Promoting spousal discussion of family planning has frequently been advocated as a viable policy tool for narrowing the gender gap in partners' fertility intentions in developing countries.¹ Discussion between spouses is expected to increase contraceptive use, because a sizable minority of women cite their husband's disapproval of contraception as the reason for nonuse, despite having never discussed family planning with their husband.² Researchers have argued that women who report infrequent discussion may, in fact, wrongly perceive that their partner disapproves of family planning, and may therefore feel inhibited from using a method.³ This line of reasoning is supported by empirical research conducted in a wide range of contexts, which shows that spouses who have discussed the topic are 2-10 times as likely as those who have not to practice contraception.⁴

The widespread assumption, then, is that discussion works to promote contraceptive use by increasing knowledge of partner's attitude. Because discussion has become a focal point for policy-making, determining the validity of this assumption is crucial. Three issues in particular need to be explored.

First, the accuracy of perceptions about spousal attitudes could increase through mechanisms other than discussion,⁵ such as through conversations with third parties or via nonverbal cues. For example, in many Sub-Saharan African cultures, spousal discussion of sexual matters is discouraged, and other persons—commonly, in-laws—act as conduits through which partners can exchange ideas on these topics.⁶ Couples in these cultures may also use other forms of communication, such as certain music, the wearing of specific waist beads, certain demeanors and the preparation of favorite meals, to convey unambiguous sex-related messages to each other.⁷ In the case of contraception, a man's use of a method (e.g., condoms) may itself be a powerful nonverbal indicator of approval. These forms of communication may not be inferior to spousal discussion in transmitting knowledge. Therefore, discussion may improve knowledge of family planning attitudes only when it is more efficient than, or augments the effectiveness of, other forms of communication.

Second, if discussion improves knowledge, the accuracy of reports of partner's attitude should improve with discussion, regardless of whether the partner approves or disapproves. However, an analysis of 1998 Kenya Demographic and Health Survey (DHS) data found that correct reporting of partner's approval increased with discussion, while correct reporting of partner's disapproval decreased.⁸ In the Kenya DHS, 90% of respondents approved of family planning in 1998. In such a context, respondents are unlikely to not know their partner's attitude toward contraception, and even those who are guessing their partner's attitude would likely report approval. Furthermore, because men are generally assumed to be uninterested in family planning, their willingness to discuss such matters may be construed as evidence of a favorable attitude. It is therefore inappropriate to conclude that discussion improves knowledge of a partner's attitude toward contraception from the study of approval alone.

Finally, the mechanisms generating the close linkage between spousal communication and family planning use are poorly understood.⁹ Predictors of contraceptive approval and use, such as high education level and urban residence, are also indicators of egalitarian marriages and spousal communication in general.¹⁰ Hence, accurate reporting of approval by a couple may derive from shared characteristics favoring approval, not from discussion per se.

The anticipated effect of spousal discussion on contraceptive uptake has strong policy relevance for Sub-Saharan Africa, where reported levels of unmet need for family planning—the discrepancy between individuals' sexual and contraceptive behavior and their stated intention to space births or end childbearing—are among the highest in the world. Understanding how spousal discussion affects the accuracy of reported partner attitudes may help policymakers shape programs designed to lower unmet need. Because analysis of DHS data from Kenya yielded counterintuitive findings about the effects of partner discussion,¹¹ we calculate the proportion of women correctly reporting their husband's disapproval for a wide range of Sub-Saharan African countries. We then examine in detail data from Chad, where, unlike Kenya, the proportion of survey respondents approving family planning is relatively low (33% of men and 32% of women),¹² thereby reducing the likelihood that reports of partner approval are biased by social norms. We conduct multivariate analyses to examine the associations between spousal discussion and accuracy of women's reports of their husband's attitude toward family planning, with controls for relevant demographic characteristics.

Data and Methods

We examined recent DHS data for 21 Sub-Saharan African countries in which both men and women were interviewed. Each survey asked respondents whether they approved of family planning (from the options "yes," "no" and "no opinion") and whether their spouse approved or disapproved of couples using a method to avoid pregnancy (or if they did not know). Respondents were also asked how often they had talked to their spouse about this subject in the past year (response options were "never," "once or twice" and "more often"). For the subgroup of respondents whose spouse was also interviewed in the survey, we matched women's reports of their partner's attitude with their partner's self-reports. We then determined, among women whose husband disapproved, the proportion correctly reporting their husband's attitude, by their reported frequency of discussion; for this calculation, women who said they did not know their husband's attitude were excluded from the denominator.*

Using multinomial logit regression analyses of data from Chad, we then examined the association between partner discussion and accuracy of women's reports of their husband's approval status, controlling for women's educational level, residence, age at the time of the survey and duration of marriage—background characteristics that are likely correlated with discussion.¹³ We excluded three couples in which either partner was sterilized, because discussion of family planning in the year before the survey may have been irrelevant, as well as two couples who had missing data for at least one of the background variables.

For couples in which the husband approved of family planning, we examined the effect of discussion frequency (as reported by women) on the likelihood that women correctly reported their husband's attitude (i.e., approval rather than disapproval) and on the likelihood that women did not know their husband's attitude rather than incorrectly reporting it as disapproval. Similarly, for couples in which the husband disapproved of family planning, we examined the effect of discussion frequency on the likelihood that women correctly reported their husband's attitude (i.e., disapproval rather than approval) and on the likelihood that women did not know their husband's attitude rather than incorrectly reporting it as approval. Because one could argue that discussion has occurred only when both partners agree it has, we also examined the association between which partner said discussion had taken place and the reporting accuracy of women whose husband disapproved of family planning.

A methodological limitation of the study is that the DHS measure of discussion does not clarify the nature, content or quality of discussion. Thus, it is impossible to determine whether a reported discussion measures an attempt by one spouse to understand the partner's views, a superficial exchange or even one partner's admonishment of the other for practicing contraception secretly. In addition, the question on approval does not ask respondents whether they are referring to their own contraceptive use or to that of other couples, and whether approval refers to method use for spacing or limiting births (or to specific methods). Another limitation is the survey's cross-sectional design, which obscures the temporal relationship between discussion and contraceptive use, as well as the stability of responses—for example, we cannot tell whether and how "don't know" responses to questions about partner attitude might change over time. Clearly, these measures are hardly optimal. Still, we use them for this study because they are the same measures that have been used to draw policy-related recommendations about discussion.¹⁴

RESULTS

Descriptive Analyses

Table 1 shows the relationship between women's reported discussion frequency and the proportion of women whose partner disapproved of family planning who correctly reported his disapproval, for 21 countries in Sub-Saharan Africa. The pattern, although counterintuitive, is strikingly consistent: In every country, proportions of women correctly reporting their spouse's disapproval were smaller if they had discussed family planning with their husband than if they had never done so.^† In 16 countries, reporting accuracy showed an inverse relationship with discussion frequency, even though these countries had differing contraceptive prevalence and approval rates. However, this analysis could not reveal whether variables other than spousal discussion contributed to reporting accuracy.

Using 1996-1997 Chad DHS data, we compared women's perceptions of their husband's attitude toward family planning with his actual attitude, by frequency of discussion (Table 2). Overall, proportions of women correctly citing their husband's attitude were larger if discussion had occurred than if it had not, regardless of whether the husband reported approval or disapproval. (The proportions of women correctly reporting spousal disapproval differ from those in Table 1 because those data are weighted and exclude couples in which the woman replied "don't know" from the denominator.) Furthermore, the proportion of women who replied "don't know" was considerably smaller among those who had discussed family planning than among those who had not.

Although the Kenya study also showed that accurate reporting of partner approval rose with discussion, the accuracy of women's reports of partner disapproval was lower the more frequently discussion occurred; however, the majority of respondents in that study approved of family planning.¹⁵ In contrast, in Chad, contraceptive approval is not the norm and reporting accuracy increases with discussion, regardless of approval status. Stereotyping—attributing normative attitudes to one's spouse—might influence both sets of findings, because it can lead to correct reports even in the absence of reliable knowledge, thus increasing accurate reporting of approval in Kenya and of disapproval in Chad.¹⁶

Multivariate Analyses

After excluding 98 couples in which the husband reported he had "no opinion" about contraceptive use, we conducted two multivariate regression analyses that controlled for selected demographic variables (Table 3, page 90). The first included only women whose husband approved of family planning and the second included only those whose husband disapproved. Women in both groups had a reduced likelihood of reporting that they did not know their husband's attitude rather than reporting it incorrectly if they had discussed family planning. However, the effects of discussion on the likelihood of reporting spousal attitude correctly (rather than incorrectly) depended on the actual attitude of the husband. Among couples in which the husband approved of family planning, the wife had an increased likelihood of correctly reporting his attitude if she had discussed the subject with him. In contrast, among couples in which the husband disapproved of family planning, the wife had a reduced likelihood of correctly reporting his attitude if discussion had taken place.^‡

The positive association between discussion and correctly reporting partner approval of contraception and the negative association between discussion and answering "don't know" support the notion that as spouses talk, they get to know each other's attitudes. Yet the negative association between discussion and correctly reporting partner disapproval raises serious questions about the validity of this notion, and suggests that stereotyping is not a factor. Stereotyping would increase the level of correct reporting of disapproval in Chad.

We also considered the possibility that respondents who had discussed family planning with their partner may have been more likely than those who had not to project their own family planning attitude onto their partner. But when we added respondents' approval status as an explanatory variable in the regression analysis (not shown), discussion remained positively associated with correctly reporting approval and negatively associated with correctly reporting disapproval.

In addition, we recognized that spouses' reports of whether discussion occurred may not always agree, suggesting that the exchange does not necessarily have the same salience for both partners.¹⁷ Presumably, a conversation that both partners say has taken place would be a more valid indication of communication and more conducive to improving mutual knowledge than a conversation reported by only one spouse. Hence, we included men's reports of discussion in multivariate analyses of couples in which the man disapproved of family planning, limiting the sample to the 420 monogamous couples to ensure that each man was referring to the woman interviewed. Because this step reduced the sample of disapproving men by 40%, creating very small cell sizes, we simplified the discussion measure to a dichotomous one (any discussion vs. none)—an approach supported by the nonsignificant difference between the results for 1-2 and at least three discussions in Table 3. Similarly, we simplified the wife's education measure to a dichotomous one (primary or higher education vs. none).

Table 4 shows the results from two regression analyses using the sample of monogamous couples. In the first analysis, which included only women's reports of whether discussion had occurred, women who reported discussion were less likely than those who did not to correctly say their husband disapproved of family planning or to say they did not know his attitude (rather than that he approved). These results held in the second analysis, which also included discussion data from husbands. However, when both spouses reported having discussed contraception, women were no longer less likely to (accurately) report that their husband disapproved. Nonetheless, jointly acknowledged discussion did not contribute to the accuracy of wives' reports.

DISCUSSION AND CONCLUSION

This study set out to examine whether spousal discussion of family planning could be linked to women's increased knowledge of their husband's attitudes. We find that discussion, as measured in the Chad DHS, raises the likelihood that women correctly report partner approval but not disapproval. This finding suggests that discussion does not necessarily improve knowledge about a partner's contraceptive attitudes.

We suspect that correct reporting of disapproval does not increase after discussion because if a husband is willing to discuss family planning, his wife may interpret this as approval of family planning. Rather than providing a clear indication of partner attitude, discussion itself may lead women to assume their spouses approve of contraceptive use. If only correct reports of approval are analyzed (as in some past studies), discussion would indeed appear to impart knowledge exchange, especially if the husband initiated or participated in a discussion on contraception in order to use a method.

Furthermore, some respondents probably find it difficult, after reporting that there has been a discussion on contraception, to acknowledge ignorance of a partner's contraceptive attitudes. These respondents may report a partner's attitude, even if they are unsure. Such an explanation is consistent with the reduced likelihood of answering "don't know" among couples who have discussed family planning.

Our results cast doubt not only on the notion that discussion imparts knowledge, but also on the influence of stereotyping and projection. Stereotyping cannot be a complete explanation because, in Chad, where about two-thirds of husbands disapprove of family planning,¹⁸ guesses made according to stereotypical assumptions would have substantially reduced the likelihood of accurate reporting of approval. Projection seems irrelevant because discussion was still associated with accuracy in describing husbands' approval, regardless of women's own attitudes. We suggest that women who have discussed family planning with their spouse are more likely than those who have not to think that their partner approves, whether or not he actually does.

One might argue that the relationship between discussion and reporting accuracy is a nonissue: If discussion makes women think their husband approves of contraception (even if incorrectly), it might lead to contraceptive use on women's part. Yet the literature provides evidence that women's adoption of family planning potentially puts them at risk of adverse outcomes (e.g., as victims of partner violence) when their husbands are opposed to contraception.¹⁹ Contraceptive discontinuation is also likely to be high for women who start using a method on the assumption that their partner approves of use and later discover he actually does not.

Perhaps we should not be overly hasty to discount women who say their husband impedes contraceptive use, while also reporting that no relevant discussion had occurred. Reports of spousal disapproval may in fact be more accurate among women who have not discussed family planning with their husband than among those who have done so, because of reliance on nonverbal and other forms of communication in some cultures. In such cultures, discussion may even confuse perceptions of partners' true attitudes. Hence, researchers need to be cautious about construing the relationship between perception of partner preference and discussion as a simple and causal one.

In addition, it is clear that we need to more fully understand the relationship between discussion and communication before implementing any policy designed to lower fertility by increasing spousal discussion. Better measures of communication than the limited discussion variable available in the DHS can be constructed using insights gleaned from qualitative data.²⁰ Similarly, longitudinal surveys would greatly enhance our understanding of the dynamics and implications of partner communication.

Although we have shown that spousal discussion does not necessarily increase mutual knowledge about contraceptive attitudes, we should not discount the value of discussion in general. Discussion appears to promote contraceptive use less ambiguously if construed more broadly than interspousal discussion. In England, public discussion of contraceptive methods during the Besant-Bradlaugh obscenity trial has been argued to have promoted fertility decline.^§21 In Sub-Saharan Africa, discussion of contraception between women and their sisters-in-law supports covert contraceptive use²² and promotes both spousal discussion and overt use.²³ Discussion within social networks also promotes contraceptive use,²⁴ and Phillips and colleagues argue that discussion plays an important role in legitimizing uptake in settings with low contraceptive prevalence.²⁵ Communications research has shown that mass media interventions work by stimulating discussion within social networks (including but not limited to discussion between spouses), which then leads to subsequent contraceptive uptake.²⁶ Therefore, context-specific understanding of discussion in social networks and between spouses should be useful in policy formation.

In conclusion, confined to the interspousal level, discussion may cause wives to perceive their husbands as more accepting of family planning than they actually are. Therefore, any anticipated reduction in unmet need for contraception through improvements in spousal discussion may be overstated. Policymakers need to bear this in mind when searching for efficient ways to reduce unmet need without potentially increasing the risk of spousal violence.