naissance sur la Mortalite. Population 3, 63, 1948. 16. See for example Carvajal M. J. and Geithman D. T, Demooraphic Profile of Costa Rica. pp. 85-93. San Jose: Direccion General de Estadistica y Censos, 1976. I% Sauvy A. General Theory of Population p. 328. Basic Books, New York 1969. 18. Willie C. V. and Rothney W. B. op. cit. 19. Income elasticity of fetal mortality is approximated by Ey/f ~ ~
all j
where Yij is the income score mean for the jth age group in the ith city, M m is the average number of fetal deaths for the jth age group in the ith city, and a~j is the least-squares estimate of the income coefficient for the jth age group in the ith city.
20. Education elasticity of fetal mortality is approximated by EE/f = Mfi~j a2ij where E 0 is the average education for the jth age group in the ith city and az~j is the least-squares estimate of the education coefficient for the jth age group in the ith city. 21. Income elasticity of child mortality is approximated by
e,j
Er/c = ~
b'liJ
where Mc~j is the average number of child deaths for the jth age group in the ith city, and bl~~ is the least-squares estimate of the income coefficient for the jth age group in the ith city. 22. Education elasticity of child mortality is approximated by E~j EElc = Mci~j b2ij where b2o is the least-squares estimate of the education coefficient for the jth age group in the ith city.