Worldwide, reproductive preferences and behavior often vary across socioeconomic and demographic subgroups. Desired family size, the intention to have more children and actual fertility are usually higher among rural women and women without formal education than among their urban and better-educated counterparts.¹ In some societies, women may marry at a relatively young age, have the children they want and then wish to stop childbearing; in others, women may marry later and seek to avoid having children before marriage. These variations not only reflect differences in patterns of family-building goals and experiences, but also are likely to necessitate differences in the means that women use to achieve their desired family size and timing of births, including contraception and abortion.

Even when differences in family formation behavior across socioeconomic and demographic subgroups narrow, as they often do when fertility declines, reliance on contraception and abortion may continue to vary for a number of reasons. The immediate causes of variations in the level of abortions are differing probabilities of unintended pregnancy across subgroups and variations in the likelihood that a woman will choose to have an abortion if she becomes pregnant unintentionally. Underlying these causes are other factors.

First is the perceived opportunity cost of having a child. Where women have opportunities for education, employment and career development, younger and unmarried women are the most likely to want to postpone marriage or childbearing, and to obtain an abortion when a pregnancy occurs. In contrast, where women marry young and are expected to have a child soon after marriage, abortion will be infrequent among young women. In many Sub-Saharan African countries, a girl must leave school if she is pregnant, and abortion tends to be most common among young unmarried women who wish to continue their education.² In developed countries where women desire to postpone the beginning of family formation, before and sometimes even after marriage, abortion is most prevalent among women younger than 25.³ Women who are employed tend to be more likely to obtain abortions than those who are not working.⁴

Second, the legal, moral and religious contexts surrounding abortion are likely to have different impacts on subgroups regarding the abortion decision. If prevailing public opinion is hostile to abortion, it is likely to affect some subgroups more than others. Differences may also be rooted in beliefs. For example, born-again Christians and Catholics may be less likely to obtain abortions than mainstream Protestants. In countries where barriers (legal or otherwise) to abortion are strong, whether a woman obtains one will depend on whether her immediate family supports her decision, whether she has information about where to seek services and her ability to pay.

Third, the pattern of contraceptive use in a country may have important effects on levels of abortion. For example, in the United States and many Latin American countries, where sterilization is a common method of family planning, abortions may be more prevalent among younger than older women. On the other hand, where premarital relationships are infrequent, where young unmarried women use effective contraceptives or where older women depend largely on less effective methods, younger women will be less likely to have abortions than older women, who may have had the number of children they want.

In this article, we examine the extent to which abortion is differentially used by women of varying characteristics, within and across countries. A better understanding of the relationship between induced abortion and women's background characteristics can improve our ability to identify subgroups in a population who have the highest levels of unintended pregnancies and induced abortion and who, therefore, are in the greatest need of effective contraceptive and postabortion services. Such information is valuable to program planners and policymakers in their efforts to facilitate women's ability to plan pregnancies.

Study Design

Measures

Whether a woman obtains an abortion may be associated with a number of her characteristics, but we examine only three in detail here: her age at the time of the abortion, marital status and number of live-born children (sometimes, number of living children). In addition, although the available data do not permit comprehensive analyses, we examine associations between abortion and education, residence, religion and ethnicity. Economic factors, such as work status and family income, are also likely to play a major role in women's decisions about whether to obtain an abortion; however, we could not examine these factors because of a lack of data.

A disproportionately high number of abortions in any subgroup may be an indication that this group has above-average difficulty in preventing unplanned pregnancy or an increased likelihood of choosing abortion to end such a pregnancy. We examine three measures that provide different perspectives on the association between abortion and women's characteristics—the percentage distribution of abortions, the abortion rate per 1,000 women and the ratio of abortions per 100 pregnancies.

The percentage distribution of abortions, which is the most commonly available of these measures, may simply reflect the distribution of women in the population and does not necessarily indicate that one subgroup is more likely to have an abortion than another. By contrast, the abortion rate allows comparison across subgroups of women's likelihood of obtaining abortions. The abortion ratio shows variation across subgroups in the probability that, once pregnant, women will choose abortion. Both the abortion rate and the abortion ratio are good measures for comparing abortion behavior between and within countries.

Data Sources

For a number of reasons, abortion data from around the world vary widely in their availability and reliability. Consequently, to present the most complete picture possible, a comparative study of this kind necessarily depends on information from a number of sources: national statistics, nationally representative sample surveys, ad hoc surveys and hospital records of women admitted for abortion or abortion complications. We place the greatest emphasis on findings from countries where we believe that the data are reasonably complete.

•National statistics. In countries where abortion was legal under broad conditions* and the total population was one million or more in 1997, we obtained official abortion data directly from national statistical offices or indirectly from local informed experts (government officials, scientists with an interest in abortion issues or family planning officials). We sent questionnaires to contact persons requesting such information as the total number of abortions and the number of procedures by women's age, marital status and parity at the time of abortion. Further details about data collection and assessment are reported elsewhere.⁵

Although abortion reporting is generally required in countries where the procedure is legal under broad conditions, it is often far from complete. If underreporting varies according to women's characteristics, abortion measures for particular subgroups will be biased; where reporting is complete or nearly complete, any bias will be minimal. Reporting is believed to be most complete where statutes prescribe a formal authorization procedure and abortions must be performed in public facilities; it is probably least complete where private practitioners are the major source of services.⁶

However, factors apart from legal status affect the quality of abortion reporting. Values, religious beliefs and the perceived consequences of obtaining an abortion are important determinants. For these and other reasons, even where abortion is legal, women who obtain abortions may be reluctant to report certain of their characteristics truthfully to providers.

In this article, we include data both for countries where abortion reporting is judged to be reasonably (i.e., at least 80%) complete and for those where it is believed to be incomplete. Patterns according to women's characteristics may be meaningful even where the overall level of abortion is underreported, though the data must be interpreted with caution.

•National fertility surveys. For a few countries where abortion is legal under broad circumstances, but official statistics do not provide data by women's characteristics, we use data from national fertility surveys that collected information on abortion—the Demographic and Health Surveys (DHS) in Kazakhstan, Kyrgyzstan, Turkey and Uzbekistan; the U.S. Centers for Disease Control and Prevention's Reproductive Health Surveys in Romania and the Czech Republic; and an independent survey in the Republic of Korea. These surveys have the dual advantages of obtaining information on characteristics of women and of representing all women in a country (except for Turkey and Korea, where the samples included only ever-married women). However, if underreporting of abortion is high, and especially if it varies according to respondents' characteristics, survey data on differences in abortion among population subgroups will be biased.

We judged survey data on abortion for Kazakhstan to be at least 80% complete, since they compare favorably with official statistics, which are believed to be reasonably complete. (Although official statistics may be underreported, this comparison provides a measure of confidence in the survey data.) However, we considered the survey data for Kyrgyzstan, Korea, Romania, Turkey and Uzbekistan to be incomplete or of unknown completeness (since, for example, there are no external statistics against which the survey data on Turkey and Korea can be compared).

The survey-based abortion measures we present in this article are based on abortions reported in the three years preceding the surveys, to minimize bias from retrospective reporting. While the surveys report women's age at abortion, they provide information on parity, marital status and education only as of the time of interview. Restricting analyses to abortions in the three years before the interview lessens the possibility of changes in characteristics since the abortion, but the potential for some discrepancy must be borne in mind in interpreting results for these characteristics.

•Ad hoc surveys and hospital reports. In countries where abortion is highly restricted by law,^† official abortion statistics and, in general, fertility surveys with data on abortion are unavailable. For these countries, any available data come from subnational sources that have substantial limitations.
The information included here is taken from published studies that are often small-scale—for example, examining specific areas of a country or a city. Community, or ad hoc, surveys typically sample a cross-section of women, but some do not include all women of reproductive age, most suffer from some degree of underreporting of abortion and some have quite small samples.

Another source of information is studies of abortion patients attending medical facilities (hospitals or clinics), for which data are collected through interviews with the women or by abstracting information from their medical records. In countries where abortion is highly restricted by law, most studies of this type include only women who are hospitalized for treatment of abortion complications and therefore do not represent all women who obtain abortions—for example, women who have complications but fail to obtain hospital treatment, and those who obtain a safe, or uncomplicated, abortion. In addition, the hospitals or clinics included in some of these studies may be different from the average facility.

Despite the limitations of subnational studies, they provide useful information for countries that would otherwise be unrepresented. We include data from such studies if the sample size is more than 100 and if the data are for women having abortions in a recent time period. Studies that present only lifetime abortion measures are not included.

Findings

Age at Abortion

•Percentage distribution. In the great majority of the 56 countries for which we have data, the highest proportions of abortions occur among women aged 20-24 and 25-29 (Table 1). Typically, the proportion is smaller among women aged 30-34, but in several countries, the difference is not great. The proportion drops notably among women aged 35-39 in all countries and generally is markedly lower among women at either end of the reproductive spectrum (those younger than 20 and 40 or older) than among other age-groups.

The proportion of abortions occurring among women younger than 20 is less than 10% in 21 countries, between 10% and 20% in 25 countries, and at least 20% in nine countries. (Information on teenagers was unavailable for Korea.) Since 15-19-year-olds make up 12-29% of women of reproductive age in these countries, these results suggest that in general, adolescents do not disproportionately obtain abortions. In some countries, however, the proportion of abortions that are provided to adolescents is substantial: It is 20% or more in three countries with complete data and six with subnational data, including five of the six Sub-Saharan African countries in this category. In Nigeria, adolescents obtain 53% of abortions, but they constitute only 24% of women aged 15-44%;⁷ however, the data are for patients with septic abortion, a group in which young women may be overrepresented, because they typically have greater than average difficulty in obtaining a safe abortion. ^‡

A number of studies of clandestine abortion in developing countries have reported that abortion and abortion morbidity are most common among young, unmarried women.⁸ In Lusaka, Zambia, for example, 60% of women hospitalized for abortion complications were aged 15-19.⁹ Among women hospitalized with abortion-related complications in Uganda, about two-thirds were aged 15-19, two-thirds were students and four out of five had never been married.¹⁰

Women aged 40-44 account for the lowest proportion of abortions in about two-thirds of the 56 countries. The proportion is lower than 10% in all but seven countries—Israel, Slovenia, Croatia, Japan, Vietnam, Malaysia and Nepal (Table 1).

•Abortion rate. We have estimated age-specific abortion rates for 34 countries with official statistics or national sample surveys that included data on abortion^§ (Table 2). These rates show differences in the annual incidence of abortion, according to women's age.

In 27 countries, abortion rates are highest among women in their 20s—in 17 countries, among women aged 20-24, and in 10, among 25-29-year-olds. (In several of these countries, rates differ very little between the two age-groups.) The rate for women aged 30-34 is slightly higher than rates for women in their 20s in six countries (Israel, Slovenia, Italy, Japan, Kyrgyzstan and Uzbekistan). Thus, in most of these countries, abortion rates show an inverted U pattern with age of women.

The incidence of abortion among adolescents is by no means negligible in many of the 34 countries. The adolescent abortion rate is extremely high (91 abortions per 1,000 adolescents per year) in Cuba, high (30-44 per 1,000) in Bulgaria, Estonia, Hungary, the United States, Romania and Russia, and moderately high (22-24 per 1,000) in Australia and Canada. However, only 10-20 abortions occur per 1,000 adolescent women in 13 countries, and fewer than 10 per 1,000 occur in 10 countries. Belgium, Germany and the Netherlands have complete national statistics and very low adolescent abortion rates; the low rates reflect very low overall levels of abortion in these countries.

Abortion rates are generally lowest for women aged 40 and older. While rates for this age-group are typically quite low, they are 10 or more per 1,000 women in 12 countries (Bulgaria, Estonia, Hungary, Kazakhstan, Singapore, Slovenia, Kyrgyzstan, Moldova, Romania, Russia, Turkey and Uzbekistan).

The total abortion rate represents the number of abortions that an average woman will have in her lifetime if she experiences the prevailing period age-specific abortion rates. As the table indicates, this rate is highest in Romania (3.4 abortions per woman) and Russia (2.4); it is less than 0.5 in 15 countries, 0.5-1.0 in 10 countries and more than 1.0 in the remainder. However, for countries with incomplete data, the true age-specific and total abortion rates will be somewhat higher than those shown here.

•Abortion ratio. In all but two of the 20 countries for which data are available (Netherlands and Sweden), the proportion of pregnancies that end in abortions is highest for women aged 40 or older (Table 3). Among the oldest women, the abortion ratio ranges from 25 abortions per 100 pregnancies in the Netherlands to 89 or more per 100 in Bulgaria, the Czech Republic and Uzbekistan; in Romania, the ratio for women aged 40 or older is the same as that for women in their late 30s (86 per 100).

In the Netherlands and Sweden, the highest ratio occurs among women younger than 20. Overall, a substantial proportion of pregnancies among adolescents end in abortions—one-third or more in 12 of the 20 countries for which data are available. The abortion ratio ranges from three abortions per 100 adolescent pregnancies in Kyrgyzstan to 65 per 100 in Sweden.

Two main patterns are found in the abortion ratios according to age-group: a U shape, in which the ratio declines after the teenage years, generally reaches its lowest point among women aged 20-24 or 25-29, and then rises to its highest level among women 40 and older; and a monotonic increase, in which the ratio is lowest among teenagers and rises steadily with age. Figure 1 shows these patterns for selected countries.

Only a few countries do not strictly exhibit either of these patterns. In the Czech Republic and the Slovak Republic, the abortion ratio increases almost monotonically with age. The Netherlands and Sweden have a reversed J-shaped pattern, with adolescents having a higher ratio than women aged 40 or older.

Marital Status

•Percentage distribution. Married women obtain a substantial proportion of abortions—40% or more in 24 of the 40 countries for which data are available (Table 4). Unmarried women obtain the majority of procedures in 17 countries.

Married women obtain a higher proportion of abortions in developing countries than in developed countries. In the developing world, married women account for very high proportions of abortions in the Asian countries, including the former Soviet republics in Central Asia, and three of the four Latin American countries represented (Colombia, Dominican Republic and Peru). By contrast, in the few Sub-Saharan African countries with data, a small proportion of abortions occur among married women; however, these results are based on studies of hospitalized women and at least partly reflect higher morbidity among unmarried, mostly adolescent women.

In the developed world, a higher proportion of abortions occur among married than among unmarried women in several Eastern European countries (Bulgaria, Czech Republic, Estonia, Slovak Republic and Albania). On the other hand, unmarried women account for close to half of abortions in five developed countries (Germany, Hungary, Israel, Italy and the Netherlands) and for a large majority of abortion procedures in many others (Belgium, Canada, England and Wales, France, Finland, New Zealand, Norway, Scotland, the United States and Spain).

•Abortion rate and ratio. For most countries, it is not possible to calculate abortion rates and ratios by marital status because the distribution of abortions, women or births by marital status is unavailable. However, we calculated rates for a few countries with survey data (Kazakhstan, Kyrgyzstan, the United States and Uzbekistan). The results confirm that for the Asian countries, the incidence of abortion is higher for married women than for unmarried women. For example, in Kazakhstan, 66 abortions occur per 1,000 married women of reproductive age, compared with 18 per 1,000 unmarried women. Similarly, in Kyrgyzstan, the rates are 61 and six per 1,000, respectively. In contrast, U.S. data show that unmarried women have a higher abortion rate (38 per 1,000) than married women (eight per 1,000).¹¹

The abortion ratio is higher among unmarried women than among married women in 15 of 16 countries with available data (Table 5). The differential is particularly marked in Canada, New Zealand, the United States and the represented countries of Western Europe. For some countries where married women account for a higher proportion of abortions than unmarried women, the abortion ratio shows a different picture: Once pregnant, unmarried women are more likely than their married counterparts to have an abortion (for example, in Czech Republic, Estonia, Hungary, Kazakhstan and Slovak Republic). Thus, the high proportion of abortions occurring among married women in these countries reflects primarily the high proportion of pregnancies that occur within marriage.

Parity

•Percentage distribution. In 40 countries with available data, the distribution of abortions by parity varies widely (Table 6). Childless women obtain fewer than 10% of abortions in 12 countries, 10-29% in nine countries, 30-49% in 14 countries and 50-68% in five countries. The lowest proportions obtained by women at zero parity are found in Eastern Europe and Asia. In general, more than half of abortions are obtained by women who have already had at least one child.

•Abortion ratios. Results on this measure, presented for 11 developed countries in a previous study, show a different perspective on the relationship between abortion and parity.¹² Although nulliparous women obtain a high proportion of abortions in several of those countries, women with two or more children end a much higher proportion of their pregnancies by abortion than do women with no children. Only in Canada and the United States is the abortion ratio for nulliparous women higher than the ratio for women with one or more children. In the Czech Republic, Denmark, Hungary and Sweden, the proportion of pregnancies ending in abortion is highest for women with two children (roughly 30-70%); in England and Wales, Finland and Norway, it is highest for women of parity three or more (close to 30%).

Other Characteristics

Characteristics other than age, marital status and parity also probably influence a woman's decision to obtain an abortion, either in themselves or as indicators of women's social and economic situation. Of particular note are residence, religion, race or ethnicity, and education.

In eight of 11 countries for which data are available, women in urban areas obtain more than 50% of abortions. This is largely because of the distribution of the population by residence, but the incidence of abortions among urban and rural women differs somewhat. Abortion rates by residence, available for five countries, also are higher among urban than among rural dwellers. In Turkey, according to DHS data, the abortion rate among ever-married women is 31 per 1,000 in urban areas, compared with 19 per 1,000 in rural regions. Similarly, a 1987 survey in South Korea reported rates of 70 per 1,000 in urban areas and 57 in rural areas.¹³ The 1994-1995 U.S. abortion rate was 27 per 1,000 metropolitan women and 14 per 1,000 nonmetropolitan women.¹⁴

A U.S. study showed that the abortion rate was lower for Protestants (17 per 1,000) than for Catholics (24 per 1,000) in 1994-1995, although Protestant women made up a larger proportion of those having abortions (37%) than Catholic women (31%).¹⁵ Survey data for Uzbekistan, where the majority of the population is Muslim, reveal that while Muslim women obtain the largest proportion of abortions (87%), they have the lowest abortion rate—17 per 1,000, compared with 39 per 1,000 among Christians. A similar pattern is found in Kazakhstan.

In some countries, abortion rates vary substantially by ethnicity. The 1995 U.S. abortion rate was much higher for black women (56 per 1,000) than for white women (17 per 1,000).¹⁶ In the Netherlands, native Dutch women obtained 57% of abortions and had an abortion rate of four per 1,000 in 1992; by contrast, Turkish-born women in the Netherlands accounted for only 5% of abortions but had the highest abortion rate—35 per 1,000.¹⁷ In Uzbekistan, while native Uzbeks have about three-quarters of abortions, the highest abortion rates are among Russians (43 per 1,000) and Kazakhs (19 per 1,000).

The percentage distribution of abortions according to women's educational attainment is available for a fairly large number of countries. In 15 of 23 countries studied, the majority of abortions are obtained by women who have had at least some secondary education (Table 7). However, the distribution of abortions according to women's educational attainment probably reflects the distribution of women according to educational attainment in a given country and therefore reveals little about differentials by education.

In Kazakhstan, Kyrgyzstan and Uzbekistan, 80-81% of abortions occur among women with a secondary education, reflecting that 82-89% of women aged 15-49 have this level of schooling in those countries. In Turkey, the proportion of abortions that are obtained by women with a primary education (59%) is exactly the same as the proportion of all women aged 15-49 with primary education. Moreover, in Bangladesh and Nepal, where the majority of women have no schooling, those with no education account for the majority of abortions. A study of women with abortion complications in eight district hospitals in Kenya found that 52% had secondary education or more, 40% had primary education and only 8% had no formal education; by contrast, according to DHS data, among all women aged 15-49, these proportions were 25%, 58% and 18%, respectively.

Abortion rates by education, available for a small number of countries, give a better indication of differences in the incidence of abortion by education and show varied patterns across countries. Rates calculated from sample survey data for three Asian countries indicate that the incidence of abortion is somewhat higher among women with more than a secondary education.** In Kazakhstan and Kyrgyzstan, 53-56 abortions occur per 1,000 women with more than a secondary education, compared with 42-49 per 1,000 women with secondary education or less. The difference is much larger in Uzbekistan; the rate for women who have more than a secondary education (32 per 1,000) is twice that of women with a secondary education (16 per 1,000). By contrast, survey data from South Korea show that among married women, those with more than a secondary education have a lower proportion of abortions and lower abortion rates than less educated women.¹⁸

Results from a study in Italy suggest that the impact of education on abortion may depend on the age of women. Among 15-19-year-olds, especially those who were married (who are a very small group), the abortion rate was highest for the most educated group. On the other hand, among married women aged 20-39, the rate of abortion declined as the level of education increased. For women aged 40 and older, the abortion rate increased with level of education.¹⁹

Discussion

Induced abortion, like unintended pregnancy, occurs in virtually all societies. Whether women terminate unintended pregnancies is likely to be determined by their background characteristics, particularly their age and parity, as well as characteristics that reflect and influence their values, attitudes and motivation to prevent an unintended birth, such as marital status, education, place of residence, religion and ethnicity.

In most countries with available data, abortion is concentrated largely around the middle of the childbearing years. Both the proportion of women who have an abortion and abortion rates show that women in their 20s are usually the most likely to obtain an abortion. This is partly because they are relatively likely to be married, sexually active and fecund, and therefore may have high pregnancy rates.²⁰ Women in their 20s seek abortions both to space births and to stop childbearing.

While the inverted U-shaped relationship between abortion rates and women's age is apparent in most countries and regions, an important regional difference is found. The proportion of abortions and the abortion rate are lowest among adolescents in many Asian countries, but in all other regions, adolescents generally account for a higher proportion of abortions than women in their 20s. Varying levels of completeness of abortion reporting may explain some of this variation, especially between developed countries (where reporting is mostly complete) and developing countries (where the reporting is largely incomplete); however, this difference is also observed among developing countries. Other possible reasons for this variation are higher levels of sexual activity (and unintended pregnancy) among unmarried young women in regions other than Asia and greater motivation among women in these regions to terminate a pregnancy (even if they are married), often because they wish to complete their education or continue working.

The higher rates of abortion at older ages in many Asian and Eastern European countries, as well as in the Asian former Soviet republics, may reflect low levels of use of modern contraceptives. For instance, in Japan, the pill is yet to be approved for contraceptive use and sterilization is rarely adopted.²¹ In the former Soviet republics and Eastern Europe, women rely largely on abortion to limit their family size because effective contraceptive methods, including sterilization, are not readily available (although availability and use are steadily improving).²²

A higher proportion of abortions occur among married than among unmarried women in about two-thirds of the countries studied. The abortion rate also is higher for married women in most of the few countries for which data are available (Kazakhstan, Kyrgyzstan and Uzbekistan), but these are all from one subregion. In contrast, U.S. abortion rates indicate that unmarried women are more than four times as likely as married women to obtain an abortion. However, the distribution of abortion by marital status varies noticeably by region. In all of the Asian countries included in this study, married women obtain by far the highest proportion of abortions. Results are mixed for the other regions, except Sub-Saharan Africa, where five of six countries with data show a higher proportion of abortions occurring among unmarried women.

In Asia, young unmarried women are not likely to be sexually active, but soon after marriage adolescents are expected to begin childbearing. Pregnancy-induced marriage is said to be common in Central and Eastern Europe, which may partly explain why the incidence of abortion is lower among unmarried women in those areas than in other parts of Europe.²³ In Sub-
Saharan Africa, where most available studies are of hospitalized abortion patients, young and unmarried women may be overrepresented because they are the most likely to obtain an unsafe abortion, given their lack of experience and resources. A study in Kenya noted that the trend has been shifting, and older, married women are making up an increasing proportion of those obtaining abortions;²⁴ this type of shift in the age pattern may be expected to occur in many countries in Sub-Saharan Africa, partly because of the increasing motivation to have smaller families.²⁵

Abortion ratios show that in both developed and developing countries, pregnancies among unmarried women are more likely to be resolved by abortion than are those among married women. This finding is reinforced by the reasons women give for obtaining an abortion—including not wanting to be a single mother, being too young, fearing their parents' objections, not wanting pregnancy to disrupt education or employment, and being unable to take care of the baby (especially without the support of the partner).²⁶

In most of the developing countries and many Eastern European countries studied, women who have had at least one live birth have the majority of abortions. This finding suggests that in these countries, abortion is obtained mainly to space or limit births rather than to delay the beginning of family formation. In the rest of Europe and in Canada and the United States, childless women account for the majority of abortions. This result is consistent with the finding that a high proportion of abortions in these regions are obtained by unmarried women.²⁷

Women in urban areas are more likely than their rural counterparts to obtain an abortion. This finding probably reflects that it is easier to obtain abortion services in urban than in rural areas.

The percentage distribution of abortions according to women's education largely reflects the distribution of women, suggesting that some women at all educational levels obtain abortions. Better-educated women may be more successful than those with less schooling in preventing unplanned pregnancy, given their higher levels of knowledge and access to contraception; but they also may have stronger motivation to achieve a smaller family size and to prevent unplanned births, given the greater opportunity costs for them (in terms of employment and income). Younger, educated women may terminate an unintended pregnancy in order to complete their education or gain work experience before starting a family; older women, who are likely to be less educated, may obtain an abortion because they already have the number of children they want. The complex relationships between education, other demographic characteristics, fertility preferences, contraception and abortion are reflected in the very different patterns of abortion rates according to women's educational attainment that are found in the few countries for which this information is available.

Knowing how demographic, social and economic characteristics relate to induced abortion can be an important means of improving understanding of the circumstances surrounding women's decision to obtain an abortion and, to some extent, of the reasons leading to unintended pregnancy and abortion.²⁸ Consequently, studies of differentials in abortion according to women's characteristics can help to identify subgroups in particular need of services and counseling to prevent unplanned pregnancy. From a broader policy perspective, analyses that demonstrate the extent to which abortion is obtained by women of all socioeconomic and demographic characteristics document the universality of women's need for abortion when faced with an unplanned pregnancy. Despite variations in the level of abortion across subgroups, these results imply that for some women in every context, personal, social or family circumstances make it unacceptable to carry a pregnancy to term.

Appendix

For the following countries, official statistics were obtained from relevant government offices (details are available from the authors): Armenia, Australia, Belgium, Bulgaria, Canada, Chile, Croatia, Cuba, Czech Republic, Denmark, England and Wales, Estonia, France, Germany, Hungary, India, Ireland, Israel, Italy, Japan, Mexico, Netherlands, New Zealand, Norway, Scotland, Slovak Republic, Slovenia, Spain and Sweden. For other countries, detailed sources are listed below.

•Bangladesh: Begum SF et al., Hospital-Based Descriptive Study of Illegally Induced Abortion-Related Mortality and Morbidity, and Its Cost on Health Services, 1991, Dhaka: Bangladesh Association for Prevention of Septic Abortion, 1991.

•Benin: Alihonou E, Goufodji S and Capo-Chichi V, Morbidity and mortality related to induced abortions, African Journal of Fertility, Sexuality and Reproductive Health, 1996, 1(1):58-65.

•Brazil: Misago C et al., Determinants of induced abortion among poor women admitted to hospitals in Fortaleza, North Eastern Brazil, paper presented at the Encuentro de Investigadores Sobre Aborto Inducido en America Latina y el Caribe, Determinantes del Aborto y Factores Asociados, Bogotá, Colombia, Nov. 15-18, 1994.

•Colombia: Special tabulations of data from prospective study of women treated for abortion complications in 36 hospitals in four countries, conducted by Federación Latinoamericana de Sociedades de Obstetricia y Ginecología (FLASOG), 1990.

•Dominican Republic: Paiewonsky D, Determinantes Sociales y Consecuencias del Aborto Inducido en la República Dominicana, Santo Domingo: Patrocinado por la Organización de la Salud, 1993.

•Ethiopia: Family Planning Guidance Association of Ethiopia, Unsafe/illegal abortion in Ethiopia, African Journal of Fertility, Sexuality and Reproductive Health, 1996, 1(1):68-71.

•Finland: Hämäläinen H, Rasimus A and Ritamo M, Abortions, 1993, Helsinki: National Research and Development Center for Welfare and Health, 1995 (in Finnish).

•Guyana: Nunes FE and Delph YM, Making abortion law reform work: steps and slips in Guyana, Reproductive Health Matters, 1997, 9:66-76.

•Kazakhstan: Kazakhstan National Institute of
Nutrition and Macro International, Kazakhstan Demographic and Health Survey, 1995, Calverton, MD, USA: Macro International, 1996; and special tabulations of the Demographic and Health Survey (DHS) data.

•Kenya: Lema V, Kamau K and Rogo K, Experience of women presenting with abortion complications in eight district hospitals, in: Rogo K, Leonard A and Muia E, eds., Unsafe Abortion in Kenya: Findings from Eight Studies, Nairobi: Population Council, 1996, pp. 17-20.

•Korea (South): Korea Institute for Health and Social Affairs,1996 National Fertility and Family Health Survey Report, Seoul: Korea Institute for Health and Social Affairs, 1996.

•Kyrgyzstan: Special analysis of 1997 DHS data.

•Malaysia: Foong CS, A study of characteristics of women seeking induced abortion, Medical Journal of Malaysia, 1982, 37(4):318-321.

•Moldova: Serbanescu F and Morris L, Reproductive Health Survey: Moldova, 1995. Preliminary Report, Atlanta, GA, USA: Centers for Disease Control and Prevention (CDC), 1998.

•Mozambique: Hardy E et al., Comparison of women having clandestine and hospital abortions: Maputo, Mozambique, Reproductive Health Matters, 1997, 9:108-115.

•Nepal: Thapa PJ, Thapa S and Shrestha N, A hospital-based study of abortion in Nepal, Studies in Family Planning, 1992, 23(5):311-318.

•Nigeria: Konje JC, Obisesan KA and Ladipo OA, Health and economic consequences of septic abortion, International Journal of Gynecology and Obstetrics, 1992, 37(3):193-197.

•Peru: Special tabulations of data from prospective study of women treated for abortion complications in 36 hospitals in four countries, conducted by FLASOG, 1990.

•Philippines: Ford NJ and Manlagnit AB, Social factors associated with abortion-related morbidity in the Philippines, British Journal of Family Planning, 1994, 20(3):92-95.

•Romania: Ministry of Health, Institute for Mother and Child Care, and CDC, Division for Reproductive Health, Reproductive Health Survey, Romania 1993: Final Report, Bucharest, Romania, and Atlanta, GA, USA: Ministry of Health and CDC, 1995.

•Russia: Goldberg H et al., 1996 Russia Women's Reproductive Health Survey: A Study of Three Sites—Final Report, Atlanta, GA, USA: CDC, 1998.

•Singapore: Chen AJ et al., Legalized abortion: the Singapore experience, Studies in Family Planning, 1985, 16(3):170-178.

•Sri Lanka: Fernando DN and Rabel RAS, Induced abortion—a hospital based study, Ceylon Journal of Medical Science, 1992, 35:1-5.

•Turkey: Akin A and Bertan M, Contraception, Abortion, and Maternal Health Services in Turkey: Results of Further Analysis of the 1993 Turkish Demographic and Health Survey, Calverton, MD, USA: Macro International, 1996; and special tabulations of DHS data.

•United States: Henshaw SK, Abortion incidence and services in the United States, 1995-1996, Family Planning Perspectives, 1998, 30(6):263-270 & 287; and Koonin LM et al., Abortion surveillance—United States, 1995, Morbidity and Mortality Weekly Report, 1998, 47(SS-2):31-68.

•Uzbekistan: Uzbekistan Institute of Obstetrics and Gynecology and Macro International, Uzbekistan Demographic and Health Survey, 1996, Calverton, MD, USA: Macro International, 1997; and special tabulations of DHS data.

•Vietnam: Hieu DT, Stoeckel J and Tien NV, Pregnancy termination and contraceptive failure in Viet Nam, Asia-Pacific Population Journal, 1993, 8(4):3-18.

•Zambia: Likwa RN and Whittaker M, The characteristics of women presenting for abortion and complications of illegal abortions at the University Teaching Hospital, Lusaka, Zambia: an explorative study, African Journal of Fertility, Sexuality and Reproductive Health, 1996, 1(1):43-49.