- Email: [email protected]
Women, men, and contraceptive sterilization
FERTILITY AND STERILITY威 VOL. 73, NO. 5, MAY 2000 Copyright ©2000 American Society for Reproductive Medicine Published by Elsevier Science Inc. Printed on acid-free paper in U.S.A.
Women, men, and contraceptive sterilization Larry L. Bumpass, Ph.D., Elizabeth Thomson, Ph.D., and Amy L. Godecker, M.S. Center for Demography and Ecology, University of Wisconsin-Madison, Madison, Wisconsin
Objective: To review the social and behavior contexts of decisions about contraceptive sterilization and to analyze factors associated with sterilization choices. Design: Multinomial logit regression of sterilization. Patient(s): Various subsamples as appropriate to specific analyses drawn from the 10,847 women interviewed in the 1995 National Survey of Family Growth, and the 5,227 men interviewed in the National Survey of Families and Households. Intervention(s): None. Main Outcome Measure(s): Tubal sterilization and vasectomy. Result(s): Surprisingly high proportions of recent tubal sterilizations were performed on unmarried women: 1 in 3 overall, 1 in 5 among white non-Hispanic women, and 2 in 3 among black women. Sterilization choice among continuously married couples also revealed large differences by race and ethnicity. Parity at the time of the last wanted birth is a major factor affecting sterilization choices, although significant effects were found as well as for a number of other variables, including age differences between spouses, education, and religion. Compared with other regions, the ratio of tubal sterilizations to vasectomies is extremely low in the Western region of the United States. Conclusion(s): Analysis of sterilization decisions must be based on time since the completion of childbearing. The findings call attention to the need for measuring variables that mediate observed associations with sterilization outcomes. (Fertil Steril威 2000;73:937– 46. ©2000 by American Society for Reproductive Medicine.) Key Words: Tubal sterilization, vasectomy, last wanted birth, decision, certainty
Received March 12, 1999; revised and accepted January 27, 2000. Supported by the Contraception and Reproductive Health branch of the National Institute of Child Health and Human Development. Presented in part at the conference entitled “Male and Female Sterilization: Medical Effects and Behavioral Issues,” which was held in Bethesda, Maryland, on June 11–12, 1998. Reprint requests: Larry L. Bumpass, Ph.D., Center for Demography and Ecology, 1180 Observatory Drive, Madison, Wisconsin 53706 (FAX: 608-262-8400; E-mail: [email protected]). 0015-0282/00/$20.00 PII S0015-0282(00)00484-2
After briefly reviewing social and behavioral aspects of the sterilization decision process, we analyze factors associated with choices among three alternatives: tubal sterilization, vasectomy, and no sterilization. Once highly stigmatized and disapproved, sterilization has become the primary contraceptive method in the United States and is adopted in three quarters of all marriages that remain intact (1–3). The profound increase in contraceptive sterilization began shortly after the introduction and rapid diffusion of oral contraceptives, and it seems likely that experience with the pill was an important catalyst for the mass acceptance of sterilization. Oral contraceptives introduced a dramatic change from the prior contraceptive regimen, because they were so much more effective and because they separated contraception from sexual intercourse. However, as concerns about the long-term safety of oral contraceptives grew, tubal sterilization and va-
sectomy became increasingly attractive as alternatives that shared the characteristics of being highly effective and unobtrusive (3). In addition to method characteristics, a number of other issues are important in sterilization decisions. It is most important that contraceptive sterilization is not appropriate for those who intend another child, nor is it attractive to those who are uncertain about whether they want another (even though operations can sometimes be reversed). For this reason, research on factors affecting the choice of sterilization ought to focus on the appropriate “risk” population, those who do not want another child, and attend to the time-dependent nature of this process over the life course. Crosssectional estimates of the proportion sterilized among all persons of reproductive age indicate the prevalence in a population (2, 4) but not the propensity to adopt sterilization once childbearing has been completed. 937
Nevertheless, we must recognize that there may be a period of uncertainty before a decision is reached to have no more children (5). This uncertainty may arise because a person is unsure of their own preference in this regard, but spouses can also be certain, but disagree, about whether they want another child. Indeed, dyadic factors affecting sterilization decisions may be indirect, e.g., through a concern about marital stability and whether another child might be desired with a future spouse (6). When couples disagree about having more children, the conceptualization and measurement of “last wanted birth” becomes even more difficult. Whose wants or intentions should we consider? Even if they eventually agree, couple disagreement should lead to the postponement of sterilization. Our present analysis is based on an estimated date of “last wanted birth” from the wife’s perspective, but both uncertainty and the husband’s attitude toward ending childbearing need to be examined in future work. One of the major puzzles in the adoption of sterilization is that tubal sterilization has become so much more common than vasectomy, when the latter is safer, less expensive, and equally effective in preventing births. This seeming paradox calls our attention to two aspects of the decision process that are often overlooked: (1) individuals may make this decision for themselves without considering sterilization for their partner, and (2) this is particularly likely if they are unmarried. Most studies of the choice between tubal sterilization and vasectomy assume a two-stage decision process in which the couples first decide to terminate childbearing with sterilization and then negotiate which spouse will be sterilized (7). Although this model may be theoretically appropriate, it has most often been structured by the data available for these studies: clinical samples including only couples who have chosen sterilization, among whom it is then sought to identify the determinants of which partner was sterilized (8 –10). For many couples, the decision about sterilization may be specific to one partner. For example, a wife may believe— correctly or not—that her husband would not consider vasectomy, so that the decision she makes (and discusses with her husband) is whether she should have a tubal sterilization. Or one of the partners may feel more strongly about preventing further pregnancies and decide on her/his own to be sterilized. One of the reasons for higher rates of female than male sterilization among couples may well be the absence of a two-stage process in many cases, i.e., women who are well aware that they do not wish to bear anymore children (because they are the ones primarily involved in child rearing) may pursue sterilization without bothering to engage in a consideration of the relative advantages and costs of vasectomy (including persuading their husband to have one). Furthermore, unmarried women are increasingly choos938
Bumpass et al.
Women, men, and sterilization
ing to be sterilized. The context of childbearing decisions has changed dramatically because marriage is being delayed, and one third of all births are to unmarried mothers (half of which are second or higher order births) (11). Hence, as a consequence of these life course changes, many women have had all the children they feel they want or can support before ever being married. Similarly, separated and divorced women who want no more children are highly likely to remain sexually active and to cohabit and, hence, to risk additional (and unmarried) childbearing. In the analysis that follows, we first estimate the extent to which sterilization occurs outside of marriage and then examine factors associated with sterilization decisions among married couples.
METHODS AND MATERIALS Data The data used come from cycle 5 (1995) of the National Survey of Family Growth (NSFG) (12). This is a periodic survey conducted by the National Center for Health Statistics with the primary goal of providing estimates of factors affecting the US birth rate and the reproductive health of US women 15– 44 years of age. Interviews averaging 105 minutes were conducted with 10,847 respondents. Key to our current analysis are pregnancy and birth histories, dates and types of sterilizing operations (and contraceptive intent), the planning status of each pregnancy, and an array of characteristics of the respondent and her husband. The NSFG is the primary data source for this article. However, because the NSFG is limited to women, estimates for men are drawn from the 1987–1988 wave of the National Survey of Families and Households (NSFH) (13). Interviews in the NSFH were conducted with 13,017 respondents and averaged about 100 minutes. Topics covered included detailed household composition, family background, adult family transitions, couple interactions, parent-child interactions, education and work, economic and psychological well-being, and family attitudes. Dates of sterilization were included in the fertility histories.
Sample Constraints The sample for the analysis of couple choices is limited to currently married couples for whom the date of last wanted birth occurred since 1980 and after the date of their marriage and for whom we have consistent dates (n ⫽ 2,158). The requirement that the last wanted birth occur after the date of the current marriage is important because among current remarriages in these data, 20% of the wives and 6% of the husbands had been sterilized before the marriage began. Currently pregnant women were also deleted unless the current pregnancy resulted from an unplanned pregnancy. For other pregnant women, the future intentions question Vol. 73, No. 5, May 2000
TABLE 1 Percent in each marital status for the total population and by race or ethnicity: women under age 40 at sterilization: tubal sterilizations 1990 –1995. Race and Hispanic ethnicity of women Marital status Married Cohabitating (formerly married) Cohabiting (never married) Formerly married Never married
Total
White non-Hispanic
Black non-Hispanic
Mexican
Other
68 5 6 8 12
79 6 3 6 7
37 4 15 14 31
73 6 9 3 10
58 9 8 19 6
Note: Data are from the National Survey of Family Growth, Cycle V (1995), weighted. Bumpass. Contraceptive sterilization. Fertil Steril 2000.
refers to births after the current pregnancy; hence, they have not yet entered the period when sterilization would be appropriate. The earlier sterilizations in these data are selective of younger ages because of the upper age limit of the sample. For example, a woman would have had to have been age 30 or younger at the time of her last wanted birth in 1980 to still be represented in a 1995 sample with an upper age limit of 45 (14). Age at last wanted birth is included in the multivariate models as a variable of interest in its own right and to provide some adjustment for this potential bias.
Definition of “Last Wanted Birth” Because sterilization is a contraceptive choice only for those who have completed childbearing, we must structure our analysis to reflect this— even given the ambiguities and uncertainties associated with the measure. Consequently, we estimated the date of last wanted birth and analyzed sterilization after this date. Only women who do not intend to have more children are relevant for this analysis. The date of the last birth is used for the 89% of our sample who do not report having had an unwanted birth, i.e., a birth after they had intended not to have another child. This measure assumes that intentions to have no more children were formed at the time of a woman’s last (wanted) birth. The relevant date is only approximated not only because of the uncertainty about future fertility preferences but also because the decision to terminate childbearing is likely a process for some women or couples whereby intentions for additional children turn into uncertainty and later into intentions not to have any more (5). For women who report unwanted births (3), we go back in their history to the last birth that was reported as wanted. For the few women who have no children and intend none, we set the date as one-half the time between age 15 and the present. Although the latter is an arbitrary coding decision, it affects too few women to matter (we could just as easily delete them). Life-table estimates indicate that a quarter of FERTILITY & STERILITY威
women complete their intended childbearing by age 25, half by age 29, and three quarters by age 34. Hence, the risk of an unwanted pregnancy persists over a substantial proportion of the reproductive years until menopause.
Statistical Method and Variables Multinomial logistic multiple regression (15) is used to compare subgroup differences in the probability of having a sterilizing operation within 5 years of the last wanted birth. Tubal sterilization and vasectomy are each contrasted to no sterilization. We also explored hazard models, but the early concentration of sterilizations in the period of observation makes the logit models more efficient estimators because of the sample sizes and because the number of events gets too thin at later durations. The independent variables include five based on life course stage at the time of the last wanted birth (ages of wife and husband, parity, duration of marriage, and whether either had married before) and the wife’s age at first birth. The next two variables are drawn from the woman’s reports on her family-planning experience: whether she reported any birth resulting from a pregnancy that she wanted to have eventually but that occurred “too soon” and whether she reported any pregnancy terminated by abortion. Social and economic variables include race or ethnicity, education of both spouses, wife’s religion, and region and size of place of residence.
RESULTS Marital Status Composition of Sterilizations We noted in our overview that a significant proportion of never-married women report that they are sterilized, even when the denominator includes all women of reproductive age. High levels of sterilization among unmarried women are even more apparent in Table 1: one third of all recent tubal sterilizations were performed on unmarried women. Sterilization while unmarried is remarkably common among all race or ethnic groups. Nonetheless, the differences 939
in level are so large that it is best to describe the numbers separately. Approximately one fifth of the tubal sterilizations on most white women were performed while they were unmarried: about half of these while never married and about half while formerly married. Approximately 9% of the sterilizations on white women were performed while they were in cohabiting relationships. Patterns are similar among Mexican-American women, but cohabitation plays a larger role, accounting for approximately 15% of sterilizations. Among black women, on the other hand, only a minority (37%) was performed on married women, and nearly as many (31%) were performed on never-married women. Almost one fifth of the tubal sterilizations in black women was performed while they were living in a cohabiting relationship. Sterilization among unmarried women appears to be part of a normal life course from the perspective of fertility. Age and parity at the time of sterilization are similar for unmarried and married women, e.g., never-married women had an average age of 30 and an average parity of 2.8 compared to an identical age and an average parity of 2.6 among married women. Furthermore, the effect of marital status becomes small and nonsignificant once parity is controlled (tables not shown). Hence, achieved fertility is far more important than whether a woman is married with respect to the decision to have a tubal sterilization. The high proportion of female sterilizations that now occur outside of marriage seems to be entirely due to the fact that many women have all the children they want, either while never married or after marital disruption. Female sterilization is so much more common than male sterilization, probably because having a partner choose vasectomy is much less an option for unmarried women. We were able to examine marital status at sterilization for men by using the 1987–1988 NSFH. Although there were too few cases of male respondents reporting on their own sterilization in these data to permit the level of detail in Table 1, we found that approximately 7% of vasectomies between 1980 and 1987 occurred outside of marriage. The much higher level of unmarried sterilization among women than among men indicates that women consider sterilization relative to alternatives in the context of having to raise children already born. Men, on the other hand, may often have little connection with children they have fathered while unmarried (and in some cases may not even know they were born) (16). Hence, it is not surprising that sterilization among unmarried men occurs much more often after rather than before first marriage. Because these findings indicate that sterilization is often an individual decision (particularly among women and particularly among black women) and that most other available descriptive data are based on “contraceptive users,” it seems useful to consider a life course description of contraceptive patterns for both men and women that is contingent on neither marriage nor factors affecting current use of any 940
Bumpass et al.
Women, men, and sterilization
FIGURE 1 Life-table estimates of vasectomy (✚) and tubal sterilization (F) by years since last wanted birth: last wanted births 1980 – 1994.
Bumpass. Contraceptive sterilization. Fertil Steril 2000.
method. Sterilization might be expected to be concentrated at older ages. Nonetheless, because of the dispersion in parity and age at last wanted birth, there is a linear increase in the proportion of women sterilized from 0 at age 20 to 36% by age 40. Among men, the increase begins later but is again linear from low levels in the early 20s to approximately 14%. (These cumulative proportions are consistent with the proportions reported among 40- to 44-year-old women who are currently using contraception: 50% and 20%, respectively, for the 71% who were contraceptive users) (2). Concerns with the “choice” between female and male sterilization must be cognizant of the extent to which female sterilizations occur outside of marriage. Nonetheless, couples’ choices in sterilization remain important, and we turn now to an analysis of factors affecting these choices.
Couple Choices of Male and Female Sterilization We focus this analysis of couple choice on the first 5 years after the birth of the last child wanted. Figure 1 shows that for both males and females the highest rates of sterilization occur in the first year after last wanted birth, with a gradual increase thereafter. Three quarters of all sterilizations occur within the 5-year period we are examining, with the first year being much higher for females, probably because of the prevalence of postpartum sterilization. Although an array of reduced-form models have been explored, the results are most easily communicated from a model that includes all of the variables considered. There are eight columns in Table 2. The first column for each of these choices (1 and 4) is the simple proportions observed to have had a sterilization of this type within 5 years of the last Vol. 73, No. 5, May 2000
TABLE 2 Tubal sterilization within 5 years and vasectomy within 5 years of last wanted birth: observed proportions and relative risk estimates from multinomial logistic regression: births 1980 –1994 (1980 –1989 for proportions). Tubal sterilization vs. none
Parameter Parity last wanted 1 2 3 ⱖ4 Age at last wanted birth (y) ⬍25 25–29 30–34 ⱖ35 Husband is younger by 2 or more years No Yes Husband is older by 5 or more years No Yes Duration of marriage (y) ⬍5 5–9 ⱖ10 Married before No Yes Husband was married before No Yes Age at first birth (y) ⬍20 20–24 25–29 ⱖ30 Had unintended pregnancy No Yes Had an abortion No Yes Race or ethnicity White non-Hispanic Black non-Hispanic Hispanic Education Did not finish high school High school graduate Some college College graduate Husband is less educated than wife No Yes Husband is more educated than wife No Yes
(1) Percent in 5 y
(2) Relative odds in 5 y
12 34 46 44
(3)
Tubal sterilization vs. vasectomy
Vasectomy vs. none
P value
(4) Percent in 5 y
(5) Relative odds in 5 y
4.1 7.7 9.6
.00 .00 .00
12 21 24 28
42 35 23 36
0.9 0.7 0.7
.66 .07 .21
35 38
1.6
35 36
(6) P value
(7) Relative odds in 5 y
(8) P value
3.4 6.8 6.0
.00 .00 .00
1.2 1.1 1.6
.49 .71 .25
15 24 21 16
1.7 1.3 0.8
.02 .35 .67
0.5 0.5 0.8
.01 .03 .62
.02
21 16
0.9
.81
1.7
.05
0.7
.03
22 18
0.6
.00
1.3
.17
35 33 42
1.0 1.3
.99 .21
18 23 19
0.9 0.7
.40 .25
1.2 1.8
.43 .04
34 45
1.2
.30
21 17
0.6
.05
1.9
.01
34 46
1.8
.00
21 17
2.4
.00
0.8
.20
52 37 29 18
1.0 1.0 0.9
.94 .93 .70
15 22 24 15
1.0 1.0 1.0
.84 .98 .96
1.0 1.0 0.9
.88 .97 .80
31 40
1.2
.20
22 19
0.9
.34
1.3
.06
37 28
0.7
.01
20 21
0.9
.64
0.7
.14
32 57 50
2.2 1.8
.00 .00
23 6 10
0.3 0.4
.00 .00
6.6 4.1
.00 .00
48 39 37 23
0.9 0.7 0.4
.75 .18 .00
8 23 20 22
2.1 2.0 1.8
.02 .04 .11
0.4 0.4 0.2
.01 .00 .00
35 37
1.2
.22
21 18
0.8
.23
1.5
.04
36 33
0.7
.01
19 24
1.1
.76
0.7
.02
Continued on following page
FERTILITY & STERILITY威
941
TABLE 2—CONTINUED Tubal sterilization vs. none
Parameter Religion Catholic Protestant No religion Other non-Catholic Region in United States Northeast Midwest South West Size of place Central city SMSA other non-SMSA
(1) Percent in 5 y
(2) Relative odds in 5 y
31 39 35 22
Tubal sterilization vs. vasectomy
Vasectomy vs. none
(3) P value
(4) Percent in 5 y
(5) Relative odds in 5 y
1.6 1.8 0.8
.00 .01 .33
18 22 15 25
31 37 40 30
1.1 1.3 0.6
.75 .08 .00
38 32 41
0.9 1.1
.43 .75
(6) P value
(7) Relative odds in 5 y
(8) P value
1.2 0.9 0.8
.26 .57 .57
1.3 2.1 0.9
.08 .02 .79
16 22 19 24
1.6 1.8 2.0
.01 .07 .00
0.6 0.9 0.3
.04 .63 .00
15 21 24
1.1 1.3
.49 .16
0.8 0.8
.22 .27
Note: A variable for years since last wanted coded 0 –5 was included to control for censoring in the 1990 –1995 period. SMSA ⫽ standard metropolitan statistical area. Bumpass. Contraceptive sterilization. Fertil Steril 2000.
wanted birth. This provides both a description of population variation and a sense of scale. The second column for each choice (2 and 5) reports odds ratios from the multinomial logit: the relative likelihood of having had an operation of the given type contrasted to having had no sterilization within 5 years of the last wanted birth. The third column (3 and 6) is simply the P value of the coefficient (the coefficient itself is not reported). In discussing these results, we do not repeat the “within 5 years” constraint each time proportions or relative risks are cited. The final two columns of Table 2 contrast the odds ratio and P value of tubal sterilization in contrast with vasectomy among couples adopting one or the other of these sterilization methods. Parity, Age, and Duration of Marriage at Last Wanted Birth The number of children a couple has, their age, and how long they have been married may affect how sure they are that they do not want more children and, thereby, affect the likelihood of sterilization. The main story of this set of variables can be told quite simply: the parity of the last wanted birth dominates the effects of the other variables, although other effects of interest remain after parity is controlled. A positive relationship between parity and the likelihood of sterilization would be expected if couples at lower parities are less sure that they have indeed completed their family, and if— because of previous planning failures—those at higher parities are more highly motivated to end childbearing. It is also possible that, despite our attempt to take into account births from unwanted pregnancies, some higher 942
Bumpass et al.
Women, men, and sterilization
parity women may have had an unwanted birth before the one coded as “last wanted,” and such experience would have raised their motivation to be sterilized even further. Finally, age differences could index cohort more than age, i.e., older women have lived more of their lives during a period when sterilization was less accepted. As noted earlier, observed differences in sterilization do not represent differential rates if groups differ in the proportions who have had all the children they want. More sterilizations occur among women with two children than at any other parity (17), but this is so because most women prefer a family size of two children. However, the likelihood of sterilization increases consistently with parity. Between parities 1 and 4, the observed proportions sterilized range from 12% to 44% for tubal sterilization, and from 12% to 28% for vasectomy. The differences in relative risks from the logistic model are large and significant, and the effect is greater for female operations than for male. Even so, among one-child families, one quarter of the couples were sterilized within 5 years: 12% each for male and female operations. As with parity, age would increase the likelihood of sterilization if it were associated with increased confidence in the decision to cease childbearing. On the other hand, age could have negative effects, if women in their late 30s or early 40s feel that they are too close to menopause to make sterilization worthwhile. These women might also see the few remaining years as low risk because of declining fecundity. Age at the birth of a woman’s last wanted child has little effect after parity is controlled, although the rate of vasectomy is significantly higher among couples in which the wife Vol. 73, No. 5, May 2000
is in her late 20s. A significant curvilinear effect is in the last two columns: couples are least likely to adopt tubal sterilization relative to vasectomy if the wife is 25–34. These results may reflect cohort rather than age effects, in conjunction with less certainty about ending childbearing among the youngest couples. The effects of husband’s age became clearest when we classified it relative to wife’s age: those 2 or more years younger than their wife and those 5 or more years older. There are significant effects of both of these variables. Wives with substantially younger husbands are more likely to have a tubal sterilization, and those whose husbands are older by ⱖ5 years less likely. Husbands who are older than their wives are also less likely to adopt sterilization for themselves. These patterns seem more consistent with a cohort hypothesis that older males are more opposed to sterilization in general than with the gender role theories that are usually brought to the interpretation of age differences (18, 19). Duration of marriage at the time of the last wanted birth has no significant net effects, other than a marginally significant higher likelihood of tubal sterilization compared with vasectomy among couples who have a sterilization and have been married the longest. Prior Marital and Reproductive History We would expect that women who were married before, those who married at an early age, and those who have experienced unintended births or abortions would be more likely to adopt sterilization. The results are only partially consistent with these expectations. It is the husband’s prior marriage that has a strong positive effect on both tubal sterilizations and vasectomies (odds ratios of 1.8 and 2.4, respectively). On the other hand, the wife’s prior marriage decreases the likelihood of vasectomy and, hence, creates a greater preponderance of tubal sterilizations among those who are sterilized. The absence of a positive effect for the wife’s prior marriage on her own sterilization and the negative effect on her husband’s are just the opposite of what we would have expected. That is, we would have expected previously married wives to be more likely to adopt tubal sterilizations because of a concern about future instability given that women overwhelmingly remain responsible for children. It is possible that the latter effect might be interpreted in terms of a desire on the part of husbands of previously married wives to keep reproductive options in case this relationship disrupts. However, the significantly higher sterilization rates for both husbands and wives when the husband has been married before does not fit with this explanation but may, instead, reflect the effects of the husband’s obligation to children from his prior marriage. The observed proportions of wives with a tubal sterilization increase dramatically the younger she was when she FERTILITY & STERILITY威
bore her first child. This finding is completely due, however, to the correlation of age at first birth with parity of the last wanted birth. We also do not find the expected effects of prior planning failures except that among couples adopting sterilization there is a marginally significant (P⫽.06) greater likelihood that the wife will have been sterilized compared with the husband. This would be consistent with a greater concern on her part about the possibility of a future unplanned pregnancy. Although the results for women who reported an abortion make sense, post hoc, we had not anticipated the effect. Women who reported an abortion are less likely to be sterilized. This could reflect their personal experience that abortion is viable. As with the other variables we have considered, male sterilization behavior is not responsive to this prior experience with family planning failure that resulted in an abortion. Social and Demographic Variables The most striking differential in choice of tubal sterilization over vasectomy is associated with race. In 1970, tubal sterilizations were more than twice as prevalent for black women as for white women, whereas only 1% of “at risk” black couples had adopted vasectomy, compared with 9% of white couples (1). Tubal sterilizations are no longer twice as prevalent among black women than among white women, but large disparities in both male and female procedures remain. In fact, although the prevalence of vasectomy has increased substantially for whites, it has increased hardly at all for blacks (18). As expected, the race or ethnicity contrasts between rates of tubal sterilization and vasectomy (next to last column in Table 2) are exceedingly large. As we developed in an earlier section, some of the predominance of female compared with male sterilization among blacks can be accounted for by the greater time spent unmarried and the marked increase in sterilization among unmarried black women. The results in Table 2 underscore race or ethnic differences in sterilization behavior, because these estimates are based on continuously married couples and count only behavior subsequent to a last wanted birth in their current marriage. Few black or Hispanic husbands adopt sterilization, and the very high rates of tubal sterilization among their wives can be seen, in part, as a response to the lack of that option. Whether the mediating link between minority status and low levels of vasectomy is the perceived higher risk of marital instability or apprehension about the consequences of vasectomy are issues to which we will return in our concluding section. Nonetheless, the higher rate for black wives more than compensates for the low rate among black husbands. The combined proportions sterilized within 5 years of their last wanted birth are 63% for black couples, 60% for Hispanic, and 55% for whites. As with age, because of the high correlation between 943
husband’s and wife’s education, we entered wife’s education into the equation and then a difference variable for the husband (based on a classification of education into the same categories as reported for wives). Confirming earlier analyses, wife’s education is negatively related to tubal sterilization. The proportion sterilized within 5 years declines from 48% among those not completing high school to 23% among college graduates, and the net relative risks show a monotonic decline (though only the college graduate contrast is significant). Part of this strong relationship may reflect ready access to tubal sterilization, but not to vasectomy, at the clinics used by low-income women (20). There is an additional depressing effect on tubal sterilization if the husband has more education than his wife.
tion are at least 40% lower in the West than in other regions, and rates of vasectomy are twice as high in the West than they are in the Northeast. This can be seen most clearly in the last set of columns, where among those adopting sterilization the relative risk of tubal sterilization to vasectomy is 70% lower in the West than it is in the Northeast. This is a very large difference, and it is consistent with our expectations about the possible effects of extensive HMO coverage. Because vasectomy costs much less than tubal sterilization, HMO physicians may recommend vasectomy more strongly to their patients than do physicians in other types of practice. It may also be the case that the West may simply be an area that is better informed in general about the risks and benefits of vasectomy.
The effects of wife’s education on vasectomy are not monotonic: the major contrast is between husbands of wives who did not complete high school and all higher levels of schooling. So the emerging positive effect of education on vasectomy observed in our earlier analysis (3) is not found here, neither in the present model nor in one considering only husband’s education. Among couples who adopt sterilization, wife’s education significantly reduces the likelihood of tubal sterilization relative to vasectomy. Compared with couples with the same level of education, tubal sterilization predominates among those in which the husband is less educated, but vasectomy rates are higher among those in which he is more educated than his wife. These patterns are consistent with our expectations that vasectomy would be more attractive to better educated husbands; however, that result only appears in this contrast with wives’ education.
The higher observed proportions with vasectomies in central cities and the lower proportions with tubal sterilizations disappear once the other variables are controlled, and size of place has no significant effect on rates of sterilization.
As expected, net of other factors, Catholic wives are substantially and significantly less likely than others to have a tubal sterilization. Even so, it should be emphasized that 31% of Catholic wives have been sterilized within 5 years of their last wanted birth. On the other hand, there are only modest and nonsignificant differences by religion in vasectomy. When vasectomy and tubal sterilization are combined, 49% of Catholic couples have adopted contraceptive sterilization within 5 years of their last wanted birth compared to 61% of Protestants. It is important to recognize that sterilization is not completely determined by individual or couple decisions. Regional variations in physicians’ attitudes toward sterilization or in the medical care delivery system may also limit or facilitate sterilization for those who have achieved their desired family size. Our earlier work noted regional differences, and the possibility that lower rates of tubal sterilization in the Northeast might reflect the influence of Catholic doctrine in that region on accessibility quite apart from an impact on individual behavior (3). We do not replicate this earlier finding, although we do find significantly higher rates of vasectomy outside of the Northeast, particularly in the West. Rates of tubal steriliza944
Bumpass et al.
Women, men, and sterilization
In conclusion, contraceptive sterilization is an extremely important factor in fertility and family processes in the United States, and analyses of the factors affecting the relevant decision processes have been too neglected in nationally representative data. The observation that sterilization is now the leading method of birth control in the United States substantially understates its importance, because it is not an appropriate method until the decision to end childbearing is held with confidence. Our high level of contraceptive sterilization is a definitive statement by men, women, and couples about their reproductive intentions. Given the long period at risk of unwanted pregnancy that follows the last wanted birth for most women, levels of unwanted pregnancy would undoubtedly be much higher if it were not for the high levels of protection afforded by sterilization; abortion rates would be higher as well. The analysis of choices in sterilization must relate to the time since sterilization becomes a viable option, i.e., after no more children are desired. Furthermore, we must recognize that such decisions are often made by individuals rather than by couples. That one third of all recent tubal sterilizations occurred outside of marriage draws our attention to a large proportion of women for whom vasectomy is not likely an option. Our multivariate analysis of factors affecting choices of tubal sterilization and vasectomy among continuously married couples documents a number of patterns that raise further questions for future research. We will need to begin to gather data that address potential mediating mechanisms in the observed differences in sterilization behavior. The dominance of parity in a study population already defined in terms of risk underscores the potential importance of the role of uncertainty in sterilization choice processes. At the same time, however, we must recall that medical practice once imposed parity constraints on access to sterilization (1), and it is possible that such rules or practices persist more than we Vol. 73, No. 5, May 2000
realize and are revealed, along with differential certainty, in the dramatic parity effects that we find. Such possible effects from the supply side—illustrated as well by the strong regional effects—make it clear that we need a provider survey to tap both the attitudes and relevant behaviors of potential providers of tubal sterilization and vasectomy. Such measures may help us better understand the prevalence of these operations relative to no sterilization and relative to each other. The effect for women who have been married before suggests a concern with the possible future marital instability, an effect we have not yet been able to definitively measure. The most dramatic differences that are yet unexplained are the low levels of vasectomy (and compensating high levels of tubal sterilization) among minority women. Part of this difference results from differing lengths of time spent unmarried, but the differential persists in our analysis of continually married couples. It seems likely that measuring concerns about marital stability and future fertility will help in understanding these differences, but it is also possible that such differences result in part from differential contact with the medical system. Minority women are more likely than other women to be in direct contact with this system, whereas minority men may have much less contact than other men. One of the largest remaining puzzles is why rates of vasectomy are so much lower than rates of tubal sterilization. Part of the answer lies in the marital status of women when they decide that they want no more children, and part lies in the avoidance of vasectomy among minorities. But even for majority whites, where the two operations are closer to equally represented, we do not understand why vasectomy is not the primary method of sterilization among stable monogamous couples. Our discussions of necessity for certainty about intentions not to have another child and couple issues in negotiating that certainty should make it clear that studying sterilization decisions will require simultaneous attention to the processes by which individuals and couples decide that they want no more children. Hence, part of the future agenda must include the incorporation of the negotiation between spouses of certainty about ending childbearing and about sterilization decisions and in the attempt to uncover, as well, the extent to which these decisions may be made individually (rather than jointly) even among married couples. In addition, however, given the importance of these methods and the role they play in the control of American fertility, we need to seriously address perceptions of the characteristics of tubal sterilization and vasectomy that may affect their adoption, and we need to collect such data in nationally representative samples. This will require attention to issues of perceived safety (both short term and long term), effects on sex life, concern with keeping reproductive opFERTILITY & STERILITY威
tions open, and the relative evaluation of the preferences for which partner should have a sterilizing operation. For example, our observations about why female operations predominate when vasectomies are cheaper and safer implicitly assume that women would prefer that their spouse have the operation if only they would—it may not be so. Attitudinal data would have the greatest potential in longitudinal designs predicting subsequent sterilization behavior as couples or individuals reach the point where the decision becomes relevant, but they may also inform our understanding of differences even if they are measured only in the cross section. An example from the early fertility studies is a case in point, 23% of white respondents and 44% of black respondents to 1965 National Fertility Study (NFS) agreed that vasectomy would impair a man’s sexual ability (21). These proportions had declined considerably by the 1970 NFS but still remained considerably higher among blacks (1). Because studies of marriage and divorce grew out of the data collected to study fertility, the study of family processes on the one hand (NSFH), and fertility on the other (NSFG), have tended to drift apart in national data collection efforts. It is time to reincorporate fertility processes as family processes. The fact that sterilization occurs increasingly outside of marriage makes this all the more important. Decisions to end childbearing and to consider sterilization are a critical stage in the family life course. References 1. Bumpass L, Presser H. Contraceptive sterilization in the U.S.: 1965 and 1970. Demography 1972;9:531– 49. 2. Piccinino LJ, Mosher WD. Trends in contraceptive use in the United States: 1982–1995. Fam Plann Perspect 1998;30:4 –10. 3. Bumpass L. The risk of an unwanted birth: the changing context of contraceptive sterilization in the U.S. Popul Stud 1987;41:347– 63. 4. Chandra A. Surgical sterilization in the United States: prevalence and characteristics, 1965–95. Vital Health Stat 23, DHHS Publication No. (PHS) 98-1996, 1998. 5. Morgan SP. Intention and uncertainty at later stages of childbearing: the United States 1965 and 1970. Demography 1981;18:267– 85. 6. Godecker AL. Gender role attitudes and marital quality in female and male sterilization: longitudinal evidence from the National Survey of Families & Households (Master’s thesis). Center for Demography and Ecology, University of Wisconsin, Madison, 1997. 7. Shain RN, Miller WB, Holden AEC. Factors associated with married women’s selection of tubal sterilization and vasectomy. Fertil Steril 1985;43:234 – 44. 8. Miller WB, Shain RN, Pasta DJ. Tubal sterilization or vasectomy: how do married couples make the choice? Fertil Steril 1991;56:278 – 84. 9. Miller WB, Shain RN, Pasta DJ. A model for determinants in married women of sterilization method choice. Popul Environ 1985– 86:8: 223–39. 10. Groat T, Neal AG, Wicks WJ. Sterilization anxiety and fertility control in the later years of childbearing. J Marriage Fam 1990;52:249 –58. 11. Wu L, Bumpass L, Musick K. Historical and life course trajectories of nonmarital childbearing (working paper 99-23). Center for Demography and Ecology, 1999. 12. Potter V, Iannacchione F, Mosher W, Mason R, Kavee J. Sample design, sampling weights, imputation, and variance estimation in the 1995 National Survey of Family Growth. Vital Health Stat 2, No. 124, 1998. 13. Sweet J, Call RA, Bumpass L. The design and contend of the National Survey of Families and Household. Center for Demography and Ecology, NSFH WP-01, 1988. 14. Rindfuss R, Palmore J, Bumpass L. Selectivity and the analysis of birth intervals with survey data. Census Forum 1882;8:5–10.
945
15. Agresti A. Categorical data analysis. New York: John Wiley and Sons, 1990. 16. Furstenberg F, Cherlin A. Divided families: what happens to children when parents part. Cambridge: Harvard University Press, 1991. 17. Bumpass L, Thomson E, Godecker AL. Women, men, and contraceptive sterilization. Center for Demography and Ecology, Working paper 99-05, 1999. 18. Forste R, Tanfer K, Tedrow L. Sterilization among currently married men in the United States, 1991. Fam Plann Perspect 1999;27:100 –7, 122.
946
Bumpass et al.
Women, men, and sterilization
19. Bean FD, Williams DG, Opitz W, Burr JA, Trent K. Sociodemographic and marital heterogamy influences on the decision for voluntary sterilization. J Marriage Fam 1987;49:465–76. 20. Haws JM, McKenzie M, Mehta M, Pollack AE. Increasing the availability of vasectomy in public-sector clinics. Fam Plann Perspect 1997; 29:185– 6, 190. 21. Bumpass L, Westoff CF. The perfect contraceptive population: the extent and implications of unwanted fertility in the U.S., 1960 – 65. Science 1970;169:1177– 82.
Vol. 73, No. 5, May 2000
Women, men, and contraceptive sterilization Larry L. Bumpass, Ph.D., Elizabeth Thomson, Ph.D., and Amy L. Godecker, M.S. Center for Demography and Ecology, University of Wisconsin-Madison, Madison, Wisconsin
Objective: To review the social and behavior contexts of decisions about contraceptive sterilization and to analyze factors associated with sterilization choices. Design: Multinomial logit regression of sterilization. Patient(s): Various subsamples as appropriate to specific analyses drawn from the 10,847 women interviewed in the 1995 National Survey of Family Growth, and the 5,227 men interviewed in the National Survey of Families and Households. Intervention(s): None. Main Outcome Measure(s): Tubal sterilization and vasectomy. Result(s): Surprisingly high proportions of recent tubal sterilizations were performed on unmarried women: 1 in 3 overall, 1 in 5 among white non-Hispanic women, and 2 in 3 among black women. Sterilization choice among continuously married couples also revealed large differences by race and ethnicity. Parity at the time of the last wanted birth is a major factor affecting sterilization choices, although significant effects were found as well as for a number of other variables, including age differences between spouses, education, and religion. Compared with other regions, the ratio of tubal sterilizations to vasectomies is extremely low in the Western region of the United States. Conclusion(s): Analysis of sterilization decisions must be based on time since the completion of childbearing. The findings call attention to the need for measuring variables that mediate observed associations with sterilization outcomes. (Fertil Steril威 2000;73:937– 46. ©2000 by American Society for Reproductive Medicine.) Key Words: Tubal sterilization, vasectomy, last wanted birth, decision, certainty
Received March 12, 1999; revised and accepted January 27, 2000. Supported by the Contraception and Reproductive Health branch of the National Institute of Child Health and Human Development. Presented in part at the conference entitled “Male and Female Sterilization: Medical Effects and Behavioral Issues,” which was held in Bethesda, Maryland, on June 11–12, 1998. Reprint requests: Larry L. Bumpass, Ph.D., Center for Demography and Ecology, 1180 Observatory Drive, Madison, Wisconsin 53706 (FAX: 608-262-8400; E-mail: [email protected]). 0015-0282/00/$20.00 PII S0015-0282(00)00484-2
After briefly reviewing social and behavioral aspects of the sterilization decision process, we analyze factors associated with choices among three alternatives: tubal sterilization, vasectomy, and no sterilization. Once highly stigmatized and disapproved, sterilization has become the primary contraceptive method in the United States and is adopted in three quarters of all marriages that remain intact (1–3). The profound increase in contraceptive sterilization began shortly after the introduction and rapid diffusion of oral contraceptives, and it seems likely that experience with the pill was an important catalyst for the mass acceptance of sterilization. Oral contraceptives introduced a dramatic change from the prior contraceptive regimen, because they were so much more effective and because they separated contraception from sexual intercourse. However, as concerns about the long-term safety of oral contraceptives grew, tubal sterilization and va-
sectomy became increasingly attractive as alternatives that shared the characteristics of being highly effective and unobtrusive (3). In addition to method characteristics, a number of other issues are important in sterilization decisions. It is most important that contraceptive sterilization is not appropriate for those who intend another child, nor is it attractive to those who are uncertain about whether they want another (even though operations can sometimes be reversed). For this reason, research on factors affecting the choice of sterilization ought to focus on the appropriate “risk” population, those who do not want another child, and attend to the time-dependent nature of this process over the life course. Crosssectional estimates of the proportion sterilized among all persons of reproductive age indicate the prevalence in a population (2, 4) but not the propensity to adopt sterilization once childbearing has been completed. 937
Nevertheless, we must recognize that there may be a period of uncertainty before a decision is reached to have no more children (5). This uncertainty may arise because a person is unsure of their own preference in this regard, but spouses can also be certain, but disagree, about whether they want another child. Indeed, dyadic factors affecting sterilization decisions may be indirect, e.g., through a concern about marital stability and whether another child might be desired with a future spouse (6). When couples disagree about having more children, the conceptualization and measurement of “last wanted birth” becomes even more difficult. Whose wants or intentions should we consider? Even if they eventually agree, couple disagreement should lead to the postponement of sterilization. Our present analysis is based on an estimated date of “last wanted birth” from the wife’s perspective, but both uncertainty and the husband’s attitude toward ending childbearing need to be examined in future work. One of the major puzzles in the adoption of sterilization is that tubal sterilization has become so much more common than vasectomy, when the latter is safer, less expensive, and equally effective in preventing births. This seeming paradox calls our attention to two aspects of the decision process that are often overlooked: (1) individuals may make this decision for themselves without considering sterilization for their partner, and (2) this is particularly likely if they are unmarried. Most studies of the choice between tubal sterilization and vasectomy assume a two-stage decision process in which the couples first decide to terminate childbearing with sterilization and then negotiate which spouse will be sterilized (7). Although this model may be theoretically appropriate, it has most often been structured by the data available for these studies: clinical samples including only couples who have chosen sterilization, among whom it is then sought to identify the determinants of which partner was sterilized (8 –10). For many couples, the decision about sterilization may be specific to one partner. For example, a wife may believe— correctly or not—that her husband would not consider vasectomy, so that the decision she makes (and discusses with her husband) is whether she should have a tubal sterilization. Or one of the partners may feel more strongly about preventing further pregnancies and decide on her/his own to be sterilized. One of the reasons for higher rates of female than male sterilization among couples may well be the absence of a two-stage process in many cases, i.e., women who are well aware that they do not wish to bear anymore children (because they are the ones primarily involved in child rearing) may pursue sterilization without bothering to engage in a consideration of the relative advantages and costs of vasectomy (including persuading their husband to have one). Furthermore, unmarried women are increasingly choos938
Bumpass et al.
Women, men, and sterilization
ing to be sterilized. The context of childbearing decisions has changed dramatically because marriage is being delayed, and one third of all births are to unmarried mothers (half of which are second or higher order births) (11). Hence, as a consequence of these life course changes, many women have had all the children they feel they want or can support before ever being married. Similarly, separated and divorced women who want no more children are highly likely to remain sexually active and to cohabit and, hence, to risk additional (and unmarried) childbearing. In the analysis that follows, we first estimate the extent to which sterilization occurs outside of marriage and then examine factors associated with sterilization decisions among married couples.
METHODS AND MATERIALS Data The data used come from cycle 5 (1995) of the National Survey of Family Growth (NSFG) (12). This is a periodic survey conducted by the National Center for Health Statistics with the primary goal of providing estimates of factors affecting the US birth rate and the reproductive health of US women 15– 44 years of age. Interviews averaging 105 minutes were conducted with 10,847 respondents. Key to our current analysis are pregnancy and birth histories, dates and types of sterilizing operations (and contraceptive intent), the planning status of each pregnancy, and an array of characteristics of the respondent and her husband. The NSFG is the primary data source for this article. However, because the NSFG is limited to women, estimates for men are drawn from the 1987–1988 wave of the National Survey of Families and Households (NSFH) (13). Interviews in the NSFH were conducted with 13,017 respondents and averaged about 100 minutes. Topics covered included detailed household composition, family background, adult family transitions, couple interactions, parent-child interactions, education and work, economic and psychological well-being, and family attitudes. Dates of sterilization were included in the fertility histories.
Sample Constraints The sample for the analysis of couple choices is limited to currently married couples for whom the date of last wanted birth occurred since 1980 and after the date of their marriage and for whom we have consistent dates (n ⫽ 2,158). The requirement that the last wanted birth occur after the date of the current marriage is important because among current remarriages in these data, 20% of the wives and 6% of the husbands had been sterilized before the marriage began. Currently pregnant women were also deleted unless the current pregnancy resulted from an unplanned pregnancy. For other pregnant women, the future intentions question Vol. 73, No. 5, May 2000
TABLE 1 Percent in each marital status for the total population and by race or ethnicity: women under age 40 at sterilization: tubal sterilizations 1990 –1995. Race and Hispanic ethnicity of women Marital status Married Cohabitating (formerly married) Cohabiting (never married) Formerly married Never married
Total
White non-Hispanic
Black non-Hispanic
Mexican
Other
68 5 6 8 12
79 6 3 6 7
37 4 15 14 31
73 6 9 3 10
58 9 8 19 6
Note: Data are from the National Survey of Family Growth, Cycle V (1995), weighted. Bumpass. Contraceptive sterilization. Fertil Steril 2000.
refers to births after the current pregnancy; hence, they have not yet entered the period when sterilization would be appropriate. The earlier sterilizations in these data are selective of younger ages because of the upper age limit of the sample. For example, a woman would have had to have been age 30 or younger at the time of her last wanted birth in 1980 to still be represented in a 1995 sample with an upper age limit of 45 (14). Age at last wanted birth is included in the multivariate models as a variable of interest in its own right and to provide some adjustment for this potential bias.
Definition of “Last Wanted Birth” Because sterilization is a contraceptive choice only for those who have completed childbearing, we must structure our analysis to reflect this— even given the ambiguities and uncertainties associated with the measure. Consequently, we estimated the date of last wanted birth and analyzed sterilization after this date. Only women who do not intend to have more children are relevant for this analysis. The date of the last birth is used for the 89% of our sample who do not report having had an unwanted birth, i.e., a birth after they had intended not to have another child. This measure assumes that intentions to have no more children were formed at the time of a woman’s last (wanted) birth. The relevant date is only approximated not only because of the uncertainty about future fertility preferences but also because the decision to terminate childbearing is likely a process for some women or couples whereby intentions for additional children turn into uncertainty and later into intentions not to have any more (5). For women who report unwanted births (3), we go back in their history to the last birth that was reported as wanted. For the few women who have no children and intend none, we set the date as one-half the time between age 15 and the present. Although the latter is an arbitrary coding decision, it affects too few women to matter (we could just as easily delete them). Life-table estimates indicate that a quarter of FERTILITY & STERILITY威
women complete their intended childbearing by age 25, half by age 29, and three quarters by age 34. Hence, the risk of an unwanted pregnancy persists over a substantial proportion of the reproductive years until menopause.
Statistical Method and Variables Multinomial logistic multiple regression (15) is used to compare subgroup differences in the probability of having a sterilizing operation within 5 years of the last wanted birth. Tubal sterilization and vasectomy are each contrasted to no sterilization. We also explored hazard models, but the early concentration of sterilizations in the period of observation makes the logit models more efficient estimators because of the sample sizes and because the number of events gets too thin at later durations. The independent variables include five based on life course stage at the time of the last wanted birth (ages of wife and husband, parity, duration of marriage, and whether either had married before) and the wife’s age at first birth. The next two variables are drawn from the woman’s reports on her family-planning experience: whether she reported any birth resulting from a pregnancy that she wanted to have eventually but that occurred “too soon” and whether she reported any pregnancy terminated by abortion. Social and economic variables include race or ethnicity, education of both spouses, wife’s religion, and region and size of place of residence.
RESULTS Marital Status Composition of Sterilizations We noted in our overview that a significant proportion of never-married women report that they are sterilized, even when the denominator includes all women of reproductive age. High levels of sterilization among unmarried women are even more apparent in Table 1: one third of all recent tubal sterilizations were performed on unmarried women. Sterilization while unmarried is remarkably common among all race or ethnic groups. Nonetheless, the differences 939
in level are so large that it is best to describe the numbers separately. Approximately one fifth of the tubal sterilizations on most white women were performed while they were unmarried: about half of these while never married and about half while formerly married. Approximately 9% of the sterilizations on white women were performed while they were in cohabiting relationships. Patterns are similar among Mexican-American women, but cohabitation plays a larger role, accounting for approximately 15% of sterilizations. Among black women, on the other hand, only a minority (37%) was performed on married women, and nearly as many (31%) were performed on never-married women. Almost one fifth of the tubal sterilizations in black women was performed while they were living in a cohabiting relationship. Sterilization among unmarried women appears to be part of a normal life course from the perspective of fertility. Age and parity at the time of sterilization are similar for unmarried and married women, e.g., never-married women had an average age of 30 and an average parity of 2.8 compared to an identical age and an average parity of 2.6 among married women. Furthermore, the effect of marital status becomes small and nonsignificant once parity is controlled (tables not shown). Hence, achieved fertility is far more important than whether a woman is married with respect to the decision to have a tubal sterilization. The high proportion of female sterilizations that now occur outside of marriage seems to be entirely due to the fact that many women have all the children they want, either while never married or after marital disruption. Female sterilization is so much more common than male sterilization, probably because having a partner choose vasectomy is much less an option for unmarried women. We were able to examine marital status at sterilization for men by using the 1987–1988 NSFH. Although there were too few cases of male respondents reporting on their own sterilization in these data to permit the level of detail in Table 1, we found that approximately 7% of vasectomies between 1980 and 1987 occurred outside of marriage. The much higher level of unmarried sterilization among women than among men indicates that women consider sterilization relative to alternatives in the context of having to raise children already born. Men, on the other hand, may often have little connection with children they have fathered while unmarried (and in some cases may not even know they were born) (16). Hence, it is not surprising that sterilization among unmarried men occurs much more often after rather than before first marriage. Because these findings indicate that sterilization is often an individual decision (particularly among women and particularly among black women) and that most other available descriptive data are based on “contraceptive users,” it seems useful to consider a life course description of contraceptive patterns for both men and women that is contingent on neither marriage nor factors affecting current use of any 940
Bumpass et al.
Women, men, and sterilization
FIGURE 1 Life-table estimates of vasectomy (✚) and tubal sterilization (F) by years since last wanted birth: last wanted births 1980 – 1994.
Bumpass. Contraceptive sterilization. Fertil Steril 2000.
method. Sterilization might be expected to be concentrated at older ages. Nonetheless, because of the dispersion in parity and age at last wanted birth, there is a linear increase in the proportion of women sterilized from 0 at age 20 to 36% by age 40. Among men, the increase begins later but is again linear from low levels in the early 20s to approximately 14%. (These cumulative proportions are consistent with the proportions reported among 40- to 44-year-old women who are currently using contraception: 50% and 20%, respectively, for the 71% who were contraceptive users) (2). Concerns with the “choice” between female and male sterilization must be cognizant of the extent to which female sterilizations occur outside of marriage. Nonetheless, couples’ choices in sterilization remain important, and we turn now to an analysis of factors affecting these choices.
Couple Choices of Male and Female Sterilization We focus this analysis of couple choice on the first 5 years after the birth of the last child wanted. Figure 1 shows that for both males and females the highest rates of sterilization occur in the first year after last wanted birth, with a gradual increase thereafter. Three quarters of all sterilizations occur within the 5-year period we are examining, with the first year being much higher for females, probably because of the prevalence of postpartum sterilization. Although an array of reduced-form models have been explored, the results are most easily communicated from a model that includes all of the variables considered. There are eight columns in Table 2. The first column for each of these choices (1 and 4) is the simple proportions observed to have had a sterilization of this type within 5 years of the last Vol. 73, No. 5, May 2000
TABLE 2 Tubal sterilization within 5 years and vasectomy within 5 years of last wanted birth: observed proportions and relative risk estimates from multinomial logistic regression: births 1980 –1994 (1980 –1989 for proportions). Tubal sterilization vs. none
Parameter Parity last wanted 1 2 3 ⱖ4 Age at last wanted birth (y) ⬍25 25–29 30–34 ⱖ35 Husband is younger by 2 or more years No Yes Husband is older by 5 or more years No Yes Duration of marriage (y) ⬍5 5–9 ⱖ10 Married before No Yes Husband was married before No Yes Age at first birth (y) ⬍20 20–24 25–29 ⱖ30 Had unintended pregnancy No Yes Had an abortion No Yes Race or ethnicity White non-Hispanic Black non-Hispanic Hispanic Education Did not finish high school High school graduate Some college College graduate Husband is less educated than wife No Yes Husband is more educated than wife No Yes
(1) Percent in 5 y
(2) Relative odds in 5 y
12 34 46 44
(3)
Tubal sterilization vs. vasectomy
Vasectomy vs. none
P value
(4) Percent in 5 y
(5) Relative odds in 5 y
4.1 7.7 9.6
.00 .00 .00
12 21 24 28
42 35 23 36
0.9 0.7 0.7
.66 .07 .21
35 38
1.6
35 36
(6) P value
(7) Relative odds in 5 y
(8) P value
3.4 6.8 6.0
.00 .00 .00
1.2 1.1 1.6
.49 .71 .25
15 24 21 16
1.7 1.3 0.8
.02 .35 .67
0.5 0.5 0.8
.01 .03 .62
.02
21 16
0.9
.81
1.7
.05
0.7
.03
22 18
0.6
.00
1.3
.17
35 33 42
1.0 1.3
.99 .21
18 23 19
0.9 0.7
.40 .25
1.2 1.8
.43 .04
34 45
1.2
.30
21 17
0.6
.05
1.9
.01
34 46
1.8
.00
21 17
2.4
.00
0.8
.20
52 37 29 18
1.0 1.0 0.9
.94 .93 .70
15 22 24 15
1.0 1.0 1.0
.84 .98 .96
1.0 1.0 0.9
.88 .97 .80
31 40
1.2
.20
22 19
0.9
.34
1.3
.06
37 28
0.7
.01
20 21
0.9
.64
0.7
.14
32 57 50
2.2 1.8
.00 .00
23 6 10
0.3 0.4
.00 .00
6.6 4.1
.00 .00
48 39 37 23
0.9 0.7 0.4
.75 .18 .00
8 23 20 22
2.1 2.0 1.8
.02 .04 .11
0.4 0.4 0.2
.01 .00 .00
35 37
1.2
.22
21 18
0.8
.23
1.5
.04
36 33
0.7
.01
19 24
1.1
.76
0.7
.02
Continued on following page
FERTILITY & STERILITY威
941
TABLE 2—CONTINUED Tubal sterilization vs. none
Parameter Religion Catholic Protestant No religion Other non-Catholic Region in United States Northeast Midwest South West Size of place Central city SMSA other non-SMSA
(1) Percent in 5 y
(2) Relative odds in 5 y
31 39 35 22
Tubal sterilization vs. vasectomy
Vasectomy vs. none
(3) P value
(4) Percent in 5 y
(5) Relative odds in 5 y
1.6 1.8 0.8
.00 .01 .33
18 22 15 25
31 37 40 30
1.1 1.3 0.6
.75 .08 .00
38 32 41
0.9 1.1
.43 .75
(6) P value
(7) Relative odds in 5 y
(8) P value
1.2 0.9 0.8
.26 .57 .57
1.3 2.1 0.9
.08 .02 .79
16 22 19 24
1.6 1.8 2.0
.01 .07 .00
0.6 0.9 0.3
.04 .63 .00
15 21 24
1.1 1.3
.49 .16
0.8 0.8
.22 .27
Note: A variable for years since last wanted coded 0 –5 was included to control for censoring in the 1990 –1995 period. SMSA ⫽ standard metropolitan statistical area. Bumpass. Contraceptive sterilization. Fertil Steril 2000.
wanted birth. This provides both a description of population variation and a sense of scale. The second column for each choice (2 and 5) reports odds ratios from the multinomial logit: the relative likelihood of having had an operation of the given type contrasted to having had no sterilization within 5 years of the last wanted birth. The third column (3 and 6) is simply the P value of the coefficient (the coefficient itself is not reported). In discussing these results, we do not repeat the “within 5 years” constraint each time proportions or relative risks are cited. The final two columns of Table 2 contrast the odds ratio and P value of tubal sterilization in contrast with vasectomy among couples adopting one or the other of these sterilization methods. Parity, Age, and Duration of Marriage at Last Wanted Birth The number of children a couple has, their age, and how long they have been married may affect how sure they are that they do not want more children and, thereby, affect the likelihood of sterilization. The main story of this set of variables can be told quite simply: the parity of the last wanted birth dominates the effects of the other variables, although other effects of interest remain after parity is controlled. A positive relationship between parity and the likelihood of sterilization would be expected if couples at lower parities are less sure that they have indeed completed their family, and if— because of previous planning failures—those at higher parities are more highly motivated to end childbearing. It is also possible that, despite our attempt to take into account births from unwanted pregnancies, some higher 942
Bumpass et al.
Women, men, and sterilization
parity women may have had an unwanted birth before the one coded as “last wanted,” and such experience would have raised their motivation to be sterilized even further. Finally, age differences could index cohort more than age, i.e., older women have lived more of their lives during a period when sterilization was less accepted. As noted earlier, observed differences in sterilization do not represent differential rates if groups differ in the proportions who have had all the children they want. More sterilizations occur among women with two children than at any other parity (17), but this is so because most women prefer a family size of two children. However, the likelihood of sterilization increases consistently with parity. Between parities 1 and 4, the observed proportions sterilized range from 12% to 44% for tubal sterilization, and from 12% to 28% for vasectomy. The differences in relative risks from the logistic model are large and significant, and the effect is greater for female operations than for male. Even so, among one-child families, one quarter of the couples were sterilized within 5 years: 12% each for male and female operations. As with parity, age would increase the likelihood of sterilization if it were associated with increased confidence in the decision to cease childbearing. On the other hand, age could have negative effects, if women in their late 30s or early 40s feel that they are too close to menopause to make sterilization worthwhile. These women might also see the few remaining years as low risk because of declining fecundity. Age at the birth of a woman’s last wanted child has little effect after parity is controlled, although the rate of vasectomy is significantly higher among couples in which the wife Vol. 73, No. 5, May 2000
is in her late 20s. A significant curvilinear effect is in the last two columns: couples are least likely to adopt tubal sterilization relative to vasectomy if the wife is 25–34. These results may reflect cohort rather than age effects, in conjunction with less certainty about ending childbearing among the youngest couples. The effects of husband’s age became clearest when we classified it relative to wife’s age: those 2 or more years younger than their wife and those 5 or more years older. There are significant effects of both of these variables. Wives with substantially younger husbands are more likely to have a tubal sterilization, and those whose husbands are older by ⱖ5 years less likely. Husbands who are older than their wives are also less likely to adopt sterilization for themselves. These patterns seem more consistent with a cohort hypothesis that older males are more opposed to sterilization in general than with the gender role theories that are usually brought to the interpretation of age differences (18, 19). Duration of marriage at the time of the last wanted birth has no significant net effects, other than a marginally significant higher likelihood of tubal sterilization compared with vasectomy among couples who have a sterilization and have been married the longest. Prior Marital and Reproductive History We would expect that women who were married before, those who married at an early age, and those who have experienced unintended births or abortions would be more likely to adopt sterilization. The results are only partially consistent with these expectations. It is the husband’s prior marriage that has a strong positive effect on both tubal sterilizations and vasectomies (odds ratios of 1.8 and 2.4, respectively). On the other hand, the wife’s prior marriage decreases the likelihood of vasectomy and, hence, creates a greater preponderance of tubal sterilizations among those who are sterilized. The absence of a positive effect for the wife’s prior marriage on her own sterilization and the negative effect on her husband’s are just the opposite of what we would have expected. That is, we would have expected previously married wives to be more likely to adopt tubal sterilizations because of a concern about future instability given that women overwhelmingly remain responsible for children. It is possible that the latter effect might be interpreted in terms of a desire on the part of husbands of previously married wives to keep reproductive options in case this relationship disrupts. However, the significantly higher sterilization rates for both husbands and wives when the husband has been married before does not fit with this explanation but may, instead, reflect the effects of the husband’s obligation to children from his prior marriage. The observed proportions of wives with a tubal sterilization increase dramatically the younger she was when she FERTILITY & STERILITY威
bore her first child. This finding is completely due, however, to the correlation of age at first birth with parity of the last wanted birth. We also do not find the expected effects of prior planning failures except that among couples adopting sterilization there is a marginally significant (P⫽.06) greater likelihood that the wife will have been sterilized compared with the husband. This would be consistent with a greater concern on her part about the possibility of a future unplanned pregnancy. Although the results for women who reported an abortion make sense, post hoc, we had not anticipated the effect. Women who reported an abortion are less likely to be sterilized. This could reflect their personal experience that abortion is viable. As with the other variables we have considered, male sterilization behavior is not responsive to this prior experience with family planning failure that resulted in an abortion. Social and Demographic Variables The most striking differential in choice of tubal sterilization over vasectomy is associated with race. In 1970, tubal sterilizations were more than twice as prevalent for black women as for white women, whereas only 1% of “at risk” black couples had adopted vasectomy, compared with 9% of white couples (1). Tubal sterilizations are no longer twice as prevalent among black women than among white women, but large disparities in both male and female procedures remain. In fact, although the prevalence of vasectomy has increased substantially for whites, it has increased hardly at all for blacks (18). As expected, the race or ethnicity contrasts between rates of tubal sterilization and vasectomy (next to last column in Table 2) are exceedingly large. As we developed in an earlier section, some of the predominance of female compared with male sterilization among blacks can be accounted for by the greater time spent unmarried and the marked increase in sterilization among unmarried black women. The results in Table 2 underscore race or ethnic differences in sterilization behavior, because these estimates are based on continuously married couples and count only behavior subsequent to a last wanted birth in their current marriage. Few black or Hispanic husbands adopt sterilization, and the very high rates of tubal sterilization among their wives can be seen, in part, as a response to the lack of that option. Whether the mediating link between minority status and low levels of vasectomy is the perceived higher risk of marital instability or apprehension about the consequences of vasectomy are issues to which we will return in our concluding section. Nonetheless, the higher rate for black wives more than compensates for the low rate among black husbands. The combined proportions sterilized within 5 years of their last wanted birth are 63% for black couples, 60% for Hispanic, and 55% for whites. As with age, because of the high correlation between 943
husband’s and wife’s education, we entered wife’s education into the equation and then a difference variable for the husband (based on a classification of education into the same categories as reported for wives). Confirming earlier analyses, wife’s education is negatively related to tubal sterilization. The proportion sterilized within 5 years declines from 48% among those not completing high school to 23% among college graduates, and the net relative risks show a monotonic decline (though only the college graduate contrast is significant). Part of this strong relationship may reflect ready access to tubal sterilization, but not to vasectomy, at the clinics used by low-income women (20). There is an additional depressing effect on tubal sterilization if the husband has more education than his wife.
tion are at least 40% lower in the West than in other regions, and rates of vasectomy are twice as high in the West than they are in the Northeast. This can be seen most clearly in the last set of columns, where among those adopting sterilization the relative risk of tubal sterilization to vasectomy is 70% lower in the West than it is in the Northeast. This is a very large difference, and it is consistent with our expectations about the possible effects of extensive HMO coverage. Because vasectomy costs much less than tubal sterilization, HMO physicians may recommend vasectomy more strongly to their patients than do physicians in other types of practice. It may also be the case that the West may simply be an area that is better informed in general about the risks and benefits of vasectomy.
The effects of wife’s education on vasectomy are not monotonic: the major contrast is between husbands of wives who did not complete high school and all higher levels of schooling. So the emerging positive effect of education on vasectomy observed in our earlier analysis (3) is not found here, neither in the present model nor in one considering only husband’s education. Among couples who adopt sterilization, wife’s education significantly reduces the likelihood of tubal sterilization relative to vasectomy. Compared with couples with the same level of education, tubal sterilization predominates among those in which the husband is less educated, but vasectomy rates are higher among those in which he is more educated than his wife. These patterns are consistent with our expectations that vasectomy would be more attractive to better educated husbands; however, that result only appears in this contrast with wives’ education.
The higher observed proportions with vasectomies in central cities and the lower proportions with tubal sterilizations disappear once the other variables are controlled, and size of place has no significant effect on rates of sterilization.
As expected, net of other factors, Catholic wives are substantially and significantly less likely than others to have a tubal sterilization. Even so, it should be emphasized that 31% of Catholic wives have been sterilized within 5 years of their last wanted birth. On the other hand, there are only modest and nonsignificant differences by religion in vasectomy. When vasectomy and tubal sterilization are combined, 49% of Catholic couples have adopted contraceptive sterilization within 5 years of their last wanted birth compared to 61% of Protestants. It is important to recognize that sterilization is not completely determined by individual or couple decisions. Regional variations in physicians’ attitudes toward sterilization or in the medical care delivery system may also limit or facilitate sterilization for those who have achieved their desired family size. Our earlier work noted regional differences, and the possibility that lower rates of tubal sterilization in the Northeast might reflect the influence of Catholic doctrine in that region on accessibility quite apart from an impact on individual behavior (3). We do not replicate this earlier finding, although we do find significantly higher rates of vasectomy outside of the Northeast, particularly in the West. Rates of tubal steriliza944
Bumpass et al.
Women, men, and sterilization
In conclusion, contraceptive sterilization is an extremely important factor in fertility and family processes in the United States, and analyses of the factors affecting the relevant decision processes have been too neglected in nationally representative data. The observation that sterilization is now the leading method of birth control in the United States substantially understates its importance, because it is not an appropriate method until the decision to end childbearing is held with confidence. Our high level of contraceptive sterilization is a definitive statement by men, women, and couples about their reproductive intentions. Given the long period at risk of unwanted pregnancy that follows the last wanted birth for most women, levels of unwanted pregnancy would undoubtedly be much higher if it were not for the high levels of protection afforded by sterilization; abortion rates would be higher as well. The analysis of choices in sterilization must relate to the time since sterilization becomes a viable option, i.e., after no more children are desired. Furthermore, we must recognize that such decisions are often made by individuals rather than by couples. That one third of all recent tubal sterilizations occurred outside of marriage draws our attention to a large proportion of women for whom vasectomy is not likely an option. Our multivariate analysis of factors affecting choices of tubal sterilization and vasectomy among continuously married couples documents a number of patterns that raise further questions for future research. We will need to begin to gather data that address potential mediating mechanisms in the observed differences in sterilization behavior. The dominance of parity in a study population already defined in terms of risk underscores the potential importance of the role of uncertainty in sterilization choice processes. At the same time, however, we must recall that medical practice once imposed parity constraints on access to sterilization (1), and it is possible that such rules or practices persist more than we Vol. 73, No. 5, May 2000
realize and are revealed, along with differential certainty, in the dramatic parity effects that we find. Such possible effects from the supply side—illustrated as well by the strong regional effects—make it clear that we need a provider survey to tap both the attitudes and relevant behaviors of potential providers of tubal sterilization and vasectomy. Such measures may help us better understand the prevalence of these operations relative to no sterilization and relative to each other. The effect for women who have been married before suggests a concern with the possible future marital instability, an effect we have not yet been able to definitively measure. The most dramatic differences that are yet unexplained are the low levels of vasectomy (and compensating high levels of tubal sterilization) among minority women. Part of this difference results from differing lengths of time spent unmarried, but the differential persists in our analysis of continually married couples. It seems likely that measuring concerns about marital stability and future fertility will help in understanding these differences, but it is also possible that such differences result in part from differential contact with the medical system. Minority women are more likely than other women to be in direct contact with this system, whereas minority men may have much less contact than other men. One of the largest remaining puzzles is why rates of vasectomy are so much lower than rates of tubal sterilization. Part of the answer lies in the marital status of women when they decide that they want no more children, and part lies in the avoidance of vasectomy among minorities. But even for majority whites, where the two operations are closer to equally represented, we do not understand why vasectomy is not the primary method of sterilization among stable monogamous couples. Our discussions of necessity for certainty about intentions not to have another child and couple issues in negotiating that certainty should make it clear that studying sterilization decisions will require simultaneous attention to the processes by which individuals and couples decide that they want no more children. Hence, part of the future agenda must include the incorporation of the negotiation between spouses of certainty about ending childbearing and about sterilization decisions and in the attempt to uncover, as well, the extent to which these decisions may be made individually (rather than jointly) even among married couples. In addition, however, given the importance of these methods and the role they play in the control of American fertility, we need to seriously address perceptions of the characteristics of tubal sterilization and vasectomy that may affect their adoption, and we need to collect such data in nationally representative samples. This will require attention to issues of perceived safety (both short term and long term), effects on sex life, concern with keeping reproductive opFERTILITY & STERILITY威
tions open, and the relative evaluation of the preferences for which partner should have a sterilizing operation. For example, our observations about why female operations predominate when vasectomies are cheaper and safer implicitly assume that women would prefer that their spouse have the operation if only they would—it may not be so. Attitudinal data would have the greatest potential in longitudinal designs predicting subsequent sterilization behavior as couples or individuals reach the point where the decision becomes relevant, but they may also inform our understanding of differences even if they are measured only in the cross section. An example from the early fertility studies is a case in point, 23% of white respondents and 44% of black respondents to 1965 National Fertility Study (NFS) agreed that vasectomy would impair a man’s sexual ability (21). These proportions had declined considerably by the 1970 NFS but still remained considerably higher among blacks (1). Because studies of marriage and divorce grew out of the data collected to study fertility, the study of family processes on the one hand (NSFH), and fertility on the other (NSFG), have tended to drift apart in national data collection efforts. It is time to reincorporate fertility processes as family processes. The fact that sterilization occurs increasingly outside of marriage makes this all the more important. Decisions to end childbearing and to consider sterilization are a critical stage in the family life course. References 1. Bumpass L, Presser H. Contraceptive sterilization in the U.S.: 1965 and 1970. Demography 1972;9:531– 49. 2. Piccinino LJ, Mosher WD. Trends in contraceptive use in the United States: 1982–1995. Fam Plann Perspect 1998;30:4 –10. 3. Bumpass L. The risk of an unwanted birth: the changing context of contraceptive sterilization in the U.S. Popul Stud 1987;41:347– 63. 4. Chandra A. Surgical sterilization in the United States: prevalence and characteristics, 1965–95. Vital Health Stat 23, DHHS Publication No. (PHS) 98-1996, 1998. 5. Morgan SP. Intention and uncertainty at later stages of childbearing: the United States 1965 and 1970. Demography 1981;18:267– 85. 6. Godecker AL. Gender role attitudes and marital quality in female and male sterilization: longitudinal evidence from the National Survey of Families & Households (Master’s thesis). Center for Demography and Ecology, University of Wisconsin, Madison, 1997. 7. Shain RN, Miller WB, Holden AEC. Factors associated with married women’s selection of tubal sterilization and vasectomy. Fertil Steril 1985;43:234 – 44. 8. Miller WB, Shain RN, Pasta DJ. Tubal sterilization or vasectomy: how do married couples make the choice? Fertil Steril 1991;56:278 – 84. 9. Miller WB, Shain RN, Pasta DJ. A model for determinants in married women of sterilization method choice. Popul Environ 1985– 86:8: 223–39. 10. Groat T, Neal AG, Wicks WJ. Sterilization anxiety and fertility control in the later years of childbearing. J Marriage Fam 1990;52:249 –58. 11. Wu L, Bumpass L, Musick K. Historical and life course trajectories of nonmarital childbearing (working paper 99-23). Center for Demography and Ecology, 1999. 12. Potter V, Iannacchione F, Mosher W, Mason R, Kavee J. Sample design, sampling weights, imputation, and variance estimation in the 1995 National Survey of Family Growth. Vital Health Stat 2, No. 124, 1998. 13. Sweet J, Call RA, Bumpass L. The design and contend of the National Survey of Families and Household. Center for Demography and Ecology, NSFH WP-01, 1988. 14. Rindfuss R, Palmore J, Bumpass L. Selectivity and the analysis of birth intervals with survey data. Census Forum 1882;8:5–10.
945
15. Agresti A. Categorical data analysis. New York: John Wiley and Sons, 1990. 16. Furstenberg F, Cherlin A. Divided families: what happens to children when parents part. Cambridge: Harvard University Press, 1991. 17. Bumpass L, Thomson E, Godecker AL. Women, men, and contraceptive sterilization. Center for Demography and Ecology, Working paper 99-05, 1999. 18. Forste R, Tanfer K, Tedrow L. Sterilization among currently married men in the United States, 1991. Fam Plann Perspect 1999;27:100 –7, 122.
946
Bumpass et al.
Women, men, and sterilization
19. Bean FD, Williams DG, Opitz W, Burr JA, Trent K. Sociodemographic and marital heterogamy influences on the decision for voluntary sterilization. J Marriage Fam 1987;49:465–76. 20. Haws JM, McKenzie M, Mehta M, Pollack AE. Increasing the availability of vasectomy in public-sector clinics. Fam Plann Perspect 1997; 29:185– 6, 190. 21. Bumpass L, Westoff CF. The perfect contraceptive population: the extent and implications of unwanted fertility in the U.S., 1960 – 65. Science 1970;169:1177– 82.
Vol. 73, No. 5, May 2000