There are very few couple-level studies of communication and decision making regarding contraceptive use in the Middle East. In Turkey—a predominantly Muslim nation that has a secular democracy and is undergoing considerable social change—fertility has declined significantly, men play a major role in contraceptive practice and women's position in society is increasingly contested. These factors make Turkey a particularly interesting country for couple-level studies.

Traditionally, the measurement of contraceptive use has been based solely on women's self-reports of current use. Although this is methodologically convenient, the individual and subjective experiences of partners may differ. A more comprehensive measure of a couple's contraceptive use requires information from each partner. Moreover, because contraceptive practice and fertility are interrelated, the married couple is the appropriate unit of analysis, especially in societies such as those in the Middle East, where the vast majority of contraceptive use and almost all births occur within marriage.

For this region, only a few fertility and family planning surveys have collected data from men or couples. The 1992 Egyptian Demographic and Health Survey (DHS) included a subsample of husbands, and the 1992 Moroccan DHS interviewed married and unmarried men aged 20–70, but we could not identify any couple-level studies using these data. The 1985 Jordan Husbands' Fertility Survey interviewed the husbands of women who were currently married when interviewed in the 1983 Jordan Fertility and Family Health Survey. Although one study presented some comparisons of spousal reports about contraceptive knowledge and family planning attitudes from these two Jordanian surveys,1 the comparisons were purely descriptive and limited in scope, and responses for husbands and wives were from different survey years. The 1988 Turkish Population and Health Survey collected limited data from men. A comparison of spousal reports revealed a high percentage of agreement about fertility measures, but only about three-quarters of the women knew their husbands' age and level of education (72% and 78% agreement, respectively, between spouses).2 A decade later, the 1998 Turkish DHS collected detailed couple-level data, including data on gender expectations and attitudes from each partner, which are analyzed in the current study.

This study has three objectives. First, it assesses the extent of interpartner concordance on self-reported aspects of reproductive decision making and behavior. Such an assessment may permit verification of the accuracy of either partner's report and help assess reliability. Second, we examine the influence of husband-wife communication and interspousal power relations on couples' current use of contraceptives, and attempt to determine if a joint decision-making model is becoming the norm. Third, after adjusting for background demographic and socioeconomic factors, we compare results from models of the relative influences of spousal attitudes and preferences, effective partner communication and gender-based power differentials on current contraceptive use, based on husbands', wives' and joint reports of use. Such an assessment is needed in the Middle East, particularly as it moves toward a more advanced stage of fertility transition. Studies on these topics are generally lacking for countries with intermediate fertility levels, defined by the United Nations as having a total fertility rate between 2.1 and 5.0 children per woman.3


Couple-Level Concurrence About Contraceptive Use

The level of spousal agreement regarding fertility and family planning remains an area of scholarly contention. Although high concurrence would be expected because of daily partner contact and common living conditions, cross-national studies of couple concurrence on contraceptive method use show frequent discrepancies between husbands' and wives' reports.4–8 In one study, the median level of partners' concordance on a range of reproductive measures in DHS countries was 76%.6 Discrepant responses may result from measurement error, social desirability biases, wives' covert use of contraceptives, and disagreement and miscommunication over spousal fertility and family planning preferences.7–10 Moreover, decisions about contraceptive use and childbearing may be compounded by unequal power relations, especially in more patriarchal societies.11,12

Until recently, researchers largely ignored the mismatches between the responses of husbands and wives and invariably assumed women's responses to be more accurate. However, the importance of assessing data from both partners has been demonstrated.8,13–15 The 1994 International Conference on Population and Development reinforced efforts to consider men in family planning and reproductive health programs and research, an imperative underscored by the HIV/AIDS epidemic. Routine data collection from men as well as women in Sub-Saharan African DHS surveys has given rise to a number of studies of spousal communication about family planning and fertility, some of which have adopted the couple as the unit of analysis.16,17 Research in Asian societies has also underscored the need to consider the preferences of both spouses.18–20 These surveys cannot be considered representative of other regions, however, particularly those where couple-level data have yet to be analyzed.

The Turkish Context

Turkey is a low-to-middle-income country located at the intersection of southeastern Europe and the Middle East. Its population was estimated at 71 million in 2007, with an annual growth rate of 1% and an adult literacy rate of 88% (80% for women).21,22 The total fertility rate declined from 4.3 children per woman in 1978 to 2.6 children in 1998;23 it fell further to 2.3 children in 2003, as Turkey moved through the middle phase of its fertility transition.24 The contraceptive prevalence rate among married women plateaued at 64% between 1988 and 1998, and rose to 71% in 2003 (Figure 1). Because male methods account for half of all contraceptive use, it is plausible to assume that a husband's report of method use is at least as reliable as that of his wife. Factors associated with the use of withdrawal, condoms and other methods have been examined elsewhere.25–28

Despite the legal codification of gender equality following the foundation of the Turkish republic in the 1920s, various laws and social traditions have reinforced the subordinate position of women. Although Turkey does not have the rigid gender stratification found in most Middle Eastern and North African societies, violence against women—including honor killings—remains a major problem,29–33 and the Turkish parliament recently strengthened laws to protect women's rights.34 In light of increased levels of education and economic development and legal reforms that support greater gender equality, one might expect spousal reports to be quite similar. However, Turkish culture and socioeconomic structures remain male-dominated. Consequently, many women may still need their husband's approval to use family planning, and husbands' attitudes and intentions may have additional effects on couples' contraceptive practice and fertility. Therefore, it is pertinent to ask whether Turkish couples have increasingly adopted joint decision making regarding family planning and fertility issues, or whether a stricter patriarchal system prevails.

The 2003 Turkey DHS did not collect data from men or couples, unlike the 1998 survey data examined here. A study on the consistency of couples' reports about contraceptive use and fertility preferences using the earlier DHS has been published, but it presented only measures of percentage agreement between spousal reports,35 which by itself is a crude measure of consensus. The current study also computes kappa values, which indicate the degree of agreement net of that expected by chance alone, and presents multivariate analyses using couple-level data to evaluate the influence of spousal communication and interspousal power relations on family planning practice.



The 1998 Turkish DHS surveyed 8,576 women aged 15–49, representing 91% of identified eligible women. The survey included 6,196 ever-married women and a subsample of husbands, drawn from about half of all surveyed households. Of the 3,043 husbands of currently married women who were identified as eligible for the survey, 1,971 were successfully interviewed.23 We matched data for these husbands with responses from their wives. The main reason for nonresponse was husbands' absence at time of interview and at repeated visits. However, differences in age, education levels and place of residence between respondents and nonrespondents were not statistically significant,35 and the total number of matched couples is higher than that for most DHS-based, couple-level studies.7

From the initial sample of 1,971 paired couples, we excluded 65 couples who reported any polygamous unions and 156 couples in which one or both partners had been married more than once. Their exclusion ensured that partners' responses would correspond to the same marital union. We further limited the sample to fecund couples in which the woman was not pregnant, because our focus was on the determinants of current contraceptive use. This led to the exclusion of 77 sterilized couples (in almost all cases because the women had been sterilized) and 127 couples who reported a current pregnancy. Thus, our final sample consisted of 1,546 matched couples for whom data were available on all variables of interest.


The social and demographic factors controlled for were age, number of living children, ethnicity, education, household wealth index, region of residence and rural-urban residence. The age variable was divided into six categories using each partner's self-reported age: husband younger than 30 and wife aged 30–39, husband aged 30–39 and wife younger than 30, husband aged 30–39 and wife 30 or older, husband 40 or older and wife younger than 40, husband aged 40–49 and wife 40 or older, and husband 50 or older and wife 40 or older (the reference category). The number of children ever born and the number of living children were treated as continuous measures. Ethnicity, a politically sensitive issue in Turkey, was determined indirectly by asking respondents about their mother tongue; couples were characterized as both Kurdish or at least one spouse not Kurdish. DHS data provide the only recently available figures on the country's ethnic composition.36

Educational attainment consisted of three categories: none or incomplete primary (uneducated), primary or incomplete secondary (basic), and secondary or higher. We categorized couples' education levels as follows: both uneducated, wife basic and husband uneducated or basic, husband basic and wife uneducated, husband or wife secondary or higher, and both secondary or higher. To gauge household economic status, we constructed a household wealth index (divided into quintiles) based on the DHS approach.*37,38 Rural-urban residence is a binary variable, and region of residence refers to the fivefold division of the country used in most social surveys. The West is the most populous, densely settled and socioeconomically advanced region; the East is the least socioeconomically developed and has the highest proportion of Kurds.

Among the intermediate variables, the number of contraceptive methods known by each partner was modeled as a continuous variable, because nearly all Turkish wives and husbands know of several methods. Attitudinal data were collected by asking respondents if they approved of couples using family planning and what they perceived as their spouse's attitude toward family planning. Respondents were also asked about their ideal family size (numerical responses were provided in nearly all cases) and if they had discussed family planning with their spouse, other relatives, neighbors or friends in the last few months. Hence, three dimensions of effective partner communication that may affect the ability of individuals and couples to negotiate contraceptive use were assessed: spousal agreement on approval, discussion between partners and spousal perception of the partner's approval of family planning. These variables are in turn related to the couple's ideal family size.

Family planning research has tended to assume women's primacy as decision makers regarding contraceptive method choice, although male prerogative and control are receiving increasing attention (e.g., in condom use and microbicide acceptability studies). We therefore constructed two indices that assess gender relations, that is, the prevailing system of socially constructed roles, norms, privileges and status between men and women: acceptance of male prerogative and of wife-beating. Separate additive indices were created from wives' and husbands' responses to the same questions; inclusion of both perspectives permits better understanding of interspousal perceptions of power differentials. Acceptance of male prerogative was assessed by asking whether respondents agreed or disagreed with four statements: that important family decisions should be made by men, that men are usually wiser than women, that a woman should not argue with her husband and that it is better for the male child to be educated than the female child. Acceptance of wife-beating was based on whether respondents agreed that the husband had a right to beat his wife in three situations: if his wife argued with him, if she talked with other men or if she spent money needlessly. Greater gender-based power differentials may be associated with lower contraceptive use; conversely, joint decision-making norms tend to characterize societies with high contraceptive prevalence and low fertility. A basic tenet in reproductive health program management and policy making is that joint decision making helps realize women's empowerment and health goals.

In our multivariate analyses, these two measures were modeled as dummy variables (i.e., for each index, whether respondents agreed with at least two of the statements). The outcome variable—current contraceptive use—refers to both modern and traditional methods used in the past year. Wives' and husbands' reports were treated as separate outcomes. A further aim of the study, therefore, is to assess if similar predictors of contraceptive use result from models based on the three sets of regression analyses.


We first assessed spousal concordance on selected survey questions: whether the wife was pregnant, contraceptive use while the wife was pregnant and knowledge of condom use. We also determined the degree of concordance between spousal reports on other variables, including specific fertility and family planning items. We calculated both the percentage of couples whose responses were in agreement and the kappa statistic.39 Kappa values may range from –1 to 1; a value of 0 indicates that agreement could occur by chance, and negative values indicate disagreement. Positive values of 0–0.20 suggest slight agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement and 0.61–0.80 substantial concurrence. Higher values indicate almost perfect agreement beyond chance.39,40

Multiple logistic regression was used to examine the influence of explanatory variables on current contraceptive use. After controlling for the effects of demographic and socioeconomic variables, we hypothesized, couples' contraceptive use is positively associated with the contraceptive knowledge of both spouses, as well as with their approval of family planning, spousal communication about family planning and husband-wife agreement regarding fertility preferences. We further hypothesized that gender-based power imbalances may constrain women's negotiation ability and affect patterns of contraceptive decision making.

We computed and compared separate multivariate models using wives' and husbands' reports of current contraceptive use, and show two sets of models for each. The first set included all the intermediate variables, and the second set also adjusted for background variables whose influence on contraceptive use has been well established in previous research. We then ran a combined model based on 1,250 couples for whom concordant reports were obtained, with the dependent variable coded as 1 if both partners reported contraceptive use and as 0 if neither reported use.


Concurrence Between Spousal Reports

Independent sources of information, such as biomarkers of semen exposure or clinic records, were not available to validate respondents' self-reports of contraceptive use. Therefore, we examined consistency of reporting both in the overall couples data set and in the study sample. Of all 1,906 monogamous couples in the data set, a husband reported current contraceptive use while his wife was pregnant in only 18 cases. This rate is lower than that for 23 DHS countries studied elsewhere6 and suggests good overall data quality. The validity of other survey items cannot be readily assessed, but measurement accuracy can be assessed by testing the reliability of both partners' re- sponses to other questions. Although some inconsistencies between husbands' and wives' reports were found, overall they were minor compared with those observed for most DHS surveys. In 92% of the cases in which wives said they were currently pregnant, husbands also reported the pregnancy. In 80% of the cases of disagreement, the wives were in the first four months of pregnancy.

Concurrence tests run on the study sample of 1,546 couples indicated generally good agreement between spousal reports. Furthermore, among respondents who said they did not know of condoms, none reported condom use. However, 68% of husbands and 83% of wives claimed to know of condoms, yet matched responses indicated that in only 59% of the couples did both partners know about this method. The kappa value for this level of agreement was poor (0.12).

Spousal reports of the two fertility measures showed high agreement and kappa values (93–99% and 0.93– 0.99, respectively; Table 1), whereas these values for the number of contraceptive methods ever used were low (38% and 0.25, respectively). For the last measure, some discordant responses may be due to sexual relations with other partners either before or outside of marriage. Reports of current method use were slightly lower among husbands than among wives (68% vs. 72%), but the level of agreement was 81% and the kappa statistic (0.56) suggests moderate concurrence beyond chance.

Discrepancies were also found in partners' attitudes toward and discussion of family planning. In 83% of couples, partners held similar attitudes toward family planning (85–90% of individual respondents expressing approval), although the kappa value (0.23) suggests only fair agreement. Spousal perceptions of each other's attitudes toward family planning showed similar levels of concordance with their spouse's actual attitudes. The percentage agreement for discussing family planning with one's spouse was 90%; however, the low kappa value (0.09) indicates that chance identical responses among spouses were high. Few discussions were reported (5–7%), but the reference period for this question was limited to the last two months rather than the past year, as is customary in DHS surveys.

Sample Characteristics

The level of agreement between spousal reports on specific fertility and family planning items was relatively high. Knowledge of contraceptive methods was almost universal in all subgroups of background characteristics. In 98% of couples, the wife and husband knew of at least one method; on average, wives knew of seven methods and husbands at least six (Table 2). About nine out of 10 wives and husbands approved of family planning, and almost identical proportions (83% and 82%, respectively) thought their spouses approved of family planning. Overall, 74% of wives and 77% of husbands accurately reported their spouse's approval; in 77% and 78% of these cases, respectively, spouses' reports were in agreement (i.e., both approved or both disapproved; not shown).

Levels of reported discussion of family planning were low, and were probably related to the unusually short reference period for this item. Such discussions were rare among spouses (5–7%), and more frequent with other relatives, neighbors or friends (23–27%). Given the low level of spousal discussion and the low kappa value noted earlier, this variable was dropped from the multivariate analyses. Wives and husbands had similar preferences regarding ideal family size, with high proportions (84% and 82%, respectively) wanting three or fewer children.

The two measures of interspousal power relations showed that about half or more of all wives and husbands agreed with at least one statement of male prerogative and agreed that in some situations, husbands have a right to assert authority through physical violence. Wives were only slightly less likely to agree with these propositions. Acceptance of wife-beating is more directly linked to interspousal power relations than is male prerogative, and it was therefore included in the regression analyses.

Spousal agreement about family planning and ideal family size was also high. In 79% of married couples, both spouses approved of family planning, and in 4% neither partner approved; in the remaining 17% of couples, only one partner approved (Table 3). Seventy-three percent of couples wanted three or fewer children, and overall 80% agreed about their fertility preference. Husband-wife inconsistencies were primarily attributable to the wife wanting fewer children than her husband would like.

Multivariate Findings

In the first model in Table 4, the outcome variable was wives' reports of contraceptive use; the model included intermediate variables only. The odds of current use rose with the number of methods known by the wife (odds ratio, 1.2). Couples in which the husband alone approved of family planning and those in which the wife perceived that her husband approved also had higher odds of current use (2.3 and 3.1, respectively). Acceptance of wife-beating was of borderline significance (1.1).

In the final model, which added background factors, the number of contraceptive methods known by the wife and the wife's perception that her husband approved of family planning were the only intermediate variables that retained significance (odds ratios, 1.2 and 3.3, respectively). As expected, contraceptive use was more likely among younger couples than in the oldest age-group (3.0–11.5), and was positively associated with number of living children (1.4). Couples in which both spouses were Kurdish, were uneducated or were living in the East had decreased odds of using a method (0.4–0.6).

The initial model using husbands' reports of contraceptive practice as the outcome showed that both the number of contraceptive methods known by the husband and his perception that his wife approved of family planning were associated with an increased likelihood of contraceptive use (odds ratios, 1.1 and 2.1, respectively), similar to findings from the wives' initial model. Couples in which one or both spouses approved of family planning had higher odds of using a method than if neither approved (5.0–8.9). Compared with couples in which both spouses wanted three or fewer children, those in which only the husband wanted three or fewer or both wanted more than three had lower odds of contraceptive use (0.6 each). The husband's acceptance of wife-beating was not associated with contraceptive use.

In the husbands' final model, the odds of use rose with the number of methods he knew (odds ratio, 1.1). Approval of family planning by one or both spouses was associated with higher odds of use than disapproval by both (3.8–5.8), and the husband's perception that his wife approved was associated with higher odds than a perception that she disapproved (2.0). Measures of ideal family size did not retain significance in this model. Except for residence in the East, the same background factors that were significantly correlated with contraceptive use in the wives' final model were also significant here, with the addition of being in either the lower-middle or upper-middle household wealth quintile.

Our last set of regression analyses combined spouses' reports on the same covariates, using concordant joint reports on contraceptive use as the dependent variable. We excluded knowledge of contraceptive methods and perception of the spouse's approval of family planning, because separate spousal reports for these items cannot be readily combined as couple variables. When both intermediate and background factors were included, couples in which both spouses approved of family planning had higher odds of practicing contraception than if only one or neither of them approved (odds ratio, 2.4—Table 5, page 134). Furthermore, couples who both wanted more than three children, and couples in which only the wife wanted three or fewer, had lower odds of using a method than did couples in which both spouses wanted three or fewer children (0.4 and 0.6, respectively). In this model, acceptance of wife-beating by either spouse was not associated with contraceptive use.

Current contraceptive use by couples was associated with age: Compared with the oldest spouses, all other age categories had elevated odds of using a method (odds ratios, 3.5–14.6). Method use was positively associated with the number of living children (1.5), higher household wealth index (2.2–3.4) and residence in the Central region (1.7). In contrast, contraceptive use was negatively associated with both spouses being Kurdish, both spouses being uneducated, living in the East and urban residence (0.2–0.6).


This study adds to the growing body of couple-level research from around the world and provides more evidence that information from one spouse cannot fully represent the views of the other partner, nor of the couple. Many problems and unresolved issues remain regarding the validity and reliability of either partner's sole reports, as well as of couples' data. The challenge remains to build on existing work and to improve the measurement, modeling and use of couple-level data.

In 1998, Turkish husbands reported a contraceptive prevalence rate that was 4% lower than that of their wives. The net difference is relatively small compared with that found in many other developing countries. However, its direction differs from the more common pattern, in which husbands report higher levels of use than do their wives because they report more use of condoms or periodic abstinence.7,10 Compared with their wives, Turkish husbands underreport use of condoms and withdrawal (though wives report less pill use and more IUD use than do their husbands). Studies from many countries indicate that use of withdrawal is frequently underreported.41,42Factors that predict the choice of withdrawal over other contraceptive methods in Turkey have been examined elsewhere.25–28

As in other studies, spouses' proxy reports on factual variables—including fertility-related items such as number of living children—showed much stronger concordance than that found for attitudinal variables, which, by definition, are more subjective. However, the widely used kappa statistic may be affected by the prevalence of the observed findings, as well as of discordant responses, and it is not unambiguously clear what constitutes good agreement. This study has employed perhaps the most widely adopted grading system of kappa, but all existing scales are based on arbitrary subdivisions.40,43 Future work should consider the suitability of alternative indices of agreement for couple-level studies of reproductive decision making.

Our multivariate analyses excluded the measure of whether family planning was discussed because of the low reported prevalence of this behavior and its low kappa value, doubtless related to the unusually short time frame considered. In actuality, such discussion is almost certainly far more common, not least because one in four Turkish couples rely on withdrawal, a method that requires active spousal cooperation. Additionally, this measure referred only to the frequency of discussion, not to its actual content, nature or quality. Although there may be some theoretical ambiguity in the relationship between contraceptive use and husband-wife agreement and communication regarding attitudes and preferences, other communication variables were associated with contraceptive use after background variables were controlled for. Such influences were not found to be moderated by interspousal power relations, although our measures were imperfect. A longitudinal analysis has confirmed cross-sectional findings that spousal communication is associated with greater contraceptive use.44

For both initial multivariate models based on separate spousal reports, intermediate variables that had positive associations with current contraceptive use included the couple's knowledge of contraceptive methods, the husband's approval of family planning and each spouse's perception of the other's approval. Wives' acceptance of wife-beating was marginally significant, but not in the final model based on their reports. After controlling for background variables, the approval of family planning retained significance only in the model based on husbands' reports of contraceptive use. However, this is an attitudinal variable for which the kappa value suggested only fair agreement when accounting for chance. Other than residence in the East region and being in either the lower-middle or upper-middle wealth quintiles, the same background factors were found to be associated with contraceptive use in the final models based on separate spousal reports.

To some extent, our results may reflect a high social desirability of reporting joint decision making and not necessarily actual joint decision making. They also cannot resolve whose reports and preferences are more valid and influential in reproductive decision making. However, consideration of separate and combined models using reports from each spouse may help detect inconsistencies between spousal reports and lead to insights not attainable in simpler, single-partner formulations, which may presume differences between spouses and genders when none exist. Although spousal reporting differences regarding family planning were not major, neither were they inconsequential. Other researchers have proposed guidelines as to which set of reports (those of wives, husbands or both) should be used to obtain the best estimate of contraceptive use,7,45 but guidance varies. The resulting inconsistency and uncertainty regarding such rules further underscore the need for more research on this topic.

Several factors may account for the difference between the results of the combined multivariate model and those from the separate models. We could not include two of our original spousal communication variables and we excluded discrepant reports. Although our two dyad members (husband and wife) are distinguishable, they are not strictly independent. Marriage implies that two individuals come together to form a couple, and they tend to stay married because they see life in a similar way; consequently, they are likely to seek a consensus on matters of mutual importance. More broadly, statistical nonindependence in dyads may be due to compositional, partner, mutual influence and common fate effects,46,47 all of which are observed within marriage. For example, compositional factors refer to the similarity of married couples across a range of variables. In Turkey, where kin-based structures remain strong, assortative mating may be more likely than in the more open, individualistic societies of Europe or North America. In addition, standard statistical methods developed for the study of individuals to address issues of stability, consistency and correlations between related quantitative measurements may not be well-suited for analyzing couple-level data.47 In the absence of greater conceptual and methodological clarity, it is not surprising that most researchers continue to use individual-centered models and statistical tools.

Treating dyadic data as if they are two samples is less than optimal, however. The phenomena we wish to study are interpersonal by definition, and our measures of contraceptive use and fertility reflect the contributions of two persons and their resolution within the couple. A key problem that remains is how to conceptualize the respective influences of each spouse on the other, while at the same time modeling the statistical interdependence that exists between partners. Much more work is needed in this area to better conceptualize and solve long-standing analytic problems in couple-level studies of contraceptive use and reproductive decision making.

Several questions in this study evaluated the influence of interspousal power relations on contraceptive use. About half of all Turkish wives and husbands accept that a husband is entitled to beat his wife should she not obey him. It may seem surprising, therefore, that our measures of power relations were not associated with contraceptive use in any of the final models. This may indicate that gender dynamics are moving toward greater equilibrium between the sexes. Although researchers often portray marriage as a power game, many spouses may negotiate with each other to resolve issues. However, our variables do not capture all the dimensions of power relations. Gender inequalities are strongly evident in the sexual sphere in Turkey, where premarital sex is disapproved for women but tolerated and even expected for young men.48

To some extent, gender inequality may be less relevant in contraceptive decision making in Turkey, given the relatively strong tradition that husbands assume some responsibility for contraception.25,49 This may partly counterbalance other gender effects arising from a husband's generalized power in marriage against the "female sphere" of influence. However, the degree of involvement by Turkish husbands in contraceptive use should not be exaggerated, as evidenced by the very low prevalence of vasectomy (less than 0.1% of husbands) as compared with tubal sterilization (about 4% of wives). Furthermore, the two variables used to capture gender-based power differentials had the expected negative associations with contraceptive use in bivariate logistic regressions, with the results highly significant in all cases (p<.01; not shown). Presumably, both of these variables are highly correlated with education level and perhaps some other factors, which reduces their significance in our multivariate logistic regression models.


Discrepant spousal reports are not unusual in Turkey, but they do appear to be less significant than in most low-to-middle-income countries with DHS data. We did not find evidence that interspousal power relations play a significant explanatory role in contraceptive use. Although our measures are imperfect, we conclude that most Turkish couples appear to have moved toward a more egalitarian mode of reproductive decision making than that believed to characterize other Middle Eastern countries. The views of both spouses matter and contribute to a fuller understanding of how couples decide on their contraceptive practice. Further research is needed to better conceptualize and assess longstanding analytic problems in couple-level studies.