Adolescent Contraceptive Method Choices

Demography, Vol. 27, No.3, August 1990

Adolescent Contraceptive Method Choices Joan R. Kahn Department of Sociology and Center on Population, Gender and Social Inequality, University of Maryland, College Park, Maryland 20742 Ronald R. Rindfuss Department of Sociology and Carolina Population Center, University of North Carolina, Chapel Hill, North Carolina 27514 David K. Guilkey Department of Economics and Carolina Population Center, University of North Carolina, Chapel Hill, North Carolina 27514 This article analyzes determinants of contraceptive method choices among adolescent women in the United States. By using data from the 1982 National Survey of Family Growth, we examine factors that differentiate users of various methods early in the sexual careers of teenaged women. We find that patterns of method choice not only vary by race and region within the United States but also change over the teenager's life course. In addition, among teenagers who did not use a method at first sex, the likelihood of adopting a method soon thereafter was low for both whites and blacks and was unaffected by social structural characteristics.

The topic of adolescent contraceptive use has become increasingly salient and simultaneously more complex. The consistently high rates of teenage fertility and the rapid spread of sexually transmitted diseases (STDs) have combined to raise the concern of parents, policymakers, health-care professionals, and researchers alike. Yet understanding adolescent contraceptive behavior is especially complex because of the variability in teenagers' contraceptive needs and motivationsas well as the desirability and effectiveness of different methods. Even though there has been an overall decline in fertility levels since the early 1960s, the United States stands out among developed societies as having one of the highest rates of adolescent childbearing, both in and out of wedlock (Cutright & Smith 1986; Jones et al. 1986; Jones, Kahn, Parnell, Rindfuss, & Swicegood 1985). Adolescent pregnancy and childbearing have been and continue to be major policy concerns at both the local and national levels. [See Hayes (1987) for a summary of research commissioned by the National Academy of Sciences on this topic.] The persistent debate over the availability of abortion continues to remind us of the importance of adolescent contraceptive use. Furthermore, the recent rise and spread of such incurable or difficult to treat sexually transmitted diseases as chlamydia, genital herpes, and AIDS have heightened general concern over the prophylactic benefits of condoms. Although recent cohorts of teens are more likely than earlier cohorts to practice contraception, I they are still much less likely than older women to do so (Bachrach 1984; Mosher & Bachrach 1987). This article analyzes the process whereby teenage women adopt contraception early in Copyright © 1990 Population Association of America

323

324

Demography, Vol. 27, No.3, August 1990

their sexual careers. We start by examining the determinants of contraceptive behavior at the time of first intercourse. In addition to focusing on whether contraception was used, we also examine choices among different methods within a multivariate context. Then by restricting our focus to teens who did not use a method at first intercourse, we analyze the determinants of the pace at which nonusers eventually become users.

Background Research on teenage contraceptive behavior suggests a variety of explanations for teens' low levels of contraceptive use. These range from their youth and inexperience to the fact that they tend to be unmarried and possibly not exposed to frequent and predictable sexual activity (Bachrach & Horn 1988; Kisker 1985). Because of their youth, teens may have less knowledge about the consequences of unprotected sex (i.e., the risks of pregnancy and STDs), the methods of contraception available, or their proper use (Cvetkovich & Grote 1981; Kantner & Zelnik 1972; Zelnik & Kantner 1980; Zelnik & Shah 1983). They may also operate under a false sense of security or a sense of invincibility by assuming that they could not possibly get pregnant or be infected with a sexually transmitted disease (Moore, Simms, & Betsey 1986). Furthermore, teenagers may be too embarrassed or lack the resources to obtain contraceptives from a pharmacy or family planning clinic. They may resist "planning" for sex because they are not yet comfortable with the notion of being sexually active or because they feel it reduces the spontaneity of a sexual encounter (Cvetkovich & Grote 1981; Zelnik, Kantner, & Ford 1981). Because teens are less likely to be in stable relationships, they may also be less able to predict when they will have sex, not perceive a great need for practicing contraception if they are not "very" active, or feel less comfortable discussing contraception and STDs with a partner they do not know well (Bachrach & Horn 1988; Thompson & Spanier 1978). Many of these same factors are likely to influence the type of method chosen by teenagers. For example, younger, inexperienced teens may prefer natural or over-the-counter methods (e.g., withdrawal, spermicides, condoms) rather than methods like the pill, which requires a doctor's visit, or rhythm, which requires sophisticated knowledge of reproductive physiology. Furthermore, to the extent that adolescent relationships are unstable, the need for protection against conception may lead to the choice of such coitus-independent methods as the pill. On the other hand, if their relationships are unstable and they have multiple sex partners, adolescents may choose condoms as protection against STDs. If disease prevention is their primary goal, then condoms would be the clear method of choice, since it is the only method other than abstinence that can protect against the spread of STDs. Yet it is not clear how many teenagers actually consider the risks of STDs. Certainly, in recent years the publicity about AIDS and "safe sex" has raised the awareness of many people. In 1982, however, when the data used in this analysis were collected, AIDS had not yet entered the scene and genital herpes had only recently been publicized in magazines aimed at teenagers (Mosher & Bachrach 1987, p. 86). In addition, a wide variety of other considerations may enter the adolescent's choice calculus. For example, the most effective methods of birth control, such as the pill or intrauterine device (IUD), may be less desirable for teens because they are costly and require a doctor's visit prior to use. Younger teens who are not fully committed to becoming sexually active may be uncomfortable with these methods because they require public confirmation of their activity (i.e., through a doctor's visit). Less effective, coital methods (e.g., condom, spermicides) may be more accessible to teenagers but may interfere with romance and spontaneity. Finally, since sexual intercourse involves two individuals rather than just one, the contraceptive decision for any given sexual encounter will reflect the motivations and preferences of both individuals.

Adolescent Contraceptive Method Choices

325

Compared with any other method, condoms are the only method to protect against both pregnancy and the spread of STDs and thus have received heightened attention in the post-AIDS environment. They are easy to obtain and use, and they do not involve longterm physiological changes. Furthermore, they have been shown to be more effective at preventing pregnancy among teens than either rhythm or withdrawal, both of which are available (in principle) to everyone (Grady, Hayward, & Yagi 1986; Jones & Forrest 1989). As teens gain more experience with sex and "adult" life in general, however, and as they enter into stable unions, we might expect them to place more emphasis on pregnancy prevention and convenience and, therefore, to use "medical" methods like the pill. 2 Results from both the 1979 National Survey of Young Women and the 1982 National Survey of Family Growth (NSFG) confirm that the condom is a very popular method at first intercourse among teenagers (more than one-third of all users, compared with only 20% who used the pill; see, respectively, Zelnik & Shah 1983, Pratt Mosher, Bachrach, & Horn 1984). After first intercourse, however, teenagers are progressively more likely to use medical methods and less likely to use condoms (Mosher & Bachrach 1987). Few studies have examined the socioeconomic determinants of patterns of teenagers' method choice. A number have looked at the correlates of contraceptive use at first intercourse and have found that teens who initiate sex at a later age, describe their first intercourse as "planned," and come from higher status, intact families are more likely to use contraceptive methods at first intercourse (Ford, Zelnik, & Kantner 1981; Guilkey, Rindfuss, & Kahn 1989; Hogan, Astone, & Kitagawa 1985; Mosher & Bachrach 1987). These studies, however, provide only a limited view of teenage contraceptive behavior because they focus almost exclusively on the first sexual act. This is unfortunate, given that contraceptive decisions seem to change so much during the early stages of a person's sexual career. The present analysis extends the past work in this area by examining the factors associated with the pace at which contraception is adopted. Another limitation of most previous studies is their focus on whether contraception was practiced rather than on which method was chosen. By not considering the types of factors that differentiate users of different methods (e.g., condom vs. pill), these studies ignore much of the complexity of contraceptive behavior. Not only may the motivations for using various methods differ (e.g., pregnancy vs. disease prevention), but the impact of the social context may vary substantially by the type of method. Although several studies have examined teenagers' method choices, they tend to limit their focus to decisions made by teens who use contraceptive methods and therefore do not consider nonuse to be an option (Bachrach 1984; Mosher & Bachrach 1987). Yet because none of the methods chosen by teens is irreversible, it is likely that teen contraceptors continue to weigh the risks and benefits of using different methods with the risk of not using any method at all. Thus models of contraceptive decision making should include nonuse as a possible choice. Although this was done in a recent study of method choices among married women (see Stephen, Rindfuss, & Bean 1988), the results are unlikely to apply to unmarried teenagers because they are likely to face very different circumstances (e.g., they tend to be younger and are less likely to be in permanent sexual relationships).

Data and Variables The data for this analysis come from the 1982 National Survey of Family Growth (Cycle III), a nationally representative sample of women of childbearing age (15-44), with oversamples of teenagers and blacks (Bachrach, Mosher, Horn, & Shimizu 1985). Unlike earlier fertility surveys, which were limited to ever-married and/or ever-pregnant women, the 1982 NSFG is representative of all women regardless of marital or childbearing status. We limit our analysis to the subsample of young women (aged 15-24 in 1982) who had

326

Demography, Vol. 27, No.3, August 1990

premarital sex as teenagers. We include women in their early 20s in part to increase the sample size but, more important, also because the overwhelming majority of these women had become sexually active as teenagers (Bachrach & Horn 1987; Hofferth, Kahn, & Baldwin 1987). We focus on premarital behavior because we expect that the nature of sexual activity as well as its consequences are different for unmarried than for married women. 3 Throughout, we examine the experiences of white and black women separately because prior research has found numerous interactions between race and the determinants of contraceptive behavior (Stephen et al. 1988). Our information on contraception COmes from several questions about the first method ever used. Respondents were also asked, relative to first intercourse, when they used their first method. Was it (1) at first intercourse? (2) within a month? (3) 1-3 months later? (4) 4-6 months later? (5) 7-12 months later? or (6) more than 12 months later? From these questions, we can determine which method was used at first intercourse, as well as the timing of first use relative to first intercourse. Our measure of contraceptive method choice is coded into four categories: condom, pill/IUD, other methods, and none. Other methods is a residual category consisting primarily of withdrawal (more than 70% of other-method users at first intercourse). Because the IUD was used by less than 1% of all young women at first intercourse, the pill/IUD category can simply be referred to as pill. By considering nonuse to be an option, we do not assume, as others have done, that individuals make a two-stage decision: first whether or not to use, and then, which method to use. Rather, we assume that individuals continue to weigh the risks of not using a method with the risksassociated with each of the other available methods (Rindfuss, Swicegood, & Bumpass 1989). Our choice of explanatory variables was guided by recent research on the sociodemographic determinants of adolescent behavior but was limited by the variables contained in the 1982 NSFG. Many of these are variables that other studies have found to be significant in predicting whether contraception is used (Ford et al. 1981; Hogan et al. 1985; Mosher & Bachrach 1987; Rindfuss et al. 1989). These include three measures of family background as indicators of socioeconomic status, family stability, and religion: mother's education (coded in years completed), whether the respondent came from an intact family (dummy-coded, where 1 = lived with both natural parents at age 14), and religious affiliation (coded as Protestant nonfundamentalist," Catholic, Protestant fundamentalist, other/none). We also include two measures of regional context: whether the respondent lives in a metropolitan area (dummy-coded, where 1 = lives in a standard metropolitan statistical area) and region of residence (coded as South, Northeast, North Central, and West). 5 Finally, as an indicator of the respondent's stage of development and pace of physical maturation, we include age at menarche (Hofferth et al. 1987).

Analytic Issues Most studies of teenage contraceptive behavior have assumed the same basic decisionmaking model: First, teenagers decide to become sexually active; only then does the decision whether to practice contraception become relevant; and finally, after having decided to use some kind of method, teenagers will then choose among the various methods. This underlying assumption is evidence in most analyses of contraceptive use at first intercourse that use age at first intercourse as an explanatory variable. It is quite likely, however, that decisions about sex and contraception are closely linked, with decisions about each affecting the other. For some people, decisions about sexual and contraceptive behaviors might in reality be a single decision (or a nondecision as some authors suggest). Guilkey et al. (1989) showed that decisions about the timing of first intercourse cannot be assumed to be causally prior to the contraceptive decision. 6

327

Adolescent Contraceptive Method Choices

In light of the foregoing considerations, we do not assume a single causal ordering of decisions about sexual activity and contraception, but rather, we allow for the possibility that each may influence the other. Thus becoming sexually active may encourage contraceptive choices, or the availability of or willingness to use different methods of contraception may influence the timing of first sex. In examining method choice at first intercourse, we therefore do not include controls for age at first intercourse, but rather, we estimate reduced form equations. Even though the full causal structure of the determinants of both sexual and contraceptive behaviors is of theoretical interest, at present we have neither the data nor the appropriate multivariate modeling techniques to press the issue further. Because the dependent variables in the analysis involve four distinct choices, we use polytomous logistic regression as the estimation technique (Hanushek & Jackson 1977; Maddala 1983; Pindyck & Rubinfeld 1981). As an extension of the dichotomous logistic regression model, the polytomous model allows for the simultaneous estimation of the logodds of more than two outcomes. The coefficients represent the reduction or increase in the log-odds of being in one outcome category relative to another. In addition to presenting coefficients, we also present predicted probabilities of each choice, derived by evaluating the regression model for different combinations of explanatory variables.

Results Method Choice at First Intercourse Previous analyses have shown that teenagers from higher socioeconomic status, intact families are more likely to use some method at first intercourse than other teenagers (Ford et a1. 1981; Guilkey et a1. 1989; Mosher & Bachrach 1987). Do these same factors differentiate users of different methods? Table 1 provides the results from the polytomous logit analysis of method choice at first intercourse for whites and blacks. In the first three columns, condoms are compared with each of the other choices, whereas in the last three columns, other choices are compared with each other. The coefficients indicate the impact of the explanatory variable on the log-odds of choosing the first relative to the second choice in the comparison (e.g., condom vs. none in the first column). It is visually apparent in Table 1 that the strongest effects are in the comparisons of condom users with each of the other method categories. In particular, among whites the largest number of significant effects are present in the condom versus none contrast. Coming from an intact family and having a well-educated mother are both associated with a greater likelihood of choosing condoms-versus either the pill or no method at all. Coming from a fundamentalist religious background is associated with a lower likelihood of choosing condoms versus any of the alternatives. Since condoms are clearly a male method and our measures of family background only reflect the woman's family, these patterns may reflect the selectivity of partners: women from higher status, intact, nonfundamentalist families may be more likely to have first intercourse with males who bring condoms or are willing to use them if provided by the female. These women may also be more likely to choose condoms because of their protective effect against both conception and the risk of STDs. 7 The polytomous logistic regression results are somewhat easier to interpret if they are transformed into probabilities. Table 2 presents predicted probabilities calculated by evaluating the regression model for different values of selected variables, holding all other variables constant at their means. These patterns must be interpreted with caution, as not all differentials achieve statistical significance. As Table 1 suggested, patterns of condom use seem to stand out from patterns of pill use or nonuse, especially in terms of family background differentials.

328

Demography, Vol. 27, No.3, Augwt 1990

Table 1. Polytomous Logit Results for Method Choice at First Intercourse, by Race, 1982 NSFG

Variable

Condom vs. none

Condom vs. pill

White (Model Age at menarche Mother's education Family status at age 14 Catholic Fundamentalist Protestant Other religion Urban residence Northeast North Central West

-.04

.06' .51** .08 -2.51*' .23 .002 .65** -.14 -.87**

Age at menarche Mother's education Family status at age 14 Catholic Fundamentalist Protestant Other religion Urban residence Northeast North Central

W9st

.003 .08** .31 .03 -.22 -.39 - .91'* .08 -.45 -.59

Pill

Other vs. none

Pill vs. none

-.13 -.14" -.22 .04 .55 .53 .24 -.26 -.03 -.02

-.02 .OSH .24 .31 -.45

- .16** -.05 .02 .32 .10 .53 .19 .21 .05 -.27

.09 -.10 -.007 .37 -.42 .60

.01 .12'* .32 -.01 .12 -.73

-1.26** -.35 .45

.42 .09 .38

vs,

other

x = 80.9**, N = 968) 2

.12 .11*' .48' -.24 -2.61** -.29 -.19 .44 -.18 -.61

Black (Model

Condom vs. other

-.02 -.03 .26 -.23 -2.06' .24 .05 .18 -.21 -.63*

-.009

-.05 .47* .08 -.24

x2 = 65.5**, N = 865)

-.10 .06 -.01 -.32 .07 -.26

-.50

.91 -.19 -1.43*'

-.01 -.04 -.02 .05 -.35

.34

-1.07**

-.34 -.54 -.98

-.58

.17

.10 .02 .32 .35

-.29

-.13 -.41 -.83' -.26 .84··

Note: The sampleincludeswomen aged 15-24 in 1982 whosefirst premarital intercoursewas as a teenager. Omitted categories are family status not intact; Protestant religion (not fundamentalist); rural residence; and Southern region. 'p < .10. •, P < .05.

Differentials in method choice by mother's educational attainment are especially striking. As mother's education increases, both white and black daughters are more likely to use condoms or some other method at first intercourse and less likely to use nothing. Pill use, however, is negatively related to mother's education for whites and unrelated for blacks. There are several possible explanations for these patterns, though unfortunately, we are unable to distinguish among them with the available data. One possibility is that attitudes about the methods may differ according to the mother's education. Daughters of better-educated mothers may be more concerned than other women with the prophylactic (i.e., disease prevention) ability of condoms, or they may be more aware of the potential side dfects of the pill. It is also possible that access to dilferent methods varies by mother's education. Since the pill is available only by prescription from a clinic or physician, patterns of pill use will reflect the likelihood of visiting a clinic or physician. The negative effect of mother's education on pill use could therefore reflect the greater likelihood that poor, lower status women go to family planning clinics for contraceptives (Torres & Forrest 1985). Daughters from higher status families may be less inclined to confirm publicly their intentions regarding sex by going to a private physician, and they may be less aware of the availability of more impersonal clinics. The pattern of differentials in method choices by religion is particularly noteworthy.

329

Adolescent Contraceptive Method Choices

Table 2. Predicted Probabilities of Method Choice at First Intercourse. by Race. 1982 NSFG Black

White Variable Age at Menarche

11 14

Mother's education 9 years 12 years 16 years Family status at age 14 Not intact Intact Religion Catholic Fundamentalist Other Protestant Rural-urban residence Rural Metro Region of residence Northeast North Central West South Total (observed)

Condom

Pill

Other

None

Condom

Pill

Other

None

22 21

10 7

21 21

46 50

14 13

10 13

7 7

69 68

19 22 24

11 9 6

18 21 26

52 48 43

12 14 18

11 11 11

6 8 11

71 67 60

17 24

9 8

21 22

53

12 15

10 12

7 8

71 65

22 3 26 23

9 11 11 7

24 18 18 20

45 68 45 50

14 12 11 14

15 8 11 11

7 9 4 8

64 71 74 67

22 22

7

9

22 21

49 48

23 11

13 10

6 8

58 71

33 21 12 23 22

8 9 9 9 11

23 22 19 20 20

36 48 60

16 10 7 15 14

5 10 24 12 13

10 8 9 6 7

69 72 60 67 66

46

48

47

Note: Probabilities are calculated by evaluating the regression equation at di1ferent values of the variable of interest. holding all othervariables in the model constant at their means.Predicted log-odds arethen transformed into probabilities and multiplied by 100 so that they can be interpreted as percentages.

Despite the Catholic ban on contraception, there is no significant difference between Catholics and Protestants who are not fundamentalists in the choice of methods. This is consistent with the findings of recent fertility studies that show the disappearance of Catholic/nonCatholic differentials in childbearing (Mosher, Johnson, & Horn 1986; Westoff & Jones 1979). One religious differential, however, persists: white fundamentalists are substantially less likely than any other religious group to have chosen condoms at first intercourse (compare 3% with more than 20% for each of the other groups). They are also much more likely than other religious groups to have used no method (compare 68% with 44%). At present, we cannot provide an explanation for these strong effects other than the possibility that white fundamentalists differ from other religious groups on unmeasured characteristics related specifically to condom use. Yet these differences would have to be restricted to whites, since black fundamentalists do not demonstrate the same aversion to condom use at first intercourse. 8 There is considerable variability in method choices across regions in the United States, and surprisingly, this variability differs for whites and blacks. Although regional patterns of condom use are similar for both races (lowest in the West and highest in the Northeast),

330

Demography, Vol. 27, No.3, August 1990

we find substantial racial differences in pill use across regions. Whereas whites in all regions have an 8% or 9% likelihood of using the pill, blacks' pill use varies from 5% in the Northeast to 24% in the West. These patterns of regional variability have been noted elsewhere and may reflect racial differences in the availability of family planning services across regions or in the ethnic or socioeconomic composition of the different regions (Mosher 1981; Mosher & Bachrach 1986).

The Transition to First Method Use Although a little more than one-half of all teenagers did not use a method at first intercourse, the majority started practicing contraception within the first three months. The specific method first chosen, however, changes as time since first intercourse elapses:Whereas the condom is the leading method at first intercourse, the pill is the most popular method for women who started practicing contraception after first intercourse (Mosher & Bachrach 1987). This shift in method choice soon after first intercourse raises interesting questions about the factors affecting the transition to contraceptive use: to what extent do the same factors predict the initial adoption of contraception after first intercourse? We examine these patterns by comparing method choices at first intercourse with subsequent first-time choices made among never-users. We ask a set of nested conditional questions: Given that an adolescent did not use a method of contraception at first intercourse, what is the probability that she will use a method within the next month? Given that she did not use a method in the first month, what is the probability that she will begin using one in the subsequent two months? And so forth. We use life table procedures to answer these questions and present the results, method-specific qx values, in Table 3. Time since first intercourse is divided into five closed intervals, defined as the response categories for the following question: "How long after first intercourse did you first use a method of contraception? (1) At first sex, (2) within the same month, (3) 1-3 months later, (4) 4-6 months later, or (5) 7-12 months later." As can be seen in Table 3, only about one-half of white and one-third of black teens used a method at first intercourse. Among those who did not use a method at first intercourse, less than one-third of whites and one-fifth of blacks used one within the first month. An Table 3.

Life Table Estimates of the Probability of Initiating Contraceptive Use by Time Since First Intercourse, Method, and Race, 1982 NSFG

Time period

All methods

Condom

Pill

Other

N*

.10 .10

.20

.09

1,030 484 336 272 240

White At first intercourse Within 1 month 1-3 months later 4-6 months later 7-12 months later

.52 .30 .19 .11 .16

.22 .11

.08

.08

.03 .03

.06

.09

.03 .01 .03

Black At first intercourse Within 1 month 1-3 months later 4-6 months later 7-12 months later

.34

.14

.19 .13 .08 .13

.06 .03 .02 .01

.11 .11 .08 .05 .12

.08

.03 .01 .01 .01

886

5n 462

403 367

• The universe for each row consists of all womenwho had neveruseda contraceptive method prior to the start of the time interval.

Adolescent Contraceptive Method Choices

331

even smaller proportion of nonusers during the first month became users during the next two months (19% of whites and 13% of blacks). This slow pace of adopting contraception may reflect a long interval between initial and subsequent sexual activity. Yet based on tabulations not shown here, we know that more than 70% of the sample of teens reported having second intercourse within three months of first infercourse. Table 3 also shows that the choice of methods shifts markedly after first intercourse. For both whites and blacks, .the likelihood of choosing either condoms or other methods drops off immediately after first intercourse, whereas the likelihood of choosing the pill remains unchanged. Black nonusers are much more likely to adopt the pill than any other method as soon as the first month after first intercourse. In contrast, white nonusers are equally likely to adopt condoms or the pill throughout the first three months. Starting with the fourth month, however, whites are twice as likely to adopt the pill as condoms. The slow pace with which nonusers become users has important policy implications because of the high risk of pregnancy faced by teens who continue to have unprotected sex (Koenig & Zelnik 1982). What type of teens have the slowest pace of adopting contraception? Do the same factors continue to predict the adoption (or nonuse) of contraception after the initiation of sexual activity? To examine these questions, we ran simple logistic regression models of the likelihood of adopting contraception during an interval after first intercourse, conditional on never having used a method prior to the interval. In other words, we predicted the transition to first contraceptive use among people at risk of using contraception for the first time. The results (see Table 4) show the expected socioeconomic determinants of use at first intercourse, but not thereafter. The first column of Table 4 summarizes the results of the polytomous analysis in Table 2 with a simple dichotomous dependent variable. If we use the number of the significant coefficients as a very rough indicator of model strength, we find that for whites, models for all intervals after fitst intercourse are very weak. For blacks, we can say the same for models after the first month. These results indicate that among teens who did not use a contraceptive method at first intercourse, the subsequent adoption of a method is not associated with their observable background characteristics. The disappearance of effects after first intercourse could be an artifact of the analysis, either because of declining sample sizes in subsequent intervals or because we are examining the use-nonuse dichotomy rather than the choice of specific methods. Yet a similar pattern holds even when we run the models on randomly selected subsamples or use a polytomous method-choice variable, as was done in Table 1 (results not shown). A more plausible explanation may be the increasing selectivity of nonusers in later intervals. That is, the samples are increasingly composed of women who are either averse to practicing contraception in principle or who are willing to take the risk of not practicing, perhaps because they feel that their risk is low. Thus as a more select group, their behavior may be less influenced by differences in family background or regional context and more influenced by their individual predispositions toward contraception. It is also possible that regardless of selectivity, as teens gain more experience, their contraceptive behaviors are less predictable on the basis of fixed background variables because of the influence of more recent situational factors, such as the peer or school environment. This is supported by our analysis of current teenage contraceptive behavior, which finds that current method choices are only marginally influenced by family background variables (results not shown).

Discussion In this article we have examined the types of contraceptive methods chosen early in the sexual careers of American teenage women. We find that the patterns of method choices not only vary by race and region within the United States but also change within the life course as teenagers gain more experience with both sex and contraception. At first intercourse

332

Demography, Vol. 27, No.3, August 1990

Table 4. Conditional Logit Results Predicting the Timing of First Contraceptive Use, by Race, 1982 NSFG Variable

At first intercourse

Within same month

1-3 months

4-6 months

7-12 months

-.003 .07 .47 .72** -.58 -.23 -.18 .35 -.16 .61

.04 .10 .20 -.16 .13 -.76 -.08 -1.24 .56 -.03

-.06 .03 .18 -.26 -.84 -.28 -.39 .70 -.42 .88**

White Age at menarche Mother's education Family status at age 14 Catholic Fundamentalist Protestant Other religion Urban residence Northeast North Central West

Na Model x2

-.04

.06** .29** .22 -.79** .31 .03 .45** -.07 -.57**

1,030 50.0**

.05 .02 .30 .02 -1.47** -.01 .08 .16 -.21 -.05 484 12.6

336 14.5

272 7.7

240 9.6

Black Age at menarche Mother's education Family statusat age 14 Catholic Fundamentalist Protestant Other religion Urban residence Northeast North Central West

Na Model x2

.06 .07** .30** .16 -.14

-.43

-.56" -.11 -.32 .28 886 34.4**

-.008 .07* .49**

.43

.86'* .22 -.03

-.48

.27 -.70 577 20.5**

.12 -.08 .37 -.30 -.36

-.44

-.36 -.98 .73* -.15 462 19.1**

-.13 .02 .001 -1.22 -1.58 -.90 .64 .10 -.06 .80 403 13.6

-.12 -.003 -.14 .02 -.09 -.13 -.09 -.22 .11 -.49 367 3.4

a The universefor each columnconsists of all those who had neverused a contraceptive method prior to the start of the time interval. 'p < .10. '* P < .05.

among whites, condom users differ from pill users and nonusers more systematically than any other comparison: they (condom users) are more likely to have better-educated mothers, come from intact families, and live in the Northeast and less likely to be fundamentalist Protestants or live in the West. It is interesting that higher social status predicts greater condom use at first intercourse, but not use of the most effective contraceptive method (i.e., the pill). In fact, we find that daughters of better-educated mothers are less likely than daughters of less-educated mothers to have used the pill at first intercourse. This may reflect their greater concern with preventing sexually transmitted disease or possibly their tendency not to attend family planning clinics, which are more likely to prescribe the pill. It is unfortunate that we do not know more about their motivations for using condoms (i,e., for disease prevention or avoiding pregnancy). We also found that among teens who did not use a method at first intercourse, the likelihood of adopting a method soon thereafter was low for both whites and blacks, though

333

Adolescent Contraceptive Method Choices

the majority of adopters chose the pill instead of the condom. These results suggest that as time since first intercourse passes, teens start to use more effective methods of contraception and they are less likely to use condoms. Still, only a small proportion of nonusers go on to adopt a method, and their decision is only marginally influenced by social structural characteristics. This suggests that there may be a persistent core of nonusers who are increasingly less likely to adopt contraception. It also points out how little we know about the social world experienced by teenagers when they are making their contraceptive decisions. How well are they doing in school? Who are their friends, and how do their friends influence their behavior? What is discussed within the family about contraception? What services are available within the community from clinics, pharmacies, and private physicians? These and related issues are the logical next step for future data collection efforts. Although the results of this study are based on 1982 data and therefore cannot reflect the recent increase in public awareness of the risk of STDs, they can serve as an important benchmark with which to compare more recent findings from data sources such as Cycle IV of the NSFG (fielded in 1988). We anticipate an increase in condom use for both whites and blacks throughout the country (Sonenstein, Pleck, & Ku 1989). Indeed, the advent of AIDS, and the evidence and controversy over its heterosexual transmission paths, has brought the issue of prophylaxis into many American homes and schools (e.g., Haverkas & Edelman 1988; Osborne 1986; Piot et al. 1988). Even though there are relatively few cases of AIDS among adolescents, increasing evidence of a very long latency period has made it clear that in many instances, behavior during adolescence may lead to an AIDS diagnosis in early adulthood (Hein 1989). Thus the issue of the prophylactic nature of various contraceptives should be considered more explicitly in analyses of adolescent method choices. Finally, we realize that we are limited in what we can infer about condom use from surveyslike the NSFG, which only interview women. Because the condom is a male method, information is needed from male respondents so that we can examine the factors related to whether they or their partners bring condoms with them, whether they decide to use them, and why they selected the condom rather than any other method (e.g., for disease prevention).

Notes I Our use of the term "contraception" includes methods that prevent either conception or the spread of STDs. Z This, of course, assumes that teenagers will no longer feel at risk of STDs. The validity of this assumption will depend in part on the nature of their sexual activity (i.e., the number of partners and the number of partners that their partners have). 3 Because we are focusing on premarital first intercourse, we drop the 79 virgin brides from our sample. 4 The following are considered fundamentalist: Adventist, Apostolic (except with Reformed Zion Union or Armenian), Bible, Free, Fundamental, Gospel, Holiness, Jehovah's Witness, Mission, Missionary, Nazarene, _ _ of God or _ _ of Living (except with Mennonite), Pentecostal, and Sanctified. Unfortunately, some of the more fundamentalist-oriented members of the Baptist denomination could not be separately identified. I We realize that both of these variables measure current residence and may not reflect the respondent's place of residence at the time of first intercourse. For many respondents, however, they are still good proxies for the prior place of residence, and indeed for many there would be no change. Further, their inclusion does not raise serious issues of causality. It would be difficult to argue that one's current place of residence is a result of one's past contraceptive behavior. Finally, based on some sensitivityanalyses, we know that none of the other coefficientsis sensitiveto the inclusion or exclusion of these two variables. 6 Results from the same analysis suggest, however, that the effects of other predictor variables on the use of contraception do not change appreciably even when a more complex joint estimation procedure is employed (see Guilkey et al, 1989).

334

Demography, Vol. 27, No.3, August 1990

7 It is unfortunate that we do not know the motivation behind condom choice. Such a question was not included in the questionnaire, and we expect that such information would not be reliably recalled retrospectively. 8 Our measure of fundamentalism is, however, based on the name of the Protestant denomination rather than on actual beliefs. Since the nature of fundamentalism varies significantly within various Protestant churches, we recommend that future data collection dforts attempt to measure beliefs as well as formal membership.

Acknowledgments This research was partially supported by National Institute of Child Health and Human Development Grant ROI-HD-22293. The general support of the Carolina Population Center and its stalf is gratefully acknowledged. Special thanks go to Erika Stone for the complex programming necessary for this analysis.

[Received August 1989. Revised December 1989.J

References Bachrach, C. A. (1984). "Contraceptive Practice Among American Women, 1973-1982." Family Planning Perspectives, 16(6), 253-259. Bachrach C. A., & Hom, M. C. (1987). Married and Unmarried Couples: United States, Vital and Health Statistics, Ser. 23, No. 15. Washington, DC: U.S. Government Printing Ofice. - - . (1988). "Sexual Activity Among U.S. Women of Reproductive Age." American Journal of Public Health, 78, 320-321. Bachrach, C. A., Mosher, W. D., Hom, M., & Shimizu, I. (1985). National Survey of Family Growth, Cycle Ill: Sample Design, Weighting, and Variance Estimation, Vital and Health Statistics, Ser. 2, No. 98. Hyattsville, MD: National Center for Health Statistics. Cutright, P., & Smith, H. L. (1986). "Trends in Illegitimacy Among Five English-Speaking Populations: 19401980." Demography, 23, 563-578. Cvetkovich, G., & Grote, B. (1981). "Psychosocial Maturity and Teenage Contraceptive Use." Population and Environment, 4(4), 211-226. Ford, K., Zelnik, M., & Kantner, J. F. (1981). "Sexual Behavior and Contraceptive Use Among Socioeconomic Groups of Young Women in the United States." Journal of Biosocial Science, 13. 31-45. Grady, W. R., Hayward, M. D., & Yagi, J. (1986). "Contraceptive Failure in the United States: Estimates From the 1982 National Survey of Family Growth." Family Planning Perspectives, 18(5), 200-209. Guilkey, D. K., Rindfuss, R. R., & Kahn, J. R. (1989). "Estimating the Determinants of Age and Contraceptive Use at First Intercourse From Censored Data." Unpublished manuscript, University of North Carolina at Chapel Hill, Carolina Population Center. Hanushek, E. A., & Jackson, J. E. (1977). Statistical Models for Social Scientists. New York: Academic Press. Hayes, C. (ed.). (1987). Risking theFuture: Adolescent Sexuality, Pregnancy, and Childbearing. Washington, DC: National Academy Press. Hein, K. (1989). "Commentary on Adolescent Acquired Immunodeficiency Syndrome: The Next Wave of the Human Immunodeficiency Virus Epidemic?" Journal of Pediatrics, 114, 144-149. Holfcrth, S. L., Kahn, J. R., & Baldwin, W. (1987). "Premarital Sexual Activity Among U.S. Teenagers Over the Past Three Decades." Family Planning Perspectives, 19(2), 46-53. Hogan, D. P., Astone, N. M., & Kitagawa, E. M. (1985). "Social and Environmental Factors Influencing Contraceptive Use Among Black Adolescents." Family Planning Perspectives, 17(4), 165-169. Jones, E. F., & Forrest, J. D. (1989). "Contraceptive Failure in the United States: Revised Estimates From the 1982 National Survey of Family Growth." Family Planning Perspectives, 21(3), 103-109. Jones, E., Forrest, J., Goldman, N., Henshaw, S., Lincoln, R., Rosolf, J., Westolf, C. F., & Wulf, D. (1986). Teenage Pregnancy in Industrialized Countries. New Haven, CT: Yale University Press. Jones, J. A., Kahn, J. R., Parnell, A. M., Rindfuss, R. R, & Swicegood, C. G. (1985). "Non-Marital Childbearing: Diverging Legal and Social Concerns." Population and Development Review, II, 677-693. Kantner, J. F., & Zelnik, M. (1972). "Sexual Experiences of Young Unmarried Women in the United States." Family Planning Perspectives, 4(4), 9-18. Kisker, E. (1985). "Teenagers Talk About Sex, Pregnancy and Contraception." Family Planning Perspectives, 17(2), 83-90. Koenig, M. A., & Zelnik, M. (1982). "The Riskof Premarital First Pregnancy Among Metropolitan-Area Teenagers: 1976 and 1979." Family Planning Perspectives, 14(5), 239-247. Maddala, G. S. (1983). Limited-Dependent andQualitative Variables in Econometrics. Cambridge. UK:Cambridge University Press.

Adolescent Contraceptive Method Choices

335

Moore, K. A., Simms, M. C., & Betsey, C. L. (1986). Choice andCircumstance: Racial Differences in Adolescent Sexuality and Fertility. New Brunswick, NJ: Transaction Books. Mosher, W. D. (1981). Contraceptive Utilization, United States, Vital and Health Statistics, Ser. 23, No.7. Hyattsville, MD: National Center for Health Statistics. Mosher, W. D., & Bachrach, C. A. (1986). Contraceptive Use, United States, 1982, Vital and Health Statistics, Ser. 23, No. 12. Hyattsville, MD: National Center for Health Statistics. - - . (1987). "First Premarital Contraceptive Use: United States, 1960-1982." Studies in Family Planning, 18(2), 83-95. Mosher, W. D., Johnson, D. P., & Horn, M. C. (1986). "Religion and Fertility in the United States: The Importance of Marriage Patterns and Hispanic Origin." Demography, 23,367-379. Pindyck, R. S., & Rubinfeld, D. L. (1981). Econometric Models and Economic Forecasts. New York: McGrawHill. Piot, P., Plummer, F. A., Mhalu, F. S., Lamboray, J. L., Chin, J., & Mann, J. M. (1988). "AIDS: An International Perspective." Science, 239, 573-579. Pratt, W. F., Mosher, W. D., Bachrach, C. A., & Horn, M. (1984). "Understanding U.S. Fertility: Findings From the National Survey of Family Growth, Cycle III." Population Bulletin, 39(5). Rindfuss, R. R., Swicegood, C. G., & Bumpass, L. L. (1989). "Contraceptive Choice in the United States: Process, Determinants and Change." In Choosing a Contraceptive: Method Choice in Asia and the United States, eds. R. A. Bulatao, J. A. Palmore, & S. Ward. Boulder, CO: Westview Press, pp. 237-256. Sonenstein, F. L., Pleck, J. H., & Ku, L. C. (1989). "At Risk of AIDS: Behaviors, Knowledge and Attitudes Among a National Sample of Adolescent Males." Paper presented at the Annual Meeting of the Population Association of America, Baltimore, MD. Stephen, E. H., Rindfuss, R. R., & Bean, F. D. (1988). "Racial Dilferences in Contraceptive Choice: Complexity and Implications." Demography, 25, 53-70. Thompson, L., & Spanier, G. (1978). "Influence of Parents, Peers and Partners on the Contraceptive Use of College Men and Women." Tournai of Marriage and the Family, 40, 481-492. Torres, A., & Forrest, J. D. (1985). "Family Planning Clinic Servicesin the United States: 1983." Family Planning Perspectives, 17(1), 30-35. Westolf, C. F., & Jones, E. F. (1979). "The End of 'Catholic' Fertility." Demography, 16,209-218. Zelnik, M., & Kantner, J. F. (1980). "Sexual Activity, Contraceptive Use and Pregnancy Among MetropolitanArea Teenagers: 1971-1979." Family Planning Perspectives, 12(5), 230-237. Zelnik, M., Kantner, J. F., & Ford, K. (1981). Sex and Pregnancy in Adolescence. Beverly Hills, CA: Sage. Zelnik, M., & Shah, F. K. (1983). "First Intercourse Among Young Americans." Family Planning Perspectives, 15(2), 64-70.